Salinity-induced mixed and barrier layers in the southwestern tropical Atlantic Ocean off the northeast of Brazil

. High-resolution hydrographic observations of temperature and salinity are used to analyze the formation and distribution of isothermal depth ( Z T ), mixed depth ( Z M ) and barrier layer thickness (BLT) in a section of the southwestern Atlantic (0 ◦ 30 0 N–14 ◦ 00 0 S; 31 ◦ 24 0 –41 ◦ 48 0 W), adjacent to the northeastern Brazilian coast. Analyzed data consists of 279 CTD casts acquired during two cruises under the Brazilian REVIZEE Program. One occurred in late austral winter (August–October 1995) and another in austral summer (January–April 1997). Oceanic observations are compared to numerical modeling results obtained from the French Mercator-Coriolis Program. Results indicate


Introduction
The southwestern tropical Atlantic Ocean is a region of prime importance to global climate change. It is an area through which oceanic signals, from intra-seasonal to decadal scales, must pass (Dengler et al., 2004;Schott et al., 2005). Moreover, this region is subjected to cyclonic and anticyclonic gyres strongly controlled by surface winds (Stramma and Schott, 1999;Lumpkin and Garzoli, 2005). These gyres drive the divergence of the southern pathway of the South Equatorial Current (sSEC) (Rodrigues et al., 2007;Silva et al., 2009), which is partially at the birth site of several current systems that flow along the Brazilian coastline. The northern branch of the sSEC termination flows northward, forming the North Brazil Current/North Brazil Undercurrent (NBC/NBUC) system. This powerful western boundary current contributes to the northward Guyana Current Stramma et al., 2005) and to the eastward North Equatorial Counter Current (NECC). It also feeds its associated complex retroflection system (Goes et al., 2005), as well as the eastward Equatorial Undercurrent (EUC). It is believed Published by Copernicus Publications on behalf of the European Geosciences Union. that NBC accounts for approximately one-third of the net warm-water flow transported across the equatorial tropical gyre boundary into the North Atlantic; this partially compensates for the southward export of North Atlantic Deep Water (NADW) (Dengler et al., 2004;Schott et al., 2005). The southern branch of sSEC forms the Brazil Current (BC), which flows poleward along the Brazilian coast.
Offshore, the upper ocean density stratification is primarily controlled by temperature variations in the thermocline depth. However, there is some evidence that salinity variations can regulate the mixed layer (e.g., Sprintall and Tomczak, 1992). The occurrence of the isohaline layer being shallower than the isothermal layer in the tropical ocean has been studied since the Meteor research cruises in 1936 (Defant, 1961). The barrier layer (BL), which is a layer between the halocline and the thermocline (Lukas and Lindström, 1991), may isolate the upper isohaline layer from the cold thermocline waters, thereby affecting the ocean heat budget and its exchanges with the atmosphere (Swenson and Hansen, 1999;Pailler et al., 1999;Foltz and McPhaden, 2009). Heavy precipitation in the western portion of a tropical ocean basin is the main mechanism accounting for BL formation (e.g., Sprintall and Tomczak, 1992;Ando and McPhaden, 1997;Tanguy et al., 2010). When the BL occurs, the energy transferred from the atmosphere to the ocean, by wind and buoyancy forcing, may get trapped in the upper mixed layer, limited by the salinity stratification (e.g., Delcroix and McPhaden, 2002). This stratification is thinner, and theoretically more reactive, than the one defined by the temperature mixed layer (Vialard and Delecluse, 1998;De Boyer Montégut et al., 2007).
In the western Pacific, the thick climatologically important BL is due to the eastward fresh jets flowing over subducted salty waters (Vialard and Delecluse, 1998). Cronin and McPhaden (2002) analyzed BL responses to westerly wind gusts over the tropical Atmosphere Ocean (TAO) array in the equatorial Pacific, and they discussed the main mechanisms by which BLs may form and spread. General charts of the seasonal variability of BL thickness in tropical oceans were obtained using Levitus climatological data (Levitus, 1982). These charts were used to determine the extent of the influence of salinity in the depth of the mixed layer (Sprintall and Tomczak, 1992). More recently, De Boyer  and Mignot et al. (2007) used instantaneous T/S profiles, including Argo data, to construct global ocean climatology of monthly mean properties of the BL phenomenon. In the western tropical Atlantic, these charts suggest that the subduction of subtropical salty surface waters toward the equator during winter, and their advection by the equatorial current system, might be the mechanism that forms the BL. This hypothesis is partially supported by the presence of subsurface Salinity Maximum Waters (SMW), as called by Defant (1936), along the western Atlantic boundary (Pailler et al., 1999;Stramma et al., 2005;Silva et al., 2005). These authors argue that this salty water, which is formed in the South Atlantic subtropical gyre, enters the region within the NBUC that flows northwestward along the Brazilian continental slope. The SMW is also known as Subtropical Underwater (Lambert and Sturges, 1977), which is characterized by a salinity maximum at about 100 m depth, at densities slightly below σ θ = 25.0.
As stated above, numerous studies about BLs have been reported in the literature for western equatorial Pacific and Indian Oceans (e.g., Sprintall and Tomczak, 1992;Delcroix and McPhaden, 2002). However, little has been done to identify formation of salinity-induced BLs at the southwestern boundary of the tropical Atlantic. This is the main topic of the present study. Our investigation is based on the analysis of hydrographic vertical profiles obtained under the Brazilian REVIZEE Program (Medeiros et al., 2009a, b) off the Northeast of Brazil. The paper is outlined as follows: data and methods are delineated in Sect. 2, where the area of study is presented along with the criteria used for determining the isothermal, mixed and barrier layers; results are presented in Sect. 3; this is followed by a summary and conclusions in Sect. 4.

