Relationship between sea surface temperature anomalies in the Southwestern Atlantic Continental Shelf and atmospheric variability on intraseasonal timescales

The intraseasonal (IS) variability of the sea surface temperature (SST) in the Southwestern Atlantic Continental Shelf (SWACS, 45–33° S—70–50° W), and its relationship with that in the atmosphere, was studied for the austral warm season. SST satellite data (11-km resolution NOAA CoastWatch Program) and data of different atmospheric variables (Reanalysis1 NCEP/NCAR and ERA-Interim) were used. Data were filtered using a 10–90 day filter to isolate the IS variability. A Principal Component analysis was applied then to the filtered SST anomalies (SSTA) and the activity of the leading modes was described through the corresponding temporal series. The first three modes are significant. EOF1 (25.7% of variance) exhibits SSTA of opposite sign to the north/south of 42° S. EOF2 (9.0%) and EOF3 (5.1%) are related with centers of SSTA of opposite sign located off the Uruguayan coast and in the middle shelf. Composites of SSTA and of key atmospheric variables were made considering the days in which the main modes were active. They show that the SSTA described by the three modes are associated with distinctive regional sea level pressure anomalies that, in turn, seem to be related to atmospheric Rossby wave trains extending from the Australia area towards South America. The corresponding atmospheric wave sources vary depending on the mode. These results show, therefore, that the SSTA in the SWACS exhibit significant IS variability that is, in part, locally and remotely influenced by atmospheric anomalies oscillating on similar timescales. These ocean–atmosphere teleconnections could help to improve ocean predictability at those timescales in the future.


Introduction
Profound knowledge of sea surface temperature (SST) anomalies distribution over the continental shelves is essential to study oceanic processes such as frontal dynamics, upwelling and downwelling events, as well as eddies and plumes evolutions, which are also important in biological applications. The knowledge of SST anomalies in the shelves is also useful for climate monitoring and prediction, since they respond to a multiple-scale organization of atmospheric circulation anomalies occurring over distant parts of the globe, commonly called teleconnections. Climate teleconnections can exhibit variability on intraseasonal (IS) time scales, which are those between 10 and 100 days (Feldstein 2003;Murakami 1981;Wang et al. 2016). Moreover, in recent years there has been a greater appreciation of the importance of the two-way interactions between the tropics and the mid-latitudes and high latitudes on IS time scales (Stan 1 3 et al. 2017). Whereas mid-latitudes are typically characterized by being energetic areas where eastward-moving atmospheric baroclinic waves dominate surface energy flux variations on synoptic timescales (3 to 10 days), quasi-stationary large-scale Rossby waves, blocking anticyclones as well as slowly moving cutoff lows contribute significantly to extratropical atmospheric variability on timescales longer than approximately 10 days (Alexander and Scott 1997;Deser et al. 2010). Berbery and Nogués-Paegle (1993), among other early studies, showed that austral summer atmospheric convection variability over the Indonesia region contributes to induce atmospheric Rossby wave trains that propagate poleward from this region towards South America and the Southwestern Atlantic Ocean (SWAO). More recently, the description and understanding of the atmospheric IS variability between 10 and 90 days has received considerable attention at both global and regional scales. Results show that atmospheric IS variability is essentially internally generated by atmospheric instability (Stan et al. 2017). The Madden-Julian Oscillation (MJO) is recognized as the leading global pattern on IS timescales, with maximum amplitude at the tropics and dominant activity between 30 and 60 days (e.g., Madden and Julian 1972). The MJO is characterized by a wavenumber 1 spatial structure propagating eastwards influencing not only atmospheric variables such as air temperature, pressure, winds, cloudiness and rainfall, but also ocean conditions like SST, ocean surface evaporation, and ocean chlorophyll (Madden and Julian 1972;Hendon and Salby 1994;Zhang 2005).
The IS atmospheric variability over South America and the surrounding oceans has received some attention from the scientific community during the last 20 years. Vera et al. (2017), and references therein, show that it exhibits considerable amplitude all year round, with large and significant influence on regional climate conditions. The South Atlantic Convergence Zone (SACZ), which is a band of active atmospheric convection that develops during the warm season (Carvalho et al. 2004), plays a key role in determining the leading patterns of regional atmospheric IS variability (e.g., Vera et al. 2017). Based on observations and reanalysis (Alvarez et al. 2016 and numerical models  it has also been confirmed that the regional IS atmospheric variability is in part influenced by the MJO activity, where the Atlantic air-sea interaction was found to be important to enhance the rainfall within the SACZ through a local warm SST anomaly forced by heat flux anomalies associated with the direct MJO impact. Air-sea interaction over the SACZ region can be directed from ocean to atmosphere, from the atmosphere to the ocean, be mutual or neutral, and each interaction is associated with distinct atmospheric anomalies (Tirabassi et al. 2014).
