This section describes hydrographic (i.e., temperature and salinity) measurements of shelf and abyssal waters collected by ships in the Southern Ocean. We begin with early expeditions from the 18th century that spanned almost three centuries of ocean explorations. We then focus on more recent observations from the second half of the 20th century to the present, which have collected measurements on the continental shelf and in the deep Southern Ocean (i.e., equatorward of the continental shelf break).

Long-term moorings have been deployed in the Weddell Sea, Ross Sea, and Adélie Coast to observe shelf and bottom waters. Here, we report results from sustained programs that span decades of observations.

While there is generally a focus on ice shelf basal melting in glaciological and sea-level-rise studies, the fate of glacial meltwater is of direct relevance to AABW production, as it helps set the conditions for the formation of (or lack thereof) Dense Shelf Waters (including ISW and HSSW; Herraiz-Borreguero et al., 2016; Williams et al., 2016; Silvano et al., 2018; Li et al., 2023). This dependence emphasizes the need for a better understanding of ice shelf basal melting, cavity circulation, and internal mixing (e.g., Foster, 1983; Stevens et al., 2020) as well as the need for year-round observations. Basal melting can be inferred from satellite estimates of ice thickness and flow (e.g., Moholdt et al., 2014) and downward-looking radar (autonomous phase-sensitive radio echosounders (ApRES)) deployed on the ice surface (Brennan et al., 2013; Nicholls et al., 2015). Direct observations of ocean conditions and basal melting beneath the ice shelf can be obtained using a hot-water drill to gain access to the underlying water column (Makinson and Anker, 2014). Typical instruments used in this environment include cameras, sediment corers, water samplers, CTD, current meters, turbulence sensors, and hydrographic time series moorings. Here, we briefly review in situ observations collected through bore holes beneath the three largest ice shelves in Antarctica (the Ross, Filchner-Ronne, and Amery ice shelves), where ice–ocean processes are pivotal to the properties of AABW spread throughout the global ocean.

Several tracers have been used to track the formation and spreading of AABW. The most widely used have been chlorofluorocarbons (CFCs), but several others have been employed (sea water isotopes, dissolved oxygen, radiocarbon, noble gases, and nutrients). Here, we provide a brief overview starting with CFCs.

Profiling floats are part of the Argo program and have revolutionized the field of oceanography and the ability to monitor the ocean since the 2000s (Riser et al., 2016). Such floats were designed to autonomously measure ocean properties (temperature, salinity, and pressure) between the surface and 2,000 m depth, usually every 10 days, and in areas not covered by sea ice. This design precluded monitoring of AABW and its source waters on the Antarctic continental shelf. However, over the last decade, new technological developments have enabled profiling floats to sample under sea ice (Klatt et al., 2007; Riser et al., 2016) on the Antarctica continental shelf and below 2,000 m depth (“Deep Argo” floats).

Profiling floats have been deployed on the Antarctic continental shelf since the 2010s (see Figure 11A ). These floats are designed not to surface when a specified threshold is reached (for example, when the ocean temperature falls below a certain value) in order to avoid sea ice encounters. This threshold can be chosen depending on the environmental setting and experimental design. As the position of the floats is unknown when sea ice is present, different interpolation schemes and acoustic tracking methods have been employed to reconstruct the under-sea ice trajectories. These include simple linear interpolation between known positions or more sophisticated methods that consider bathymetry, potential vorticity, sea level, and ocean density (Wallace et al., 2020; Yamazaki et al., 2020; Oke et al., 2022). It is also common to park floats at the seafloor between consecutive profiles to avoid strong drifting along and away from the continental shelf (e.g., Porter et al., 2019; Silvano et al., 2019; Wallace et al., 2020; van Wijk et al., 2022). These floats have been deployed in many locations around Antarctica, providing new insights into the seasonality of the mixed layer and its impact on shelf properties (e.g., Porter et al., 2019; Silvano et al., 2020).

