Energetic overturning flows, dynamic interocean exchanges, and ocean warming observed in the South Atlantic

Since the inception of the international South Atlantic Meridional Overturning Circulation initiative in the 21st century, substantial advances have been made in observing and understanding the Southern Hemisphere component of the Atlantic Meridional Overturning Circulation (AMOC). Here we synthesize insights gained into overturning flows, interocean exchanges, and water mass distributions and pathways in the South Atlantic. The overturning circulation in the South Atlantic uniquely carries heat equatorward and exports freshwater poleward and consists of two strong overturning cells. Density and pressure gradients, winds, eddies, boundary currents, and interocean exchanges create an energetic circulation in the subtropical and tropical South Atlantic Ocean. The relative importance of these drivers varies with the observed latitude and time scale. AMOC, interocean exchanges, and climate changes drive ocean warming at all depths, upper ocean salinification, and freshening in the deep and abyssal ocean in the South Atlantic. Long-term sustained observations are critical to detect and understand these changes and their impacts. The overturning circulation in the South Atlantic Ocean is driven by winds, pressure and density gradients, interocean exchanges, and eddies that vary spatially and temporally. A synthesis of observations reveals that waters that engage in this circulation are experiencing pronounced warming.

subtropical South Atlantic across the equator towards high latitudes in the North Atlantic where they sink and flow equatorward as cold deep water. Beneath the upper AMOC cell, a weaker abyssal cell exists, that is sourced by the sinking of colder, saline, waters along the ice edge of Antarctica 24 (Fig. 1). Together, these cells carry a maximum of 25% of the net global ocean and atmosphere energy (heat) transport 1,21,25,26 . The upper AMOC cell exerts a tremendous influence on regional weather and global climate both via air-sea exchanges [27][28][29][30][31][32][33][34][35][36] and via exchanges with the cryosphere, for instance, through sea-ice melting 37 . AMOC regulates excess anthropogenic heat and carbon storage from the surface into the deep ocean 21,[38][39][40] , which modulates the timing and regional impacts of anthropogenic climate change. All greenhouse gas emission scenarios predict that the upper AMOC cell is expected to weaken over the 21st century in response to anthropogenic climate forcing 41 , with consequential impacts on heat and carbon uptake and transport, sea level change, the water cycle, changing patterns of atmospheric circulation, extreme events, and marine ecosystems 17,22,[42][43][44][45][46][47] . Extreme freshening in the North Atlantic can trigger a weakening, or a "shutdown" of the upper AMOC cell, which paleoclimatic studies indicate can have vast consequences on the global climate [48][49][50] .
The South Atlantic Ocean plays a unique role in the global overturning circulation because it is the only ocean basin with a net equatorward heat transport 2,7,16,[51][52][53][54][55][56][57] , and it is the basin where the AMOC upper and abyssal cells interact the most (Fig. 1). Furthermore, the South Atlantic acts as the primary conduit for the deeper cold dense water masses formed in the North Atlantic to exit the Atlantic and interact with water masses in other ocean basins 8,10,[58][59][60][61][62][63][64][65] (blue pathway in Fig. 1). Interocean exchanges from the Pacific and Indian basins alter the thermohaline and chemical properties of the surface and intermediate water masses (red and yellow pathways, respectively, in Fig. 1) as they are formed, transformed, and transported through the South Atlantic on their way to the North Atlantic 65,66 . Cold and fresh water masses from the Pacific Ocean flow through Drake Passage 6,58,65,67 and largely traverse eastward to meet the warm and salty water masses from the Indian Ocean flowing through the gap between South Africa and the Antarctic continent (Fig. 1). These flows primarily enter the South Atlantic through the Cape Basin region via mesoscale eddies referred to as Agulhas Rings and the Benguela Current System (BCS) 59,63,68,69 to feed into the upper limb of the AMOC 59,60,65-67,70,71 . Interocean exchanges induce two regions of high mesoscale variability in the South Atlantic, the Confluence of the Brazil Current (BC) and Malvinas Current (MC) on the western boundary and the Agulhas Leakage on the eastern boundary (areas with elevated sea surface height variance in Box 1). These regions of high mesoscale activity and air-sea interactions contribute to the formation of new water masses [9][10][11]65 . At the deepest layers (purple pathway in Fig. 1), newly formed abyssal waters of Antarctic origin flow northward into the Argentine and the Brazil Basins [72][73][74][75][76] , cross the equator, and ultimately mix with other deep and abyssal waters.
While many of the constituent flows of AMOC are meridional (north-south) in orientation, zonally oriented (east-west) flows can play an important role in linking different AMOC pathways. This is particularly apparent in the South Atlantic, where the lower limb of the upper AMOC cell, marked by North Atlantic Deep Water (NADW) carried southward by the Deep Western Boundary Current (DWBC), splits around 20°S and loses a small portion of its flow eastward across the basin toward the African coast 12,64,77,78 (blue pathway in Fig. 1), although it remains in question whether this eastward NADW flow continues into the Cape Basin or recirculates around the Angola Basin 79 . The upper limb of the AMOC transports surface and intermediate water masses northwestward across the basin via Agulhas Rings and through the BCS and South Equatorial Current (SEC) 10,62,[80][81][82][83][84][85] (red and yellow pathways, respectively, in Fig. 1).
Deep water formation associated with the AMOC at high latitudes in the North Atlantic has been linked to changes in local stratification and in the thermohaline properties of the upper limb of the AMOC in the South Atlantic 10,16,65 . Theory and models have suggested the direction of the meridional freshwater transport by the upper AMOC cell at the southernmost border of the South Atlantic (i.e., 34.5°S) is a control parameter on the mono-or bi-stability of AMOC [86][87][88][89][90][91] . In a monostable system, AMOC imports freshwater into the South Atlantic. In a bistable system, where AMOC exports freshwater poleward out of the South Atlantic, it is possible to abruptly switch from an "AMOC on" to "AMOC off" state causing abrupt regime shifts in the water cycle and regional weather patterns leading to planetary climate impacts 47 .
Over the past two decades, the international scientific community has made numerous advances in the study of the AMOC due to the growth of long-term observing programs at various latitudes in the Atlantic basin 15,16,19,[92][93][94][95][96][97][98] . These observational programs combined with models have revolutionized our understanding of the mean structure and pathways of the AMOC, its variability on time-scales ranging from days to decades, and its drivers. While the development of a comprehensive AMOC observing system in the South Atlantic has lagged behind the North Atlantic, there is now a concerted international effort focused on studying the role of the South Atlantic dynamics in driving AMOC changes and regional climate variability. This international South Atlantic Meridional Overturning Circulation (SAMOC) initiative began in 2007 10,[99][100][101] . The SAMOC observing system consists primarily of purposely designed trans-basin and boundary current mooring arrays and synoptic ship-based hydrographic observations in the South Atlantic and in the regions of interocean exchanges (Box 1; Box 2). SAMOC also leverages global satellites, Argo profiling floats, and surface drifting buoys, which provide key observations that are useful for overturning studies. The goal of the SAMOC initiative is to monitor climatically relevant oceanic fluxes of mass, heat, and freshwater, and provide observations critical to validate and improve numerical models and climate predictions [102][103][104] .