The study of SST variability in the Southwestern Atlantic has received considerable attention on annual (e.g., Podestá et al. 1991;Lentini et al. 2000;Rivas 2010;Simionato et al. 2010;Delgado et al. 2014;Luz Clara et al. 2019), interannual (Combes and Matano 2018;Bodnariuk et al. 2021) and decadal timescales (e.g., Venegas et al. 1997;Paegle and Mo 2002;Sterl and Hazeleger 2003). Performing numerical experiments with both atmospheric and ocean global models, Rodrigues Chaves and Nobre (2004) showed the predominance of negative thermodynamic feedback between the atmosphere and the ocean over the southwest tropical Atlantic during strong SACZ events. They conclude that negative SST anomalies often observed underlying the SACZ represent an oceanic response to atmospheric forcing.
The surface ocean conditions in the Southwestern Atlantic Continental Shelf (SWACS) present a very marked SST annual cycle, explaining over 90% of the total variance (Luz Clara et al. 2019). It is characterized by SST amplitude diminishing offshore and southward, and with maximum amplitudes in the inner Río de la Plata and El Rincón areas (Podestá et al. 1991;Rivas 2010;Simionato et al. 2010;Delgado et al. 2014;Luz Clara et al. 2019). Regarding the SST variability on shorter time scales (synoptic to intra-seasonal), it is known that the wind variability has a large impact on coastal dynamics (Guerrero et al. 1997;Framiñan 2005;Pimenta et al. 2008;Simionato et al. 2006Simionato et al. , 2007Simionato et al. , 2010Saraceno et al. 2014). Dominant winds, that vary from the northeast to the southwest in a scale of a few days (Simionato et al. 2005(Simionato et al. , 2007, produce alternating events of downwelling and upwelling that cool and heat surface waters along the Uruguayan coast (e.g., Framiñan 2005;Pimenta et al. 2008;Simionato et al. 2010;Trinchin et al. 2019). In addition, other atmosphere-ocean interaction processes such as marine heat waves have been lately studied in the SWACS. Marine heat waves in the Atlantic Ocean are caused by persistent anticyclonic circulation associated with atmospheric blocking, which is in turned triggered by tropical convection in the Indian and Pacific oceans (Rodrigues et al. 2019), probably associated with the MJO (Manta et al. 2018), being this mode of variability a key source of long-term variability in subtropical South America blocking frequency (Rodrigues and Woollings, 2017).
However, SST variability on IS timescales on the SWACS and its relationship with atmospheric circulation anomalous patterns have not been fully studied yet.
The aim of this paper is, therefore, to describe the IS variability of the SST anomalies in the SWACS and to explore the associated atmosphere mechanisms. The study focuses particularly on the warm season, that is when regional atmospheric IS variability is strong in association with the SACZ activity ) and when IS variability of SST in the SWACS presents higher activity, as will be shown below.

Data and methods
SST daily data used for this study (hereafter also referred to as satellite data) is a satellite product obtained from the NOAA CoastWatch Program. Data are mapped into a regular grid of 0.1° × 0.1°, and available on daily composites derived from 5-day running averages for the period extending from July 2002 to March 2014.
SST data were estimated by optimal interpolation methods merging SST observations coming from infrared and microwave sensors installed on multiple platforms. These merged, or "blended", products have good spatial coverage and a low total error compared with the individual errors for each instrument. Reynolds et al. (2002) showed that those errors can be dramatically reduced in infraredmicrowave sensor combinations. In areas where data are scarce, such as the SWACS region, those blended products provide an opportunity of evaluating SST variability even on relatively short time scales (Simionato et al. 2010). The data processing is described in Powell et al. (2008).
The seasonal cycle was calculated from the SST data, first adjusting minimum squares sinusoidal functions (harmonic analysis) to the raw daily observations (lineartrend-removed) at every grid point, and then computing the combination of the corresponding annual and semiannual oscillations. The residues after removing the seasonal cycle from the raw data are regarded as the temporal anomalies that describe the transient variability. The interannual anomalies were determined by simply subtracting the annual means from the seasonal cycle. The signals obtained after subtracting both seasonal and inter-annual anomalies from the raw data were considered as the SST anomalies (SSTA) on sub-annual time scales. SSTA were then filtered to retain only IS variations using a 10-90day bandpass Lanczos filter with 101 weights (Duchon, 1979). An Empirical Orthogonal Functions (EOF) analysis was then performed to identify the leading patterns of SSTA variability over the SWACS region. In this case, the S-mode covariance matrix of SSTA was used as input for the EOF analysis. A Fisher test was implemented to evaluate the statistical significance of the variances using a confidence level of 95%.
