A global coupled model is used to examine pathways of freshwater transport in the Southern Ocean. On the background of a strong zonal freshwater transport along the pathway of the Antarctic Circumpolar Current (ACC), there are meridional freshwater flows distributed nonuniformly around the globe, including in the upper ocean. The analysis does not support a simple two-dimensional scheme of Antarctic Intermediate Water (AAIW) formation, according to which the fresh AAIW forms uniformly around the circumpolar ocean. Rather, a more complex three-dimensional picture of the freshwater transport in the Southern Ocean is revealed, with enhanced AAIW formation in the southeast Pacific Ocean both north and south of the Drake Passage latitudes. Freshened by intense precipitation and surface waters from around Antarctica, the ACC transports freshwater from the northwest to the southeast toward Drake Passage. There, a fraction of this freshwater is transported southward across 60°S with the subsurface ACC and the eddy-induced flow. West of the Antarctic Peninsula, the freshwater subducts to intermediate depths and turns northward, following the ACC and contributing to the formation of AAIW. This analysis supports previous results of enhanced subduction localized to the southern tip of South America.
A remarkable feature of Southern Ocean hydrography is a salinity minimum at intermediate depths associated with Antarctic Intermediate Water (AAIW). The salinity minimum can be seen from observations at any longitude, giving the impression that AAIW forms in a circumpolar manner because of the northward transport of Antarctic surface water (Sverdrup et al. 1942). In contrast, McCartney (1977) suggested that the majority of AAIW forms in a localized region near Cape Horn from waters to the north of the Antarctic Circumpolar Current (ACC). Modeling studies based on OGCMs, regardless of the dominant mechanism that brings fresh surface water down to intermediate depths, seem to support both the circumpolar (Sorensen et al. 2001) and the localized (England et al. 1993) hypotheses.
There is one aspect of the original idea of Sverdrup et al. (1942) that appears robust; namely, that at least some fraction of AAIW is due to a circumpolar mixture of fresh Antarctic and Subantarctic surface waters. However, it remains unclear whether the freshwater escapes high southern latitudes evenly around the globe, that is, in a circumpolar manner.
Using a global coupled model we aim to illustrate that the north–south fluxes of freshwater in the Southern Ocean and, in particular, at latitudes near 60°S are considerably nonuniform around the globe. The largest reversal of the freshwater transport across 60°S is found in the Pacific Ocean sector of the Southern Ocean. In a vast Pacific region to the west of about 90°W, the general direction of freshwater transport near 60°S is from the northwest to the southeast, mainly in the upper ocean below the Ekman layer. This freshwater then escapes the subpolar southern latitudes in a relatively narrow region west of the Antarctic Peninsula, predominantly at intermediate depths. Model experiments with passive tracers released from the sea surface indicate that the regions west of the Antarctic Peninsula and around the southern tip of South America are associated with the most rapid penetration of the surface signal down to intermediate depths.
2. The coupled model and experimental design
The coupled model we use comprises an ocean GCM (Pacanowski 1996), an energy–moisture balance atmosphere model (Weaver et al. 2001) and a dynamic–thermodynamic sea ice model (Bitz et al. 2001). The coupled model is described in detail by Weaver et al. (2001). All model components have the same horizontal resolution of 3.6° longitude × 1.8° latitude. The ocean model uses isopycnal mixing and eddy-induced advection (after Gent and McWilliams 1990) with the coefficients of thickness diffusivity and isopycnal diffusivity set to 1.0 × 107 and 2.0 × 107 cm2 s−1, respectively. The use of the Gent and McWilliams mixing scheme considerably improves the simulated vertical salinity structure (Fig. 1a), particularly in the Southern Ocean. The vertical diffusivity in the model follows the profile of Bryan and Lewis (1979), with values ranging from 6 × 10−5 m2 s−1 in the upper ocean to 1.6 × 10−4 m2 s−1 in the deep ocean. There are 19 vertical levels in the ocean model that vary smoothly in thickness from 50 m at the surface to 518 m at the deepest level.
At the surface the model calculates fluxes of heat and freshwater. The former is represented by shortwave and longwave radiative fluxes, sensible fluxes, and latent fluxes, whereas the latter is due to precipitation, evaporation, runoff, and sea ice growth and melt. Precipitation occurs when the relative humidity exceeds a threshold value of 85%. The fraction of precipitation that falls as rain on land is returned instantaneously to the ocean. Snow precipitated on land returns to the ocean upon melt. The catchment area of the land precipitation is divided into 32 river discharge basins (see Weaver et al. 2001 for details).