Several recent review articles have summarized what we have learned from AMOC observing systems, but they primarily focused on the North Atlantic 19,95,96 . The only review article focused solely on the South Atlantic circulation precedes major findings from the SAMOC initiative 10 . The present review will highlight recent advances revealed by the SAMOC observing system along with newly developed observational analysis techniques. The review will serve to demonstrate the critical role of long-term sustained observations for detecting and understanding circulation variability in this region across a multitude of timescales, and our present ability to model some of the changes. Specifically, we will focus on the insights gained into changes in circulation, water masses, transports, boundary currents, and eddy-rich interocean exchanges. These changes must be monitored and studied to continue to assess and understand their known (and still unknown) impacts on climate phenomena such as precipitation patterns and extreme events (e.g., marine heat waves, hurricanes), sea level rise, and temperature and salinity trends. Finally, we present a summary and our perspective on the remaining key outstanding research questions and identify observational gaps that challenge our current ability to answer these questions.
Recent progress in understanding South Atlantic water masses formation, transformation, and pathways Properties of South Atlantic water masses have been studied since the earliest oceanographic surveys, conducted by Germany 51,105,106 , and have been actively coordinated internationally since the late 1950s [107][108][109] . Since then, the World Ocean Circulation Experiment (WOCE) in the 1990s, the Climate Box 2 | Observing South Atlantic interocean exchanges from the Pacific and Indian basins Drake Passage has been uniquely well-sampled in the Southern Ocean and has been the focus of a number of field campaigns beginning with the International Southern Ocean Studies (ISOS) program 253 . The cDrake study of Antarctic Circumpolar Current (ACC) transport and dynamics deployed 41 bottom-moored current and pressure-recording inverted echo sounders (CPIES) for a period of 4 years (2007-2011) 206 (Box 1). The cDrake array comprised a transport line spanning the full passage and a local dynamics array situated in the northern passage eddy kinetic energy maximum. Sustained observations in Drake Passage include a long-term (20+ years) time series of upper ocean temperature, salinity, and current observations, as well as concurrent shipboard meteorological and CO 2 sampling, from the year-round transits of the U.S. Antarctic Program supply vessel, ARSV Laurence M. Gould 254 . This ship crosses Drake Passage on a spatially near-repeated transect on average 20 times a year and provides a unique data set of underway continuous SADCP, meteorological, and carbon measurements, and~6 transects per year, including discrete XBT profiles at relatively high temporal and spatial resolution (designated AX22 in Box 1). The WOCE and GO-SHIP repeat-hydrography line SR1b has been occupied almost annually since 1993-1994, providing one of the longest Southern Ocean time series of full-depth measurements of temperature and salinity, enabling investigation of interannual to decadal changes in water masses of both the upper and lower cells of the overturning circulation 189 .
There have been several experiments and programs to measure contributions to Indo-Atlantic interocean exchange from the Agulhas Leakage and its variability, although sustained and comprehensive observing programs have proven difficult. These include the GoodHope program 230,255,256 , repeated GO-SHIP survey line A12, the high-resolution AX25 XBT transect, mooring arrays in the Atlantic Ocean, including SAMBA 16,62,84,85,99,100 (Box 1), and the earlier Agulhas South Atlantic Thermohaline Transport Experiment (ASTTEX) 257 and BEnguela Source and Transport (BEST) 258,259 arrays, as well as mooring arrays in the Agulhas Current that include the Agulhas Current Time-series (ACT) and Agulhas System Climate Array (ASCA) programs 260-262 . and Ocean-Variability, Predictability and Change (CLIVAR) project in the 2000s, and more recently, the Global Ocean Shipbased Hydrographic Investigations Program (GO-SHIP) have continued with full-depth sampling and establishing best practices for repeated decadal hydrographic transects 110 (Box 1). With the expansion of the international Argo program in the 2000s 111,112 , as well as research cruises in support of the SAMOC initiative (Box 1), hydrographic observations in the region have increased dramatically in the last 20 years 12,56,85,[113][114][115] . As of March 2022, there were over 540 active Argo floats (including 210 biogeochemical Argo floats and a dozen deep Argo floats) collecting water property profiles (mostly in the upper 2000 m) in the South Atlantic sector between the equator and 60°S. Beyond estimating AMOC transports (Box 1), these observations have been key in assessing the distributions, properties, and large-tomesoscale (1000 to 50 km) ocean dynamics in the basin. Moreover, by acquiring repeated hydrographic data that spans multiple decades, it is possible to assess how these waters are affected by climate change.
A simple picture of AMOC water mass pathways in the South Atlantic involves alternating northward and southward flows along surface/intermediate, deep, and abyssal layers which compensate one another, such that the total flow across any fulldepth, trans-basin line approximately balances at long timescales. These flows either move in the same density layers (adiabatic) or they eventually cross them (diabatic). Other ocean circulation systems can impact, and can be impacted by, the overturning circulation, such as the South Atlantic subtropical gyre 32,116 and the subtropical cells 115 . Due to diverse air-sea interaction processes, the upper layer is composed of water masses originating at the surface from the tropics (Surface Water), subtropics (Central Water), and the circumpolar (Antarctic Intermediate Water, AAIW) regions. These waters can be formed locally and can be influenced by waters from the Pacific, Indian, and Southern Ocean, and their complex distributions and pathways are regulated by the overturning circulation ( Fig. 1). This constitutes the time-mean background, with which the upper limb of the upper AMOC cell interacts as it makes its way to the North Atlantic Ocean. The lower limb of the upper AMOC cell involves the export of NADW to the Southern Ocean where it joins the Antarctic Circumpolar Current (ACC). NADW export is compensated by a northward flow of surface, central, and intermediate waters through the upper layers in the South Atlantic and across the equator, and by the northward flow of Antarctic Bottom Water (AABW) carried by the abyssal cell 58,69,74,76,114,[117][118][119] . Heat and freshwater transported by these interocean exchanges substantially impact the Atlantic Ocean as a whole 58,65,66,69,120 .
How AMOC water mass pathways in the South and North Atlantic Ocean are linked, and in particular how water masses cross the equator, is also a topic of considerable interest. For deep water masses, the main route from the North Atlantic to the South Atlantic is thought to be via the DWBC with some zonal detours within the deep equatorial current system 67,121 (Fig. 1). For warm surface and intermediate waters flowing northward from the South to the North Atlantic, the dominant pathway along the western boundary via the North Brazil Under Current (NBUC) and North Brazil Current (NBC) system is often referred to as the "direct" pathway [122][123][124][125] . The "direct" pathway reaches the Caribbean Sea via the NBC retroflection region, where NBC rings are spun off (Fig. 1). The "indirect" interior pathway consists of the complex zonal equatorial current system, fed in part by the NBC retroflection, and interior northward transport in the subtropical cells in the tropical Atlantic which also plays an important role in the AMOC 115 (Fig. 1).