The associated dynamic processes were explored by computing daily mean composites (averages) of the 10-90-day bandpass Lanczos filtered anomalies of different variables taken from the ERA Interim database (Dee et al. 2011): SST, 925-hPa air temperature, sea level pressure (SLP), wind at 10 m and geopotential height (GPH) at 300 hPa. Anomalies were obtained for every variable by first removing the linear trend and then the first three harmonics of the daily average of the period 1979-2016. Daily stream function (STR) at 250 mb derived from NCEP/NCAR reanalysis dataset (Kalnay et al. 1996), were also considered for the period 1999-2018. Anomalies were computed following the same method as for the GPH though using the available reference period.
Composite significance was tested using a Monte Carlo approach by resampling events to perform the composites: the day and year of the beginning of each event was randomized 10 thousand times preserving the length of each event, and these random composites were computed to build a random distribution. In this way, the persistence of anomalies is considered. The 2.5 and 97.5 percentiles were used to consider the 5% significant values for each variable and grid point.

Leading patterns of SST IS variability
Variability on the intra-seasonal time scale accounted for up to 50% of the total SSTA variance. Highest values of IS variance were observed over the Río de la Plata Estuary (RDP; Fig. 1), in the middle continental shelf north of 40° S, and in coastal areas, particularly along the coast of El Rincón (ER; Fig. 1) area. Instead, minimum values were observed south of 40°S in the outer shelf.
The EOF analysis applied to the daily filtered SSTA indicated that nearly 40% of the variance is explained by the first three EOFs (Fig. 2); they accounted for 25.7%, 9.0% and 5.1% of the variance, respectively.
The leading pattern, EOF1 (Fig. 2a), exhibited scores of opposite signs north and south of 40°S, approximately, though the magnitude is greater in the north; it showed The temporal evolution of the three leading patterns was analyzed to better understand their characteristics. A positive (negative) event of any of the three patterns was defined as the sequence of more than three consecutive days associated with loading values above 0.4 (below − 0.4), thresholds Continental Shelf associated with the EOF1 (upper panels), the EOF2 (middle panels) and the EOF3 (lower panels) of IS variability in its positive (a, c and e) and negative (b, d and f) phases. Coloured areas indicate significance to the 95% confidence level that correspond to the mean value of the positive (negative) loadings plus (minus) one standard deviation. A summary of the positive and negative events identified for the three leading patterns is presented in Table 1. Events are described separately for the warm (ONDJFMA) and the cold (MJJAS) austral seasons of the year. The seasonal separation is motivated by the fact that atmospheric circulation anomalies exhibit on IS timescales distinctive features associated with the presence of the SACZ during the warm season or with its absence in the cold season . Table 1 also includes the dates of occurrence of the shortest and longest events identified. The summary shows that around 60% of total EOF1 events (warm and cold seasons) were negative; for the cold season, 75% of EOF1 events were negative, while for the warm season, 53% of those events were negative. Whereas the number of positive and negative EOF1 events was similar during the warm season, the negative ones dominated during the cold months with an average duration of 6 days (4 days for the shortest event and 12 days for the longest). Most of the events for both EOF2 and EOF3 occurred during the warm period, with an average duration of 9 days (31 days for the longest event) and 6 days (15 days for the longest event), respectively. Thus, in terms of the mean time span of the events, EOF2 seems to have greater persistence, a fact possibly related to the persistence of the anomalies in the atmosphere. Additionally, this EOF2 case covers a more extensive area over the ocean (Fig. 3c, d).
Therefore, as (i) the warm season shows the largest activity of these IS modes of variability, that (ii) it is when regional atmospheric IS variability is intense due to the SACZ activity , and (iii) the ocean-atmosphere interaction has been detected in summer (Rodrigues Chaves and Nobre 2004), in what follows we focus our analysis on the warm season. For the interested reader, the time series of the principal components (PC or loadings) associated to the first 3 SST EOFs are presented in the supplementary Fig. 1, for the ONDJFMA season.