Data collection and numerical results
The area of study comprised a section of the western tropical Atlantic (0 • 30 N-14 • 00 S; 31 • 24 -41 • 48 W), adjacent to the Brazilian NE coast ( Fig. 1). High-resolution Conductivity, Temperature, Depth (CTD) hydrographic data were collected onboard the R/V Antares under the REVIZEE Program (Brazilian Program for Assessing the Sustainable Potential of the Live Resources of the Exclusive Economic Zone). The dataset gathered during this program (Medeiros et al., 2009a, b) comprised 279 continuous CTD casts −146 were obtained during the late austral winter (2 August 1995-26 October 1995 and 133 during the austral summer (20 January 1997-17 April 1997.
Vertical and continuous profiles of temperature and salinity were obtained using a Sea Bird Electronics SBE911plus CTD probe. It was equipped with a centrifugal pump and high-resolution sensors for conductivity, temperature and pressure measurements. During all of the cruises, the CTD was operated at up to a depth of around 500 m, with a descending speed of 1 m s −1 and a sampling frequency of 24 Hz. The equipment was connected to an SBE 11plus boarding platform during the profiling, which permitted realtime monitoring of data acquisition and quality control.
Oceanic observations were compared to numerical modeling results obtained from the French Mercator-Coriolis Program (http://www.mercator-ocean.fr). The numerical product used herein (PSY3V2) is the result of the global ocean model OPA (ORCA025 configuration) with 1/4 • of horizontal resolution and 50 levels of vertical grid. It is refined at Ocean Sci., 7, 63-73, 2011 www.ocean-sci.net/7/63/2011/ the surface with a discretization of 1 m up to 20 m depth, and ending in the bottom of the ocean with 500 m layer steps. Atmospheric forcing derived from ECMWF (European Centre for Medium-Range Weather Forecasts) with precipitation coming from GPCP (Global Precipitation Climatology Project) data set. The simulation used here was run during the 1992-2006 period.

