On the Role of Wind-Driven Sea Ice Motion on Ocean Ventilation

3376 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 On the Role of Wind-Driven Sea Ice Motion on Ocean Ventilation OLEG A. SAENKO, ANDREAS SCHMITTNER, AND ANDREW J. WEAVER School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada (Manuscript received 1 November 2001, in final form 10 May 2002) ABSTRACT Simulations with a coupled ocean atmosphere sea ice model are used to investigate the role of wind-driven sea ice motion on ocean ventilation. Two model experiments are analyzed in detail: one including and the other excluding wind-driven sea ice transport. Model-simulated concentrations of chlorofluorocarbons (CFCs) are compared with observations from the Weddell Sea, the southeastern Pacific, and the North Atlantic. The authors show that the buoyancy fluxes associated with sea ice divergence control the sites and rates of deep- and intermediate-water formation in the Southern Ocean. Divergence of sea ice along the Antarctic perimeter facilitates bottom-water formation in the Weddell and Ross Seas. Neglecting wind-driven sea ice transport results in unrealistic bottom-water formation in Drake Passage and too-strong convection along the Southern Ocean sea ice margin, whereas convection in the Weddell and Ross Seas is suppressed. The freshwater fluxes implicitly associated with sea ice export also determine the intensity of the gyre circulation and the rate of downwelling in the Weddell Sea. In the North Atlantic, the increased sea ice export from the Arctic weakens and shallows the meridional overturning cell. This results in a decreased surface flux of CFCs around 65 N by about a factor of 2. At steady state, convection in the North Atlantic is found to be less affected by the buoyancy fluxes associated with sea ice divergence when compared with that in the Southern Ocean. 1. Introduction Wind is one of the major forces driving sea ice, and the transport of sea ice, in turn, influences the climate system in a number of ways. Pollard and Thompson (1994) and Holland et al. (2001b), for example, showed that accounting for sea ice dynamics can considerably moderate the sensitivity of a GCM to global warming induced by increasing CO 2. Recently, Saenko and Weaver (2001) demonstrated the importance of winddriven sea ice motion for the formation of Antarctic Intermediate Water. The fact that sea ice forms close to the regions of formation of global-scale water masses suggests that sea ice motion could play an important role in determining the formation rates and sites of these water masses. However, many coupled GCMs still use thermodynamic-only sea ice models or models with simplified advection schemes for sea ice [see Kreyscher et al. (2000) for a discussion]. Here it is our objective to evaluate the role of wind-driven sea ice motion in ocean ventilation. Ocean ventilation is defined as the renewal of ocean interior waters by surface waters that has been in contact with the atmosphere (England and Maier-Reimer 2001). It is one of the most important processes relevant to the ocean s role in climate, intimately connected to water mass formation in polar and subpolar regions. The rate Corresponding author affiliation: Dr. Oleg Saenko, School of Earth and Ocean Sciences, University of Victoria, P.O. Box 3055, Victoria, BC V8W 3P6, Canada. E-mail: oleg@ocean.seos.uvic.ca of ventilation affects the ability of the ocean to uptake CO 2 and heat from the atmosphere, thereby affecting climate (e.g., Sarmiento and Le Quere 1996; Gregory 2000). Of particular interest is the effect of wind-driven sea ice motion on the ventilation of the global-scale water masses, such as North Atlantic Deep Water (NADW), Antarctic Intermediate Water (AAIW), Subantarctic Mode Water (SAMW), and Antarctic Bottom Water (AABW). In this study we attempt to obtain a better understanding of how the formation of these water masses and the associated uptake of CFC-11 are connected to the high-latitude atmospheric circulation that forces sea ice motion. Chlorofluorocarbons (CFCs) are purely anthropogenic tracers. The industrial release of CFC began in the early 1930s and their atmospheric concentrations increased rapidly until the last decade. The recent extensive use of CFCs in ocean models [e.g., England et al. 1994; Robitaille and Weaver 1995; Caldeira and Duffy 1998; Dutay et al. 2002; see also England and Maier- Reimer (2001) for a review] was motivated by a number of reasons. First, the uptake of CFC by the ocean is qualitatively similar to the uptake of anthropogenic CO 2. Second, CFCs are conservative tracers and their modeling, unlike CO 2, does not require biogeochemical subcomponents. Third, there is a growing number of CFC observations available for model evaluation. Dutay et al. (2002) compared the 13 models participating in the Ocean Carbon Model Intercomparison Project (OCMIP-2) with regard to their skill in simu- 2002 American Meteorological Society

