Deep ocean inventories of dissolved chlorofluorocarbon-11 (CFC-11) along representative sections off Antarctica provide the first estimate of the overall strength of all dense water sources in the Southern Ocean. Their formation rates are reported for three density layers that span the main water masses involved in the lower limb of the Thermohaline Circulation (THC). The bottom layer is supplied via sinking of Antarctic Bottom Water (AABW) produced at a few continental shelves. The middle layer receives the offshore injection of ventilated Modified Circumpolar Deep Water (MCDW) produced along much of the lengthy Antarctic Slope Front. The top layer is ventilated by northward export of Antarctic Surface Water into the Upper Circumpolar Deep Water of the Antarctic Circumpolar Current. Average Southern Ocean inputs to the upper two layers of the deep ocean for the 1970–1990 period are derived on the basis of the CFC-11 distributions along meridional sections, the mean CFC-11 saturations of all water mass ingredients, and the inferred mixing leading to production of dense source waters. About 5.4 ± 1.7 Sv (1 Sv = 106 m3s−1) of near-freezing Shelf Water ventilate the bottom layer, and 4.7 ± 1.7 Sv and 3.6 ± 1.3 Sv of cold Antarctic Surface Water ventilate the middle and top layers. Therefore the total contribution of ventilated Southern Ocean waters to the lower limb of the global THC is about 14 Sv. This is close to the about 17 Sv estimated for North Atlantic near-surface sources from CFC-11 inventories. Their entrainment of 7.4 ± 2.4 Sv of CFC-poor subsurface Lower Circumpolar Deep Water during the formation and sinking of AABW and MCDW raises the total Southern Ocean input to the deep ocean to about 21 Sv.
 The global thermohaline circulation is driven by sinking of cold upper waters at high latitudes, where complex interactions with the polar atmosphere and ice result in net buoyancy loss. While near the sea surface, waters acquire chlorofluorocarbons (CFCs) from the atmosphere, where concentrations have been increasing since anthropogenic release started in the early 1930s. But air-to-seawater transfer of CFCs at high latitudes often does not reach solubility equilibrium, since it strongly depends on sea-ice cover and other physical factors. Nonetheless, upon mixing with subsurface waters and sinking to great depths, only these dense polar surface waters can inject dissolved CFCs into the previously CFC-free deep ocean reservoir. Deep western boundary currents then carry the cold, recently ventilated waters toward lower latitudes [Weiss et al., 1985; Smethie, 1993; Orsi et al., 1999]. The combined sinking, entrainment and equatorward spreading of high-latitude upper waters is generally referred to as the ventilation of the deep ocean. Because the polar overturning provides the coldest source waters involved in the global meridional redistribution of heat, it is a critical regulator of Earth's climate. Determining the overall strength of Southern Ocean overturning is therefore relevant to better understanding the present state of climate. A global (northern plus southern) input of 25 Sv to 30 Sv (1 Sv = 1 × 106 m3s−1) associated with deep water formation has been used for decades, which seems consistent with the general notions of abyssal upwelling and diffusivity in the ocean interior [Munk, 1966; Munk and Wunsch, 1998].
 Recent attempts to determine the total rate of sinking of near-surface ventilated waters into the deep ocean have resulted from basin to global scale assimilations of a suite of deep ocean properties. Broecker et al.  analyzed the radiocarbon budget for the deep Atlantic ocean north of 30°S to conclude that about 15 Sv of surface water must sink below 2200 m in the northern North Atlantic. When combined with the phosphate-oxygen relationship (PO4*) in the deep (3000 m) Indian and Pacific oceans, an equivalent input of surface waters was inferred within the vast interior regions to the south of the Antarctic Circumpolar Current (ACC). A similar regional partitioning (13 ± 4 Sv in the north and 15 ± 4 Sv in the south) of the global overturning to depths greater than 1500 m was confirmed by [Peacock et al., 2000], using a simple box model with temperature and salinity as additional constraints. Analyses of CFC measurements have provided independent estimates of Antarctic Bottom Water (AABW) as 8.1 Sv [Orsi et al., 1999, hereinafter referred to as OJB1999] and of North Atlantic Deep Water (NADW) production as 17.2 Sv [Smethie and Fine, 2001; hereinafter referred to as SF2001] averaged over the past few decades. The optimal solution to a recent global box model inversion [Ganachaud and Wunsch, 2000] of selected hydrographic data collected during the World Ocean Circulation Experiment (WOCE), which also requires conservation of silica within neutral density layers, renders a global thermohaline circulation driven by a net deep water formation of 15 ± 2 Sv in the northern North Atlantic (γn > 27.72 kg/m3) and 21 ± 6 Sv in the Southern Ocean (γn > 28.11 kg/m3).
 Comparison among the estimates of the strength of southern source waters is not straightforward. Aside from their use of slightly different upper limits to the deep ocean, e.g., a level surface (1500 m) in Broecker et al.  versus an isopycnal surface (γn = 27.72 kg/m3) in Ganachaud and Wunsch , difficulties also arise from incompatible definitions of water masses as well as the location and spatial extent where sinking takes place in the Southern Ocean. Both Broecker et al.  and Peacock et al.  considered the production of extremely cold (θ < −1.9°C) well-ventilated Shelf Water (SW) prior to their descent around the Antarctic perimeter. Ganachaud and Wunsch  estimated the conversion of a much warmer (1°C < θ < 2°C) water mass, the Circumpolar Deep Water (CDW), over the entire region south of about 20°S to 30°S. Neither of these two definitions for the southern source water matches the characteristics of the classic AABW (θ < 0°C [Gordon, 1966; Carmack, 1977], γn > 28.27 kg/m3 (OJB1999)) filling the abyssal layer of the Southern Ocean basins.