Criteria for determining isothermal, mixed and barrier layers
Barrier layer thickness (BLT) is calculated as a difference between mixed (Z M ) and isothermal (Z T ) layers (BLT = Z M − Z T ). The definitions for determining Z M and Z T are arbitrary, and different parameters can be used as criteria. For example, Brainerd and Gregg (1995) based their criteria on a difference in temperature T (or density, σ t ) from the surface value. For evaluating Z T , deviations varying from 0.1 • C to 1.0 • C from sea surface temperatures are normally considered (Monterrey and Levitus, 1997;Kara et al., 2000;De Boyer Montégut et al., 2004). Z M is estimated as the depth at which density is equal to its sea surface value, plus an increment σ t , which is equivalent to a desired net decrease in temperature. Spall (1991), for example, uses σ t = 0.125σ t (0) for determining the mixed layer depth, while Sprintall and Tomczak (1992), as well as Ohlmann et al. (1996) The criteria used in this paper for determining isothermal and mixed layers are the same as those defined by Sprintall and Tomczak (1992). These authors considered that the depths of the isothermal and mixed layers are evaluated in terms of temperature and density steps -T = 0.5 • C and σ t = 0.5 • C(∂σ t /∂T ) -from the sea surface temperature and density (T (0) and σ t (0)) obtained from CTD vertical profiles: where ∂σ t /∂T is calculated as a function of the surface temperature and salinity (Blanck, 1999). Since the SBE911plus CTD has two thermometers, with an accuracy of about 0.001 • C, the error in computing Z T for a T = 0.5 • C is around 0.2% for a local Z T . Moreover, previous studies in the western tropical Atlantic also used the 0.5 • C criterion for the isothermal layer and its equivalent in density change (Silva et al., 2005), whose are found to be the most adaptable due to their robustness (Tanguy et al., 2010). When density stratification is exclusively controlled by temperature, the isothermal layer depth becomes equivalent to the mixed layer depth, and BLT = 0. A particular situation occurs when the near-surface distribution of salinity is sufficiently strong enough to induce a pycnocline inside of the isothermal layer, or Z M < Z T . In such a case, BLT > 0 and warm surface waters may be maintained in isolation from cool thermocline waters.
As in De Boyer Montégut et al. (2007), and , an analysis of individual T/S profiles, combined with a kriging interpolation (within a radius of 400 km with at least 5 grid points), is used to construct isothermal, mixed and barrier layers charts. Wilson et al. (1994) and Bourlès et al. (1999) identified three different origins for the water masses at the upper western equatorial Atlantic: North Atlantic Water (NAW), South Atlantic Water (SAW), and Eastern tropical Atlantic Water (EAW). The NAW originates in the subtropical region of the northern hemisphere and is advected towards the equator by the North Equatorial Current (NEC). This water mass is mostly characterized by high salinity values and relatively low dissolved oxygen (O 2 ) concentrations below the thermocline. The SAW exhibits high salinity values above the thermocline, as well as low salinity values and high O 2 concentrations below it, in relation to the NAW. It is advected to the study region through subduction processes from subtropical areas by the southern branch of the SEC. The EAW reaches the area of interest via the southern edge of the NEC, along with the central and northern branches of the SEC. This EAW is characterized by the lowest concentrations of oxygen and salinity around the thermocline, when compared to the SAW. The origin of the water masses entering the area of study is identified using T/S diagrams of the easternmost boundary hydrographic stations (Fig. 1). The T/S references used to identify the water masses were obtained from Wilson et al. (1994) and Bourlès et al. (1999). Particular attention is given to those areas where SMW cores are observed between the 24.5 and 26.25 isopycnals. Three distinct signatures are found in late winter (Fig. 2a-c) and summer (Fig. 2d-f) seasons: (i) EAW at 2 • -4 • S whose representative station (black full circles) is located at longitude 38.0 • W and latitude 1.6 • S ( Fig. 2a and d); (ii) EAW-SAW transitional zone at 4 • -8 • S whose representative station (black full circles) is located at longitude 31.5 • W and latitude 6.5 • S (Fig. 2b and e); and (iii) SAW at latitudes higher than 8 • S whose representative station (black full circles) is located at longitude 34.5 • W and latitude 12.5 • S (Fig. 2c and f). These figures show clearly the southern incoming of SAW brought by sSEC NBC/NBUC current system during winter vs. summer (Fig. 2b vs. e and 2c vs. f).

