This study is based on the analysis of the Copernicus Marine Environment Monitoring Service (CMEMS) global ocean reanalysis (GLORYS12V1) during 1993–2018. GLORYS12V1 is based on the NEMOv3.1 ocean model, described in Madec et al. (2017). The model is driven at the surface by the ERA-interim atmospheric reanalysis from the European Centre for Medium-Range Weather Forecasts. Satellite-derived along-track sea level anomaly (SLA, Pujol et al., 2016), sea surface temperature (SST) from NOAA's Optimum Interpolation Sea Surface temperature (0.25°, Reynolds et al., 2007), sea ice concentration from CERSAT and in situ vertical profiles of temperature (T) and salinity (S) from the CORA4.1 database (Cabanes et al., 2013) are jointly assimilated. The assimilation procedure employs a reduced-order Kalman filter with a 3D multivariate modal decomposition of the background error. The assimilation procedure includes an adaptive-error estimate and a localization algorithm. In addition, a 3D-VAR scheme corrects for the slowly-evolving large-scale biases in temperature and salinity (Lellouche et al., 2018). The reanalysis does not incorporate tidal forcing. The model output is provided daily on a standard regular grid at 1/12° (~8 km) and on 50 standard vertical levels, with 22 levels in the upper 100 m. Consequently, the vertical resolution is 1 m at the surface, decreases to 33 m at the bottom of the continental shelf (200 m), and reaches 450 m near the bottom in the deep ocean (5,700 m).

To evaluate the model performance, the model was compared with in-situ data collected during 12 hydrographic surveys over a cross-shelf transect off southern Brazil, hereafter referred to as the Albardão transect (Figure 2 and Table 1). The data consists of high-resolution, cross-shore sections of conductivity-temperature-depth (CTD) profiles occupied between August 2003 and October 2018. In addition, the alongshore volume transport between the coast and the 200 m isobath near 39°S (hereafter referred to as the MDP transect, location shown by the orange dashed line in Figure 1) has been recently estimated, hereafter referred to as T_Obs, based on direct current observations during 11 months from December 2014 to November 2015. Details of these transport estimates are provided in Lago et al. (2021). Note that the in-situ observations used herein were not assimilated to the ocean reanalysis.

The altimetry data (sea level anomalies and associated geostrophic currents) used in this analysis were the delayed-time merged all satellites Global Ocean Gridded SSALTO/DUACS Sea Surface Height L4 product from Copernicus Marine (Pujol et al., 2016). The gridded product has a spatial resolution of 1/4° on a regular grid and a daily temporal resolution.

Wind stress data was obtained from the ERA5 global atmospheric reanalysis (Hersbach et al., 2020), which covers 1979–2022, produced by the European Center of Middle-range Weather Forecast (ECMWF). This product is available hourly over a regular grid of 1/4°. Daily averages were used in this work.

Since the WBCs may act as barriers to exchanges between the shelf and the open ocean, the possible impact of the BC and MC intensity is studied. The analysis of the BC variability is based on the depth-averaged model velocity in the upper 0–200 m at 34.5°S projected in the local along-slope direction (orientation of 36° from the true north). Likewise, the MC velocity was defined at 42°S, where the model velocity is projected to 45° from the true north. Since the BC presents substantial mesoscale variability (see Chidichimo et al., 2021), it is difficult to determine its precise position in the cross-shore direction along 34.5°S in the daily model outputs. Therefore, the position of the current was determined using the mean velocity field between 1993 and 2018 considering the current as a uniform flow with intensity >10 cm.s⁻¹. Once the position of the current is established, the along-slope velocity component is averaged horizontally to obtain a single velocity time series. The same procedure is followed to determine the MC intensity. The sensitivity of the definition of both currents to changes in the velocity threshold and depth over which the velocity is integrated was analyzed (Supplementary Figure 1). The resulting velocity magnitude is not significantly sensitive to changes in depth of integration considering a range of 200–400 m. Although the velocity magnitude displays variations associated with changes in the threshold used to define the BC and MC, tested in the range of 5–20 cm.s⁻¹, these changes are small and the resulting velocities are highly correlated (R²>0.85^*). The significance of the statistical correlations was calculated at a 95% confidence level. Hereafter, significant correlations are indicated with an asterisk (^*), and correlations with R² < 0.15 will not be addressed. In addition, the BMC location was analyzed as a possible driver of cross-shelf exchange. The position of the BMC is determined daily as the latitude where the 10°C isotherm at 200 m intersects the 1,000 m isobath (Goni and Wainer, 2001; Matano et al., 2014).

In this section, we evaluate the model's ability to capture the patterns of the shelf circulation and its time variability. To this end, we compare the model outputs with available in-situ and satellite observations.

