Western boundary currents (WBCs) and adjacent continental shelves display two distinct circulation regimes separated by a strong vorticity barrier and therefore exhibit distinct water-mass properties and ecosystem composition (Roughan et al., 2011; Franco et al., 2018). Nevertheless, at the separation of WBCs from the shelf-break, the instabilities of the currents generate meanders and eddies that induce strong interactions between the shelf and the deep ocean generating intense cross-shelf exchanges (Arruda and da Silveira, 2019; Ismail and Ribbe, 2019; Malan et al., 2020; Roughan et al., 2022). These cross-shelf exchanges have important consequences for marine life, as they provide nutrients to the oligotrophic deep ocean and offshore advection of larvae from species that live on the shelf (Franco et al., 2017, 2018). Historically, these systems have been difficult to model accurately, due to a combination of numerical constraints such as small-scale highly energetic dynamics over steep and complex topography, and also due to the lack of observations to compare with, making the sampling of these sporadic events very valuable.

In the South Atlantic, the WBCs are the poleward flowing warm and salty south Brazil Current (BC), which is a relatively weak western boundary current associated with the subtropical gyre, and the cold and relatively fresh northward-flowing Malvinas Current (MC), which is a branch of the Antarctic Circumpolar Current (Gordon, 1989). While the BC extends down to about 500 m, the MC presents a predominantly barotropic structure and extends to the bottom (e.g., Vivier and Provost, 1999). Both currents converge along the South American slope at about 38°S at the so-called Brazil-Malvinas Confluence (BMC), generating one of the most energetic mesoscale regions of the global ocean (Chelton et al., 1990). On average, the BC begins to separate from the shelf break at 36.5°S, and the main core of the MC retroflects at about 38.8°S where the 200 and 1,000 m isobaths diverge at Ewing terrace. At that point, the MC is no longer constrained by the topographic steering and retroflects, while a shallow branch of the MC continues northward over the shelf-break (Matano et al., 2010). Between 36.5 and 38.8°S, the ocean area over the slope is characterized by high eddy kinetic energy with changes in current speed and direction (Figure 1A). Satellite-derived salinity, in situ observations, and modeling studies suggest that this region is also where most of the cross-shelf exchanges in the Southwestern Atlantic take place (Guerrero et al., 2014; Matano et al., 2014; Franco et al., 2018; Berden et al., 2020; Orúe-Echevarría et al., 2021).

The Uruguayan Economic Exclusive Zone (EEZ) is located in the northwestern edge of the Argentine Basin covering the slope between 34.6 and 37.3°S. The EEZ spans an area of about 140 × 10³ km² (Figure 1) and includes part of the region of high kinetic energy and intense cross-shelf exchanges (e.g., Guerrero et al., 2014; Matano et al., 2014; Franco et al., 2018; Berden et al., 2020). The water column over the slope region is characterized by at least six water masses from various sources, a micro-tidal regime, and large seasonal variability. Sea surface temperature (SST) variability presents a seasonal amplitude of 12°C and the wind stress curl reverses seasonally (Piola et al., 2018). Most prominent in the shelf circulation is the northward flow of Subantarctic Shelf Water (SASW) originating from the Patagonia continental shelf (Piola et al., 2000). The northeastward flow of SASW is dynamically controlled by the along-shore wind stress variability (Lago et al., 2021) and in the outer shelf by variations of the intensity of the MC (Matano et al., 2010; Artana et al., 2019). Numerical simulations indicate that in early summer and early winter about 1 Sv of SASW is exported offshore close to the BMC (Franco et al., 2018), while the mid-shelf portion of SASW continues flowing northeastward up to near 32°S. There, SASW meets the subtropical shelf waters (STSW), forming a density compensated front referred to as the Subtropical Shelf Front (Piola et al., 2000). This frontal system extends from the inner shelf at 32°S to the outer shelf near 38.5°S, at the initiation of the BMC (Figure 1B; Gordon, 1989; Piola et al., 2000).