 Caution must be exercised while equating total sinking and ventilation in the Southern Ocean with rates estimated for just the AABW component since there is compelling evidence that injection of less dense Antarctic mixtures also takes place at intermediate levels near the continental margins [Carmack and Killworth, 1978; Whitworth et al., 1994; Rintoul and Bullister, 1999].
1.1. General Approach
 In this study we expand upon a previous work on the AABW production (OJB1999) to estimate the total rate of sinking in the Southern Ocean by accounting for all recently ventilated waters exported along the Antarctic margins to the rest of the deep ocean, and then we compare southern rates for specific density layers to CFC-based estimates of sinking in the North Atlantic (SF2001). After a brief review of the main mechanisms producing dense waters in the northern North Atlantic and along the Antarctic margins (section 2), we describe the meridional circulation and ventilation of the CDW of the ACC in section 3. In section 4 we estimate the magnitude of the lateral contribution to CDW by Southern Ocean waters based on CFC-11 inventories calculated on selected meridional sections in each basin (Figure 1), the volumetric CFC-11 inventory in the AABW calculated by OJB1999, the CFC-11 saturation levels of surface and shelf waters found near the Antarctic continental slope, and the recent definitions of Southern Ocean water masses [Whitworth et al., 1998; OJB1999]. Thus, for the first time, we are able to make an independent estimate of the total rate of deep and bottom water production in the Southern Ocean and to compare it to production rates of deep waters in the North Atlantic based entirely on CFC budgets.
 We have divided the world's deep ocean into three density layers (Table 1) that span the main deep water masses of the Southern Ocean, namely the AABW and the Lower and Upper portions of the CDW (LCDW and UCDW). The bottom layer (γn > 28.27 kg/m3) is filled exclusively by the sinking of new AABW types down the Antarctic continental slope, since there is no corresponding source of water in the North Atlantic at this density (OJB1999). As discussed in section 3, the much thicker middle layer above (27.98 < γn < 28.27 kg/m3) receives inputs of ventilated CDW from the south and of Nordic Seas overflows (SF2001) from the north. The top layer (27.7 < γn < 27.98 kg/m3) captures the northward injection of Antarctic Surface Water (AASW) into the ACC, and the end product of winter convection in the Labrador Sea from the north (SF2001).
Table 1. Deep Ocean Layers and Water Mass Acronyms Definition
Neutral Density, kg/m3
27.7 < γn < 27.98
27.98 < γn < 28,27
γn > 28.27
Antarctic Bottom Water
Antarctic Intermediate Water
Antarctic Surface Water
ACC Bottom Water
Circumpolar Deep Water
High Salinity Shelf Water
Ice Shelf Water
Lower Circumpolar Deep Water
Low Salinity Shelf Water
Modified Circumpolar Deep Water
Upper Circumpolar Deep Water
Atlantic Surface Water
Classic Labrador Sea Water
Denmark Strait Overflow Water
Ice-Scotland Overflow Water
Labrador Sea Water
North Altantic Deep Water
North Indian Deep Water
North Pacific Deep Water
Upper Labrador Sea Water
1.2. CFC Data
 CFC measurements from various sources (Table 2) were used in this study. Southern Hemisphere inventory computations presented in section 4 are derived from the AJAX (1983–1984)-SAVE (1989) cruises in the Atlantic Ocean, in the Indian Ocean from WOCE cruises I9S (1995) and SR3 (1991), and in the Pacific Ocean from WOCE cruises S4 (1992)-P15S (1995) and from Palmer (1994)-WOCE P18S (1994) cruises. Except for the Palmer 1994 cruise, all of these CFC measurements were used by OJB1999 to compute the AABW layer inventory for the Atlantic and the Indian-Pacific sectors, separated near 80°E by the Princess Elizabeth Trough (PET, Figure 1). Following OJB1999, our new CFC inventories are also normalized to 1987 in the Atlantic sector (where the data time interval was 1984–1990), and to 1993 in the Indian-Pacific sector (where the data time interval was 1991–1996). Additional WOCE and pre-WOCE CFC data were selectively analyzed to determine percent saturations in all the major ingredients producing dense Antarctic water mixtures near the Antarctic continental margins (Table 3).
Table 2. Cruises of Analyzed CFC and Hydrographic Data
 In a sharp contrast to the Southern Ocean, open ocean convection is a major mechanism ventilating the northern North Atlantic, where it contributes to both the upper and lower components of NADW. Although the densest, coldest waters of the North Atlantic are trapped to the north of the Greenland-Iceland-Scotland Ridge, some is able to enter the world's deep ocean through gaps in the ridge. Iceland-Scotland Overflow Water (ISOW) enters the eastern basin and then continues as a boundary current into the western basin via the Charlie Gibbs Fracture Zone. Denmark Strait Overflow Water (DSOW) enters the western basin, filling the bottom layer. While sinking down the slope, both overflows entrain surrounding waters; but only the ISOW incorporates a component with relatively higher CFC concentrations [Smethie, 1993; Doney and Bullister, 1992]. This is because ISOW is drawn from a poorly ventilated density horizon in the Norwegian Sea, whereas winter open ocean convection in the Greenland and Iceland seas produces CFC-rich waters that extend down to depths of 1000 m or greater and contribute to DSOW.