The connection to the neighboring ocean basins adds complexity to the South Atlantic's water mass configuration. In the subtropical South Atlantic, between 35°S and 11°S, recent hydrographic surveys have provided more detailed descriptions and sources of the water masses 12,56,74,84,85,114,126 (Fig. 2c; Table 1). South Atlantic Central Waters (SACW) are produced locally as they originate from the subduction of subtropical Mode Waters in hotspots of air-sea interaction and eddy activity in the southwestern Atlantic at the Brazil-Malvinas Confluence (BMC) and the southeastern Atlantic 127,128 (Box 1). These two regional hotspots also encompass pathways whereby relatively fresh and highly oxygenated intermediate waters enter the South Atlantic. In the BMC, a regional variety of Subantarctic Mode Water is subsequently transformed into a fresh Atlantic variety of AAIW [129][130][131] , and a saltier AAIW variety that has Indian origin enters the South Atlantic through the Agulhas Leakage and BCS 56,68,80,131,132 (Fig. 2a-c).
Between~1100 and 1600 dbar (~1100-1600 m depth), an oxygen minimum layer of Upper Circumpolar Deep Water (UCDW; Fig. 2c) originating from Drake Passage flows equatorward in the Cape Basin and southward in the Argentine Basin 56 , while the trans-basin zonally-integrated UCDW transport is southward at 34.5°S. At greater depths, two different veins of NADW flow southward. Near the western boundary, NADW has more pristine properties (saltier and warmer) due to a more direct pathway from its North Atlantic source region, than the vein hugging the African continental slope (fresher and colder), which has followed a more complex and longer route connected through an interior pathway 64,67,77,78,85,117 .
Lower Circumpolar Deep Water (LCDW) and Weddell Sea AABW both flow northward along the western side of the Brazil Basin. However, the northward spreading of LCDW and AABW is blocked on the eastern side of the Mid-Atlantic Ridge by the Walvis Ridge, and only older varieties of abyssal waters are found with no clear interior pathway 7,56,60,80,114,129 (Figs. 1, 2). Abyssal water masses are mixed on their way to the north and become harder to track and distinguish from other water masses near the tropical region. Although there are some inter-study differences as to which isopycnal marks the boundary between NADW, LCDW, and AABW 12,56,133 , waters with neutral densities larger than 28.15 kg m −3 are classified as AABW, with younger waters in the western portion of the basin and older waters in the eastern portion (Table 1).
Warming and salinity changes detected from the surface to the abyssal ocean Recent observational and modeling studies suggest that a reduction in the strength of the AMOC could lead to an additional warming of South Atlantic sea surface temperatures relative to cooling of North Atlantic sea surface temperatures 35,134,135 and a weakened salinity divergence between the South and North Atlantic, resulting in a net salt gain in the South Atlantic 136 .
Tracking water mass characteristics and their changes over space and time can help to determine water mass pathways and understand their origins and impacts. As an example, a 21stcentury climate projection predicts increased Agulhas leakage in response to the poleward shift of the Southern Hemisphere westerlies, which will cause an increased transport of warm and salty water from the Indian into the South Atlantic Ocean 81 . An upper ocean warming and salinification trend have been traced by observations through the South Atlantic 137 (Fig. 3a, c) up to the western boundary at 11°S 122 from the early 2000s to the early 2010s, apparently consistent with the above-mentioned scenario. Adding more nuance, a more recent modeling study 65 showed that the combined effect of a weakening fresh South Pacific inflow and increasing salty Indian inflow on the northward AMOC transport in the NBUC region might be driving the observed salinification from 2000-2009 (Fig. 3c).
Near the surface, the observed ocean warming associated with ongoing planetary warming is fairly ubiquitous in the South Atlantic from 1980 to 2020 138 (Fig. 3a), with maximum warming, on the order of 1.2°C decade −1 in the upper 700 dbar (~700 m depth) (Fig. 3a), found along the BC and in the BMC region [138][139][140] . Upper ocean salinity has also increased from the 1950s to the 2000s (Fig. 3c), with large increases in high salinity regions such as in the sea surface salinity maximum found between 0°and 30°S [141][142][143] and in the BC and BMC region where an increase on the order of 0.15 psu decade −1 (comparable to 0.15 g kg −1 decade −1 ) in the upper 700 m is observed (Fig. 3c).
The thermocline waters most affected by long-term warming signals are the mode and intermediate waters since they are formed at the surface in both the subpolar and subtropical regions and then penetrate the South Atlantic subtropical gyre at depth [144][145][146][147] . A recent study found that about 1.5 × 10 22 J of heat was converging in the South Atlantic in the σ θ = 26.0-27.0 kg m −3 potential density layer (roughly between 200 and 700 m depth) from 2006 to 2015 148 (Fig. 3a). Despite its small area, upper ocean warming in the South Atlantic accounts for roughly 10% of global upper ocean warming. Some of this excess heat has been taken up by subtropical Atlantic mode waters since nearly 1980, which then ventilate the main thermocline and transport heat away from the surface 128,138,148,149 . South Atlantic AAIW properties also show significant warming, and their salinity seems to be decreasing over the past few decades 148,150 (Fig. 3b, d), due to competing influences from increased Agulhas Leakage and declining South Pacific inflow (salinification) vs. the intensification of the hydrological cycle (freshening) (Fig. 3d). In the mid-depth layer between AAIW and NADW, warming on the order of 0.044°C decade −1 and freshening on the order of 0.006 g kg −1 decade −1 has been observed from 2006 to 2015 151 (Fig. 3b, d). Although temperature and salinity changes in the ocean interior can be due to isopycnal heave (i.e., vertical movement of density layers), this was not the case for this mid-depth layer. Rather, changes were attributed to anomalous southward advection by the AMOC 151 .
Decadal and longer time-scale temperature and salinity trends associated with deep (2000-4000 m) and abyssal (4000-6000 m) water masses are relatively small and can be obscured by subdecadal time-scale variability 38,[152][153][154][155][156] . However, in these deep and abyssal layers, noticeable warming and freshening trends are emerging in the South Atlantic that is either expression of decadal and multi-decadal variability or of anthropogenic climate change 12,73,[157][158][159][160] . More studies are needed to determine whether these trends, should they persist, are due to changes in the properties of NADW and AABW, or their relative abundance and their vertical distributions 158,161 . The observed AABW warming is strongest in the Southwest Atlantic where the younger variety  of AABW has to cross through the Vema Channel near 31°S, 39°W. Steady and significant AABW warming has been observed from moored observations and hydrographic data in the Vema Channel of about 0.02°C decade −1 since the 1980s, which has accelerated to approximately 0.026°C decade −1 since 2005 76,162 . Abyssal warming of 0.02 ± 0.01°C decade −1 was also observed over the period 2009-2019 from continuous hourly measurements from two moored instruments located near the 4500 m and 4800 m isobaths in the northwestern Argentine Basin along the SAMBA line at 34.5°S 75 (Fig. 3e). Closer to the tropics, abyssal warming on the order of 0.02 to 0.03°C decade −1 was observed in the Brazil Basin between 0°-30°S, 40°W-10°W from recent deep Argo data (2019-2020) and hydrographic data spanning multiple decades 163 , and in the western tropical Atlantic from multi-decadal repeat hydrographic cruises 12 .