Deep Argo floats, with the ability to profile to either 4,000 m or 6,000 m depth and under sea ice, can provide year-round observations of the abyssal ocean. Presently, the floats have been deployed in regional pilot arrays in specific basins where changes in bottom water had been previously observed through traditional shipboard hydrographic surveys. The arrays of Deep Argo floats simultaneously sample AABW throughout these basins at a significantly higher frequency than repeat hydrographic sections (every 10 days compared to years apart), thus providing a new perspective of AABW variability and circulation. The first array was deployed in the Southwest Pacific basin, with original test deployments in 2014 and deployment of a more complete array in 2016. The resulting data allowed for the quantification of recent AABW warming (Johnson et al., 2019) and abyssal transport (Zilberman et al., 2020) in that basin. In 2018, the first Antarctic pilot array was deployed in the Australian Antarctic Basin (see Figure 11A ), and in 2019, another pilot array was deployed in the Brazil Basin. These floats are already providing new information about the pathways and variability of AABW (Johnson et al., 2020; Thomas et al., 2020; Foppert et al., 2021). The Deep Argo floats have offered a new perspective on AABW, despite the technology still being relatively new, and they are poised to transform our understanding of AABW variability and circulation as the pilot arrays transform into a global fleet.

Marine mammals, in particular Elephant (Southern Mirounga leonina) and Weddell (Leptonychotes weddellii) seals, equipped with biological and physical electronic sensors, such as CTD instruments (Nicholls et al., 2008; Roquet et al., 2013), have provided unique observations in the Southern Ocean (see McMahon et al. (2021) for a full discussion). Seals have the ability to dive to the seafloor (up to ~2,000 m depth for elephant seals) and under sea ice, providing year-round observations in regions hard to reach (such as the Antarctic continental shelf; see Figure 11B ), especially in the winter months. Temperature and salinity profiles collected by seals have delivered new insights into the formation of dense waters on the Antarctic continental shelf (e.g., Nicholls et al., 2008; Ohshima et al., 2013; Kitade et al., 2014; Williams et al., 2016; Portela et al., 2021). These data have highlighted factors limiting dense water formation, including intrusions of warm CDW onto the shelf (Ribeiro et al., 2021; Herraiz-Borreguero and Naveira Garabato, 2022) and glacial meltwater input from melting ice shelves (Williams et al., 2016; Silvano et al., 2018; Ribeiro et al., 2021).

Recent advances in long-endurance autonomous robotic technologies are beginning to open up new avenues for AABW observation. In particular, the advent of deep ocean gliders (Osse and Eriksen, 2007; Testor et al., 2019) and autonomous underwater vehicles (AUVs; Furlong et al., 2012; Roper et al., 2021) capable of measuring in waters as deep as 6,000 m over distances of up to several thousand kilometers and periods of several months are enabling the controlled, spatially targeted observation of AABW properties and flow with fine spatio-temporal resolution—reaching areas that are difficult to access with ship-deployed instrumentation or drifting platforms. Examples include AUV (Jenkins et al., 2010; Davis et al., 2022; Davis et al., 2023) and glider (Nelson et al., 2017; Friedrichs et al., 2022) missions under Antarctic ice shelves, as well as glider campaigns on the continental shelf (Kohut et al., 2013) and AUV deployments in the abyssal Southern Ocean (Naveira Garabato et al., 2019).

The potential of these technologies to advance understanding of processes relevant to AABW is illustrated by measurements of the flow through the Orkney Passage, a key chokepoint in the equatorward export of the AABW formed in the Weddell Sea (Naveira Garabato et al., 2002; Jullion et al., 2014). During a 3-day mission, the AUV Autosub Long Range ( Figures 11C, D ) acquired hydrographic, velocity, and turbulent microstructure data within 100 m of the seafloor along a ~180-km track, with an along-track resolution of a few meters (Naveira Garabato et al., 2019). This new mode of AABW observation unveiled the occurrence of important, previously undocumented submesoscale dynamics, suggested to be centrally implicated in the wind-forced regulation of the deep ocean heat budget of the South Atlantic (Meredith et al., 2011; Naveira Garabato et al., 2019). Autonomous robotic missions have also been deployed under ice shelves, including the Filchner-Ronne Ice Shelf (AUV; Davis et al., 2022) and Ross Ice Shelf (gliders; Nelson et al., 2017). As the endurance and robustness of autonomous robotic assets expand in coming years, it is expected that the AABW observing system will increasingly rely on these technologies for process-understanding and monitoring AABW characteristics across key circulation pathways.