A distinct characteristic of the regional shelf circulation is the influence of the discharge of the Rio de la Plata (RdlP) estuary, which drains the second largest hydrological basin in South America (Framiñan et al., 1999; Piola et al., 2005). With a mean outflow of about 25,000 m³s^–1, the RdlP significantly dilutes and supplies nutrients to the neighboring shelf waters (Piola et al., 2000). The RdlP buoyant plume, associated with absolute salinities lower than 33.65 g kg^–1, presents a strong seasonal variability, with northeastward spreading in austral fall-winter, and southward retraction in summer (Piola et al., 2000). These waters spread over the entire Uruguayan shelf and are often exported to the deep ocean (e.g., Provost et al., 1996; Guerrero et al., 2014). Large interannual variability has also been observed mainly driven by a positive relationship between El Niño Southern Oscillation (ENSO) and precipitation in the RdlP basin during spring, summer and fall (e.g., Grimm et al., 2000; Barreiro, 2010).

The Uruguayan EEZ, although known to be highly variable, is poorly sampled, with historically no national oceanographic research vessel capable of performing full-depth CTD profiles nor underway velocity measurements. During April and May 2016, the Spanish R/V Sarmiento de Gamboa conducted 82 full-depth CTD stations in the Uruguayan EEZ, covering along-slope transects, thus providing ideal measurements to study cross-shelf exchanges. Here we analyze the cruise data, together with other in situ and satellite observations and an eddy tracking algorithm, to provide a full water column description of the water masses and circulation, with particular focus on the intense mesoscale variability and cross-shelf exchanges.

We analyzed the hydrographic data collected by the Spanish R/V Sarmiento de Gamboa cruise conducted between 9 April and 10 May 2016 (henceforth “the cruise”). The sampling consisted of 12 along-shelf transects, separated by approximately 25 km, and a nominal distance between stations of 50 km. The sections were occupied from the deep ocean toward the coast. In this way, an almost regular grid was created over ∼320 km along-slope and 370 km across-slope polygon. In total, 82 full-depth stations (between 18 and 4,147 m) were occupied using a Sea Bird Electronics (SBE) 32 Carousel Water Sampler rosette and a SBE 911 plus Conductivity, Temperature and Depth (CTD) profiler, fitted with a SBE43 oxygen sensor and an ECO-AFL/FL Wet Labs Fluorometer. The CTD was fitted with dual temperature and conductivity sensors. Water samples for salinity, dissolved oxygen, nutrients and chlorophyll-a determination were taken in all odd-numbered transects at 5 m depth, and at additional depths depending on the maximum station depth. Two additional water samples were collected within the euphotic depth for chlorophyll-a, and for nutrients at 100, 500, 1,000, and 1,500 m, and near the bottom when the station depth allowed it. At stations shallower than 100 m, a mid-depth sample was collected. The CTD sensors were laboratory calibrated in April 2015. In addition, salinity and dissolved oxygen were calibrated based on the water samples collected during the cruise.

Additionally, underway measurements of temperature, salinity and fluorescence were sampled at 1 min intervals using a Sea-Bird SEACAT SBE 21 Thermosalinograph (TSG) and a Turner 10 fluorescence system. Temperature and salinity from the TSG were calibrated with the nearest CTD measurements. Wind measurements, air temperature, and humidity were also collected underway at about 10 m above the sea surface with a Geonica Meteodata automatic weather station. A hull-mounted Teledyne RD Ocean Surveyor Acoustic Doppler Current Profiler (ADCP) operating at a frequency of 75 and 150 kHz recorded underway ocean currents with 16 m and 8 m depth bins, starting at 10 m and reaching up to 700 and 400 m depth, respectively. The ADCP was operated at 75 kHz over the deep ocean and at 150 kHz over the shelf, recording one profile approximately every 1 min.