Composites of high-resolution SST anomalies
Composites of the SSTA from satellite data (Fig. 3) were computed considering the positive and negative events identified during the warm season. EOF1 positive events were associated with SST negative anomalies over the shelf north of 41° S, with values cooler than − 1 °C in and around the RDP and the Uruguayan coast (Fig. 3a). Instead, EOF1 negative events were related to positive SSTA in most of the northern shelf, maximizing in and around the RDP (Fig. 3b). The composite for EOF2 positive events exhibited positive SSTA over 0.3 °C along the PDE region (Fig. 3c) and around Península Valdés (PV; Fig. 1). SSTA composite for negative EOF2 events displayed strong positive anomalies (over 0.5 °C) in the MSF area, with a strong center (> 1 °C) in the northern shelf area, centered at 37° S (Fig. 3d), and negative anomalies centered in PDE area. SSTA composite for EOF3 positive events (Fig. 3e) showed low significant SSTA values (~ 0.3 °C), positive anomalies in PDE and negative ones around 36.5° S and south of 42° S. The negative phase of EOF3 was characterized by an SSTA composite field with negative anomalies (~ 0.4 °C), with low significance, over the MSF region, between 38° and 42° S, and around PDE, and positive anomalies (~ 0.7 °C) south of 42° S, maximizing close to the coast (Fig. 3f).

Dynamics
Composites for different variables were also computed based on the positive (PosE) and negative (NegE) events identified for the three leading EOFs during the warm season. For each EOF, composites are described first on a regional domain and then over a large-scale domain to assess the role of atmospheric variability in explaining the regional atmospheric circulation anomalies. The latter was motivated by previous results (e.g., Vera et al. 2017, and references therein) showing the significant influence of tropical atmospheric convection variability as well as of the atmospheric jet stream instabilities, on regional atmospheric circulation anomalies on IS timescales.

EOF1
Composites of SSTA from ERA-Interim datasets for EOF1-PosE (Fig. 4a) showed negative values dominating over the SWAO between 30° and 44° S and further north along the Brazilian coast. Positive SLP anomalies dominated the oceanic region, with its center over the Malvinas Islands and maximum values zonally extended at around 55° S, as well as over Argentina (Fig. 4e). Low northwestward wind anomalies were observed over central and north Argentina coast and more intense off the coast (Fig. 4c). Air-temperature anomalies are shown in Fig. 4g: cold anomalies were present over eastern Argentina north of 39°S, and warm anomalies appeared south of 43° S over the coast. Composite map of 10-90 day filtered anomalies of daily STR at 250 mb and the wave activity fluxes (Fig. 7a) over the large-scale domain showed the presence of intense alternating values indicating that energy is propagated along a wave train coming from west to Argentina region, also noticeable in 300 mb ISfiltered geopotential height anomalies composites (Fig. 8a).
Composites for the EOF1-NegE showed that positive SSTA prevailed in the SWACS region north of 44° S (Fig. 4d). Negative SLP anomalies were found to be large and zonally distributed at mid-latitudes, centered around 55° S (Fig. 4f). Air temperature composites showed positive anomalies over eastern Argentina and the Atlantic Ocean north of 45° S (Fig. 4h). STR composites showed upperlevel cyclonic anomalies over the southern tip of South America embedded in a Rossby wave train (Fig. 7b). Composites for GPH anomalies in polar stereographic projection (Fig. 8b) exhibited a cyclonic anomaly associated with a zonally symmetric pattern of upper-level atmospheric circulation anomalies with cyclonic anomalies at middle latitudes and anticyclonic ones at higher latitudes.

EOF2
The composites of SSTA from ERA Interim dataset for EOF2-PosE showed large negative anomalies with maximum magnitudes over the SWACS (Fig. 5a). Wind at 10 m composites exhibited a cyclonic circulation over an extensive area of the SWACS (Fig. 5c) and, in agreement, SLP composites showed a negative anomaly centered at around 45°S of latitude and 35°W of longitude (Fig. 5e). Negative air temperature anomalies were on average observed farther away from the coast, excepting the southeastern coast of Buenos Aires province area (Fig. 5g). At the large-scale, composites of STR anomalies (Fig. 7c) showed weak activity over the South Pacific Ocean, while the composite of GPH anomalies (Fig. 8c) revealed a cyclonic circulation anomaly over the SWACS.
For EOF2-NegE, positive SSTA (ERA-Interim) prevailed in the southwestern Atlantic, with larger values between 35° and 45° S (Fig. 5b). Negative SLP anomalies extended over the southeastern Pacific and the continent's southern tip, while a center of positive SLP was located eastward centered in 45° S-30° W (Fig. 5f). In agreement, wind anomalies described a large anticyclonic circulation, with a southwestward component and larger magnitude at the subtropics and southward further south (Fig. 5d). Air temperature anomaly composites showed warm values over the region (Fig. 5h). STR composites showed upper-level cyclonic anomalies over the southern tip of South America weakening towards the ocean region. Composites of GPH anomalies (Fig. 8d) showed a regional pattern of atmospheric circulation anomalies in upper levels associated with a wavenumber-4 structure at middle latitudes, more intense over the south Atlantic and Indian Oceans as in the EOF2-PosE cases.