To study the cross-shore transport between 30 and 40°S and its temporal and spatial variability, the 200 m isobath was divided into 20 zonal bands each spanning 0.5° of latitude (Figure 5). The time-averaged offshore transport through each outer shelf segment was estimated as the sum of the transports entering through the northern and southern boundaries of each zonal band. The total offshore transport between 30 and 40°S, hereafter referred to as T_Cross (Figure 6C), is defined as the sum of the transports entering through the northern boundary at 30°S (T_N, positive southward, i.e., entering the domain; Figure 6A) and the southern boundary at 40°S (T_S, positive entering the domain; Figure 6B). We estimated the changes in the volume of shelf water based on the changes in the model-derived SLA. The resulting rate of daily volume change has a standard deviation of 0.16 Sv, which is an order of magnitude smaller than the typical daily mass convergence between 30 and 40°S. Thus, we assumed the mass storage plays a minor role in the cross-shore transports estimated based on mass conservation. We also investigate extreme events which are defined as those periods when the transport is higher/lower than the record-mean ±2 standard deviations. In addition, possible physical mechanisms modulating the variability of the cross-shore transport such as the intensity of the BC and the MC, the BMC location, the alongshore transport variability, and the winds are analyzed.

The variability of the shelf transports (T_N, T_S, and T_Cross) can be caused by circulation changes occurring on a variety of time scales. Thus, it is useful to split the model-derived transport variability by filtering the daily transport signals as follows:

where T_mean is the record-mean transport and T_weekly, T_subannual, T_seasonal, and T_interannual are associated with weekly (<10 days), subannual (10 days–5 months), seasonal (5–23 months), and interannual (>23 months) time scales, respectively (Hirschi et al., 2007). To compute each component the daily transport time series is filtered with a Butterworth window of order 5. The weekly scale is included to study the high-frequency variability, which is particularly relevant to understand the extreme offshore transport anomaly events. Table 2 shows the standard deviation of each transport and for each of the above-defined time scales. The larger variability is concentrated on the subannual and weekly time scales, with the latter being the largest. Although the transport power spectrum shows a marked peak around 1 year (2.7–9.2% of energy), there is a substantial amount of energy (>85%) distributed in the high frequencies (subannual and weekly), without a characteristic frequency (Figure 6D).

In this section, the extreme cross-shore transport events are studied based on the daily, unfiltered T_N, T_S, and T_Cross. The monthly distribution of the number of positive extreme events is shown in Figure 9A. Recall positive transports imply northward flow over the shelf through 40°S (T_S), southward flow through 30°S (T_N), and offshore flow through the outer shelf between 30 and 40°S (T_Cross). Extreme southward (positive) T_N events (red bars in Figure 9A) occur more frequently in spring, last <3 days, and reach 6.32 Sv, eight-fold the mean value (0.78 Sv; Table 4). Extreme northward (positive) T_S events (blue bars in Figure 9A) are more frequent in austral winter (June–July) and infrequent in summer. Typically, the extreme transport events across the southern boundary last <4 days and reach transports of up to 8.36 Sv (Table 4), six-fold the mean value (1.31 Sv). The occurrence of positive T_Cross extreme events (green bars in Figure 9A) has two maxima, in fall and spring, and are less frequent in summer. The extreme export events have a shorter duration (<2 days) and present a maximum export transport of 9.45 Sv (more than four-fold the mean export transport, 2.09 Sv). During the entire period, the offshore transports across the BMC sector and S4 account for 73 and 20% of the total transport across the 200 m isobath between 30 and 40°S, respectively. However, during the extreme events, the transports across BMC and S4 decrease to 43 and 12% of the record-mean T_Cross. Therefore, the offshore transport during the extreme export events is spatially more uniformly distributed. During the average positive T_Cross export event (T_Cross mean of 6.08 Sv), 4.27% of the total volume of shelf waters between 30 and 40°S is exported to the open ocean.

The monthly distribution of the number of negative extreme events is shown in Figure 9B. Negative (northward) extreme T_N events are more frequent in June–July and uncommon in summer, last <6 days, and reach −6.54 Sv. Negative (southward) extreme T_S events occur more frequently between August and October and infrequently in summer, last <3 days, and reach transports of up to −4.92 Sv. The occurrence of negative extreme T_Cross events is maximum during winter and minimum during summer. The extreme cross-shore transport events have a short duration (between 1 and 4 days) and the maximum onshore transport reaches −6.53 Sv. During these negative cross-shore transport events the BMC sector continues to be an export location with offshore transports of shelf waters of 0.48 ± 0.64 Sv (Table 3). Therefore, the on-shelf intrusion of open ocean water is distributed over the remaining of the shelf break domain (Figure 5B), including S4. These negative extreme events represent on average 1.79% of the total shelf volume between 30 and 40°S.