Turbidity from water samples was determined on board with a Multiparameter Senso Direct 150 Lovibond. In situ chlorophyll-a was also determined on board by acetone extraction and fluorometer reading following Yentsch and Menzel (1963). Fluorescence values from the TSG were calibrated using 32 measurements of chlorophyll-a from water samples collected at 5 m depth by empirically computing a scale factor to minimize the root square mean error. Fluorescence values from the CTD were previously adjusted with a “dark offset” calculation for each profile, by subtracting the mean value of the aphotic zone in values deeper than 200 m to each profile. The adjusted chlorophyll-a data compare well and are significantly correlated (Supplementary Figure 1), although absolute values of chlorophyll-a are out of the scope of this study, which focuses on identifying low and high-productivity regions. Water samples for nutrients were fixed on board and measured at the Institute of Marine Sciences (ICM-CSIC) in Barcelona, Spain. At the laboratory nitrate, nitrite, and phosphate concentration were determined using a Seal Analytical MT19 Auto Analyzer with the colorimetric technique and an automated system by continuous flow analysis applying the methodology of Grasshoff et al. (2009).

To complement the cruise hydrographic data we used all Argo profiles available for the Southwestern Atlantic between 32–42°S and 58–49°W (Argo, 2021). Moreover, to gather more information on the regional dynamics, we use satellite data sets with daily resolution: (I) (SST with 0.25° horizontal resolution from the National Oceanic and Atmospheric Administration Optimum Interpolation Sea Surface Temperature for analyzing the long-term characteristics between 1993 and 2019 (OI SST V2; Reynolds et al., 2007), (II) The Multi-scale Ultra-high Resolution (MUR) SST analysis (Chin et al., 2017). This data set has a horizontal resolution 0.01°, making it more useful to identify (sub) mesoscale structures. These data are only available since 2002 and were therefore only used to study the variability of SST within the period of the cruise; (III) Surface chlorophyll-a concentration L4 (gridded and gap-filled) product from Marine Copernicus (CMEMS) with 4 km resolution (Garnesson et al., 2019) within the period of the cruise; (IV) L4 Sea Surface Salinity (SSS) v03.21 from the European Space Agency Sea Surface Salinity Climate Change Initiative consortium at 50 km and a weekly resolution, spatially sampled on a 25 km EASE (Equal-Area Scalable Earth) grid and 1 day of time sampling. This dataset combines the satellite measurements from SMOS, SMAP, and Aquarius covering the period January 2010–September 2020 (see Boutin et al., 2021); (V) Ssalto/Duacs Multimission Altimeter derived absolute dynamic topography (ADT) and derived surface geostrophic velocity fields at a 0.25° horizontal grid for the period 1993–2019 distributed by CMEMS (Capet et al., 2014; Pujol et al., 2016). This latter dataset is also used as input for the Tracked Ocean Eddies (TOEddies) automatic detection algorithm developed by Chaigneau et al. (2011), Pegliasco et al. (2015) and Laxenaire et al. (2018, 2019, 2020). The TOEddies algorithm detects eddies as ADT extrema, corresponding to the center of an eddy, within closed ADT contours. By restricting the detection to eddies large enough to be in geostrophic equilibrium, these ADT contours correspond to streamlines and the largest closed contour is identified as the eddy boundary. Eddies are then tracked over time by computing their surface overlap over time taking advantage of their large size relative to their daily displacement. A detailed description of TOEddies can be found in Laxenaire et al. (2018).

Furthermore, we used ETOPO1 bathymetry with 1 min (about 2 km) horizontal resolution (Amante and Eakins, 2009). Finally, we also accounted for the daily discharges of the Rio de la Plata between 1988 and 2019 from Borús et al. (2017) and ERA5 reanalysis (Hersbach et al., 2020) hourly surface winds and sea level pressure from 1993 to 2019 in the Southwestern Atlantic (32–42°S 60–45°W) with 0.25° horizontal resolution to study the climatological context during the period of the cruise. Table 1 summarizes the datasets used.