 Relatively young, CFC-rich DSOW is sampled at the bottom of station 567 (Figure 3a). The bottom CFC-11 maximum (Figure 2a) at the foot of the northern continental slope (>2 pmol/kg at Z > 3000 m) indicates the southward flowing ISOW/DSOW within the North Atlantic deep western boundary current and lateral mixing and circulation into the interior. It extends farther to the south as an abyssal (∼4000–4500 m) CFC maximum (>0.05 pmol/kg) above relatively denser and older bottom water, the ACC bottom water (ACCbw; 28.18 < γn < 28.27 kg/m3), which is derived from the densest portion of the ACC's LCDW rather than directly from the Antarctic continental margins (OJB1999).
 Winter cooling in the Labrador Sea provides sufficient buoyancy forcing to transform low-salinity surface waters into relatively denser water types. Classical Labrador Seawater (CLSW) forms by open ocean convection [Lazier, 1973; Talley and McCartney, 1982] as indicated (Figure 2a) by the uniformly high CFC-11 concentrations greater than 4 pmol/kg (1 pmol/kg = 1 × 10 −12 mole/kg) extending from the sea surface to about 2000 m at the northern end of the Atlantic section (north of 50°N). The subsurface signature of CLSW is captured by stations 42 and 567 (Figures 3a and 3d), which show a relative salinity minimum (<34.9) at potential temperatures near 3°C. A tongue with relatively high CFC concentrations (>0.05 pmol/kg, Figure 2a) along the top density layer (γn = 27.70–27.98 kg/m3) depicts the prominent lateral spreading of CLSW to the south, reaching past the Equator with CFC-11 > 0.01 pmol/kg. This tongue also includes Upper Labrador Seawater (ULSW), a shallower contributor to the upper NADW only recently distinguished from CLSW [Pickart, 1992; Smethie et al., 2000]. SF2001's integration of CFC measurements in the North Atlantic used the shallow (∼1000 m) and deep (∼2500 m) CFC minima as the upper and lower limits to ULSW and CLSW respectively; these minima lie roughly along the γn = 27.7 and 27.98 kg/m3 isopycnals (Figure 2a) defining the top density layer.
2.2. Southern Ocean Deep Waters
 A variety of ventilating water mixtures are produced and exported along much of the Antarctic continental margins (OJB1999). All of them involve a cold surface component, and are classified relative to the density of the local remnant of CDW. Circumpolar Deep Water is the world's most voluminous water mass [Worthington, 1981], and conceivably the ultimate source for all other water masses produced within the Southern Ocean (see Figure 5). Within the ACC it occupies the thick meridionally tilted layer (densities between 27.7 and 28.27 kg/m3) found below the Antarctic Surface and Intermediate Waters and above the AABW (Figure 2). Such extreme poleward shoaling of isopycnals across the ACC has important implications for the effective exchange of waters and properties between the Southern Ocean and the deep basins farther to the north. Deep waters entering the circumpolar current regime from the north have access to gas exchange near the sea surface in regions south of the ACC. Conversely, near-surface and intermediate waters entering the ACC from the south can rapidly descend and ventilate the deep and abyssal waters of the world ocean. Prevailing wind patterns along the southern edges of the ACC can force, at some locations, a net northward advection of ventilated Antarctic waters within the same broad density range of the CDW. Lateral ventilation and inputs to the CDW have been qualitatively inferred at the Weddell-Scotia Confluence [Whitworth et al., 1994]. This process is also expected to occur at other locations around Antarctica where there is northward advection of subpolar surface and near-surface waters. Estimating the amounts of this broad export of ventilated Antarctic waters is quite difficult by means of direct current measurements or classical hydrographic data. As demonstrated in section 4, simple quantitative analysis of transient tracers like the CFCs offer an attractive alternative.
3. Circulation and Ventilation of Circumpolar Deep Water
 Determining the mixing history of CDW over the vast region to the south of the ACC, the Subpolar Regime, and the rate at which it is recycled back into the ACC are two critical aspects toward fully understanding the meridional overturning circulation in the Southern Ocean and its impact on global climate.
3.1. Northern Influences
 CDW incorporates middepth (1000–3000 m) waters from all ocean basins to the north. Characteristic property cores that indicate these contributions are more clearly seen near the northern fringes of the ACC. There, input of lower NADW (lNADW) imprints the distinctive salinity maximum of LCDW, whereas separate inputs of North Indian (NIDW) and North Pacific Deep waters (NPDW) imprint the characteristic oxygen minimum of UCDW.
 A deep salinity maximum (S-max) at the northernmost stations of the Atlantic western section, 42 and 567 in Figure 3a, roughly coincides with a deep CFC-11 minimum (Figure 2a). By the time this core enters the South Atlantic it is only slightly diluted, and salinities are still as high as 34.95. A progressive erosion of this northern high-salinity signal is observed in the South Atlantic along its transit toward the ACC. This attenuation is illustrated by the red crosses in Figure 3a, which correspond to the S-max characteristics at stations located between 15°S and station 266 near the SAF. The core's salinity decreases from 34.95 to 34.78 along that southward path, but it shows a relatively narrow neutral density range (γn = 27.98–28.04 kg/m3) within the middle layer (Table 1). The observed density-compensated cooling and freshening of the S-max in NADW result from progressive mixing with northward flowing AAIW above and ACCbw below. Its entrance to the circumpolar regime is marked by a prominent high-salinity signal within the deep CFC-11 minimum tongue shown in Figure 2a: water with the lowest concentrations at sta. 266 (CFC-11 < 0.05 pmol/kg at 1500–2500 m) is also more saline (S > 34.74) than the saltiest CDW entering the Atlantic Ocean via Drake Passage. Therefore the ultimate source of the high-salinity signal in the LCDW of the ACC is water that lies at levels between the LSW and the ISOW/DSOW in the Northwestern North Atlantic. Once within the northern rim of the ACC, the S-max fades gradually to the east, as indicated by the freshening downstream of Atlantic station 266 where the highest salinity in LCDW is 34.78, compared to 34.75 at Indian station 118, and 34.74 at Pacific station 60 (Figure 3d).