EOF3
EOF3-PosE were associated with negative SSTA (ERA-Interim dataset; Fig. 6a) in regions of the Argentinean coast between 36° and 38° S, and around 45°S. SLP composites showed a core of negative anomalies centered in the SWAO ( Fig. 6e) which promoted south of 39° S large northward wind anomalies along the coast, particularly northeastward anomalies at that latitude (Fig. 6c). Cold air temperature anomalies in low levels (Fig. 6g) were observed over the ocean region, particularly with more intensity south of 42° S. Composites of STR anomalies (Fig. 7e) showed cyclonic circulation over the SWAO and anticyclonic anomalies, though weak, were observed over south Brazilian coast and northwest of the Antarctic peninsula; this was also seen in GPH anomaly composites (Fig. 8e).
EOF3-NegE were associated with positive ERA-Interim SSTA dominating further south of 44° S and around 38° S (Fig. 6b). A large and intense center of positive SLP anomalies was located at around 45°W-50°S (Fig. 6f), which is associated with strong southwestward wind anomalies (Fig. 6d). Warm air temperature anomalies were present in Fig. 6h in coincidence with warm SSTA over the SWACS. The upper-level atmospheric circulation present in STR anomalies composites (Fig. 7f) showed a wave train coming westward from the Indian ocean, with a cyclonic anomaly Fig. 4 Composite maps of ERA-interim 10-90-day filtered anomalies of: SST (°C, a and b), zonal and meridional wind at 10 m (vectors) and its magnitude (colors, in ms-1, c and d), SLP (mb, e and f), and 925 hPa air temperature (°C, g and h) for EOF1 positive events (left) and negative events (right). Shading indicates statistically significant regions at 95% (see text for details) west of the Antarctic Peninsula and a strong anticyclonic anomaly downstream over the southwestern Atlantic Ocean. This pattern is also clear in the GPH anomalies in Fig. 8f.

Discussion
In the previous sections the three leading modes of SSTA IS variability over the SWACS were described together with the associated atmospheric variability over both regional and large-scale domains. We now discuss the potential mechanisms of atmospheric forcing of the leading modes on IS time scales.
For positive (negative) phases of EOF1, SSTA composites in the SWAO showed negative (positive) values along the eastern coast of southern South America to the north of 41°S, with more intensity in the RDP area (Figs. 3, 4a,b). The wind composite anomaly for the positive phase showed weak east/southeast anomaly directions over the Argentinean coast, and barely more intense anomalies from the  , a and  b), zonal and meridional wind at 10 m (vectors) and its magnitude (colors, in ms-1, c and d), SLP (mb, e and f), and 925 hPa air temperature (°C, g and h) for EOF2 positive events (left) and negative events (right). Shading indicates statistically significant regions at 95% (see text for details) east direction were observed all over the SWACS around 45° S (Fig. 4c). This wind forcing, despite its low significance, was possibly added to the average north ocean shelf current, favoring cold advection and the retreat of the Brazil current to the north, in agreement with the positive SLP observed over the oceanic region. Composites for SLP anomalies corresponding to EOF1-PosE showed the largest magnitude along the 50° S latitude, with maximum values extending to the east of the continent (Fig. 4e). Thinking in terms of Coriolis balance and Ekman transport, the wind and SLP patterns obtained for EOF1-PosE should have shown an SSTA in the coastal region different from that shown in Fig. 3a, also because an anticyclonic anomaly of the atmospheric circulation would induce an oceanic circulation coming from the south, which, in summer, would transport the warm water from the estuary northwards on the continental shelf. Therefore, since composites for air temperature anomalies at 925 hPa for EOF1-PosE showed negative values over the northeastern region of Argentina, including the RDP area and the continental shelf north of 30° S, the great cooling of Fig. 6 Composite maps of ERA-interim 10-90-day filtered anomalies of: SST (°C, a and b), zonal and meridional wind at 10 m (vectors) and its magnitude (colors, in ms-1, c and d), SLP (mb, e and f), and 925 hPa air temperature (°C, g and h) for EOF3 positive events (left) and negative events (right). Shading indicates statistically significant regions at 95% (see text for details) the atmosphere observed over the region (Fig. 4g, air temperature) might be the cause/responsible of the cold SSTA observed throughout the RDP. Thus, the cold air-temperature anomalies might influence the SST anomalies of the RDP estuary, as its temperatures are usually warmer than those of the continental shelf during summer. But in this case, the water might be losing heat to the atmosphere through sensible heat fluxes. For the opposite phase, negative SLP anomalies associated with the EOF1-NegE extended from the southeastern Pacific eastward into the Atlantic, were large and zonal at middle latitudes, centered around 55° S (Fig. 4f) and favoring north/northeast wind anomaly directions at low levels along the Brazilian coast. Therefore, in contrast to the composites for EOF1-PosE, warm waters of the Brazil current can extend farther south explaining the positive SST anomalies observed for EOF1-NegE. Moreover, the corresponding air temperature composites showed positive anomalies over eastern Argentina and the Atlantic Ocean north of 45° S, which may also favor downward sensible heat fluxes and warmer-than-normal SSTs over the whole area (Fig. 4h).