DECEMBER 2002 SAENKO ET AL. 3377 FIG. 1. Model domain on the rotated grid geometry. Also shown are three ocean regions for which budgets of CFC have been diagnosed. lating the observed distribution of CFC-11. The participating models differed in complexity. In particular, some of them were ocean-only models, whereas others incorporated sea ice components. It was found that the models differed substantially in simulating the observed CFC-11. Probably the most dramatic differences were found in the models ability to simulate the locations of AABW formation. Some models formed AABW only in Drake Passage, contrary to observations which show the location of source regions of AABW around Antarctica (Orsi et al. 1999). It was found by Dutay et al. (2002) that the ocean models which were coupled to a sea ice model offered an improved representation of AABW. As we show below, it is an underestimation of FIG. 2. Zonally and annually averaged surface flux of CFC-11 for year 1990 in the (a) CTR simulation (solid line) and (b) WSO simulation (dashed line). the effect of sea ice motion that could result in the simulation of AABW formation within Drake Passage. We compare two model experiments: in one of them (CTR), both the sea ice and the ocean are forced by the wind stress, whereas in the second experiment (WSO) only the ice-free ocean is driven by the wind stress. These same experiments have also been considered by Saenko and Weaver (2001). Here we apply a more extensive analysis of ocean ventilation processes in both hemispheres, which includes overturning, convection, and budgets of simulated CFC-11. Detailed budgets of CFC-11 are presented for three regions: the Weddell Sea, the southeast Pacific, and the North Atlantic. These regions are associated with the formation of global-scale water masses. We show that internal and external exchanges of CFC in these regions, as well as overturning and convection, are considerably affected by the winddriven divergence of sea ice. The remainder of the paper is organized as follows. Section 2 provides a brief description of the model and the design of the numerical experiments. Results of the experiments are presented in section 3, starting with a discussion of the zonally and hemispherically averaged diagnostics (section 3a), which is followed by a separate discussion of the effect of the wind-driven sea ice motion on the uptake of CFC in the model and the ventilation processes in the Southern Ocean (section 3b) and in the North Atlantic (section 3c). Section 4 compares simulated and observed distributions of CFC-11. A discussion and conclusions follow in section 5. 2. The coupled model and experimental design In this work we use a climate model of intermediate complexity (Weaver et al. 2001) with an additional trac-

3378 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 3. Zonally and annually averaged latitude depth section of (a) CFC concentration simulated by the CTR model, (b) CFC concentration simulated by the WSO model, and (c) difference CTR minus WSO models for year 1990. Contours are drawn at 0.3 pmol kg 1 intervals in (a) and (b) and at 0.1 pmol kg 1 intervals in (c). In (c) regions of negative differences are shaded. er representing CFC-11. The basic model components include the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM) version 2 (Pacanowski 1995), an updated version of a 2D energy moisture balance model (Weaver et al. 2001), an energyconserving thermodynamic sea ice model with a subgrid-scale ice thickness distribution allowing for six ice categories within each model grid cell (Bitz et al. 2001), and an elastic viscous plastic representation of sea ice dynamics (Hunke and Dukowicz 1997). The static stability criteria of ocean columns is checked separately under each ice category within a grid cell. A description of coupling, as well as results of the model application for simulations of past, present, and future climates can be found in Weaver et al. (2001). The ocean model uses isopycnal mixing after Gent and McWilliams (1990) with the coefficients of both thickness and isopycnal diffusivities set to 2.0 10 7 cm 2 s 1, and with mixing surfaces limited to a maximum slope of 0.01. The horizontal viscosity is 2.0 10 9 cm 2 s 1, and the coefficients of vertical diffusivity and viscosity are set to 0.8 and 10 cm 2 s 1, respectively. There are 19 vertical levels in the ocean model and up to 8 layers in the sea ice model. The atmospheric model diffusively transports heat and moisture. Precipitation occurs when the relative humidity exceeds a threshold value of 0.85. All model components employ a rotated grid geometry [see Weaver et al. (2001) for details] with horizontal resolution of 3.6 1.8 in longitude and latitude, respectively. Figure 1 shows the configuration of the model domain, as well as the ocean regions for which the budgets of CFC have been diagnosed and are discussed in the next section. In the model version used here, the external forcing of the coupled system is the solar radiation at the top of the atmosphere and a specified wind stress field. The wind stress data are taken from the National Centers for Environmental Prediction National Center for Atmospheric Research reanalysis of Kalnay et al. (1996), averaged over the period 1958 97 to form an annual cycle from the monthly fields. The model was initially spun up for 1000 years. Then, given the reconstructed history of the atmospheric CFC- 11 from the 1930s to 1990s (e.g., see England and Maier-Reimer 2001), two experiments were conducted in which the uptake of CFC-11 (hereafter CFC) was incorporated. The surface flux of CFC into the ocean is parametrized following England et al. (1994) [their Eq. (8)], except we use model-predicted sea ice concentrations here. In the control experiment (CTR), both sea ice and the ocean surface are forced by wind stress. In the second experiment (WSO), wind stress is applied only to ice-free ocean surface. Two additional experiments are discussed in section 4. In one of them, we restricted the area of open water in the ice-covered regions of the Southern Ocean to be at least 5% for the surface flux of CFC. The other experiment was the same as WSO, except we use sea ice concentration from the CTR experiment to modulate the CFC flux. 3. Results First the impact of the wind-driven divergence of sea ice on the zonally and/or hemispherically averaged distributions of CFC simulated by the CTR and WSO models is addressed. This is followed by more detailed analysis