Quantifying energetic South Atlantic meridional volume, heat, and freshwater transports As described in Box 1, estimates of AMOC and associated meridional heat transport (MHT), and/or freshwater transport from the surface to the seafloor across the South Atlantic have been derived from repeated CTD sections, XBT transects, inverse models, blended in situ and satellite data, and moorings at multiple latitudes in the South Atlantic from 35 to 11°S with different record lengths, time and space scales, and accuracies. Key recent findings are synthesized and described below ( Table 2, Fig. 4).
Time-mean estimates of the upper AMOC cell and full-depth MHT at various latitudes determined from trans-basin shipboard data and also a~30-year (1993-present) blended product of in situ and satellite observations (Table 2) exhibit a wide range of northward volume transport values ranging from 11.7 to 28.0 Sv 126,164 with MHT ranging from 0.27 to 0.92 PW 56,164 and large standard deviations ranging from 1.4 to 4.6 Sv and from 0.10 to 0.22 PW, respectively 56,164 . Concurrent AMOC estimates at multiple latitudes reveal that both the variability of AMOC and MHT decrease northward from 34.5 to 20°S 19,57,164,165 (Fig. 4d, e).
Using data from the TRACOS mooring array at 11°S in the tropical Atlantic, the seasonal and shorter time-scale variability of the upper AMOC cell was investigated over the period 2013-2018 98 (Table 2; Fig. 4a). A pronounced annual (seasonal) cycle has been observed at 11°S with an amplitude of 8 Sv attributed to the variability of both the upper ocean geostrophic and Ekman transport. Based on Argo data and repeated shipboard velocity measurements combined with satellite winds, the time-mean upper AMOC flow at 11°S was estimated as 16.4 Sv 115 .
In the subtropical South Atlantic, the first estimates of the daily upper cell AMOC transport were obtained using measurements from a pair of PIES moorings near the 1300 m isobath on either flank of the South Atlantic at 34.5°S deployed as part of the pilot configuration of SAMBA 15,100 . These analyses have demonstrated that the AMOC is highly variable at time-scales ranging from a few days to a year. Using nine of the SAMBA PIES moorings that had continuous daily observations from September 2013 to July 2017, a more recent study resolved the daily AMOC transports from the surface down to~4700 dbar (~4700 m) 97 , and also found a highly variable upper AMOC cell (black solid curve in Fig. 4b). This improved and updated AMOC estimate allows for better accuracy of the volume transport estimates as well as the ability to attribute the sources of the signals. A key new result from this study was the estimation of the daily time-varying strength of the abyssal cell (defined as the maximum southward flow between 3000 dbar and the seafloor beneath the upper AMOC cell; black dashed curve in Fig. 4b); previously, this cell has only been determined from once-a-decade snapshot estimates from basin-wide sections 126,[166][167][168] or combinations of hydrographic sections with inverse models 61,169 . The flows in the upper and abyssal cells vary independently, and the upper cell circulation (time-mean of 17.3 Sv northward transport, standard deviation of 15.5 Sv) is more energetic than the abyssal-cell circulation (time-mean of −7.8 Sv southward transport, standard deviation of 6.2 Sv) at all time-scales resolved over this 4-year observational period 97 (Table 2; Fig. 4b). Of interest, historical observations yielded~7 Sv of northward-flowing AABW 170 (~4 Sv through the Vema Channel 171 and~3 Sv through the Hunter Channel 172 ), which matches remarkably well with the independent SAMBA estimate of −7.8 ± 2.7 Sv (time-mean ± bias error) for the southward upper limb of the abyssal cell 97 , assuming that this flow is compensated by the AABW in the lower limb of the abyssal flow. In addition, the amplitude of the upper AMOC seasonal cycle was 7 Sv using nine moorings 97 , compared to an amplitude of only 4 Sv when using two moorings 15 . SAMBA has continued to expand over time and now has on the order of twenty moorings 97 , but an AMOC estimate has yet to be generated using more than nine mooring sites.
Measurements from SAMBA have revealed that the variations in the upper AMOC cell at 34.5°S are somewhat more complex than what has previously been observed in the subtropical North Atlantic on intraseasonal, seasonal, and interannual timescales 13,18,19,[92][93][94]153,[173][174][175] . This is because density and pressure variations on both sides of the Atlantic at 34.5°S are important to the AMOC variations at all time-scales, from daily to interannual, based on the observations collected to date. Furthermore, at 34.5°S it is critical to independently observe the Ekman, baroclinic, and barotropic components of the AMOC volume and heat transports, as all contribute roughly equally to the semi-annual AMOC cycle, and both the baroclinic and barotropic components can dominate interannual variations in different years 15,55,97 . In contrast to the importance of both the Ekman and geostrophic contributions at both boundaries to the AMOC at 34.5°S, at 11°S the annual (seasonal) AMOC variations are primarily forced by geostrophic variations with the largest seasonal pressure variability on the eastern boundary, and secondarily by winds 98 . Despite some dynamical differences at 11°S and 34.5°S, the amplitude of the seasonal cycles observed by the  expanding our understanding of the latitudinal connectivity of the AMOC system 57 . It is plausible that MHT variations on interannual and shorter time-scales are dominated by regional processes with low coherence across the gyres and hemispheres, as has been suggested by numerical simulations 176 . Longer observational records will improve analyses of AMOC/MHT meridional coherence within the Atlantic basin on seasonal, interannual, decadal, and longer time-scales and across a range of processes.
On shorter time-scales ranging from a few days to weeks, moored observations from TRACOS (11°S) and SAMBA (34.5°S) have also revealed large amplitudes of variability and rapid changes of the boundary current transports by the NBC/NBUC, the BC, and the DWBC near the western boundary 113,122,177 and by the Angola Current, BCS, and Deep Eastern Boundary Current (DEBC) near the eastern boundary 85,178,179 . These results highlight the strength of continuous observing systems to resolve these highly variable flows near the continental slope, which in turn are key contributors to the upper and abyssal AMOC transports variability across the basin.
Moored instruments and regular hydrographic sections provide unprecedented temperature, salinity, velocity and/or bottom pressure information near the western and eastern boundaries of the South Atlantic 74,84,85,98,113,114,122,[177][178][179][180][181] . However, observations from which meridional heat or freshwater transports can be computed in the interior of the basin are still very sparse. A new technique for estimating interior full-depth temperature profiles by combining satellite, CTD, and Argo profile measurements determined an average MHT during 2013-2017 of 0.5 PW, with a daily standard deviation of 0.8 PW, for the SAMBA array at 34.5°S 55 . This time-mean estimate falls within the range of the previous results from snapshot shipboard measurements and blended products ( Table 2) and demonstrates that moored arrays on the boundary combined with ancillary data in the interior can be used to continuously observe heat transport.