Upper-level stream function anomalies (Fig. 7) showed the presence of a corresponding large-scale pattern of alternating positive and negative circulation anomalies, coming from western tropical South Pacific. Those alternating anomalies are evidence of atmospheric Rossby wave trains that can either be excited by instabilities in the midlatitudes westerly jets or by the atmospheric tropical convection.
In particular, the anticyclonic anomaly observed over the southwestern Atlantic in the EOF1-PosE events is associated with a planetary wavenumber 3 structure extended at around 50° S in superposition with a Rossby wave train extended along an arch from Australia towards southeastern Pacific and then southern South America. The wave activity fluxes showed that energy is propagated along this wave train, suggesting that the regional anomalies are remotely influenced by tropical variability (Fig. 7a). This wave train is the most prominent feature of the southern hemisphere circulation as seen in the 300 mb IS-filtered geopotential height anomalies composites (Fig. 8a). On the other hand, the EOF1-NegE showed that the cyclonic STR anomaly meridionally extended along the southern tip of South America is associated with a more zonally symmetric pattern of upper-level atmospheric geopotential height anomalies being cyclonic at middle latitudes and anticyclonic at higher latitudes, resembling a negative phase of the Southern Annular Mode (Fig. 8b). Moreover, when computing the average of the SAM index for each negative event, it was found that more EOF1 events occur in association with negative SAM index values (not shown). Accordingly, previous studies (e.g., Carvalho et al. 2005;Pohl et al. 2010;Flatau and Kim 2013) show evidence of SAM activity on circulation anomalies in the Southern Hemisphere on IS timescales.
EOF2 composites exhibited statistically significant SSTA over almost the entire study area, including the RDP and PDE regions, for both modes (Fig. 3c, d). The composites of SSTA from ERA Interim dataset for EOF2-PosE showed large negative anomalies with maximum magnitudes over the SWACS (Fig. 5a). These anomalies can be explained in terms of the wind force over the SWAO. A negative SLP anomaly is centered at around 35° W-45° S, which is mainly related to southwest wind anomaly directions off the coast of Uruguay, forcing the retreat of the Brazil current to the north. Negative air temperature anomalies for EOF2-PosE are on average observed farther away from the coast directly related to the location of the SLP anomalies more to the east. The patterns for EOF2-PosE can be explained in terms of cold air temperature and wind forcing over the shelf: according to the barotropic vorticity balance in the ocean, since the oceanic circulation in the region is dominated by the wind curl, this circulation gets the sign of the wind stress curl; in this case this would produce advection of cold waters from the south to the north, cooling the area. Note also that the atmosphere was cold over the region; this could favor the loss of sensible heat by the ocean, contributing to the observed cooling as well (Fig. 5g). At the large-scale, composites of STR anomalies (Fig. 7c) showed weak activity over the South Pacific Ocean, and the hemispheric perspective of GPH anomalies (Fig. 8c) revealed that the cyclonic circulation anomaly observed over the southwestern Atlantic is part of a planetary wavenumber-4 structure extending along the middle latitudes in the Southern Hemisphere but that is more defined over the Southern Atlantic and the Indian Ocean regions.