DECEMBER 2002 SAENKO ET AL. 3379 FIG. 4. Depth profiles of CFC in the CTR run in (a) Northern Hemisphere and (c) Southern Hemisphere. Depth profiles of CFC difference between the CTR and WSO runs in (b) Northern Hemisphere and (d) Southern Hemisphere. Shown in (b) and (d) also are the approximate depths of the NADW, AAIW, and AABW water masses. of the physical processes involved and the CFC uptake in the Southern Hemisphere and the North Atlantic. a. Global uptake of CFC Both model versions, CTR and WSO, simulate the largest flux of CFC into the ocean in subpolar regions, peaking around 65 N and 55 S (Fig. 2), in general agreement with previous modeling results (e.g., Dutay et al. 2002). The large fluxes of CFC are associated with intensive ocean ventilation around these latitudes. In order to maintain the enhanced surface flux, the CFCs need to be effectively pumped from the surface into the ocean interior. Thus, differences in CFC fluxes are typically indicative of the different intensity of water column ventilation below, except in ice-covered regions. From the zonally averaged FIG. 5. Seasonal cycle of (a) sea ice volume anomaly relative to the annual mean and (b) ice area (space integral of ice concentration) simulated by the CTR model (solid line) and by the WSO model (dashed line) for the Southern Hemisphere. Also shown in (b) with is observed sea ice area climatology from 1973 to 1990 (see Walsh and Johnson 1979).

3380 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 6. Concentration of CFC in the Southern Ocean at depth 2872 m for year 1990 simulated by the (a) CTR model and (b) WSO model. In (b) also shown is a maximum concentration (units are pmol kg 1 ). CFC fluxes in Fig. 2, it can be preliminarily concluded that the ocean simulated by the WSO model is likely to be more effectively ventilated around 65 N and 55 S as compared to the CTR simulation. Moving downward from the surface into the ocean interior (Fig. 3), it can be seen that the latitudes of large surface fluxes are indeed associated with subpolar and polar regions where the surface CFC signal quickly penetrates to the intermediate and deep ocean. In the Northern Hemisphere, both models show a sinking of CFC down to about 3000 m, associated with the formation of NADW. In the Southern Hemisphere, enhanced CFC concentrations at intermediate depths reflect the formation of AAIW and SAMW in the CTR model (Fig. 3a). The WSO model also simulates high CFC at intermediate depths of the Southern Hemisphere (Fig. 3b); but as shown by Saenko and Weaver (2001), AAIW is too saline in this simulation. This suggests that the penetration of the CFC signal to intermediate depths may not necessarily be associated with the formation of AAIW. In fact, we will show that the CTR and WSO models simulate considerably different vertical structures and penetration mechanisms of CFC in the southeast Pacific, believed to be a dominant region of AAIW formation (McCartney 1977; England et al. 1993). At high southern latitudes, the distribution of CFC reflects the formation of AABW in the CTR model, which can be seen in both the surface flux of CFC (Fig. 2) and in its oceanic distribution close to Antarctica (Fig. 3a). These features are largely suppressed in the WSO run. The hemispherically averaged vertical profiles of CFC (Fig. 4) and their differences provide additional useful information for the preliminary analysis. Even though the concentrations of CFC are largest in the upper 1000 m (Figs. 4a,c), the differences between the vertical distributions of CFC simulated by the two models below 1000 m are of comparable magnitude to, or even larger than, the differences in the upper ocean (Figs. 4b,d). This is particularly the case at depths associated with horizontal propagation of AAIW and NADW. Thus, from the globally and hemispherically averaged CFC concentrations simulated by the models, it can be concluded that the high-latitude atmospheric circulation is able to considerably affect the ventilation of intermediate, deep, and bottom water through driving sea ice in both hemispheres. In the following, we try to understand in more detail at which locations the sea ice divergence influences ocean ventilation, particularly in the Southern Ocean and North Atlantic. Because the simulated ocean in the Southern Hemisphere is able to accommodate much more CFC than that in the Northern Hemisphere (see Figs. 4a,c), we begin our discussion with the Southern Ocean. b. Southern Ocean ventilation Before going into the analysis of the oceanic circulation and CFC uptake, it is useful to make a few remarks on how the simulated sea ice in the Southern Ocean is affected by the wind forcing and how this compares with other model results. Fichefet and Morales

DECEMBER 2002 SAENKO ET AL. 3381 FIG. 7. Zonally integrated meridional overturning streamfunction in the Southern Ocean (calculated using Eulerian mean velocity, which advects velocity) simulated by the (a) CTR model and by the (b) WSO models (units are Sv). Maqueda (1997) and Hewitt et al. (2001) found that the thermodynamic-only sea ice components of their models considerably overestimate ice mass and compactness and underestimate their annual cycles. In particular, Hewitt et al. (2001) show that without sea ice dynamics their model simulates a steady increase of Southern Hemisphere sea ice thickness from 0.8 to 1.8 m over the 20-yr run of their GCM. They point out that the increase occurs mainly over a small area around the Antarctic coast, which is supported by our results. We find that without a transfer of momentum from the winds to sea ice, the sea ice velocities are considerably reduced