Callahan  traced the origin of the characteristic oxygen minimum (O2-min) found in the UCDW to the oxygen-poor deep waters in the North Pacific and Indian oceans. The red dots in Figures 3b and 3c indicate the T-S characteristics of the O2-min at stations located to the north of the ACC in the Indian and Pacific western sections. This northern core is found within the top density layer (γn = 27.70–27.98 kg/m3). Oxygen concentrations along the O2-min increase to the south, away from its origins, as is expected from the progressive mixing with more recently ventilated, northward flowing Antarctic Intermediate Water (AAIW) above and LCDW below. Along the northern edge of the ACC, the most prominent low-oxygen signal in UCDW is gained in the southeast Pacific sector [Orsi et al., 1995]. The minimum oxygen concentration in UCDW at Indian station 118 and Pacific station 60 is about 175 micromoles/kg; yet at the northern Drake Passage, e.g. at station 103 from WOCE A21 cruise [Roether et al., 1993], the O2-min of UCDW shows a much lower oxygen concentration of 159 micromoles/kg.
3.2. Regional Transformations
 The extent to which the low-oxygen signal of UCDW can reach to the south is marked by a subsurface water mass boundary used to define the poleward limit of the ACC [Orsi et al., 1995]. There, UCDW is transformed into much colder AASW. On synoptic transects with closely spaced stations like the three sections shown in Figure 2, the southern boundary of the ACC (bndry) is also apparent in θ-S space (Figure 3) by a void of samples in the top density layer (γn = 27.70–27.98 kg/m3). A full discussion of the multiple crossings (see Figure 1) of the southern boundary of the ACC captured by the Atlantic (AJAX) western section can is given by OJB1999. Cyan curves in Figure 3 correspond to the three stations located just south of this boundary, and show the relatively cold and fresh southern source water at the same densities of UCDW. Over the oceanic domain of the Subpolar Regime, the thin AASW layer lying between the summer subsurface temperature minimum and the intermediate temperature maximum is often referred to as thermocline or transitional water [Carmack, 1977].
 The high-salinity signal of the LCDW lies deep enough to continue poleward beneath the surface waters of the ACC and the Subpolar Regime. This upward (southward) intrusion is clearly depicted by the shoaling of the deep CFC minimum in the Atlantic western section (Figure 2a). This deep core shows its lowest concentrations (CFC-11 < 0.01 pmol/kg) at about 3000 m at the Subtropical Front, located near 40°S [Orsi et al., 1995]. It reaches to about 1500 m (CFC-11 < 0.1 pmol/kg) in the middle of the ACC (Figure 2a), and to only 1000 m near the current's southern boundary in the Pacific western section. LCDW enters the subpolar cyclonic circulations at their eastern limbs. The isolated CFC minimum (CFC-11 < 0.05 pmol/kg) found above the γn 28.27 kg/m3 isopycnal and south of 60°S in Figure 2a indicates the LCDW recirculating within the interior of the Weddell Gyre [Orsi et al., 1993].
 Part of the LCDW volume originally drawn from the ACC is upwelled within the interior of the subpolar circulations, where it is slowly transformed into upper waters [Gordon and Huber, 1984]. Some of its relatively warm and saline volume continues toward the Antarctic continental margins, where the ultimate transformation of these subpolar remnants of LCDW takes place at the Antarctic Slope Front (ASF; [Jacobs, 1991]). Along the Front's ∼18,000 km length, LCDW intensively mixes with the AASW above to produce Modified CDW (MCDW) [Newsom et al., 1965; Whitworth et al., 1998], but at several locations around Antarctica it also mixes with the underlying SW to form AABW (OJB1999). The three blue θ-S curves in Figure 3 correspond to slope stations numbered 94, 85, 777, located at depths of 2034 m, 2920 m, and 1683 m, respectively. They all show MCDW between the AASW and AABW, which is much fresher and colder than the regional LCDW remnant found offshore (Figure 3d).
3.3. Antarctic Influences
 Northward injection of CFCs into the top layer (γn = 27.7–27.98 kg/m3) is indicated in all three western sections of Figure 2. A relatively high-CFC signal (CFC-11 > 0.1 pmol/kg) originating in the AASW of the Subpolar Regime, crosses the ACC near 1600 m in the Atlantic (at sta. 266) and 2000 m in the Indian and Pacific (at sta. 118 and 60). The latter sections show a stronger penetration of this signal partly because their CFC data was collected in more recent times (1992–1996) compared to the Atlantic section (1984–1989).
 At several locations around Antarctica, the path of the ACC lies in close proximity to the continental margins, resulting in prime regions for the export of recently ventilated MCDW into the middle layer offshore (γn = 27.98–28.27 kg/m3). Intense lateral ventilation of the deep Southern Ocean is also indicated in all three western sections of Figure 2. CFC-11 concentrations higher than 0.1 pmol/kg within the middle layer are found throughout the ACC in the Atlantic western section (62°–50°S), south of 60°S in the Indian, and south of 65°S in the Pacific. A remarkable example of such a northward injection of MCDW is indicated by the CFC maximum of station 266 (Figure 2a), at depths between 3400 m and 4000 m. The source of this prominent high-CFC signal is the Weddell-Scotia Confluence located just upstream (60°–35°W) of this section [Whitworth et al., 1994]. Another location where offshore lateral ventilation within the middle layer has been reported is that off the coast of Wilkes Land, near 120°–160°E [Foster, 1995; Rintoul and Bullister, 1999]. Stations located in the southern ACC regime of Figure 2c (along 170°W), also reveal a relative CFC maximum extending downward from about 1000 m at 65°S to near 1500 m at 60°S. This high CFC signal results from the lateral input of MCDW over the northwestern corner of the Ross Gyre.