For EOF2-NegE, positive SSTA (ERA-Interim) prevailed in the southwestern Atlantic, with larger values Fig. 8 Composite maps of 10-90 day filtered anomalies of 300 mb geopotential heights presented in stereographic projection. Negative contours are dashed and 0 contour is omitted. Coloured areas indicate significance to the 95% confidence level. a EOF1 positive, b EOF1 negative, c EOF2 positive, d EOF2 negative, e EOF3 positive, and f EOF3 negative between 35 and 45° S (Fig. 5d) (in particular, it displayed an over 0.5 °C positive SSTA in the MSF area north of 41°S). Negative SLP anomalies extend over the southeastern Pacific and the continent's southern tip, while a center of positive SLP is located eastward centered in 45° S-30° W (Fig. 5f). In agreement, wind anomalies described a large anticyclonic circulation, with a northeast component and larger magnitude at the subtropics and southward farther south (Fig. 5e). The anticyclonic atmospheric circulation would lead in the ocean an anomalous anticyclonic circulation along the coast (southward), driving warm waters and giving rise to the generalized warm SSTA pattern observed. In this case, the atmosphere was warmer over the region; this could reduce the loss of sensible heat by the ocean, contributing to the observed warming as well (Fig. 5h). In turn, the northeast wind anomaly over the RDP will produce a retraction of the freshwater plume to the southeast and an increase in upwelling producing the cold signal that is seen along the coast of the RDP in the composites of the satellite data (Fig. 3d). In this sense, in a recent study, performed with more accurate satellite data resolution at the Uruguayan coast, this upwelling is revealed (Trinchin et al. 2019). For this phase of the EOF2, stream function anomalies (Fig. 7d) showed that regional wind anomalies are part of a Rossby wave train coming from the Indian/Pacific Ocean. GPH anomaly composites showed (Fig. 8d) that the regional atmospheric circulation anomalies were in upper levels associated with a wavenumber-4 structure at middle latitudes for the EOF2-NegE, and as in the EOF2-PosE, more intense over the south Atlantic and Indian Oceans. Therefore, EOF2 seems to be related to the wavenumber 4 that, according to its phase, favors a positive or a negative EOF2.
EOF3-PosE are associated with negative significant SSTA (as described by ERA-Interim dataset) in regions of the Argentinean coast between 36° and 38° S and around 45° S, (Fig. 6a). Moreover, a center of negative SLP anomaly values is centered in the SWAO (Fig. 6e) which promotes south of 39° S large south wind anomaly directions along the coast, particularly southwest anomalies at that latitude (Fig. 6c). These wind anomalies along with the cold air temperature anomalies in low levels (Fig. 6g) may be forcing cold SSTA over the mentioned areas through advection and loss of sensible heat to the atmosphere. The regional circulation anomaly is, in upper levels, part of a coherent Rossby wave train emanating from northwestern Australia, with a strong positive anomaly west of the Antarctic Peninsula and a cyclonic anomaly downstream over the Atlantic Ocean. Furthermore, anticyclonic anomalies, though weak, were observed over Antarctica, which might be associated with a SAM negative phase (Fig. 8e). It was also found that most of EOF3-PosE are related to negative SAM index values (not shown).
EOF3-NegE were associated with positive SSTA dominating further south of 43° S and between 36° and 38°S (Fig. 6b), showing anomalies around − 0.5 °C over the outer MSF. A large and intense center of positive SLP anomalies was found located at around 45° W-50° S (Fig. 6f), which is associated with strong northeast wind anomaly directions (Fig. 6d). As for the EOF3-PosE, advection of warmer waters to this region might explain the SSTA, as well as sensible heat fluxes from the atmosphere to the ocean in association with the warm air temperature anomalies (Fig. 6h). The upper-level atmospheric circulation related to the EOF3-NegE (Fig. 7f) showed a Rossby wave train emanating from western Australia in opposite phase than in the EOF3-PosE, with a cyclonic anomaly west of the Antarctic Peninsula and a strong anticyclonic anomaly downstream over the southwestern Atlantic Ocean. This pattern was also clear in the GPH anomalies in Fig. 8f, and, on average, SAM phases during these events are positive (not shown).

Conclusions
This work examines the main features of SST variability in the SWACS in the intraseasonal time scale and assesses the relationship of this local variability with that observed in the SWAO and the Southern Hemisphere. Results are based on EOF analysis of IS-filtered SST daily anomalies, where three main patterns were obtained, and compositions of oceanic and atmospheric variables were made considering the days in which those modes were active.
SSTA composites for EOF1 in the SWAO showed significant values along the eastern coast of southern South America, north of 41° S, with high intensity around the RDP region. This pattern is possibly caused by sensible heat fluxes, as suggested by the associated air temperature composites over eastern Argentina and the Atlantic Ocean north of 43-45° S. EOF2 composites showed an alternating pattern between the northern RDP/PDE region and the rest of the study area, with SSTA of magnitude greater than ± 0.5 °C in the MSF area north of 41° S. These anomalies can be explained in terms of the wind force over the SWAO, and the associated water temperature advection. In the case of EOF3, the SSTA composites showed significant values in outer MSF and the RDP mouth southernmost area with opposite anomalies, more noticeable in the negative case. ERA-Interim SSTA showed additionally a significant core over the coast around 45° S. These two patterns appear to be related to both wind distributions and air temperature conditions.