CTD profiles were generated in 1 dbar bin average. All the gridded fields were linearly interpolated and previously smoothed with a 5 dbar centered moving average to remove spiciness from the plots. ADCP data was processed with Cascade V7.2 (Le Bot et al., 2011) with tide correction and a 16 m vertical resolution. Along and across slope velocities were computed by assuming a 40° inclination of the slope with respect to the zonal axis. Conservative temperature (Θ, °C) and absolute salinity (S_A, g kg^–1) were calculated from observations from the Thermodynamic Equation of SeaWater 2010 (McDougall and Barker, 2011) and used for the water mass characterization. Water masses in the deep ocean were classified based on neutral density (γn) ranges and thermohaline properties following Valla et al. (2018). Mixing of SACW and TW with shelf waters leads to substantial freshening over a wide neutral density range and are labeled as “others.” On the shelf (stations shallower than 200 m depth) thermohaline properties were used to identify shelf water masses following Berden et al. (2020; specified in Table 2). The data were sorted as shelf (0–200 m) and the slope (500–2,500 m) based on interpolated ETOPO1 bathymetry (Amante and Eakins, 2009), and retaining only the values within the desired ranges. In the case of Argo floats, we also filtered duplicated profiles and retained only those with good quality (flagged as 1 or 2), and the co-localized values between 1,000 and 2,500 m. To study the vertical distribution of the shelf waters with Argo floats, we interpolated the irregular sampling to a regular 10 m interval, which is the coarser resolution of the Argo profiles, in order to get the same number of observations per profile. For all the observations, we colocalized each measurement to obtain the closest satellite measurement in space and time to compare.

Using satellite altimetry, we followed the criteria proposed by Lumpkin and Garzoli (2011) to detect the location of the confluence over the slope “as the latitude where the surface current vectors interpolated to the 1,000 m isobath change direction from southward (Brazil Current) to northward (Malvinas Current) in 77 day low−passed velocity fields.” However, we modified it as follows. Instead of interpolating the meridional velocity to a unique latitudinal point over the 1,000 m isobath, we localized the change in the meridional currents averaged between the 500 and 2,500 m isobaths over the slope. Though no substantial differences are found in the results, our indicator is less sensitive to slight zonal displacements of the currents.

Following earlier work (e.g., Barré et al., 2011; Ferrari et al., 2017) we used three specific ADT contours as (i) a proxy for the position of the BMC (ADT 0.25 m; correlated with R² = 0.94; p < 0.001), (ii) the retroflection of MC (ADT 0 m), and (iii) the separation of BC from the shelf-break (ADT 0.5 m). Note that ADT is not a useful tool to study long-term variations because both, steric height and shift in the currents have induced variations (e.g., Fang and Zhang, 2015; Ruiz-Etcheverry and Saraceno, 2020), but it can be a useful tool to identify fronts in the past decade.

Results are divided into four subsections. First, we describe the interannual contextualization of the period of the cruise and show that 2016 was a period of strong anomalous circulation conditions. Secondly, we analyze the oceanic variability during the period of the cruise. Then, we present a general description of the in situ observations, followed by the characterization of the shelf water export to the adjacent deep ocean and its relationship with the position of the BMC and mesoscale activity. Finally, based on extended Argo profiles and satellite observations, we explore the frequency of the shelf water distribution over the slope and the utility of altimetry to identify the shelf water export events.

The hydrographic cruise that took place in April–May 2016 in the Uruguayan EEZ provided an unprecedented dataset to investigate local water masses, ocean dynamics, and cross-shelf exchanges at the Brazil Malvinas Confluence. Despite the survey covering a relatively small area, our study shows that it is characterized by a highly variable and dynamic environment. Water properties can fluctuate from below 0°C to over 25 °C in conservative temperature and from almost freshwater to over 37 g kg^–1 in absolute salinity. The region hosts intense western boundary currents and mesoscale activity developing offshore, displaying large oceanic space-time variability at all scales (from synoptic, to interannual and interdecadal) driving cross-shore exchanges.

The first half of 2016 was an anomalous period in the southwestern Atlantic in terms of atmospheric and oceanic circulation. It was characterized by intense southwesterly winds, negative SST anomalies, and large Rio de la Plata discharges. By extending the Brazil Malvinas Confluence latitudinal position index of Lumpkin and Garzoli (2011), we showed that the confluence was located anomalously north during the first half of 2016, occupying the Uruguayan EEZ when the in situ survey was conducted.