The atmospheric model transports heat and moisture. However, the model does not calculate winds and windstress. These are prescribed from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996), averaged over the period 1958–97 to form an annual cycle from the monthly fields. Of importance here is our use of advective transport of moisture (using the prescribed winds), recently incorporated into the model. This considerably improves the simulation of the hydrological cycle (Fig. 1b; see also Weaver et al. 2001). The model was integrated for 2500 years from an idealized distribution of temperature T and uniform salinity S. No T–S restoring is incorporated into the model nor do we employ a salinity enhancement around Antarctica. The model's ability to capture the observed distributions of chlorofluorocarbons (CFCs), including in the Southern Ocean, is discussed by Saenko et al. (2002).
The final year of the model integration is used to diagnose oceanic freshwater fluxes at each model grid box. The fluxes include all processes transporting salt in the model and are calculated relative to a salinity of 35 psu. We also employ passive age tracers, south of 60°S and between 60° and 45°S, to identify regions of most rapid ventilation in the Southern Ocean. The definition of the age tracer is given, for example, in England and Rahmstorf (1999).
A large amount of relatively fresh water circles the globe in the Southern Ocean, forced by strong ocean currents, enhanced precipitation, and sea ice melt. If integrated from the surface to the bottom and across the ACC, the model gives a total eastward freshwater transport on the order of 1 Sv (Sv ≡ 106 m3 s−1). In comparison with this number, the total northward freshwater transport across latitudes near 60°S is an order of magnitude smaller. However, this transport increases northward from about 0.05 Sv across latitudes near 60°S to about 0.5 Sv across 45°S. This increase compensates for a net excess of precipitation (including runoff) over evaporation (P − E) in that latitude band. In other words, the net northward freshwater transport across 45°S must balance (at steady state) the net northward freshwater transport across 60°S combined with the freshwater gain at the surface due to P − E between 45° and 60°S.
In the upper 50 m, this surface freshwater input (i.e., about 0.45 Sv) is enhanced by the relatively fresh Antarctic surface waters, transporting about 0.2 Sv across 60°S. On the other hand, the net exchange of freshwater at 45°S removes about 0.15 Sv from the surface layer, with the rest (i.e., 0.5 Sv) penetrating into the subsurface ocean. These integral freshwater fluxes, however, do not reflect the actual pathways of freshwater in the Southern Ocean. We seek to quantify (i) whether the freshwater transport across 60°S is relatively uniform around the globe and (ii) where any enhanced formation of AAIW occurs. Also, we consider (iii) where the freshwater crosses the ACC to escape into subtropical latitudes.
To address these questions, we calculate total freshwater fluxes and their northward components at several model depths (Fig. 2). We also calculate meridional sections of northward freshwater flux at 61°S (Fig. 3). In Fig. 2, a barotropic streamline corresponding to a mass transport of 60 Sv is shown in order to roughly represent the axis of the ACC so that the regions of enhanced freshwater fluxes across ACC can be identified.
In the model, the enhanced freshwater flux at each depth closely follows the pathway of the ACC (Fig. 2), decreasing downward but still seen at depth of about 2500 m. Both the axis of the ACC and the maximum freshwater flux have a general direction from the northwest to the southeast almost everywhere (Fig. 2). Similar to the ACC, the freshwater flux has a wavelike pattern because of the north–south steering of the current by large-scale topographic features (Fig. 2; see also Fig. 4b). At latitudes near 60°S, this contributes to the nonuniform structure of the meridional freshwater flux, both around the globe and with depth. In the upper 50 m there is a strong northward freshwater flux from the regions just south of 60°S. This northward flux is dominated by Ekman dynamics and is considerably nonuniform around the globe (Fig. 2a). Out of the globally integrated value of 0.2 Sv transported in the upper layer across 61°S, as much as 0.13 Sv of freshwater is transported between about 90°W and 30°E (i.e., over 1/3 of the latitude circle 2/3 of the freshwater transport is found). The value of 0.2 Sv in the upper layer is comparable to the freshwater gain between 45° and 60°S due to the net P − E flux. However, below the upper layer the southward component of the ACC and the eddy-induced flow (due to the Gent and McWilliams mixing scheme) redirect the freshwater flux near 60°S to the south (Fig. 2b). The eddy-induced transport becomes more important between about 150° and 80°W (see Figs. 3b,c).