Regardless of the observational platform, studies consistently show that the upper cell AMOC volume and heat transports vary in a coherent manner (i.e., the AMOC and MHT time series are strongly correlated with one another with significant correlation coefficients r greater than 0.7 (Table 2; Fig. 4b, c). Furthermore, the characteristics of their variability are very similar on intraseasonal, seasonal, and interannual time-scales. Linear regressions of AMOC volume and heat transport find that a + 1 Sv change in AMOC at 34.5°S is equivalent to approximately a + 0.05 PW increase in MHT 55,182 .
From SAMBA, the AMOC upper cell and MHT appear to be strengthening during 2013-2017 by 2.5 ± 3.8 Sv year −1 and at a rate of 0.14 ± 0.18 PW year −1 , respectively 55 . These short-term trends, while large, are not statistically significant relative to the large interannual modulation observed during this 4-year time period at 34.5°S. However, other studies suggest significantly smaller trends in transport when a longer time period is considered. For example, significant positive AMOC upper cell transport trends are observed from 1993 to 2020 at 25°S and 34.5°S (0.44 ± 0.38 Sv decade −1 and 0.48 ± 0.36 Sv decade −1 , respectively), while a significant MHT trend is detected only at 34.5°S (0.03 ± 0.01 PW decade −1 ) from the blended in situ and satellite altimetry data 57 (Fig. 4c). More years of continuous measurements are likely necessary to detect significant longer period signals and trends in the moored transport records given the large variability on short time-scales (e.g., Fig. 4b).
Cross-comparisons of the in situ/satellite AMOC/MHT estimates with that from moored arrays indicate that while the time means agree well, they usually yield low correlations and different seasonal-to-interannual variability even after subsampling the daily mooring records at monthly resolution 57,183 (Fig. 4b-e).
The lack of coherence suggests that the estimates are sensitive to the methodology and/or sampling characteristics of the observations used to derive them. While the CTD, XBT and blended records have unquestionably expanded our understanding of the oceanic fluxes across multiple latitudes, the lower temporal sampling rates from synoptic (snapshot) observations and/or monthly gridded products (which combine observations with different temporal resolutions) may miss the high-frequency AMOC and MHT fluctuations captured by the mooring arrays. Moreover, moored arrays that use bottom pressure information to provide information about the barotropic contribution produce more energetic AMOC and MHT time series 15,55,97 than the other observational methods that use an assumption of net-zero volume (mass) transport (Box 1). To better determine the state of the AMOC/MHT and understand its variability, it is necessary to understand the physical reasons for these observational disparities (taking into account the uncertainties associated with each technique) when longer time series from all of these observing systems become available.
Estimates of the meridional freshwater transport by the overturning circulation, computed either from the sea surface to the seafloor or from the sea surface to the base of the upper AMOC cell (nominally 3000 dbar), have been limited to full-depth CTD sections 56,166-168 , XBT transects 16 , and inverse models 61,184,185 . Most trans-basin observations collected to date have shown that the freshwater transport associated with the overturning circulation (M ov ) between 11°S and 34.5°S is southward 16,166,167,184 , and may have increased in strength over the past 30 years associated with increasing salinity trends in the upper ocean 184 (Fig. 3c). In contrast, a recent study found northward M ov from a 2017 trans-basin section along 34.5°S 56 . As mentioned earlier, the sign (i.e., direction) of the freshwater transport by the AMOC upper cell at this latitude may control whether AMOC exists in a bistable or monostable state, with far-reaching climate implications. Given their importance, more M ov estimates from observations are clearly needed. Continuous M ov estimates from the RAPID/MOCHA/WBTS array in the subtropical North Atlantic (26.5°N; Fig. 1; Box 1) and Argo data suggest that a linear relationship describing 91% of the variance exists between AMOC and M ov 186 and so a + 1 Sv increase in AMOC volume transport at 26.5°N is roughly equivalent to a −0.05 Sv decrease in M ov . Thus it seems feasible that future blended in situ and satellite products and moored observations can be used to construct M ov time series similar to approaches used for MHT 55,57,164,165 .

Dynamic inflows to the overturning circulation in the South Atlantic from the South Pacific and Indian basins
In the South Atlantic, branches of the AMOC are carried by currents at the western and eastern boundaries of the basin 6,65,69,119 (Fig. 1). In the west, cold fresh waters carried by the MC, a branch of the ACC, enter the basin through Drake Passage to flow northward along the coast of South America and then turn eastward as they encounter the warm salty southward flowing BC. Together, these waters flow eastward and join the warm water masses in the BCS, entering via the Agulhas Leakage, to flow northward off South Africa and Namibia. Both of these water mass routes feed the SEC, which crosses the South Atlantic south-east to north-west. Ultimately, these waters reach the equatorial regions, feeding into the NBC.
The Antarctic Circumpolar Current in Drake Passage. As well as influencing Atlantic circulation, the ACC transport through Drake Passage serves as a valuable metric for validating ocean and climate models and constraining inverse models. Few numbers have been more cited for this reason than the estimate of 134 Sv of time-mean ACC transport from the International Southern Ocean Studies (ISOS) program conducted in the 1970s 187,188 . However, a reassessment of the ISOS data estimated the uncertainty of that mean ACC transport as 27 Sv 189 . Recent observational estimates of the total ACC transport using different methodologies can differ by more than 30 Sv, with 141 Sv estimated from the DRAKE experiment 190 , 173 Sv estimated from the cDrake experiment 191,192 , and 175 Sv determined from an analysis of Argo float displacements and historical hydrography 193 . The attribution of the differences between the Drake Passage observational estimates is primarily due to differences in the estimates of the barotropic component of the ACC transport 102,192 .
Recent high-resolution global numerical simulations are providing useful context in interpreting these observational transport estimates. A multi-decadal global ocean and sea-ice simulation based on the HYbrid Coordinate Ocean Model (HYCOM) 194 with 1/12°horizontal resolution was analyzed to diagnose differences between the DRAKE and cDrake transport estimates 102 . The total time-mean HYCOM-modeled transport was 157 Sv which falls between the DRAKE and cDrake estimates, and also closely agrees with estimates from other recent dataconstrained models (e.g., 155 Sv; 195 149 Sv 196 ). In general, good agreement between the model and observations is found for the baroclinic component of transport; the HYCOM estimates 102 (Fig. 5a) track the 20-year time series of baroclinic transport estimated from annually repeated GO-SHIP hydrographic sections along SR1b 197 . The daily continuous observational estimates from cDrake permit the study of the variability of the full-depth ACC baroclinic transport in detail 191 . The energetic baroclinic transport had a standard deviation of 8.1 Sv with 65% of the transport variance concentrated at periods shorter than 60 days, highlighting the high-frequency variability in the record, while the annual means were steady during 2007-2011 191 (magenta curve in Fig. 5a). The contemporaneous five samples of baroclinic components from SR1b and cDrake agree well 191 (black dots and magenta curve in Fig. 5a). The main challenge for observations appears to be the barotropic component, which requires accurate velocity measurements across the entire passage at high spatial resolution. Shipboard ADCP (SADCP) measurements, although limited to the upper 1000 m, provide direct velocity measurements at the required horizontal resolution, on the order of 5 km 198 . The time-mean ACC transport in the upper 1000 m estimated from repeat SADCP transects of Drake Passage is 95 Sv 198 , in agreement with the 93 Sv estimated from HYCOM in the upper 1000 m (Fig. 5c). Finally, although the structure and strength of winds over the Southern Ocean have been changing over the past several decades 199,200 , there is at present little evidence of a significant trend in the recent ACC transport estimates, and no long-term trend was found in the HYCOM Drake Passage transport over 40 years 102 .