3382 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 8. Downwelling in the Weddell Sea region as a function of depth: in CTR simulation (solid line) and WSO simulation (dashed line). so that the simulated sea ice is basically motionless (not shown). The Antarctic annual mean sea ice volume in the WSO run is about 30 10 3 km 3, which is much larger than that of about 8.5 10 3 km 3 simulated by the CTR model and that of about 10 4 km 3 estimated from observations (Weatherly et al. 1997). Even though in both models (CTR and WSO) the simulated sea ice edge in the Southern Ocean is in reasonable agreement with observations (see Saenko and Weaver 2001), the WSO model tends to attenuate the seasonal variations of sea ice volume and area (Figs. 5a,b). Both models, however, overestimate the sea ice area (i.e., space integral of ice concentration) compared to the data (Fig. 5b), which means that the fraction of leads and polynyas is underestimated. The onset of sea ice melt is delayed in the WSO model by more than a month (Fig. 5a). The Northern Hemisphere sea ice is, in general, found to be less affected by the wind forcing (not shown). Figures 6a and 6b show the distributions of CFC simulated by the two models in the deep Southern Ocean. There is a strong CFC signal in Drake Passage when sea ice is not forced by the winds (Fig. 6b), indicating enhanced deep ocean ventilation there. On the other hand, the CFC signal is very weak around the Antarctic continent. This contradicts the common view and the recent compilation of CFC data, showing that the sources of AABW are located around Antarctica (Orsi et al. 1999). In general agreement with these data, the CTR model simulates the formation AAWB in the western Weddell Sea and in the Ross Sea (Fig. 6a). Qualitatively the same result was obtained by setting sea ice concentration to zero for the CFC surface flux (not shown). This suggests that it is weak ocean ventilation, rather then the higher sea ice concentrations in the WSO model, that prevents a penetration of CFC into the deep Weddell and Ross Seas. The two experiments above show a different rate of meridional overturning near Antarctica (Figs. 7a,b: note that the shape of the Deacon cell is also affected by sea ice dynamics). Thus, the large-scale wind-driven divergence of sea ice around Antarctica controls the locations and rates of bottom-water formation in the Southern Hemisphere. Dutay et al. (2002) also found formation of bottom water in Drake Passage simulated by some of the models participating in the OCMIP-2 project. They concluded that those models which are coupled to sea ice components show much better skills in capturing the proper location of AABW formation. The offshore motion of sea ice from regions around Antarctica exposes open water to cold air and leads to thinner sea ice. This enhances heat loss by the ocean, regrowth of new ice and the associated brine release into the ocean, enabling the production of AABW. In turn, the northward exported sea ice feeds, upon melting, the AAIW with freshwater (Duffy et al. 2001; Saenko and Weaver 2001). In ocean-only models, the described coupling between sea ice dynamics and thermodynamics and its effect on the freshwater flux in the Southern Ocean is sometimes accounted for through the use of restoring to enhanced salinity off the Antarctic continent (e.g., England 1992). However, it is possible that oceanonly models, which tend to simulate AABW formation in Drake Passage, may underestimate the above described effect of sea ice dynamics. We now take a closer look at the simulated vertical and horizontal circulations in the Weddell Sea, which is believed to be a major region of AABW formation. The different rate of meridional overturning simulated by the two models close to Antarctica (Fig. 7) is mainly due to the different rate of downwelling in the Weddell Sea region (Fig. 8; see Fig. 1 for the region location). It should be noted that the range of estimates for the rate of AABW production is rather large. The generally accepted estimate for the net AABW production is of the order of 10 Sv (Sv 10 6 m 3 s 1 ; e.g., Orsi et al. 1999), whereas some other estimates give a value of the order of 1 Sv for the AABW production in the Weddell Sea (Fahrbach et al. 2001). In our experiments, less than 1.5 Sv downwells in the Weddell Sea region in the WSO case at depths of around 2000 m versus about 6 Sv in the CTR run (Fig. 8). This suggests that the local winds implicitly (through sea ice divergence followed by brine rejection and enhanced heat loss) control the thermohaline circulation in the Weddell Sea. It might be speculated that the aforementioned large degree of uncertainty in the rate of AABW production in the Weddell Sea is due to high variability of the local winds. For example, it is possible that the variability of AABW production is largely controlled by katabatic winds that rush downward off the ice sheet onto the coastal areas of Antarctica. The horizontal structure of the Weddell Sea circulation is also considerably affected by sea ice motion. From hydrographic data it is known that the Weddell

DECEMBER 2002 SAENKO ET AL. 3383 FIG. 9. (a) (b) Velocity and (d) (e) potential density in the Weddell Sea region at depth 793 m in (a) (d) CTR simulation and (b) (e) WSO simulation. Also shown are velocity (c) and potential density (f) in the experiment when ocean under sea ice is directly driven by the wind.