Latitudinal analysis of the vertical stratification
Vertical distributions of salinity and temperature along the easternmost boundary hydrographic stations (Fig. 1) are presented in Fig. 3. The vertical sections of salinity indicate a salinity maximum at a depth of around 120 m -a signature of subtropical underwaters that is due to the presence of SAW. Salinity maximum cores vary seasonally around the 24.5 isopycnal. In the late austral winter of 1995, low latitude (≤6 • S) salty surface waters were observed, except for a single surface point around 4 • S. A maximum salinity core (≥36.5) was noticed at the thermocline level (Fig. 3a) between the latitudes of 6 • S and 12.3 • S, and between 11 • S and 12 • S, at the surface. The temperatures of these cores ranged from around 24 • C to 26 • C reaching 27 • C in the southern stations.
During the austral summer of 1997, persistent high salinity values (≥36.0) were found in the upper ocean layer (down to 80 m depth) at 2 • -10 • S (Fig. 3b). It was combined with a shallow core of maximum salinity (≥36.5) at the 7 • -14 • S latitude range. In contrast to late winter, a wider area of high salinity waters was noticed at the surface layer, between 10 • S and 14 • S (Fig. 3b).The temperatures of the cores extended similarly to the winter period's (24 • C to 26 • C) and southern stations attained 28 • C. The latitudes higher than 6 • S observed for salinity maximum cores in both seasons are in accordance with the ones related to SAW signature represented by Fig. 2c and f. As a matter of comparison to the experimental results, interannual simulation data of salinity and temperature parameters are also presented in Fig. 3. These parameters confirm that model results of salinity and temperature (Fig. 3eh) agree with REVIZEE in situ data (Fig. 3a-d) from the late winter and summer seasons. These data were obtained in scope of the Mercator Project "Transport de masse et de chaleur dans la zone de divergence du CSE dans l'Atlantique -TransAt" (2008-2009). High Z T values, limited by the 90 m depth isoline, were detected in two offshore subregions in the study domain: (i) at N-NE from 2 • -5 • S, and (ii) at S-SE from 7 • -12.3 • S (Fig. 4a). A thick mixed layer (70-80 m) was also seen offshore, between 2 • and 10 • S (Fig. 5a). This resulted in a thin barrier layer widely observed over the study area north of 8 • S (Fig. 6a). However, a thick BL prevailed south of 8 • S, as a result of the combination between a deep isothermal layer (Fig. 4a) and a salt-induced (Fig. 5a) shallow mixed layer during late winter.

Winter barrier layers
Southeastern trade winds are more intense during late winter, and surface salinity values low, because of the Inter-Tropical Convergence Zone (ITCZ) precipitations at the equator. Meanwhile, subsurface salinity values are relatively high due to the presence of NBUC, whose equatorward transport increases during this period, as stated by Rodrigues et al. (2007) and Silva et al. (2009). For that reason, isothermal layer deepens and low surface salinity values induce shallow mixed layer resulting in thick BLs in the southern study Ocean Sci., 7, 63-73, 2011 www.ocean-sci.net/7/63/2011/ area (south of 8 • S). Such results are partially in agreement with Sato et al. (2006), which observed that thick BLs in the subtropical gyres occur almost exclusively in the winter hemisphere during July-September, comprising latitudes from 15 • to 5 • S and longitudes from 40 • W to 0 • in the South Atlantic.

Summer barrier layers
Isothermal layers during summer were thinner than during winter. The former showed a median value of 65 m, ranging from 5 to 105 m. Furthermore, shallower Z M (5-50 m) were observed, with a median value of 30 m. The BLT differed from that in late winter, showing a lower range of 5-70 m (median = 35 m) -the highest frequency of occurrence, at 91.0% of the CTD stations. When compared to the winter period, the highest median of BLT (obtained within the lowest BLT range) was justified by the highest frequency of occurrence during this summer period, with 7.5% more CTD stations than during the other season. A qualitatively similar distribution of winter isothermal layer depth can be found during the austral summer (January-April 1997). High Z T values, limited by the 60 m contour, were observed in the N (1 • -3 • S; 37 • -42 • W) to NE (4 • -9 • S; 32 • -35 • W) subareas, and from 9 • to 14 • S (Fig. 4b). The difference between the late winter and summer periods relies on computed values of mixed layer depths from the latter, with Z M not exceeding 50 m. These shallow mixed layers are due to the near-surface intrusion of salty SAW (Fig. 3b), which is transported westward from subtropical regions by the intensified SEC flow verified during this period (Rodrigues et al., 2007;Silva et al., 2009). Consequently, high BLT values during austral summer are not only concentrated in the southernmost part of the study area, as observed during late winter (Fig. 6a), but are also present from 1 • to 10 • S (Fig. 6b).
www.ocean-sci.net/7/63/2011/ Ocean Sci., 7, 63-73, 2011 The BLs found by Mignot et al. (2007), equatorward of the subtropical gyre, might be the same ones found in this study. However, it is not possible to confirm it due to the different BL formulation criteria used in each study. De Boyer  and Mignot et al. (2007) defined the BLT as the difference of two depths: one whose temperature has decreased by 0.2 • C (reference depth of 10 m); and one whose potential density has increased from the reference depth by a density threshold equivalent to the same temperature change 0.2 • C at constant salinity. In the current study it was considered that the 0.5 • C criterion for the isothermal layer and its equivalent in density for defining the base of the BL (Sprintall and Tomczak, 1992;Tanguy et al., 2010).