When integrated along 61°S and between 50 and 350 m, the net freshwater flux generates a transport of about 0.2 Sv to the south, mostly in the Pacific (Fig. 3a). However, to the west of the Antarctic Peninsula in the southeast Pacific the ACC veers to the north (Fig. 4a) and so does the meridional freshwater flux (Fig. 2 and Figs. 3a,b). The area of enhanced northward flux is rather localized and centered around 500-m depth west of the Antarctic Peninsula (Fig. 3b). Integrated between 60° and 100°W and over the depth range from 350 to 1100 m, this northward freshwater flux injects about 0.35 Sv into intermediate depths north of 60°S, contributing to the final stage of formation of AAIW. A large fraction of this AAIW escapes through Drake Passage (see Figs. 2c,d); it then may take a while for it to cross the ACC and finally escape to the subtropics. It should be noted again that this freshwater that feeds the AAIW formation west of the Antarctic Peninsula is composed mostly of water brought in from the northwest with the subsurface ACC flow and the southward eddy-induced transport.
As noted by Read et al. (1995), a southward loop and a dramatic broadening of the ACC in the southeast Pacific is required to conserve potential vorticity after the current passes the topographic barrier of the Pacific–Antarctic Ridge. The ACC, in turn, is freshened by enhanced P − E along its pathway and by the northward transport of freshwater in the Ekman layer from polar and subpolar regions around Antarctica, resulting from ice melt. Further, it can be seen that the freshwater at intermediate depths tends to escape into the subtropical latitudes over the broad regions of the eastern Pacific and Indian Oceans, as well as in the Atlantic (Figs. 2c,d). In the latter case, at least some fraction of the freshwater appears to be transported from the Indian Ocean.
In order for this freshwater pathway to exist at around 60°S, the region to the west of the Antarctic Peninsula should be favorable for subduction. This is illustrated below using passive tracers. It appears that the enhanced subduction in that region, as well as in the region around the southern tip of South America, is preconditioned by the convergence of the ACC before it enters the Drake Passage. Another factor contributing to the intense subduction in the southeast Pacific is an enhanced region of heat loss (Fig. 5; also suggested by McCartney 1977). It is, however, possible that some fraction of the heat loss in the southeast Pacific might be a consequence, rather than a cause, of the intense mixing and subduction in that region. We note that over open ocean areas, the model-simulated surface heat flux (Fig. 5) is in good agreement with the flux derived by Keith (1995), who employed the residual method to the European Centre for Medium-Range Weather Forecasts (ECMWF) analyses, Earth Radiation Budget Experiment (ERBE) data, and National Meteorological Center (NMC) analyses. In particular, both fields show an enhanced heat loss near the southern tip of South America.
In order to find the regions of most rapid subduction in the Southern Ocean we conducted two experiments with passive age tracers. In one of them (E1), the tracer was released at the ocean surface south of 60°S, whereas in another experiment (E2) the tracer was released between 60° and 45°S. Experiment E2 will thus track both Subantarctic Mode Water (SAMW), formed north of the ACC, and AAIW. Both experiments were run for an additional 100 years.
The distribution of the age tracer in experiment E1 has several minima at depths just below the surface (Fig. 6a) and at intermediate depths (Fig. 6c). These minima are indicative of regions of relatively rapid ventilation. Two of the minima are due to the formation of Antarctic Bottom Water (AABW) in the Weddell and Ross Seas and can be tracked down to the ocean bottom. Another minimum is in the southeast Pacific sector of the Southern Ocean, west of Antarctic Penensula (Figs. 6a,c). In this region the tracer sinks only to intermediate depths, having a characteristic AAIW tongue shape (Fig. 6e). One more minimum can be seen in the Indian Ocean sector, close to the Antarctic coast around 80°–90°E (Fig. 6a). This, however, considerably weakens with depth (Fig. 6c). Accordingly, the tongue of tracer subduction is less pronounced there (not shown).
In experiment E2, the most rapid mode and intermediate water subduction occurs in the Pacific (Fig. 6b) with particularly enhanced subduction around the southern tip of South America (Fig. 6d). There the signal propagates to the north at somewhat shallower depths and reaches the subtropical latitudes faster (Fig. 6f) when compared with the signal originating south of 60°S (Fig. 6e). It thus appears that both regions, that is, north and south of the ACC in the southeast Pacific, are important for ventilating the ocean at a depth between about 500 and 1500 m. The former has been suggested in previous observational and model studies (McCartney 1977; England et al. 1993), though it contrasts with the finding of Sorensen et al. (2001).