3384 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 10. Annual mean difference of near-surface air temperature between the CTR and WSO runs (contour interval is 0.5 C). Regions of negative difference are shaded. gyre is a large cyclone, located south of the Antarctic Circumpolar Current (Orsi et al. 1993; Orsi et al. 1999). The gyre is captured by our CTR model (Fig. 9a), whereas it is very weak in the WSO run (Fig. 9b). The large-scale horizontal circulation of the Weddell gyre is modified by the pressure gradients, associated with buoyancy fluxes (brine rejection and heat loss) due to the offshore sea ice transport. This can be seen from the simulated horizontal density structure (Figs. 9d,e). TABLE 1. Cumulative budgets of CFC-11 for the Weddell Sea region (Fig. 1) from 1930 to 1990, simulated by the CRT and WSO models. The region is subdivided by three layers, 0 228 m, 228 1128 m, and 1128 m bottom. Positive numbers indicate positive contribution due to a particular process to the net budget of CFC-11 in the corresponding layer; H stands for horizontal and V stands for vertical transports. In the middle layer there are two vertical transports, from layers above 1 and below 2. (Units are mol 10 5.) Layer Interface CTR WSO 0 228 m Surface H transport V transport Convection Subtotal 228 1128 m H transport V transport 1 V transport 2 Convection Subtotal 1128 m bottom H transport V transport Convection Subtotal 46 45 64 9 18 5 64 45 6 30 22 45 2 25 4 42 18 3 25 27 18 10 2 37 2 10 1 9 Total 73 71 There is an increase of density toward the western part of the Weddell Sea in the CTR run, which is not captured by the WSO model. To further illustrate the indirect effect of high-latitude winds on the oceanic circulation around Antarctica, we conducted an additional experiment in which the wind stress was applied to the ocean under the sea ice so that the ice is only driven by the resulting upper ocean currents. Such an approach is used in some climate models (e.g., Hewitt et al. 2001). Figures 9c and 9f shows that, in this case, the Weddell Sea circulation is still rather weak and, accordingly, its density structure still does not resemble that simulated by the CTR model. This suggests that the cyclonic circulation of the Weddell gyre is, at least in part, driven by the buoyancy fluxes implicitly associated with the wind-driven ice motion. The same conclusion applies to the Ross Sea gyre (not shown). It should be noted that the near-surface oceanic circulation, which is another force driving sea ice, is also considerably affected by the buoyancy fluxes due to the sea ice divergence. As might be expected, the different general circulation of the Weddell and Ross Seas simulated by the two models also affects budgets of heat and freshwater in these regions. It is known that the Weddell gyre entrains heat laterally, mainly from the Antarctic Circumpolar Current (e.g., Orsi et al. 1993), and moves it toward the nearshore regions of deep- and bottom-water formation. This lateral heat gain is balanced by heat lost through the surface. In the CTR run, the surface heat loss in the Weddell Sea region (see Fig. 1) amounts to about 76 TW (1 TW 10 12 W), whereas it is only 20 TW in the WSO run. About the same estimates apply to the Ross Sea, where the surface heat loss is about 63 TW in the

DECEMBER 2002 SAENKO ET AL. 3385 FIG. 11. Section of CFC-11 simulated by (a) the CTR model and (b) the WSO model at 85 W for year 1990. (c) Difference between the CTR and WSO models. (Contour interval is 0.3 pmol kg.) Negative values in (c) are shaded. CTR run versus 16 TW in the WSO. The larger heat loss through the surface in the regions around Antarctica due to the wind-driven sea ice divergence results in higher air temperatures there, particularly over the Ross Sea (Fig. 10). The proximity of the Weddell Sea to Drake Passage, where the CTR model simulates less FIG. 12. Annual mean section of salinity at 85 W (a) simulated by the CTR model, (b) simulated by the WSO model, and (c) observed (Levitus and Boyer 1994). (Contour interval is 0.05 psu.) heat lost to the atmosphere and lower air temperatures (Fig. 10), partly offsets the increase of air temperatures over the Weddell Sea due to the sea ice divergence. The oceanic freshwater budget in regions around Antarctica also responds to the export of sea ice. The CTR model simulates about 0.069 and 0.037 Sv of net annual freshwater lost by the ocean though the surfaces within

3386 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 13. Downwelling in the Drake Passage region as a function of depth in the CTR simulation (solid line) and the WSO simulation (dashed line). the Weddell Sea and Ross Sea, respectively. In contrast, the WSO model gives close to zero surface freshwater budgets in both seas. Table 1 summarizes budgets of CFC simulated by the two models for the Weddell Sea region (see Fig. 1). It can be seen that both models give about the same total amount of CFC in the Weddell Sea box by year 1990, however due to different processes. The penetration of CFC through the surface into the upper layer (0 228 m) is about 11 times lower when sea ice does not move. In the middle layer (228 1128 m), the lower penetration of CFC from above in the WSO run is compensated by larger horizontal transport. More CFC penetrates and escapes into the deep layer (below 1128 m) because of more intensive vertical and horizontal transports as a result of sea ice motion. Another region of particular interest is the southeast Pacific, close to Drake Passage (see Fig. 1). Our attention to this region is motivated by two reasons: First, it is the region where the WSO model tends to ventilate the deep and bottom waters (Fig. 6b). Second, it has been suggested that most of the AAIW forms in the southeast Pacific (McCartney 1977; England et al. 1993). As noted above, our model distinctly simulates a tongue of fresh AAIW only when the sea ice around Antarctica is forced by winds (Saenko and Weaver 2001). On the other hand, as we illustrated in Fig. 3b, the CFC signal still penetrates into the intermediate depths between 50 and 60 S even without moving sea ice by the wind. Moreover, the WSO model shows even larger concentrations of CFC at intermediate depths of the Southern Hemisphere (Fig. 4d), particularly to the north of about 65 S (Fig. 3c). Thus when winds do not force sea ice to move, the penetration of CFC into the intermediate depths of the Southern Hemisphere seen in (Fig. 3b) is not associated with the formation of low- FIG. 14. Maximum depth of convective ventilation and the sea ice edge (dashed line) in the Southern Ocean in the (a) CTR and (b) WSO runs. Numbers show maximum depths of convection.