Westward transport of salty subtropical waters
Salinity cores observed to the south of 6 • S (Fig. 3), during both seasons, are associated with the SEC penetrating into the region from the south. They are also associated with the westward transport of SAW formed in the South Atlantic subtropical gyre Rodrigues et al., 2007;Silva et al., 2009). Specifically, this salinity maximum water must be transported to the study area, first by the sSEC, and near the coast by the NBC/NBUC system (Stramma et al., 1995Pailler et al., 1999). This system carries salty and warm waters toward the equator (Fig. 7). According to this figure, the subsurface salty layer in western Atlantic was more pronounced during late austral winter (Fig. 7a) than summer (Fig. 7b). Table 1 presents the main contribution from horizontal transport up to 200 m and surface salinity budget for both seasons from model output (Fig. 7). Results indicate a horizontal transport 100-fold greater than the surface budget. Thus, the increased thickness of the BL appears more influenced by horizontal advection of salt than the surface freshwater balance. This conclusion was also obtained by Silva et al. (2005) for the northern area of Brazil as well as by Zeng et al. (2009) at the southeast of Vietnam.
The differences between summer and winter BLTs are displayed in Fig. 8. According to that figure, the winter BLT (solid lines) is outstanding in the southern region of the Northeastern Brazil due to deep Z T . The overcoming summer BLT is clearly noticeable mainly in the northern and central regions (thin dashed lines) of the study area. Thicker summer BLT areas are also observed between 0.5 • -3 • S and 4 • -6 • S due to the shallow Z M during this season.
As observed by Rodrigues et al. (2007) and Silva et al. (2009), sSEC bifurcation occurs at lower latitudes during the summer, and maximum NBUC transport happens at 6 • S, which brings it within the SMW (see Fig. 7). In this period, evaporation overcomes precipitation, causing saltier waters at the surface/subsurface layers. In addition, northeasterly winds are present once ITCZ moves southward. Consequently, Z T and Z M become shallower, and the presence of SMW in the NBUC leads to the occurrence of BLs with an equal spatial distribution. This horizontal homogeneity found during the summer period suggests that the South Atlantic western boundary current has a strong influence. This was not observed by other authors (Sato et al., 2006;De Boyer Montégut et al., 2007;Mignot et al., 2007) due to the wide grid used in their study, its goal being the measurement of BLT around the global oceans.

Summary and conclusion
Global climate is strongly linked to upper tropical Atlantic dynamics and their exchange with the atmosphere. Thus, its forecasting tends to improve with an increased understanding of the processes that govern the relative distribution of the ocean's thermodynamic properties. This paper has focused on the isolation of warm surface waters from cool deep waters, by salinity-induced, mixed and barrier layers in the southwestern tropical Atlantic (0 • 30 N-14 • 00 S; 31 • 24 -41 • 48 W). Thicker barrier layers (BL), higher than 60 m depth, were noticed in the southernmost offshore portion of the area of study during late austral winter. During the austral summer, shallower depths of isothermal and mixed layers occurred, when compared to the late winter; however, barrier layer thickness (BLT) ≥30 m was observed over the Ocean Sci., 7, 63-73, 2011 www.ocean-sci.net/7/63/2011/  entire region. In both cases, the formation of the BL seems to be associated with the advection of Salinity Maximum Waters formed in the region of the subtropical gyre. This salty water is transported to the study area by both the sSEC and the western boundary current found in the NBC/NBUC system . The predominance of larger BLT in the south (north) during late winter (summer) indicates that it is associated with the southward (northward) position of sSEC (Figs. 3, 7 and 8).
Historically, it was thought that the halocline significance in the surface layer was only meaningful when dealing with higher latitude regions. However, our results suggest that www.ocean-sci.net/7/63/2011/ Ocean Sci., 7, 63-73, 2011  the inclusion of salinity dynamics and its variability are also necessary for studying mixed and BL layer behaviors in the southwestern tropical Atlantic. This variability is mostly driven by horizontal advection/vertical diffusion terms, as well as the westward transport of these salty, subducted waters by the sSEC. Therefore, the BL genesis in the southwestern tropical Atlantic needs further assessment. The contribution of each mechanism that controls the thickness of the BL should be scrutinized. These scientific tasks can only be achieved through the acquisition of long-term time series of in situ data (e.g., temperature, salinity and current velocity). These data are derived from mooring arrays and Argo floaters deployed at key sites along the western tropical Atlantic, combined with numerical modeling efforts.