Inverse modeling has, to date, been the most widely used indirect approach to quantitatively determine AABW circulation on regional-to-basin scales. In their most common form, inverse models (Wunsch, 1996; McIntosh and Rintoul, 1997) enable the assessment of the “steady state” ocean circulation that is consistent with both observations (e.g., measurements of hydrographic and biogeochemical variables along transects bounding a closed box) and fundamental dynamical and thermodynamical constraints (e.g., hydrostatic and geostrophic balance; conservation of mass, heat, and salt). Inverse models of this type (commonly referred to as “box-inverse models”) are formally underdetermined (Wunsch, 1996). As they have an infinite number of possible solutions, generating a realistic and physically insightful circulation requires that the models’ initial conditions and uncertainties be defined judiciously under firm guidance from observations and sound physical principles.

The view of AABW circulation emerging from box-inverse models spanning the circumpolar domain (Ganachaud and Wunsch, 2000; Sloyan and Rintoul, 2001; Lumpkin and Speer, 2007; Naveira Garabato et al., 2014) is one of significant AABW production within, and equatorward export along, each of the three major Southern Hemisphere basins. AABW denser (neutral density) than 28.27 kg/m³ is assessed by these models to form at a circumpolar rate of approximately 7 to 20 Sv, with the factor of three variations being attributed to uncertainties associated with differing model formulations (e.g., different assumed flow-governing physics or locations of observational transects). At the regional scale, inverse modeling has been performed in the Weddell Sea, providing an estimate of AABW production in the south-western Weddell Sea of 8 Sv (Jullion et al., 2014; Naveira Garabato et al., 2016; see Figure 10B ).

The many observational challenges associated with direct observation of AABW, noted in previous sections, have motivated efforts to indirectly observe AABW, for example, using remote sensing technologies. To date, the only aspect of AABW that has been systematically estimated via remote measurements is the rate of sea ice formation in coastal polynyas around Antarctica (Tamura et al., 2008; Tamura et al., 2016). These studies developed a thin ice thickness algorithm using Special Sensor Microwave Imager (SSM/I) and estimated sea ice formation from a heat budget analysis with the assumption that all the heat loss is used for ice formation. Higher spatial resolution estimates by Nihashi and Ohshima (2015) used the Advanced Microwave Scanning Radiometer-EOS (AMSR-E) data.

Recent work highlights the importance of underwater frazil ice production in Antarctic coastal polynyas (Thompson et al., 2020; Ohshima et al., 2022). The underwater frazil ice formation prevents heat-insulating surface-cover ice from forming, thereby enabling efficient ice production. A thin ice algorithm that detects active frazil, a mixture of frazil/pancake ice and open water, has been developed for AMSR-E (Nakata et al., 2019) and SSM/I (Kashiwase et al., 2021), which provides a more accurate estimation of sea ice formation. Figure 2B shows the updated map of sea ice formation based on that algorithm. The three polynyas exhibiting the highest sea ice formation rate (Ross Ice Shelf Polynya, Cape Darnley Polynya, and Mertz Polynya) correspond to three of the four major AABW formation sites. Sea ice formation in the Mertz Polynya, however, abruptly decreased due to the glacier calving in 2010 (see the lower panels in Figure 2B ), resulting in a decrease in AABW formation (Aoki et al., 2013; Snow et al., 2018). This example demonstrates that the variability of AABW is closely linked to sea ice formation in coastal polynyas. In addition, recent changes in AABW formed in the Ross Sea have also been suggested to be associated with variability in sea ice formation (Silvano et al., 2020).

While satellite-derived estimates of sea ice formation provide useful information on AABW formation and variability, surface freshwater/buoyancy fluxes do not directly quantify the rate of dense water formation, as the transformation rate also depends on the surface density gradients (Abernathey et al., 2016; Groeskamp et al., 2019). The rate of AABW formation also depends critically on the entrainment/mixing occurring over the continental shelf and slope (e.g., Akhoudas et al., 2021; Bowen et al., 2021). After forming and descending the Antarctic continental slope, AABW spreads at depths that are not directly observable using remote sensing techniques. However, there is precedent for observing components of the ocean’s overturning circulation indirectly via satellite measurements of sea surface height (SSH; Chelton et al., 2001), sea surface temperature (SST; Deser et al., 2010), and ocean bottom pressure (OBP; Tapley et al., 2019). Indeed, there is now a substantial body of research devoted to the identification of “fingerprints” of Atlantic Meridional Overturning Circulation (AMOC) variability in SSH and SST and to the use of SSH and OBP measurements for reconstruction of subsurface meridional transports in the North Atlantic (see Frajka-Williams et al., 2019, and references therein).