DECEMBER 2002 SAENKO ET AL. 3387 TABLE 2. As in Table 1 but for the southeast Pacific region (Fig. 1). Layer Interface CTR WSO 0 228 m Surface H transport V transport Convection Subtotal 228 1128 m H transport V transport 1 V transport 2 Convection Subtotal 1128 m bottom H transport V transport Convection Subtotal 497 70 353 4 70 96 353 72 4 189 32 72 0 40 629 9 521 33 66 186 521 185 22 172 77 185 11 119 Total 299 357 salinity AAIW. Close to Drake Passage, the tongue of CFC at intermediate depths in the WSO run no longer resembles that in the CTR run (Fig. 11). Unlike in the CTR, there is no resemblance with the observed vertical salinity structure there (Fig. 12). The two experiments simulate a considerably different rate and depth of downwelling and convection near Drake Passage (Figs. 13 and 14). At around depth 1200 m, the downwelling reaches its maximum of approximately 5 Sv in the WSO run, whereas it is close to zero in the CTR case. This, in turn, is a result of weaker stratification in the WSO run around 60 S, especially close to Drake Passage, seen also in the convection depth (Fig. 14). The weaker stratification is caused by less freshwater input from the melting of Antarctic sea ice. The aforementioned more intensive heat loss by the ocean within Drake Passage in the WSO experiment also contributes to enhanced downwelling there (see Fig. 10). An interesting feature to note is that in the WSO run, convective mixing has a circumpolar structure, closely following the sea ice edge (Fig. 14b) and, unlike in the CTR run (Fig. 14a), there is no convection in the Weddell and Ross Seas. Thus, the reduction of winds and the associated northward transport of sea ice in the Southern Ocean increases the depth of convection along the sea ice edge during Southern Hemisphere winter, particularly in the regions close to Drake Passage (cf. Fig. 14a and Fig. 14b). This effect of Southern Ocean winds on the ocean ventilation may not always be properly captured by ocean-only models with restoring boundary conditions at the surface and/or prescribed (i.e., decoupled from the winds) salinity enhancement around Artarctica. For example, Ribbe (1999, 2001) does not mention this effect. To summarize, the WSO model shows enhanced downward mixing of CFC around 60 S, which particularly intensifies and reaches the deep ocean near Drake Passage. The more intensive downward mixing of CFC FIG. 15. Meridional overturning streamfunction in the North Atlantic (calculated using Eulerian mean velocity, which advects velocity) in (a) CTR simulation, (b) WSO simulation, and (c) the difference WSO minus CTR. (Units are Sv.) Contour interval is 2 Sv in (a) and (b) and 1 Sv in (c). Negative values are shaded.

3388 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 16. Annual-mean difference (WSO minus CTR) of the simulated CFC and ocean velocity in the North Atlantic at the depth of 2872 m for year 1990. around 60 S contributes to a larger surface CFC flux at these latitudes when sea ice is not forced by winds (see Fig. 2). Quantitative estimates of the external and internal exchanges of CFC in the southeast Pacific region (see Fig. 1) are given in Table 2. Vertical processes dominate the budgets. The WSO model simulates a considerably larger vertical penetration of CFC from the atmosphere to the surface layer and down to intermediate and deep layers in this region. Also note that, unlike in the CTR run, the CFC penetrates convectively to depths below 1100 m, which can also be concluded from Fig. 14. The mass transport through Drake Passage also shows considerable sensitivity to the sea ice motion. It has been suggested (e.g., Gent et al. 2001) that there is a strong correlation between the thermohaline circulation off the Antarctic shelf and the mass transport through Drake Passage. As we showed above, the winds at high southern latitudes intensify the thermohaline circulation off Antarctica through driving sea ice divergence (see Figs. 7a,b). The mass transport through Drake Passage simulated by the WSO model is about 36% less than in the CTR model (62 versus 97 Sv, respectively). This is consistent with Gent et al. (2001), who found that an increase in the thermohaline overturning off the Antarctic shelf of 1 Sv produces an increase of about 7 Sv in Drake Passage transport. c. North Atlantic ventilation The northern North Atlantic is another region, where the CFC signal quickly penetrates into the deep ocean due to formation of NADW. The northern North Atlantic is connected to the Arctic, which makes it possible for the high-latitude winds to considerably affect the variability of deep-water formation there through sea ice export from the Arctic (Holland et al. 2001a; Saenko et al. 2002, hereafter SWW). The freshwater introduced into the North Atlantic with sea ice in the CTR run contributes to a weakening of the meridional overturning there. On the other hand, the increase of the freshwater flux to the North Atlantic due to an increase of sea ice export can be partly offset by reduced liquid freshwater transport from the Arctic due to an associated smaller Arctic liquid freshwater content (SWW). At steady state, the net effect, however, results in a weaker and shallower overturning cell in the CTR run (Fig. 15). The more intense North Atlantic overturning in the WSO run enhances the uptake of CFC

DECEMBER 2002 SAENKO ET AL. 3389 FIG. 17. Maximum depth of convective ventilation and the sea ice edge (dashed line) in the North Atlantic in the (a) CTR and (b) WSO runs. Over the land areas, the dashed line shows winter snow edge. Numbers show maximum depths of convection. from the atmosphere and the intensified flow of NADW advects more CFC out of the North Atlantic (Fig. 16). Convection in the North Atlantic does not change as dramatically as in the Southern Ocean. The sites of maximum deep convection are shifted from the west to the northeastern part of the basin (Figs. 17a,b). It should be noted that both models overestimate the southward extend of the sea ice edge (Figs. 17a,b), which is a common problem in coarse-resolution climate models (see also Holland et al. 2001a). Table 3 summarizes the budgets of CFC for the North Atlantic region (see Fig. 1), where both models simulate most of the NADW formation. The net contribution of convection to the internal redistribution of CFC between the layers of the North Atlantic water column is about the same in the two models. One of