 Near-freezing waters that are much denser (γn >28.27 kg/m3) than any inshore flowing remnant of the ACC's LCDW are found over several Antarctic shelves [Whitworth et al., 1998]. This Shelf Water (SW) is an indispensable near-surface ingredient to produce AABW at the Antarctic Slope Front. Even though distinct types of SW (Figure 3d) are readily identifiable by their temperature extremes (θ < −2°C, the Ice Shelf Water: ISW) or salinities (S > 34.8, the High Salinity Shelf Water: HSSW; S < 34.8, the Low Salinity Shelf Water: LSSW), the exact processes and rates of SW formation are largely unknown. It is most likely that SW results from progressive transformations (buoyancy loss) of AASW (and/or of LCDW) over the continental margins, in particular within coastal polynyas; further thermohaline changes occur on SW that manages to transit underneath glacial ice shelves.
 The best documented export of dense Antarctic waters to the rest of the World Ocean is that of cold AABW (γn > 28.27 kg/m3) spreading underneath the ACC. Supplied by the Weddell Gyre circulation, one of these outflows extends northward in the deep western boundary current that connects the Scotia, Georgia, Argentine and Brazil basins in the Atlantic Ocean (OJB1999). But all of the boundary currents in the Southern Hemisphere carry the densest portion of the ACC's LCDW, referred to as ACCbw (28.18 < γn < 28.27 kg/m3), to as far north as the Equator. ACCbw results from the continuous mixing of NADW and ABBW along the lengthy circumpolar path of the ACC. Near 30°S the estimated net northward transport of waters below the γn = 28.11 kg/m3 isopycnal is reported to be about 21 Sv [Ganachaud and Wunsch, 2000], which includes both the ACCbw and AABW.
4. Strength of Southern Overturning
4.1. Bottom Layer: CFC Inventory and Input Rate
Orsi et al.  calculated a total of 6.2 million moles (Mmoles) to be the CFC-11 (normalized to 1987) content in the AABW layer (γn > 28.27 kg/m3) in the Atlantic sector and 4.8 Mmoles of CFC-11 (normalized to 1993) in the combined Indian-Pacific region (Table 4) to the east of the Princess Elizabeth Trough (Figure 1). These inventories, together with models of the CFC content of newly formed AABW as a function of time, were used to estimate the formation rate of AABW. The observed CFC saturation in new AABW types after entrainment of LCDW is 35%, and the regional CFC-11 inventories for the AABW layer supported an abyssal input rate of 8.1 Sv (4.9 Sv in the Atlantic and 3.2 Sv in the Indian-Pacific sectors, see Table 5) (OJB1999).
Table 5. Production Rate of Source Water Ingredients Calculated Using the Adopted CFC-11 Saturations Listed in Table 3 and the Mixing Recipes Explained in the Text: Those for the NADW Components Were Estimated by Smethie and Fine a
In million cubic meters per second.
2.2 ULSW + 7.4 CLSW
2.9 AASW + 2.9 LCDW
1.8 AASW + 1.8 LCDW
2.4 DSOW + 2.4 ISOW + 1.8 CLSW + 1AW
3.3 SW + 1.6 LCDW
2.1 SW + 1.1 LCDW
 Only minor (<2% each) uncertainties in the AABW inventory estimate (and AABW formation rate) are expected to derive from errors in CFC concentration measurements [Bullister and Weiss, 1988], CFC solubility coefficients [Warner and Weiss, 1985] in seawater, and from the errors in the reconstructed atmospheric histories of CFC concentrations for the past three decades [Walker et al., 2000]. In calculating the inventories from data sets collected at different times, all of the CFC-11 measurements given by OJB1999 were first normalized to a common date (1987 or 1993). The common dates were chosen to be near the midpoint of the cruise dates for each sector, so about the same number of samples had CFC concentrations adjusted upward as has concentrations adjusted downward by the normalization process. The normalization term for each sample was proportional to the difference between the common date and the measurement date, multiplied by the annual rate of increase of atmospheric CFCs at the time the water sample would have been in equilibrium with the atmosphere. Since all of the CFC measurements used in each of the two ocean sectors were collected within three years of the adopted common dates, the temporal adjustment of CFC data was at most three years. The time that a water sample would have been in equilibrium with the atmosphere is based on its pCFC apparent age. The annual rate of increase of CFCs in young (<10 years) high concentration samples was only a few percent per year, so the maximum adjustment (for a three year difference) for these high concentration samples was <10%. The annual rate of increase of CFCs in older (∼10–30 years) samples was ∼10–20% yr−1, but since the CFC concentrations in these older waters was low, the effect of these adjustments on the total CFC inventory in AABW was minor. The overall estimated error due to normalizing CFCs to a common date is <∼12%.