While the ACC transport has remained steady, trends in ACC eddy energy appear to be correlated with interannual and longer time-scale changes in the prevailing winds 201,202 . This phenomenon, predicted by theory and observed in eddy-permitting models, is termed eddy saturation, whereby increased winds result in increased eddy kinetic energy rather than increased circumpolar transport 203 . A related phenomenon for the meridional overturning circulation is eddy compensation, whereby poleward eddy fluxes compensate for increases in the equatorward wind-driven component of overturning transport 202 . Thus, poleward eddy fluxes are a key dynamical process for closing the AMOC in the Southern Ocean. Drake Passage is one of the few polar regions where simultaneous measurements of temperature and velocity have made it possible to estimate eddy heat fluxes. Moored observations during the ISOS program yielded consistently poleward eddy heat fluxes, although estimates varied with depth and location and were also sensitive to time series length 204,205 . Longer time series from the recent cDrake experiment 206 provided eddy heat flux estimates that were statistically stable over 2-year subsets. These fluxes were predominantly poleward, with 50% of the heat flux surface intensified above 1000 m and uniformly distributed below 207 . The strongest heat fluxes observed by cDrake were between the Subantarctic and Polar Fronts, but the 50 km spacing in southern Drake Passage may not have captured the full eddy heat flux of the narrow Southern ACC Front. The high horizontal resolution is required in southern Drake Passage where the radius of deformation is of the order of 15 km. Using seven years of highresolution SADCP and AX25 XBT temperature data from spatially near-repeated underway surveys in Drake Passage Agulhas Leakage in the Cape Basin. An interocean exchange route that is fundamental for the AMOC lies in the gap between the southern shores of Africa and the ACC. Lagrangian experiments using eddy-resolving ocean and, more recently, climate, models are able to quantify the simulated Agulhas Leakage from the Indian Ocean and allow estimates of associated water mass properties and their transformation 62,65,67,81,119,210 . In the real ocean, however, Agulhas Leakage waters are difficult to trace and the interocean fluxes, which occur as eddies and filaments spun off the Agulhas retroflection, represent an extreme challenge to quantify because of the highly energetic environment 131,[211][212][213][214] .
Recently, it has been possible to disentangle some elements of the Agulhas Leakage complexity using eastern boundary observations from the SAMBA observing system 85 (Box 1). While the total Agulhas Leakage is estimated to range from~5-15 Sv 132 to 21.3 ± 4.7 Sv 215 , the SAMBA observations quantify a time-mean northward transport of thermocline waters across 34.5°S within the Cape Basin of 24 Sv with the large variance (standard deviation of 17 Sv) due to the passing of Agulhas Rings 85 . This transport is associated with the upper limb of the AMOC and contributes to the northward transport of heat and salt (Fig. 1). The SAMBA observations demonstrate that water mass exchange in the Cape Basin occurs primarily through advection by mesoscale eddies and via broad water mass intrusions brought about by intense dipoles 84 . Importantly, the study also highlights the need to continuously sample the full water column, as inter-basin exchanges occur intermittently and can affect the buoyancy and currents at multiple depths and timescales.
Agulhas Rings act as the loci of intense water masses transformations; large amounts of Indian Ocean central waters captured in the ring cores are transformed into subtropical mode waters across a relatively large temperature range 128,216 . Earlier transport estimates of 2-10 Sv have been made by observing longlived Agulhas Rings from ships or satellites 59,[217][218][219][220][221][222] . However, simulations consistently underestimate the amount of leakage that is carried in the rings and it is mostly the non-eddy part of the flux that plays the dominant role in the modeled Agulhas Leakage transport 210,223,224 . This is most likely because the models do not adequately represent the full spectrum of mesoand submeso-scale (less than 50 km) variability, hence biasing their estimates low and introducing large uncertainty into the estimates. Recent studies that combine satellite altimetry with in situ hydrographic data have enabled a better assessment of the impact of Agulhas Rings in efficiently connecting the eastern and western boundary systems of the South Atlantic 216,222,225,226 . Despite the complex dynamics associated with the passage of the Agulhas Rings (with many splitting and merging in the Cape Basin) some rings remain as coherent subsurface structures while crossing the South Atlantic, taking about three years to reach the other side of the basin on average 216,222,225,226 . There is potential for observations of the Agulhas Current in the Indian Ocean to shed light on leakage variability that is thought to be driven by the upstream control of leakage by wind and meanders 227 , although recently, these controls have come into question 228,229 .
The SAMBA array has also led to a better assessment of the underlying time-mean southward transport of −12 Sv (standard deviation of 17 Sv) associated with the DEBC in the South Atlantic 85 . This deep current includes the southward flow of an eastern vein of NADW 64,77,78,230 . The SAMBA observations further indicate that much of the DEBC must recirculate within the Cape Basin, with a smaller portion of the flow participating in the global overturning circulation 85 .
Applications for SAMOC observations for improved ocean and climate models used for weather and climate forecasts The considerable expansion of the in situ SAMOC observational network over the past 15+ years (Box 1), particularly along the SAMBA line at 34.5°S and the TRACOS line at 11°S, have provided unprecedented information about the structure and variability of the AMOC transports at these two latitudes. However, continuous in situ observational AMOC records are still limited in space and time, and models are needed to study AMOC mechanisms and the impact of AMOC on weather and climate. Model-based studies 30,32,103,104,125,231,232 have provided important insights into AMOC behavior and impacts in the South Atlantic, but the value of those insights depends on the fidelity of the models to the real ocean. In order to assess the climate impact of the AMOC and predict future variability, it is critical that climate models used for weather and climate forecasts accurately simulate the sensitivity of the AMOC in the South Atlantic in response to forcing.
Observations from both the North and South Atlantic Ocean suggest that the geostrophic contribution to the volume and heat transports of the AMOC plays an equal or greater role compared to that played by the wind-driven Ekman transport in terms of the total AMOC and MHT seasonal to interannual variability 13,26,57,153,174,233,234 . However, studies based on earlier eddy-permitting (¼°to 1°horizontal resolution) numerical models [235][236][237][238] indicated that the seasonal to interannual variations in the modeled AMOC and MHT are predominantly controlled by the Ekman transport. The AMOC estimates at 34.5°S from monthly climatologies of temperature and salinity constructed from observations and numerical models suggested that the seasonality of the geostrophic transport from observations is largely controlled by density variations at the western boundary, but in the models, the eastern boundary dominates due to poor representation of the MC strength at the western boundary 239 . The weak seasonal cycle in the modeled geostrophic transport was primarily attributed to excessively strong baroclinicity below the surface mixed layer, whereas the observations show a strong vertical coherence in the velocity down to 1200 m.