3390 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 TABLE 3. As in Table 1 but for the North Atlantic region (Fig. 1). Layer Interface CTR WSO 0 228 m Surface H transport V transport Convection Subtotal 228 1128 m H transport V transport 1 V transport 2 Convection Subtotal 1128 m bottom H transport V transport Convection Subtotal 297 30 247 41 39 18 247 184 39 120 103 184 2 83 341 15 278 40 38 42 278 249 39 110 150 249 1 100 Total 243 248 the major differences between the two budgets is the amount of CFC transported vertically in and out of the intermediate layer (228 1128 m). In the WSO experiment, the vertical uptake of CFC from the surface layer (228 1128 m) is about the same as loss to the deeper layer (not accounting for convection). However, in the CTR experiment more CFC penetrates into the intermediate layer from above than exits below to the deep layer (Table 3). This is due to the shallower meridional overturning in the CTR run (Fig. 15). The increased penetration of CFC through the surface, simulated by the WSO model, is almost balanced by the net increase in horizontal export of CFC out of the North Atlantic box. As a result, the two models simulate about the same net amount of CFC in the North Atlantic water column by year 1990. 4. Comparison with observations There is a growing number of CFC observations in the ocean that are being extensively used to test simulated uptake and transport of CFCs by models (e.g., England et al. 1994; Robitaille and Weaver 1995; Dutay et al. 2002), which is also the objective of this section. Here, three sections have been chosen for comparison, where the two models, CTR and WSO, show considerable differences in deep-water formation. Two of the sections (S4P and the western part of AJAX) are in the Southern Ocean (see Figs. 18 and 19). The third section (NA20W) is in the North Atlantic (see Fig. 20). The observations from the AJAX and NA20W sections have beed published in Weiss et al. (1990) and Doney and Bullister (1992), respectively. The S4P CFC data are unpublished (J. Bullister and M. Warner 2001, personal communication). In order to achieve an objective comparison with the observations, the modeled CFCs were interpolated into the locations of the observations. Then, both modeled and observed CFCs were gridded, using the same mapping routine. Preliminary comparison with the observations indicated that both models underestimate the CFC concentration in the deep Southern Ocean. Also, as we noted above, the fraction of open water in the icecovered regions of the Southern Ocean is underestimated, even in the CTR experiment. These motivated an additional experiment with the CTR model, in which we constrained the fraction of open water to be at least 5% in the ice-covered grid cells of the Southern Ocean in order to account for the presence of leads and polynyas. In this experiment (denoted CTR ICE05), the restriction on the minimum fraction of open water was applied only to the CFC flux, but not to the heat and freshwater fluxes, so that it did not affect ocean circulation. One more experiment was conducted in order to separate the effects due to the internal ocean dynamics and ice coverage on the simulated distribution of CFC. In this experiment (denoted WSO ICE CTR) we use the WSO model, but with ice concentration for the CFC flux diagnosed from the CTR model. The comparison of this experiment with CTR isolates the effect of changed ocean dynamics, while its comparison with WSO yields the influence of sea ice coverage on the CFC uptake. In the Southern Ocean, accounting for wind-driven sea ice divergence results in a dramatic improvement of the models ability to capture the observed distribution of CFC, particularly close to Drake Passage (Figs. 18 and 19). There, the large concentrations of CFC in deep and intermediate waters predicted by the WSO model (Figs. 18d and 19d) completely disagree with the observations (Figs. 18e and 19e). Another conspicuous property of the CTR and CTR ICE05 models is their ability to capture the correct location of bottomwater formation in the Weddell and Ross Seas, seen as filaments of enhanced concentration of CFC around 170 E in Figs. 18a,b and around 50 W in Figs. 19a,b. However, the simulated concentrations of CFC in the filaments and near bottom are underestimated compared to the observed (Figs. 18e and 19e). Note that allowing at least 5% of open water in the Southern Ocean gives better agreement with the observations. Using ice coverage from the CTR run in the WSO model to modulate the CFC flux (WSO ICE CTR run; Figs. 18c and 19c) slightly improves the distribution of CFC along the two Southern Ocean sections. However, toward Drake Passage the CFC concentrations are still too high. This suggests that the differences in the internal ocean dynamics dominate the changes in the CFC uptake simulated by the CTR and WSO models along the two sections, particularly close to Drake Passage. Close to the Antarctic coast, the more divergent and less compact sea ice cover increases CFC uptake and concentration in the deep ocean in the WSO ICE CTR run compared to the WSO run (not shown). In the North Atlantic, both the CTR and the WSO models capture the general structure of the observed uptake of CFCs (Fig. 20). The differences between the

DECEMBER 2002 SAENKO ET AL. 3391 FIG. 18. Concentration of CFC-11 along the 1992 S4P section in the Pacific sector of the Southern Ocean (see right top corner for the section location) (a) simulated by the CTR model with at least 5% of open water in the Southern Ocean allowed for the CFC flux; (b) simulated by the CTR model; (c) simulated by the WSO model, but with ice concentration for the CFC flux diagnosed from the CTR model; (d) simulated by the WSO model; and (e) observed. Units are pmol kg 1.