Recent studies have explored the possibility of using satellite measurements to observe the circulation of AABW. All of these studies are founded on the dynamical link between the meridional geostrophic transport of AABW and the zonal momentum balance, as indicated schematically in Figure 12A . Zonal momentum input from surface winds is removed at the sea floor via topographic form stress (TFS) (see, e.g., Olbers et al., 2004). The geostrophic meridional transport of AABW, T _AABW, is exactly related to the difference between the interfacial form stress (IFS) at the isopycnal upper bound of the AABW layer and the TFS at the sea floor (Stewart and Hogg, 2017; Stewart et al., 2021).

where f is the Coriolis parameter and ρ is a reference density. Assuming that IFS varies slowly, Eq. (1) implies that variations in TFS, expressed as changes in OBP (Olbers et al., 2004), control AABW transport. The equivalent barotropic structure of the ACC means that these TFS changes may have an expression in SSH.

For example, Mazloff and Boening (2016) used the Southern Ocean State Estimate (SOSE; Section 4.3) to show that variability in the deep, geostrophic flow of AABW into the Pacific could be inferred from OBP measurements. Mizobata et al. (2020) combined SSH measurements with in situ hydrography to compute the export of AABW by a series of standing eddies in the Indian sector of the Southern Ocean and showed that the velocity structure is amenable to indirect inference from satellite observations alone. Stewart and Hogg (2017) performed controlled experiments using an idealized high-resolution channel model of the ACC to show that changes in the multi-annual-mean export of AABW lead to potentially measurable signatures in SSH and OBP. Figure 12C schematically illustrates the suggested correspondence between deep AABW outflows across the Southern Ocean and anomalies in SSH and OBP that might be leveraged to indirectly observe its transport. Changes in the surface circulation of the standing meanders in the ACC may also be expected to produce a signature in SST, but the potential for indirectly observing AABW transport via SST remains to be explored.

It may also be possible to observe dense water overflows at the shelf break closer to the source from satellite-observed SSH. Modeling work has shown that simulated dense water overflow pulses in cross-slope canyons in the Ross Sea are associated with a negative steric height anomaly (Morrison et al., 2020). A similar signal is seen in the North Atlantic, where dense overflow variability has been detected in concurrent observations of satellite SSH and in situ mooring data (Høyer and Quadfasel, 2001). It has been proposed that the Denmark Strait overflow transport variability may be able to be monitored solely from satellite data (Lea et al., 2006). However, Haine (2010) concluded that SSH measurement noise in altimetry data available in 2010 was too large to allow for accurate detection of Denmark Strait overflow variability. An additional challenge for observing Antarctic overflows from satellites is the presence of sea ice. However, with the development of methods to detect SSH beneath sea ice (e.g., Armitage et al., 2018; Auger et al., 2022) and additional recent altimetry missions, the possibility of detecting Antarctic overflows from satellites warrants further investigation.

Recent works have drawn additional links between the deep circulation of AABW and near-surface winds. Tandon et al. (2020) showed that winds drive substantial fluctuations (several Sverdrups) in the Indo-Pacific overturning streamfunction, due to the long time scale over which the stratification adjusts to wind forcing, associated with the westward propagation of baroclinic Rossby waves. Stewart et al. (2021) showed that a similar phenomenon occurs at high southern latitudes (south of 30°S). Regarding time scales shorter than a few years, the baroclinic (IFS) component of the zonal momentum balance responds negligibly to wind fluctuations, whereas the barotropic (TFS) component adjusts to compensate for the wind changes (see Figure 12B ). This leads to AABW northward transport fluctuations that are approximately equal in magnitude and opposite in sign to wind-driven surface Ekman transport fluctuations on time scales shorter than a few years. This finding potentially allows high-frequency variability in AABW transport to be indirectly observed/reconstructed from the zonal surface wind stress, which can be calculated from reanalysis products or surface pressure measurements. However, such high-frequency fluctuations pose a challenge for efforts to measure AABW export from the Antarctic continental shelf across the Southern Ocean, which likely manifests as interannual-to-decadal fluctuations in AABW transport (Abrahamsen et al., 2019; Silvano et al., 2020).

Data assimilation products have the potential to add value to sparse observations by filling data gaps with dynamically realistic models. Data assimilation acts to “nudge” a model toward observational constraints. Similarly, a state estimate is produced by obtaining the least squares fit of a forward-running ocean model to a wide range of available observations to produce a temporally evolving three-dimensional field (e.g., Wunsch and Heimbach, 2007). In contrast, inverse models, as discussed in Section 4.1, usually lack temporal resolution and primarily incorporate only hydrographic section data.