Another well-known disagreement between observations and models is the direction of the freshwater transport induced by the overturning circulation (M ov ) near the southern boundary of the Atlantic 16,90,91,120,166,167 . Contrary to the positive values of the M ov from numerical models 35,50,[240][241][242] , most recent observationbased estimates of the M ov in the South Atlantic 16,120,166,167,184 gave negative values ranging from −0.28 to −0.05 Sv (with the exception of the recent snapshot estimate from a trans-basin cruise at 34.5°S 56 ), suggesting that changes in the AMOC would amplify the freshwater anomalies in the convection sites. As a result, contrary to the monostable state of the AMOC from models, observations suggest that the AMOC is bistable and might collapse with a large enough freshwater perturbation. The model-data difference in freshwater transports has been attributed to the biases in the modeled salinity fields: too fresh in the northward-flowing upper layer and too salty in the southward flowing lower layer 35,50,91,[240][241][242] .
Eddies play an important role in transporting heat and freshwater in the South Atlantic, and improved eddy parameterizations and/or higher resolution numerical models are required to properly simulate eddy characteristics and their impacts on regional and global climate 66,201,243 . A recent study to investigate the impact of model resolution on the subtropical South Atlantic variability demonstrated the dominant role geostrophic transport plays in controlling the MHT variability in an eddy-resolving (0.1°horizontal resolution) simulation while both the geostrophic and Ekman transports influence MHT variability in a non-eddy resolving (1°horizontal resolution) simulation 244 . Interestingly, while model MHT at the northern boundary of the South Atlantic (8°S) drives heat content variability in the non-eddy-resolving simulation, in the eddy-resolving simulation ocean heat content variability is predominantly driven by MHT at the southern boundary of the South Atlantic (35°S). Recently, a 1/20°horizontal resolution ocean model was compared to AMOC observations at multiple latitudes in the North and South Atlantic, and was able to reproduce well the AMOC volume transport variations at 34.5°S and AMOC, NBUC, and DWBC transport variations at 11°S 104 . Thus, better agreement between models and observations may be found as more models move towards eddyresolving horizontal resolution.
The increased observations in the last 15+ years in the South Atlantic are invaluable for use in the initialization, validation, and evaluation of newer generations of ocean general circulation and coupled climate models 103,104 . Models can provide information on multi-decadal to millennial time scales; however, the differences in the AMOC volume, heat and freshwater transports between numerical models and observations suggest that the AMOC variations in models respond differently to forcing than in the real ocean. It is challenging to implement model improvements based on comparatively short observational records. Thus, existing model-data differences point once again to the need for longer-term measurements in the South Atlantic from moorings to better assess the realism and representation of processes that control AMOC strength and variability in numerical models.
Summary and future perspectives Through its long-term storage and transport of heat, freshwater, and carbon, ocean circulation strongly influences the magnitude, pace, and regional impacts of anthropogenic climate change. These processes are of particular importance in the South Atlantic Ocean, as they influence, and are influenced by, AMOC variability, the development of key water masses, and interocean exchanges. This review provides a comprehensive overview of recent advances in understanding AMOC variability in the South Atlantic, enabled by the tremendous growth of the SAMOC observing system since its inception in 2007.
As a result of the excess heat absorbed from the atmosphere associated with ongoing planetary warming, AMOC variability, interocean exchanges from the Pacific and Indian basins, and other factors elucidated in this review, the South Atlantic Ocean has warmed from the surface to the deep and abyssal ocean 12,75,76,147,148,151,163 (Fig. 3a, b, e). Upper ocean South Atlantic salinity has increased, while intermediate, deep, and abyssal water masses are freshening 148,151,158 (Fig. 3c, d). Recent observational and modeling evidence suggests that a reduction in the strength of the AMOC associated with anthropogenic warming is linked to additional surface South Atlantic warming and salinification 135,136 . Continuing to monitor these water mass changes and their causes is an ongoing priority for South Atlantic research.
Tracking water masses and their changes by combining models and observations has allowed us to refine our understanding of AMOC pathways in the South Atlantic. Quantifying the relative contributions of interocean exchanges from the Indian and Pacific basins into the South Atlantic, allows for a better understanding of the warming and salinification trends observed in upper layers 65,81,122,137 (Fig. 3a, c). Observational evidence of NADW in the DEBC 85 and in the interior 64,77,78,230 in the South Atlantic has elucidated new deep water mass pathways. However, many open questions about AMOC water mass pathways remain, particularly the zonal interior paths, and the pathways between the well-observed latitudes. The South Atlantic Gateway Array (SAGA) array deployed along 10°W in 2021 measures the zonal flows carried by the BCS and Agulhas Rings and the underlying NADW interior pathway from the west to east in the mid-latitude central South Atlantic (Box 1) and will help to answer some of these questions in the near future.
The AMOC arrays have expanded over time to include more moorings to better resolve boundary current processes associated with the AMOC variability, such as wind and buoyancy forcing, advection, and/or wave propagation 85,113,122 . Even the relatively short trans-basin mooring records available to date have transformed our understanding of the energetic South Atlantic circulation, its dynamics, and its contribution to upper and abyssal AMOC variability on daily to interannual time scales 15,97,98 , illustrating the need for high-frequency measurements to avoid aliasing rapid short time scale signals. The dominant mechanisms causing changes in the AMOC strength varies between wind forcing, buoyancy forcing, and internal ocean dynamics on seasonal, interannual, and decadal time scales between 11°S and 34.5°S 15,57,97,98 (Fig. 4; Table 2). Because upper cell AMOC and MHT vary in a consistent manner 16,55,57,97 (Fig. 4b, c), changes in AMOC can impact ocean heat content, pointing to the importance of sustained AMOC monitoring at various latitudes in the South and North Atlantic.
The AMOC volume and property (e.g., heat) transports appear to be somewhat sensitive to the methodology, assumptions, and/ or sampling characteristics of the observations used to derive them. For example, the phasing (semi-annual vs. annual) and amplitude of the seasonal cycles currently vary between different studies using moored arrays or blended in situ and satellite measurements at the same latitude, with moored arrays producing records with larger transport variations than other methods. It is possible that the temporal lengths of some of the observational records are still too short to produce directly comparable estimates and/or ascertain whether any of the observing systems miss or alias important contributions to the AMOC transport variability. Future research plans include reconciling the different estimation methods when longer overlapping time series become available. It is important to acknowledge that while infrequent or monthly proxy or partial-basin observations will never be as accurate as full-depth trans-basin continuous moored measurements, these observations are cost-effective and available on a longer time-scale and at a wider range of locations, thus providing high scientific value despite the inherent limitations.