 Perhaps the most significant error associated with the calculation of the CFC inventory for the AABW layer given by OJB1999 originates in the subjectively contoured spatial distribution of the layer's mean CFC concentration (OJB1999, Figure 7a), prior to its objective mapping onto the uniform grid used to compute the volumetric inventory. That pattern unavoidably relies on a rather small number of control points, i.e. the total number of stations with available CFC data. However, that CFC pattern was guided by patterns of highly correlated tracers like dissolved oxygen and silicate, for which a much larger number of recent and historical data points were utilized. To test the sensitivity of the CFC inventories to the contouring, alternative patterns were drawn to fit the sparse CFC data set but deliberately trying to minimize and maximize the CFC inventories. Even when these extreme patterns (which started to render somewhat inconsistent circulations for the AABW) were integrated, the associated CFC-11 inventories did not differ by more than 20% from that reported by OJB1999. Note that the highest CFC concentrations in Figure 7a of OJB1999 appear along narrow boundary currents, whereas vast interior areas show significantly smaller concentrations for a thicker AABW layer. Thus, although estimating the largest error in the AABW CFC-11 inventory (Table 4) is necessarily subjective and hard to quantify, we believe that 25% is a conservative estimate that includes all of the smaller errors discussed above.
 Among the assumptions made by OJB1999 during their calculations was that both the relative saturation levels of new AABW types and the input rates of these AABWs to the abyssal ocean remained constant at their input sites during the previous ∼30–40 years, when measurable levels of CFC were present in the environment. Observed CFC-11 saturation levels in shelf waters at a number of sites (Table 3) around Antarctic averaged ∼0.5 (+−0.1) during the mid-1980s to mid-1990s. The atmospheric CFC-11 concentration increased in a quasi-linear manner during the 1–2 decades prior to 1990, so it is likely that the relative saturation levels in the high-CFC bearing shelf-water components involved in AABW formation would not have varied drastically during this period. The CFC-11 saturation in LCDW components involved in AABW formation averaged ∼0.05 (+−0.04) when observed in the late 1980s and early 1990s (Table 3), so variation in this saturation level during the earlier period would have only a minor impact on the estimated overall CFC content of new AABW. No CFC data are available in the Southern Ocean prior to the mid-1980s, so potential decadal variability in AABW production rates cannot be resolved. In any case, the AABW inventories (and derived average AABW production rates) are most strongly influenced by the 1–2 decades prior to the early 1990s, the period which had the highest concentration CFC waters entering into the AABW reservoir. From this (and in the absence of CFC data in the earlier period) we estimate an error of about 10–20% in the CFC content of new AABW as a function of time. Therefore the overall uncertainty in the AABW production rate (8.1 Sv) reported by OJB1999, albeit based on subjective estimates of its larger error terms, is expected to be 32%, which is similar in magnitude to that estimated by SF2001 for the NADW production rate (17.2 Sv).
4.2. CFC Inventories for the Middle and Top Layers
 Inventories for the two shallower density layers considered in this study (Table 1) were not reported by OJB1999. To expand our spatial coverage using CFC data currently available to us, the 11 million moles of normalized CFC-11 inventory for the AABW (bottom) layer are extrapolated into the two overlying layers (middle and top) based on the computed fractional (layered) CFC-11 inventories along six meridional sections in the Southern Hemisphere (Figure 1). As in the precursor work of OJB1999, these section inventories are of normalized CFC-11, to 1987 in the Atlantic and to 1993 in the Indian-Pacific. Figure 4 shows the stacked CFC-11 inventories for the three density layers (Table 1) computed along these southern meridional sections.
 Stations located close to source sites of new deep and bottom waters reveal the highest CFC content. Total station CFC-11 inventories above 5000 pmol.m/kg are observed at the southernmost stations of the western Pacific (Figure 4e) and eastern Indian sections (Figure 4d), which are located over the Antarctic continental slope offshore of well-documented sources of AABW types in the western Ross Sea and along the Adelie Coast (OJB1999). The western Atlantic section (Figure 4a) shows the overall maximum CFC-11 inventory of 5245 pmol.m/kg near 59°S, at station 107 located in the South Sandwich Trench (Figure 2a). This trench is filled with outflowing Weddell Sea Deep and Bottom Waters (see OJB1999, Figure 10), whose relatively high CFC concentrations and thickness result in a large inventory.
 Western section inventories in the Atlantic (Figure 4a) and Pacific (Figure 4e) oceans reflect the sampling of relatively young water masses carried northward by boundary currents, whereas the eastern section inventories include relatively older deep waters. The cumulative total inventories (areal integration) for the western Atlantic (8.3 millimol.m2/kg) and western Pacific (3.4 millimol.m2/kg) sections are at least fifty percent greater than their eastern counterparts. The opposite pattern is observed in the Indian sections, where the eastern inventory (3.8 millimol.m2/kg) is larger than the western inventory (2.4 millimol.m2/kg), since the major sources of AABW in this ocean sector, located off the Adelie Coast and Wilkes Land (135°–155°E), are very close to the eastern Indian section. Orsi et al.  found that the AABW in the Australian-Antarctic Basin has the highest volume-weighted CFC-11 concentration (0.47 pmol/kg), in addition to almost half the volume, when compared to the AABW in the Atlantic (0.17 pmol/kg) and Pacific (0.18 pmol/kg) basins, which suggests a more rapid turnover in the Indian basin. Figure 4 also lists each layer's relative contribution, in percentages, to the total inventory of these sections.