A recent intercomparison project compared the global performance of 26 ocean data assimilation products (Balmaseda et al., 2015), highlighting the recent advancement of such products. However, only a few of these data assimilation models are evaluated for AABW formation and circulation. For example, Azaneu et al. (2014) showed that AABW in the ECCO2 reanalysis is formed by deep convection in the Weddell Sea rather than by on-shelf dense water production and subsequent descent down the continental slope (Menemenlis et al., 2008). A range of data assimilation products (ECMWF-ORAS4 (Balmaseda et al., 2013), CFSR (Saha et al., 2010), MyOcean-UR025.4 (Ferry et al., 2012), ECCO2 (Menemenlis et al., 2008), ECCOV4r4 (Forget et al., 2015), SODA 2.2.4 (Carton and Giese, 2008; Giese and Ray, 2011, and SOSE (Mazloff et al., 2010)) are compared in Dotto et al. (2014); Aguiar et al. (2017), and Bailey et al. (2023). While some of these reanalyses formed AABW at realistic rates (e.g., ECCOV4r4, SODA, and SOSE), all of the models did not well capture the spatial distribution and variability of AABW compared with observations. Despite this discrepancy between data assimilation models and observations, a few studies have utilized data assimilation products to study the hydrography and circulation of AABW (e.g., Mazloff et al., 2013; Van Sebille et al., 2013; Abernathey et al., 2016; Cessi, 2019; Rousselet et al., 2021; Stewart et al., 2021).

The physics of AABW formation and export are thus rarely well captured in forward ocean models (Renner et al., 2009; Kerr et al., 2012; Heuzé, 2021), whether or not they assimilate hydrographic observations. In the rare cases where AABW processes are better captured (e.g., ACCESS-OM2-01; Kiss et al., 2020; Solodoch et al., 2022), it is clearly dependent on high resolution (Mensah et al., 2021) and thus comes at a computational expense that restricts the length of simulations. These limitations hamper our ability to characterize the role of AABW in global ocean ventilation. The Ocean Circulation Inverse Model (OCIM; DeVries and Primeau, 2011), which is a three-dimensional dynamical ocean model optimized to fit a wide range of relevant hydrographic observations, provides an alternative tool to characterize AABW pathways. This circulation estimate is steady and has a relatively low spatial resolution (2° in the horizontal, 48 layers in the vertical; Holzer et al., 2021). However, because it assimilates six tracers (temperature, salinity, CFC-11, CFC-12, radiocarbon, and Helium-3) whose distributions are reproduced with high accuracy, it allows faithful quantification of ventilation rates and pathways in the global deep ocean (DeVries and Holzer, 2019; Holzer et al., 2021). For example, the Ocean Circulation Inverse Model has provided revised estimates of the strength (approximately 8 Sv) and structure (confined to depths > 2.5 km) of the abyssal overturning in the Pacific Ocean (Holzer et al., 2021). These estimates reflect a modern mean state; changes in AABW cannot be assessed with this tool.

Q1) What is the contribution of AABW to the lower limb of the meridional overturning circulation?

Box-inverse models provide a wide range of AABW formation rates (7–20 Sv; Section 4.1). The weaker estimates (<10 Sv) are consistent with diagnostics from CFCs (Orsi et al., 1999) but are difficult to reconcile with direct measurements of the northward export of AABW at selected chokepoints, such as the Orkney Passage (through which 4–6 Sv of AABW has been shown to leave the Weddell Sea; Naveira Garabato et al., 2002; Jullion et al., 2014; Abrahamsen et al., 2019) and the deep boundary current to the east of the Kerguelen Plateau (via which ~12 Sv has been observed to flow equatorward; Fukamachi et al., 2010). These comparisons suggest that AABW production is potentially in line with, or may possibly even exceed, the highest inverse model estimates and signal the robust quantification of circumpolar AABW formation as well as mixing as a major outstanding challenge.