In this work, we focused on cross-shelf exchanges as they have important consequences for carbon export, water masses transformations, and marine biodiversity. Previous studies provide evidence that, regionally, shelf water export is closely related to the confluence position (e.g., Matano et al., 2014; Berden et al., 2020; Orúe-Echevarría et al., 2021; Figure 11A). This control is somehow expected since the Brazil and Malvinas currents generate a barotropic pressure gradient that extends into the shelf controlling the outer shelf circulation (Palma et al., 2008; Matano et al., 2010). Our analyses provide evidence of the 3-D structure and evolution of two different kinds of shelf water export associated with the BMC dynamics: one related to Rio de la Plata waters that remain confined to the surface layers; the other involves SASWs that subducts under the subtropical thermocline while crossing the slope. To understand the dynamical setting associated with these processes and to assess their significance in time we used a combination of different observations that include the cruise hydrographic data, Argo profiles, satellite maps of SSS, SST, ADT, and chlorophyll-a, and a recently developed eddy detection algorithm.

A near-surface export of shelf water consisted of a thin off-shelf layer of Rio de la Plata waters following the path of the retroflection of the Brazil Current after an anticyclonic eddy shedding that dissipated after 10 days due to strong winds that mixed the stratified plume (Figure 11B). In addition, the hydrographic observations reveal a subsurface export of SASW. In this case, the driving mechanism also appears to be associated with mesoscale dynamics triggered by the merging of a cyclonic eddy with the Malvinas Current at the confluence with Brazil Current (Figure 11C). This generated intense off-shelf velocities at the shelf break. Evidence of subsurface export of shelf waters based on hydrographic and lowered ADCP observations near the BMC has been recently reported by Berden et al. (2020, their Figures 4C, 8E). The novelty of this study relies on the availability of closely spaced along-slope sections which reveal the detailed structure of the cross-shelf fluxes. Combining these in situ and satellite observations we reconstruct the 3-D field, quantify the export, and associate mesoscale ocean dynamics as the driving mechanism.

Using numerical simulations, Matano et al. (2014) showed that a time-averaged convergence of volume of about 1.21 Sv in the along-shore flow of SASWs (1.15 Sv) and STSW (0.038 Sv) and the inflow of Rio de la Plata waters (0.024 Sv) between 34 and 38°S must be exported offshore. The 1.21 Sv could hardly be achieved by the exported low salinity RdlP plumes, but the sum of both surface and subsurface export as observed in this work is closer to. In addition, Franco et al. (2018) using numerical simulations showed that released neutrally buoyant particles equally distributed in the vertical at 38°S over the shelf were exported all along the shelf-break between the retroflection of the Malvinas Current and the Subtropical Shelf Front at about 32°S. Most of the neutrally buoyant particles were exported at the BMC, in particular within the 50–200 m layers. They also showed that as SASW was exported, it rapidly became saltier. The intense mixing observed over the slope together with the fact that the difference between SASWs and the surrounding ocean waters can be less than 0.2 g kg^–1, makes it difficult to detect this water mass based only on salinity differences with the surrounding waters, as the shelf water freshwater signature rapidly fades as shown by Franco et al. (2018).

By using Argo profiles located within the southwestern Atlantic slope, we provide evidence of the presence of Rio de la Plata waters and SASWs between the separation of the Brazil Current from the shelf-break and the confluence and at the same depths as observed during the 2016 cruise. At specific values of absolute dynamic topography of 0.35–0.45 m, we found shelf waters over the slope in more than half of the Argo profiles, further suggesting the export of shelf waters at this location is relatively frequent. Also, by inspecting Argo profiles together with satellite surface salinity we show that SASWs is often found over the slope, especially within the shallow branch of the Malvinas Current. Because of the similarity between SASWs and surrounding waters thermohaline properties and the intense mixing taking place over the slope, detection of shelf waters within this current is challenging. Nonetheless, by combining in situ water velocity measurements and concomitant satellite altimetry observations we found evidence of SASWs along the BMC. In contrast, Rio de la Plata waters are more easily detectable in the open ocean due to their low salinity characteristics.