4. Discussion and conclusions
Using a coupled model we have examined the pathways of freshwater transport in the Southern Ocean. On the background of a strong zonal freshwater transport along the pathway of the ACC, there are meridional freshwater flows. A zonal mean freshwater budget across the latitudes of the ACC in the upper ocean layer suggests that about 0.5 Sv of freshwater is available to subduct into AAIW. This 0.5 Sv comes from surface freshwater input (0.45 Sv) and a quasi-northward freshwater flux of Antarctic surface waters across the ACC (0.2 Sv, with 0.13 Sv concentrated between 90°W and 30°E), less the advection/mixing of salinity from adjacent water masses north of the ACC (−0.15 Sv).
A detailed analysis of the spatial distribution of the freshwater transports within the ocean interior, as well as the distribution of passive tracers, does not support a simple two-dimensional view of AAIW formation in which intermediate water subducts uniformly around the globe. Rather, a more complex three-dimensional picture of the freshwater pathway in the Southern Ocean is revealed from our analyses, with enhanced AAIW formation in the southeast Pacific both north and south of the Drake Passage latitudes. Freshened by enhanced precipitation at high latitudes and Antarctic surface waters due to ice melt, the ACC transports the freshwater southeastward toward Drake Passage. There, a fraction of this freshwater is transported farther south with the subsurface ACC and the eddy-induced flow. West of the Antarctic Peninsula this freshwater subducts to intermediate depths and turns northward, following the ACC and contributing to the formation of AAIW. A large fraction of this AAIW appears to escape through Drake Passage, from where it eventually crosses the ACC and penetrates into subtropical latitudes. Farther north near Drake Passage, around the southern tip of South America, there is another region of enhanced subduction, which has been discussed previously (McCartney 1977; England et al. 1993) and is supported by observed oxygen data (Talley 1996). However, the region just west of the Antarctic Peninsula was given little attention until now.
Climatological data (Levitus and Boyer 1994) clearly shows a two-core structure of low salinity in the southeast Pacific (Fig. 7). These two salinity minima originate from two different regions, that is, north and south of Drake Passage latitudes, spanning a few degrees in latitude. In reality, however, the low-salinity region west of the Antarctic Peninsula appears to have smaller north–south extent. For example, several recent observational studies have found a banded structure of upper-ocean salinity west of the Antarctic Peninsula (Bellingshausen Sea). In particular, Pollard et al. (1995) report on salinities as low as 33.7 psu in a quasi-zonal band 15–20 km wide between about 86.5° and 84.5°W centered around 67°S; north and south of this band, the salinities increase. An abrupt deepening of this low-salinity water to intermediate depths can be seen in the observed salinity section along 85°W, presented by Read et al. (1995). Pollard et al. (1995) argue that this freshwater in the top 70 m could only have reached 85°W by advection from the west. This is supported by our freshwater budget analyses, though the model has too coarse a resolution to explicitly capture these directly observed features.
Some of our results contrast with those obtained in a study by Sorensen et al. (2001), even though the oceanic component of our coupled model is similar to the OGCM used in that study. In particular, we found that enhanced formation of AAIW in our model occurs in the southeast Pacific, both north and south of the ACC core. In contrast, Sorensen et al. (2001) argue for a circumpolar mechanism of AAIW formation in their model. This is a little surprising because they simulate some zonal inhomogeneity in ventilation rates for AAIW, with the most rapid renewal in the southeast Pacific and Atlantic Oceans. Also, their model appears to produce no direct AABW ventilation in the Ross and Weddell Seas, in contrast with observed tracers (e.g., CFCs: Orsi et al. 1999). One possible explanation for this discrepancy could be that Sorensen et al. (2001) employ a model salinity enhancement around Antarctica in order to account for the effects of sea ice and to capture the salinity minimum associated with AAIW. Such a procedure can imply the introduction of an artificial salt flux distribution and may distort ocean ventilation. Instead, we use a dynamic–thermodynamic sea ice model, which links the formation of dense AABW and fresh AAIW in a coupled system (Duffy et al. 2001; Saenko and Weaver 2001; Saenko et al. 2002). Further efforts are needed, both observational and modeling, to establish dominant regions and mechanisms for AAIW formation.
The authors are grateful to the Meteorological Service of Canada/Canadian Institute for Climate Studies, International Arctic Research Centre, Canadian Climate Change Action Fund, NSERC, Canadian Foundation for Climate and Atmospheric Sciences, the Killam Foundation, and the Canada Research Chair Program for supporting this research. Matthew England was supported by the Australian Research Council. We thank E. Wiebe for discussions and assistance.
Corresponding author address: Dr. Oleg A. Saenko, School of Earth and Ocean Sciences, University of Victoria, P.O. Box 3055, Victoria, BC V8W 3P6, Canada. Email: email@example.com