 To account for the observed regional differences in the deep layers section inventories listed in Figure 4, basin inventories are computed as weighted average percentages proportional to the corresponding section inventory. We assume that the inventory partitioning observed along these paired (western-eastern) meridional sections are representative of the overall partitioning for each basin. This assumption, in contrast to just using the western or eastern ratios alone, implies an uncertainty of about 17% in the estimated inventories for the middle and top layers combined. For the southern Atlantic basin, the bottom and middle layers have very similar inventory ratios of 37% and 38% of the total inventory, and the smallest portion (25%) corresponds to the top layer. In contrast, the total inventory for the southern Indian-Pacific basin is partitioned, from bottom to top layers, as 37%, 29%, and 34%. Using the computed CFC-11 inventory for the entire AABW volume (OJB1999), we estimate the corresponding volume inventories for the two less dense layers above it. Since the inventory in the bottom layer of the Atlantic sector (6.2 Mmoles) is 37% of the total in all three layers combined, as much as 6.2 Mmoles are expected in the middle layer and 4.1 Mmoles in the top layer. Similar inventories are estimated in the bottom (4.8 Mmoles) and top (4.5 Mmoles) layers of the Indian-Pacific sector (Table 4), with a minimum inventory of 3.9 Mmoles in the middle layer. It is interesting to note the small range of the calculated total CFC-11 inventories from (circumpolar) southern sources in all three density layers (11, 10.1 and 8.6 Mmoles, respectively).
 We estimate that the CFC inventories (10.1 and 8.6 Mmoles) for the middle and top Southern Ocean layers analyzed here have each a compounded error of ∼30%. This derives from the ∼25% uncertainty in the bottom layer (AABW) inventory calculation reported above and from the ∼17% sensitivity of the upper two layers combined inventory to our selection of paired (western-eastern) meridional sections.
4.3. Mixing Recipes and Production Rates
 The volumetric inventory in each of these three deep ocean layers reflects the cumulative gain of atmospheric CFCs during the past few decades, and thus the integrated input of surface and near-surface waters over that time. Layer CFC inventories (Table 4) are used in combination with the saturation data for the near-surface and subsurface ingredient waters presented in Table 3 to calculate the formation rates of source dense waters (Table 5). All major assumptions made during the AABW abyssal input estimate are also applied here, rendering an estimated uncertainty of 36% for the source water input rates to the middle and top layers. In the following calculations we have also neglected any transfer of CFCs between density layers, since to a first order approximation it is lateral advection that redistributes CFCs within deep layers more effectively and rapidly than cross-isopycnal mixing. Therefore these inventory-based estimates are somewhat conservative, specialy for the bottom layer where a minor loss of CFCs in recent years across the top of the AABW isopycnal is expected (OJB1999).
 The CFC content (11 Mmoles) of the bottom layer is almost entirely supplied by the SW component. However, OJB1999 noted that newly formed AABW types sinking at the bottom of the slope show relatively low CFC saturation (about 35%). Since it is not clear in what relative proportion the different types of SW participate in AABW formation, or if any one in particular dominates at a certain site, we adopt their average saturation of 50% based on observations from ISW, LSSW and HSSW listed in Table 3. This low saturation is the result of extensive ice cover over the shelf regions, which restricts air-sea gas exchange. Near the Antarctic continental shelf break, the subsurface LCDW available for mixing with SW to produce AABW has very low CFC concentrations (saturations that range from near zero to about 10%), indicating this water's minor and slow mixing with the AASW above. We adopt an average saturation of 5% for the LCDW ingredient in all of our production rate calculations. Based on these characteristic saturations (and the observed ranges) of its ingredient water masses, we infer that the 8.1 ± 2.6 Sv production rate of AABW estimated by OJB1999 must result from the mixing of 5.4 ± 1.7 Sv of SW (50% saturation) and 2.7 ± 0.9 Sv of LCDW (5% saturation) at the upper slope.
 The CFC content (10.1 Mmoles) of the middle layer is provided by the AASW, whose intense dyapycnal mixing with the regional variety of LCDW offshore from the slope forms MCDW. Based on their temperature-salinity relationship near the shelf break, an average mixing scenario of equal volumes from each ingredient is assumed while estimating the MCDW production rate [Newsom et al., 1965; Gill, 1973; Foster and Carmack, 1976]. MCDW is found to spread northward along isopycnal surfaces into the relatively older deep waters offshore [Carmack and Killworth, 1978; Whitworth et al., 1994]. AASW saturations can be as high as 90% in the interior of the Weddell Gyre, but the observed range near the continental slope is from 60% to 70% (Table 3). Thus we adopt an average 65% saturation for the AASW component ventilating both the middle and top density layers. Based on the adopted and observed ranges of CFC saturations in its ingredient water masses, we infer that 9.4 ± 3.4 Sv of MCDW must be produced along the Antarctic Slope Front from the mixing of 4.7 ± 1.7 Sv of AASW (65% saturation) and 4.7 ± 1.7 Sv of LCDW (5% saturation) prior to lateral injection to the oceanic domain.
 The CFC content (8.6 Mmoles) of the top layer is entirely provided by the AASW. This water extends down to the summer subsurface temperature minimum over the vast oceanic domain of the Subpolar Regime. At certain locations, in particular where the ACC extends close to the Antarctic continental margins, wind-induced northward transport of AASW carries out the entire ventilation of the UCDW layer [Sloyan and Rintoul, 2001; Speer et al., 2000]. We estimate a northward export of 3.6 ± 1.3 Sv of AASW (65% saturation) into the top layer. Table 5 shows the estimated regional distribution of these production rates in the Atlantic and Indian-Pacific sectors.