The extreme events have a typical duration in the weekly time scale. As discussed in section 3.2.2.1, the variability on this scale is mostly driven by the local alongshore wind stress. Therefore, it is useful to further explore the possible association between variations in regional wind patterns and extreme transport events. Recall that southwesterly wind stress is referred to as positive and northeasterly wind stress as negative. A composite (average fields during extreme events) of positive T_N extreme events is characterized by negative (northeasterly) local alongshore wind stress (~−0.27 N.m⁻²) at 30°S associated with a high atmospheric pressure center located near 40°S–35°W (Figure 10A). Consequently, a negative SLA develops along the coast, which is most intense north of the RDP estuary. The resulting cross-shore pressure gradient leads to a southwestward geostrophic flow north of 38°S. The negative SLA and the associated geostrophic flow weaken near the southern boundary. The negative T_N extreme events are characterized by positive local alongshore wind stress (~0.16 N.m⁻²) at 30°S, associated with a high atmosphere pressure center located near 32.5°S–59.5°W (Figure 10D). Consequently, a positive SLA develops along the coast north of 38°S, which leads to a northeastward geostrophic flow at 30°S.

The positive extreme events in T_S are associated with positive (southwesterly) local alongshore wind stress (~0.21 N.m⁻²) at 40°S, associated with a low atmospheric pressure center located south of 50°S (Figure 10B), that generates southwesterly wind stress over the continental shelf south of 30°S. This intense wind stress creates a positive SLA along the coast in the entire domain, which leads to a relatively strong northeastward geostrophic flow at 40°S. This northeastward flow extends northeastward, also leading to northeastward (negative) flow across the northern boundary. Conversely, negative T_S extreme events are associated with negative (northeasterly) local alongshore wind stress at 40°S (~−0.13 N.m⁻²) associated with a high atmospheric pressure center located at 45°S–47.5°W (Figure 10E). In response to this wind stress pattern, there is a negative inner shelf SLA south of 30°S and an associated southwestward flow across the southern boundary (Figure 10E).

The positive T_Cross extreme events are characterized by southwesterly wind stress at 40°S (~0.22 N.m⁻²) and northeasterly wind stress at 30°S (~−0.08 N.m⁻²). A low atmospheric pressure center located near 50°S–50°W (Figure 10C) generates southwesterly wind stress over the continental shelf south of 40°S and a positive SLA along the coast south of 34°. Thus, this atmospheric circulation pattern leads to relatively intense northeastward transport (4.03 ± 1.69 Sv, Table 3) over the shelf across the southern boundary (T_S, 40°S). Conversely, a high atmospheric pressure center located near 35°S–28°W generates northeasterly wind stress over the continental shelf north of 30°S. In response to this wind stress pattern, a weak negative SLA develops in the inner shelf north of 33°S, this negative anomaly leads to southwestward transport (2.02 ± 1.57 Sv, Table 3) over the shelf across the northern boundary (T_N, 30°S). The resulting mass convergence over the shelf between 30 and 40°S is compensated by a strong offshore flow of shelf waters. The negative extreme events of T_Cross are associated with a high atmospheric pressure center located over the RDP (Figure 10F). This feature leads to northwesterly wind stress at 40°S (~−0.04 N.m⁻²) that generates a negative SLA over the coast south of the RDP and southwesterly wind stress at 30°S (~0.14 N.m⁻²) that generates a positive SLA north of RDP. As a result, a weakening of northeastward flow across the southern boundary (T_S, 0.14 ± 1.28 Sv) and a northeastward transport across the northern boundary are observed (T_N, −2.16 ± 1.42 Sv). Analogous to the positive extreme events, the mass divergence over the shelf between 30 and 40°S is compensated by an influx of deep-ocean waters to the shelf. These import events can have an important biogeochemical impact on the shelf.

Intense Malvinas Current intrusions over the shelf, causing substantial cooling, increased salinity, and elevated surface chlorophyll concentrations have been reported near 41°S (Piola et al., 2010). A recent analysis of in-situ observations collected in late April 2016 between 34 and 38°S presents relatively cold, salty, and high-nutrient waters in the outer shelf near 36.5°S (Manta et al., 2022, their Figures 2C,F, 5C). These observations suggest onshore flow of subantarctic water from the Malvinas Current. Our model analysis indicates that T_Cross reached −2.55 Sv in the previous days. About 30% of this inflow occurred across the BMC (T_BMC ~−0.81 Sv). As might be expected, on average T_BMC is positive and peaks in the BMC sector of the shelf-break (Supplementary Figure 4). However, there are a few instances (about 2.20% of the time) when T_BMC is negative (onshore), as observed in late April 2016. In the model, the relatively intense import of slope waters was associated with northeastward flows of −4.09 Sv across 30°S and 1.54 Sv across 40°S, which peaked on 27 April 2016. Thus, the model circulation is consistent with the observation of subantarctic waters in the outer shelf south of the BMC reported by Manta et al. (2022). Moreover, the model results suggest that the onshelf flow observed in late April 2016 between 30 and 40° was a response to divergence of the along-shore circulation, which was mostly driven by the pattern of alongshore winds. However, in agreement with the observations of Manta et al. (2022), mesoscale eddies that approach the outer shelf may create a fine scale structure of relatively intense cross-shore flows (see Supplementary Figure 5). In addition, highly non-linear submesoscale features of characteristic scale <10 km, not resolved by the GLORYS reanalysis, are ubiquitous in the Brazil/Malvinas Confluence (Orúe-Echevarŕıa et al., 2019), and may further contribute to cross-shelf exchanges.