The observing systems at the gateways from the Pacific and Indian basins monitor interocean exchange processes such as buoyancy exchange, winter mixing, convection, and upwelling. Estimates of the total ACC transport through Drake Passage from a number of recent targeted field campaigns show higher ACC transports  than historical values, with increased transport largely due to improved estimation of the barotropic component of the flow 191,192 . The annual mean ACC transport appears to have been steady from observations 191,192 , and no longterm trend has been detected evaluating ACC transport in a dataconstrained model 102 . Maintaining long-term observations of the ACC barotropic component remains a challenge (e.g., the highlyresolved cDrake experiment ended in 2011), making it hard to directly and continuously monitor the total ACC transport. Thus, it becomes important to have realistic data-constrained models of ACC transport and to continue monitoring via the annual fulldepth SR1b hydrographic cruises and repeat XBT/SADCP transects. In addition, observations and models suggest that the baroclinic and barotropic ACC transports are not correlated; therefore, monitoring either transport component alone seems insufficient to assess the temporal variability of the total ACC transport 102 .
In comparison to Pacific-Atlantic, Indo-Atlantic ocean exchanges are less well observed because they occur intermittently since they are largely accomplished through eddies and/or filaments 84 . The eastern SAMBA moorings measure highly variable northward transports of thermocline waters at 34.5°S, associated with the passage of Agulhas Rings that contributes to transports by the upper limb of the AMOC. Developing the techniques and methods to better estimate the eddy flux contributions to interocean exchange, and hence reconcile the disparate estimates of transport by the Agulhas Rings 85,132,215 remains an ongoing challenge. As a step in this direction, the number of eastern SAMBA PIES moorings was doubled in 2021 from five to ten sites between Walvis Ridge and the African coast in order to better resolve the variability associated with the Agulhas Leakage into the South Atlantic.
Topographically constrained narrow boundary current flows and mesoscale features and processes associated with eddies are often inadequately represented in low-resolution numerical models, which can lead to an inadequate representation of the AMOC. SAMOC observations in the past 15+ years have provided unprecedented information about the structure and variability of the AMOC and water mass distributions and pathways, which are invaluable for validating and constraining models 65,103,104,234 . Model-data intercomparisons are helping to reconcile differences in key processes driving the AMOC in the South Atlantic, such as the role of eddies, AMOC pathways, the relative roles of geostrophic and wind-driven Ekman transports in setting AMOC variability, and the direction of the meridional freshwater transports. Recently, an ocean model with more realistic atmospheric forcing and eddy-resolving horizontal resolution was able to better reproduce the observed AMOC volume transport variations at 11°S and at 34.5°S, indicating the need for models with better representation of boundary flows over sloping topography and mesoscale eddies to accurately represent AMOC variability in the South Atlantic 104 .
Continuous evaluation and improvements of the existing observing systems are essential for further scientific advances. For example, three additional moored instruments in the interior of the basin were added to the SAMBA array in 2019-2021 to further improve estimates of the interior and abyssal flows. Future plans include extending the western and eastern SAMBA measurements to the shelf-break regions to better observe processes at the continental boundaries, where boundary currents intensify, and to better constrain basin-wide AMOC estimates. At the western edge of SAMBA in the Southwestern Atlantic two tall dynamic height moorings will be deployed in 2022 at the continental shelf break and at the base of the continental slope to add observations of the vertical structure of the BC and of the DWBC, respectively. Similarly, the eastern SAMBA array will soon include additional moorings on the South African shelf.
Most continuous observing systems in the South Atlantic have focused on measuring physical properties (velocity, temperature, and salinity) to determine volume and heat transports.
More time series estimates of AMOC freshwater transport are needed to determine whether volume and freshwater transport co-vary, and whether or not observed freshwater transports are consistently positive or negative, suggesting AMOC is monostable or bistable, respectively. Given the important role of the AMOC in modulating carbon uptake by the ocean from the atmosphere, future moored array improvements could include new biogeochemical sensors, which would allow for the estimation of oceanic nutrients and carbon fluxes. Some of the South Atlantic arrays are starting to incorporate dissolved oxygen sensors (e.g., some TRACOS and SAMBA moorings) to better track water masses and to study combined biogeochemical and physical properties changes relevant for marine ecosystems.
In addition to augmenting the measurements made by existing platforms, it is important to develop pilot observing systems to study AMOC-related processes in undersampled regions and shed light on other remaining unknowns. Such observational arrays could also include surveying from newer autonomous underwater vehicles and uncrewed surface vehicles. Pilot and process studies like the SAGA program can help to increase our understanding of the zonal component of the AMOC, and deploying more bottom-mounted moorings and deep Argo floats in the South Atlantic is important for quantifying ongoing deep and abyssal temperature and salinity changes. Expendable data pod technology systems are being tested at a few SAMBA and SAGA mooring sites to more routinely (once or twice a year) collect data recorded by bottom-mounted moorings, reducing ship time requirements and lowering the risk of data loss.
Pan-Atlantic observational efforts are crucial for improving climate models and tracking the projected decline of AMOC in the next century in response to anthropogenic forcing 19,41 . Sustained and continuous monitoring of AMOC changes in the South Atlantic is needed to assess their impacts on climate patterns, such as shifts in regional surface temperature and precipitation patterns 32 and extreme events (e.g., marine heat waves, hurricanes), the variability of the key boundary current systems 85,113,122,177 , coastal sea level rise, warming/cooling or salinification/freshening of the ocean 75,76 , deoxygenation and acidification processes. All of these phenomena have large societally-relevant planetary impacts on agriculture, fisheries, infrastructure, and health around the globe 103,245 .
Sufficiently long in situ AMOC time series in the North and South Atlantic, together with high-resolution numerical models, will enable a synthesis of the observed transports to elucidate how each key component contributes to the AMOC's mean state and variability, the forcing processes, the influence of meridional freshwater transport on AMOC stability, and the mechanisms responsible for the meridional coherence and propagation/pathways of AMOC signals on multi-decadal time scales. These studies require international coordination, collaboration, and data and code sharing among scientists within the entire Atlantic Ocean basin.
The SAMOC initiative is the result of a concerted international effort where observing components are combined towards a common objective taking advantage of the available resources. Sustaining a long-term AMOC observing system in the South Atlantic for decades will be challenging but is necessary to address the many outstanding research questions. The ability to maintain these observing systems in the South Atlantic depends not just on institutions and funding agencies in many countries, but also on the availability of resources (research infrastructure, trained personnel). Reducing the costs of these observing programs by determining which sites are key for sustained monitoring, and reducing the need for ship time to collect data from the moorings (i.e., through technological innovations), are high priorities for all of the ongoing AMOC arrays. The SAMOC observing system can learn from assessments ongoing in the North Atlantic 96 , but must be mindful that a "one-size-fits-all" AMOC observing strategy may not be possible because of the many unique features of the South Atlantic that have been highlighted in this review.

Data availability
The datasets analyzed in this review are cited in the appropriate sections of the manuscript text. Data from the South Atlantic MOC Basin-wide Array (SAMBA) can be found at: https://www.aoml.noaa.gov/phod/SAMOC_international/samoc_data.php. The gridded Argo climatology 246 used in this study is freely available (https://sio-argo.ucsd. edu/RG_Climatology.html). The Argo data were collected and made freely available by the International Argo Program and the national programs that contribute to it (http:// www.argo.ucsd.edu, http://argo.jcommops.org). The Argo Program is part of the Global Ocean Observing System.