We also provided evidence that, although satellite surface salinity reveals shelf water exports (Guerrero et al., 2014), the coarse resolution of satellite data limits the ability to detect the fine structure and temporal evolution of the exported waters. This suggests that several export events might not be properly captured by satellite surface salinity. Chlorophyll-a, which has higher resolution especially under clear sky conditions, could be a complementary tool to detect these freshwater export events, but need to be contrasted with other salinity observations, as high chlorophyll-a filaments could be associated with several other processes not related to shelf-waters. Combination of different observations in the future (e.g., satellite Chlorophyll-a and salinity observations) could provide more insights on detecting and characterizing shelf water export events.

Another useful approximation to study shelf water export processes are high-resolution numerical models, as they reproduce the shelf-water export, especially when the confluence and the separation of the Brazil Current from the shelf break are along the same front (Berden et al., 2020; Orúe-Echevarría et al., 2021), but probably also to better simulate the impact of mesoscale eddies, like those observed during the 2016 cruise (Figure 11). Recent numerical simulations highlight the role of along-shelf winds in modulating the magnitude of along-shelf transports and the resulting shelf-open ocean exchanges (Berden et al., in press). In agreement with our findings, the numerical models indicate that a majority of the cross-shore flow occurs close to the confluence. Our retrospective analysis indicates seasonal fluctuations of the export location between 35.5 and 40°S which are modulated by displacements of the confluence, and also suggest a secondary export route near 33°S. This latter export location is also apparent in ocean reanalyses (Berden et al., in press). One of the several advantages of studying cross-shelf exchanges with numerical models is the ability to be able to track the water being exported (Franco et al., 2018; Combes et al., 2021).

Finally, our analyses suggest that geostrophic currents derived from satellite altimetry are a robust indicator for the regional dynamics and a useful tool to track cross-shore water export events. By providing relatively accurate steric height estimates, satellite altimetry shows a high correlation with absolute velocities at 100 m of depth. Hence, satellite altimetry seems to be more useful than satellite surface salinity to identify subsurface shelf water export events. As altimetry is relatively accurate in identifying the location of the confluence (ADT contour 0.25 m over the slope) and the separation of Brazil Current from the shelf-break (ADT contour 0.5 m), across-slope velocity maxima confined between the 0.25 and 0.5 m ADT contours over the shelf break might be a good indicator for subsurface shelf water export occurrences. However, satellite altimetry does not resolve smaller scales than the largest mesoscale. Hence, several filaments of a few tens of km and small eddy-like structures near the BMC (diameter about 50 km) such as the merging of eddy C1 with the MC are not well-resolved as suggested by Ferrari et al. (2017). In addition, intense (> 1Sv) wind-driven high frequency (1–3 days) cross-shelf exchanges were detected by Berden et al. (in press) from high-resolution reanalyses not properly resolved by altimetry.

In conclusion, in situ hydrographic and ADCP observations combined with long-term satellite observations, an eddy detection algorithm, and Argo float profiles allowed us to identify and characterize the 3-D structure and evolution of surface and subsurface shelf water export to the adjacent deep ocean. The two types of shelf water export pathways may contribute to balancing the mass convergence associated with along-shore flows. Future oceanographic surveys covering the entire width of the continental slope and achieving high-resolution synoptic along-slope sections at the BMC might help to better constrain the structure of subsurface shelf water export.

The full processed data necessary to reproduce the reported finding is available at https://doi.org/10.17882/86206. Raw TSG and meteorological underway measurements are available at http://data.utm.csic.es/viewer/?vessel=sdg&cruise=20160408&dataset=tss. Raw CTD and ADCP data is available for research under the previous authorization of the Uruguayan national oil company (https://www.ancap.com.uy). ADT and chlorophyll-a data were downloaded from http://marine.copernicus.eu/, SST from https://psl.noaa.gov/, SSS from https://catalogue.ceda.ac.uk/uuid/fad2e982a59d44788eda09e3c67ed7d5 and eddy tracking from https://vesg.ipsl.upmc.fr/thredds/catalog/IPSLFS/rlaxe/catalog.html?dataset=DatasetScanIPSLFS/rlaxe/Database_South_Atl.zip.

GM processed the hydrographic data and made the transport calculations, and prepared the manuscript with contributions from all co-authors. All authors participated in the scientific interpretation of the results.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.