Coupled with this problem, there is a long-standing discussion on the relative contributions of different Southern Ocean sectors to AABW production and northward export, ranging from a view of the dominance of the Weddell region (e.g., Lumpkin and Speer, 2007) to another of approximate equipartition between the Atlantic, Indian and Pacific sectors (e.g., Naveira Garabato et al., 2014), as also suggested by recent modeling work (Solodoch et al., 2022). Settling this issue will entail gaining a greater quantitative understanding of zonal, inter-sector exchanges of AABW. It is plausible that the Weddell’s disproportionate importance as an AABW formation hotspot suggested by some works may reflect the accumulation in, and subsequent export from, the Weddell Sea of AABW produced in the Indian and Pacific sectors (Jullion et al., 2014), which may be promoted by the westward flow prevalent along much of the Antarctic continental slope (Naveira Garabato et al., 2014; Solodoch et al., 2022).

Q2) What is the role of AABW in ocean heat storage and biogeochemical cycles?

AABW plays an important role in storing heat in the ocean abyss, thus affecting the planetary heat budget as well as the sea level (Purkey and Johnson, 2013). Moreover, AABW supplies oxygen and “preformed” nutrients to the global ocean abyss, thus affecting marine ecosystems and the carbon cycle (Marinov et al., 2006; Henley et al., 2020). However, very little is known about the impact of AABW changes on deep ocean heat and carbon storage as well as deep ocean oxygenation. Historically, abyssal biogeochemical properties have been less observed than physical properties, hampering our knowledge of the full climatic impact of AABW and its changes.

Q3) What are the temporal changes in AABW formation and how are they driven?

Observations during the past decades have detected temporal changes in AABW formation. However, relatively short, sparse, and intermittent observations prevent us from capturing all time scales of AABW variability. At interannual to multidecadal time scales, several forcing mechanisms have been shown to regulate AABW formation variability. These include atmospheric variability associated with climate modes, sea ice formation, ice sheet melting, iceberg calving, and tides (e.g., Gordon et al., 2009; Snow et al., 2018; Silvano et al., 2020; Hattermann et al., 2021; Jacobs et al., 2022; Bowen et al., 2023). However, the paucity of observations prevents us from 1) fully capturing the dynamics behind these forcing mechanisms, 2) quantifying their relative roles at different time scales, and 3) verifying whether their impact changes spatially.

Q4) What regulates AABW transport through the Southern Ocean?

Even less is known about what controls AABW transport through the Southern Ocean and its changes. Only a few mooring-based estimates of flow are available (Fukamachi et al., 2010; Abrahamsen et al., 2019). In addition, a limited understanding of AABW mixing processes in the deep Southern Ocean hampers quantification of AABW’s contribution to tracer transports and budgets.

Q5) What is the role of seasonal processes on AABW formation and circulation?

Seasonal variability (in particular wintertime processes) remains poorly understood. Our knowledge of key processes such as sea ice formation, convection, cross-shelf exchange, and gravity currents relies on very few observations collected in rare winter expeditions, mooring deployments, (deep and ice-capable) Argo floats, and seals.

Q6) How do AABW anomalies form and propagate?

AABW has warmed, freshened, and contracted over past decades within the Southern Ocean and to its north (e.g., Purkey and Johnson, 2010). However, the mechanisms responsible for these trends are not fully understood. The relative roles of changes in the source waters of AABW, waves, advection, mixing, and past offshore polynya events (e.g., the Weddell Polynya in the 1970s) in regulating multidecadal trends in AABW properties need further examination.

Q7) What is the role of small-scale ocean dynamics on the formation and export of AABW?

While recent advances in microstructure measurements, AUVs, and gliders have started to show the importance of sub-mesoscale and turbulent processes in AABW-related processes (Naveira Garabato et al., 2019; Spingys et al., 2021), a global quantification of their impact is missing. The multiscale nature of the lower limb of the global overturning circulation is only starting to emerge now from an observational point of view.

To address the outstanding questions highlighted above, an internationally coordinated and strategically designed observing system is required. Here, we suggest some priorities for an AABW observing system (summarized in Figure 13 ), taking into consideration what has been measured in the past:

1) Continue long-term summertime time series of HSSW and ISW properties in the Ross Sea, Weddell Sea, and Adélie Land (Section 2.1.2) and initiate repeat measurements in Prydz Bay and Cape Darnley (Q3).

2) Design targeted campaigns in poorly sampled coastal regions using the latest technologies in measuring ocean properties, including biogeochemical parameters, under ice. These areas include the south-western Weddell Sea, eastern Ross Sea/western Amundsen Sea, and several regions around East Antarctica where only seals have been able to provide in situ observations (Q1, Q2, Q3, and Q5).