 The estimated total input of all southern near-surface waters to the deep ocean based on CFC-11 inventories is 13.7 Sv of SW and AASW combined. Roughly there is an average error (weighted according to each layer inventory) of 28% in the total 29.7 Mmoles listed in Table 4. An additional error component in the total input rate (13.7 Sv) is expected from the uncertainty in the adopted CFC-11 saturation for each source water, which we estimate by contrasting cases that would result from the mixing of least (most) ventilated upper waters with most (least) saturated subsurface waters. Using a SW saturation of 40% (70%), AASW saturation of 60% (70%), and LCDW saturation of 10% (1%) renders a total input of Antarctic near-surface water of 16.3 Sv (11.5 Sv). Thus the uncertainty in the total Southern Ocean input of upper waters due to the observed CFC saturation ranges in SW, AASW and LCDW could be as large as 18%. A final error of about 33% is estimated for the total input of 13.7 Sv (7.4 Sv) of southern near-surface (subsurface) waters to the deep ocean. This is the first time that all of the Southern Ocean contributions, estimated basically from the same CFC approach, are readily comparable to those from the northern North Atlantic.
4.4. Northern Versus Southern Ventilation
Smethie and Fine  calculated the CFC-11 content of the abyssal density layer of the combined ISOW/DSOW in the North Atlantic in 1990 to be 10.9 million moles. Almost double that amount (18.9 Mmoles) was integrated for the ULSW/CLSW layer (Table 4). Production rates derived from the CFC-11 budgets are 2.4 Sv each for the DSOW and the ISOW, which also entrain about 1 Sv of CFC-rich northeast Atlantic Surface Water (AW) and 1.8 Sv of interior CLSW as they sink to great depths. The rates of ULSW and CLSW production are 2.2 and 7.4 Sv (Table 5). Thus the total input rate to the deep ocean from all components produced in the North Atlantic is ∼17.2 Sv, which has an error of about 30%.
 The estimated total southern input of 29.7 Mmoles of CFC-11 is remarkably close to the northern input of 29.8 Mmoles in 1990, but injected by a somewhat smaller volume of near-surface components (13.7 Sv). The differences in volume inputs of northern and southern sources is due to differences in CFC saturations for the various source waters and differences in normalization dates. Analysis of the deep ocean's radiocarbon and PO4* data [Broecker et al., 1998] and temperature and salinity data [Peacock et al., 2000] lead these authors to conclude that surface waters must sink at a rate of about 15 Sv in both the northern North Atlantic and around the Antarctic margins. Our total rate of 30.9 Sv of the near-surface water overturning in the global thermohaline circulation for the last couple of decades (1970–1990), as well as the northern to southern ratio, are remarkably consistent with those previous independent inferences, which in turn represent averages over the past few hundred years.
 CFC-11 inventories for three deep density layers (Table 1) are used to calculate the total input rate of upper waters from the Southern Ocean. SW input to the bottom layer (γn > 28.27 kg/m3) is estimated as 5.4 Sv ± 1.7 Sv. The middle layer (27.98 < γn < 28.27 kg/m3) receives an estimated AASW input of 4.7 ± 1.7 Sv. AASW input to the top layer (27.7 < γn < 27.98 kg/m3) is estimated as 3.6 ± 1.3 Sv. Comparison of the estimated CFC content in the deep Southern Ocean and North Atlantic suggests that up to about 1990, they have received similar amounts of CFC-11 (29.7 and 29.8 million moles, respectively) sustained by an average input of 13.7 Sv and 17.2 Sv of surface and near-surface waters from all sources in each polar region.
 Sinking of near-surface waters along the Antarctic continental margins involves entrainment of a relatively older subsurface component. Unlike the minor role in building up the CFC inventories of the middle and bottom layers, the subsurface contribution of 7.4 ± 2.4 Sv of LCDW to the total southern input to the deep ocean is quite significant. We estimate that 17.5 ± 6 Sv of MCDW and AABW combined are injected into the middle and bottom layers of the deep ocean. Ganachaud and Wunsch's  optimal solution indicates a conversion of 21 ± 6 Sv of CDW with densities between 27.72 < γn < 28.11 kg/m3 into denser bottom water over the entire region south of about 20°–30°S. That scenario is compatible with our results (Figure 5), which provide further details on the formation of dense southern waters near the Antarctic margins and their injection to the deep ocean, i.e. the linkage between the water mass structure and meridional circulation inferred within the Subpolar Regime.
 In contrast to the topographically constrained input of 8.1 ± 2.6 Sv of AABW to the bottom layer, we estimate that as much as 9.4 ± 3.4 Sv of ventilated MCDW and 3.6 ± 1.3 Sv of AASW participate in the broader southern contribution to the middle and top layers of the deep ocean. Quantifying this southern input at intermediate levels has been extremely difficult to make based on classical hydrography or direct measurements, thus simple oceanic budgets of transient tracers like the CFCs may provide an important tool for estimating those inputs. Although all of our estimates are based on fairly sparse data coverage and carry significant errors (∼33%), they demonstrate that CFC inventories provide independent information on the strength of the global thermohaline circulation. We expect that when the complete WOCE and other modern CFC data sets become available, the extrapolations used here for the upper density layers in the Southern Ocean will no longer be necessary and more accurate computations of deep volumetric inventories will be feasible.
 We thank H. Lee for his assistance with data reduction and analysis. A. H. Orsi was supported by grants (OCE-9811481, OCE-9811460) from the National Science Foundation. Support for W. M. Smethie was provided by cooperative agreements from the National Oceanographic and Atmospheric Administration (NOAA) through grants NA47GP0188 UCSIO PO 10075411 and NA77RJ0453 UCSIO PO 10156283 (CORC I and CORC II). Support for J. L. Bullister was provided by NOAA, NOAA's Office of Global Programs and NOAA's OACES Program. This work is a World Ocean Circulation Experiment contribution.