There are different processes related to atmospheric forcing that could be involved with SST variability at different time scales. Kara et al. (2009) showed, from climatological monthly global means, that SST variations at all latitudes are generally strongly and positively correlated with increases in near-surface air temperature, and with vapor mixing ratio and net shortwave radiation at the sea surface, although they are often moderately and negatively correlated with increases in near-surface wind speed. Particularly, variations in the net shortwave radiation and vapor mixing ratio are found to have more influence in driving the seasonal cycle of SST than other atmospheric variables. A study of the variability of the SWAO shelf circulation based on a high-resolution ocean model, revealed that there is an abrupt change of the dynamical characteristics of the circulation at 40° S (Combes and Matano 2018). South of that latitude, the seasonality is driven by onshelf fluxes coming from austral currents, and interannual variability of the shelf circulation is principally wind driven. North of 40° S, the local wind forcing drives the seasonal variations of the shelf transport, while the interannual variability of the flow is driven by RDP runoff combined with local wind stress and the Brazil-Malvinas confluence (Combes and Matano 2018). Over the SWACS, SAM-induced along-shore wind stress anomalies modulate the along-shore transport variability. This mechanism holds also at synoptic scales, highlighting the dominant role of the wind on the along-shore transport (Lago et al. 2021). Scasso and Piola (1998) showed that there is an exchange of net heat between the continental shelf water and the atmosphere through latent heat flux, resulting in excess evaporation (E) over precipitation (P). While flowing towards lower latitudes, the water experiences an increase in salinity along its path over MSF area and, as the process of evaporation will absorb heat from the surrounding waters, a cooling of SST. The net air-sea exchange of water in the SMG area shows similar seasonality as the E rate (i.e., with a minimum in austral spring and a maximum in the fall (Scasso and Piola 1988)) while the P rate is nearly constant throughout the year. On the other hand, Luz Clara et al. (2019) identified the presence of a warm water mass in the same region (where E > P) in the transition seasons: minimum E rates in austral spring suggest less ocean heat loss, and thus, warmer SSTs; whereas in the fall, the SMG water begins to cover the ER area and the excess of E is at its peak, thus, the loss of heat through latent heat flux is greater, resulting in cooler SSTs. This exchange of net heat mechanism could be involved in the SST variability observed in the results of the present work. We believe the present study is the starting point to a profound budget analysis in the future.
Throughout this work, we showed that the SST anomalies in the SWACS exhibit significant IS variability that is, partially, influenced by that in the atmosphere on similar timescales, not only on the ocean in mid-latitudes but also extended in the tropical zones of the Pacific and South Indian Oceans. These ocean-atmosphere teleconnections could help in the future to improve ocean predictability at those timescales, particularly in the SWACS, where ecosystems are highly productive and diverse, with socioeconomic relevance for Argentina, Uruguay, and Brazil, and have a complex oceanographic dynamic (Lercari et al. 2019). For example, the demersal fish distribution is mainly associated with environmental gradients along the SWACS, where temperature and salinity are environmental parameters that have physiological importance and have a direct influence on fish distribution at all spatial scales (e.g., Jaureguizar et al. 2006Jaureguizar et al. , 2016Jaureguizar et al. , 2020Jaureguizar et al. , 2021Menni et al. 2010;García et al. 2010). The relative influence of temperature is higher within marine environment, where the spatial-temporal variability affects the distribution (Cortés et al. 2011a, b;Jaureguizar et al. 2020), reproductive pattern (Colonello et al. 2014;Elisio et al. 2017), and large-scale horizontal movements De Wysiecki et al. 2020) of elasmobranchs. The poleward shift of the SACZ in recent years, associated with the poleward expansion of the South Atlantic subtropical high (Zilli et al. 2019) would also force the southward displacement of warmer waters into the SWAO, which in turn would favor the occurrence of the intra-seasonal events associated with EOF1 negative phase. This could explain the higher "tropicalization" of the coastal marine community (i.e., the occasional occurrence of more than 30 tropical and subtropical species) observed along the northern Argentinean coast during the last decade (Milessi et al. 2018). Therefore, as a relevant predictor of fish availability along the SWACS, the predictability of temperature spatial distribution is an essential instrument to develop management strategies within the climatic change scenarios going on in the South Atlantic Ocean.