3) Maintain monitoring by moorings in key regions of AABW formation: Ross Sea, Weddell Sea, and Adélie Land (Section 2.2). Sustain mooring observations in Prydz Bay and Cape Darnley. Moorings should ideally cover continental shelf, shelf break, and slope to capture processes involved in AABW formation and could be extended to also include biogeochemical sensors that measure additional ocean properties and provide new insight (Q2, Q3, and Q5).

4) Sustain long-term mooring observations of AABW variability and export to other ocean basins (Weddell Gyre and Orkney Passage; Section 2.2) and establish new long-term (i.e., decadal) mooring systems in key chokepoints such as east of the Kerguelen, South Sandwich Trench, and Campbell Plateau (Q1, Q2, Q4, Q5, and Q6).

5) Repeat Southern Ocean GO-SHIP transects (Section 2.1.3) at least once every 10 years and prioritize the annual/biannual occupations of more frequent sections (e.g., A23, SR1b, SR3, and SR4) (Q2, Q3, Q4, and Q6).

6) Sustain under-ice shelf measurements at the Ross and Filchner-Ronne Ice Shelves and possibly re-establish a program at the Amery Ice Shelf (Section 2.3; Q3, Q5, and Q7).

7) Tracer measurements should become standard in every oceanographic campaign (Section 2.4). In particular, CFCs and oxygen isotope measurements should be prioritized given their ability to provide essential information on AABW ventilation, freshwater sources, and anthropogenic imprint. Careful consideration is needed to avoid mismatches between different laboratories when processing oxygen isotope samples (Q1, Q2, Q3, Q4, Q5, Q6, and Q7).

8) Further, strengthen the array of (ice-capable and deep) profiling floats and seals measuring ocean physical and biogeochemical properties on the Antarctic continental shelf and in the abyssal Southern Ocean (Sections 3.1, 3.2). Development of more sensors to be installed on these platforms (e.g., microstructure) would be highly beneficial (Q2, Q3, Q5, Q6, and Q7).

9) Design and deliver new missions with AUVs and gliders (Section 3.3) for multi-month, high vertical and horizontal resolution observations in the Southern Ocean (Q2, Q4, Q5, and Q7).

10) Further develop box-inverse and data assimilation models, including at a regional scale. The combination of different observationally guided modeling approaches is essential given their strengths and weaknesses (Sections 4.1, 4.3; Q1, Q2, Q3, Q4, Q5, Q6, and Q7).

11) Further refine remote sensing approaches to estimating sea ice formation rates from satellites (e.g., Tamura et al., 2008; Ohshima et al., 2022; see Section 4.2). Further work on indirect observations of AABW circulation through satellite-derived SSH, SST, and OBP needs to be progressed beyond theory, except in a few specific regions (Mazloff and Boening, 2016; Mizobata et al., 2020; see Section 4.2) (Q3, Q4, Q5, and Q6).

12) Develop and deploy new technologies. One example of (relatively) low-cost, autonomous tools that can be used to monitor conditions (e.g., on the Antarctic continental shelf) is recording “pop-up” buoys (e.g., Long Term Underwater Sensing bottom landers, LoTUS, and T-pops). The buoys register temperature during a pre-set time interval (years) after which they let go of their anchor and “pop up” to the surface. They are designed to survive in and below sea ice and to transfer data via satellite once out of the ice. Similarly, ice-tethered profilers employed in the Arctic (Toole et al., 2011) can be deployed in Antarctic coastal regions to collect year-round high-resolution observations near the surface.

A further opportunity for sustained low-cost observations at high space-time resolution (meter to centimeter, minute to millisecond) is presented by distributed fiber optic sensing (Hartog et al., 2018). This technology can be utilized in a wide range of marine environments using dedicated and legacy cables to measure temperature (Selker et al., 2006), including in Antarctic environments (e.g., Kobs et al., 2014). Additionally, recent studies have demonstrated the potential of utilizing strain to observe ocean properties (Williams et al., 2022; Spingys et al. ² ). Finally, further development and deployment of sound sources for under-ice navigation are envisaged, especially on continental shelves and under-ice shelves.

13) Machine learning can be used in several ways to help address AABW-related questions. For example, a possible means of circumventing the limitations of satellite-derived proxies of AABW circulation (SSH, SST, and OBP; see Section 4.2) is to leverage machine learning techniques, which have been used, e.g., to infer subsurface velocity fields (Chapman and Charantonis, 2017) and subsurface meridional heat transport/overturning from SSH (George et al., 2021; Solodoch et al., 2023).