3392 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 FIG. 19. As in Fig. 18 but for the western part of the 1984 AJAX section.

DECEMBER 2002 SAENKO ET AL. 3393 FIG. 20. Concentration of CFC-11 along the 1988 NA20W section in the North Atlantic (see right top corner for the section location) (a) simulated by the CTR model, (b) simulated by the WSO model, and c) observed. (Units are pmol kg 1.) two models are much smaller than in the Southern Ocean. In the CTR run, the core of CFC rich NADW penetrating southward is confined to levels above 2500 m, while in the WSO model the core is broader and reaches greater depths. This leads to too large CFC concentrations in waters north of 40 N and depths between 2000 and 3000 m in the WSO model when compared to the observations. However, comparison with the observations does not lead to an obvious unequivocal preference of one of the model versions. 5. Discussion and conclusions We have simulated the distributions of CFC-11 to evaluate the effect of high-latitude winds on ocean ventilation in polar and subpolar regions. It was shown that the uptake and distribution of simulated CFCs in the ocean s interior can be considerably affected by the buoyancy fluxes associated with the wind-driven divergence of sea ice through its impact on downwelling and convection.

3394 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32 In the Southern Hemisphere, the high-latitude winds control the formation of AABW close to the Antarctic continent through driving sea ice motion. Without transfer of momentum from the winds to the sea ice, bottom water tends to form within Drake Passage. This explains the tendency for some ocean-only models to form bottom water within Drake Passage as found by Dutay et al. (2002). We suggest that ocean-only models, which do not show formation of AABW close to the Antarctic continent, may have underestimated the effect of largescale divergence of sea ice. Northward transport of sea ice from around Antarctica, coupled to sea ice thermodynamics, produces a negative annual freshwater flux around the Antarctic continent by increasing sea ice formation around Antarctica and its melt elsewhere. This, as well as the associated enhanced heat loss though open water and thiner ice, appear to be critical for the simulation of AABW formation in the Weddell and Ross Seas. The northward advected sea ice, in turn, has two important effects: First, it feeds AAIW with freshwater, contributing to the relatively low salinities of this water mass. Second, it leads to the freshening of the upper ocean around 60 S preventing deep convection there. In contrast, when sea ice is not driven by winds, the depth and area of convection increases all along the sea ice edge in the southern hemisphere, whereas no convection occurs in the Weddell and Ross Seas. Ribbe (1999), using an ocean-only model, found that his model simulates an increase of the depth of midlatitude convection (around 50 S) when wind velocities in the momentum transfer to the surface ocean south of 50 S were doubled. Our results suggest that an increase of wind velocities forcing sea ice in the Southern Ocean would tend to decrease the depth and area of convective activity at latitudes of maximum sea ice extent (around 60 S). This effect of Southern Ocean winds on ocean ventilation, however, cannot be simulated in ocean-only GCMs in which the surface salinity flux (particularly around Antarctica), is decoupled from winds and the associated ice motion. In the Weddell Sea region, both the vertical and horizontal circulations are considerably affected by the sea ice motion. We have shown that the negative buoyancy flux due to sea ice growth, encouraged by the winddriven sea ice divergence there, considerably intensifies the Weddell gyre and the rate of downwelling in the Weddell Sea region. It is illustrated that, even when the ocean below sea ice is directly driven by the winds as though there is no sea ice, the Weddell gyre is still much weaker compared to the case when momentum is transfered to sea ice wherever it is present. The overall reduced intensity of the vertical and horizontal circulations considerably affects the budgets of CFC, heat and freshwater in the Weddell Sea, as well as in the Ross Sea. When sea ice is not driven by the winds, there is much less uptake of CFC and the heat and freshwater loss through the surface within the Weddell and Ross Seas are much weaker. Without the direct forcing of sea ice by winds, our model fails to simulate a low-salinity tongue of AAIW, whereas it still captures a tongue of enhanced CFCs at intermediate depths of the Southern Hemisphere to the north of about 65 S, except in the southeast Pacific. This latter regions is believed to be a dominant region of AAIW renewal (McCartney 1977; England et al. 1993). There, the CFC signal penetrates down to the deep and abyssal ocean, unless sea ice is forced by winds. Thus, a simulation of zonally averaged penetration of CFC into the intermediate depths north of 65 S may not necessarily be associated with a proper simulation of low salinity AAIW. The more rapid penetration of CFC to the subsurface ocean in the WSO run results in a considerably larger CFC flux at 55 60 S. In the North Atlantic, wind-driven sea ice export from the Arctic weakens and shoals the meridional overturning cell. This results in a decreased flux of CFC around 65 N by about a factor of 2. The enhanced sea ice export from the Arctic also affects the spatial pattern of convection in the North Atlantic. In general, however, North Atlantic convection is less affected by the wind-driven sea ice transport than that in the Southern Hemisphere. A comparison of modeled and observed CFC concentrations shows that accounting for the momentum transfer from the wind to sea ice considerably improves the agreement with observations, particularly in the Southern Ocean. Acknowledgments. The authors are grateful to the Meteorological Service of Canada/Canadian Institute for Climate Studies, the International Arctic Research Center, CFCAS, and NSERC for supporting this work. The CFC data have been kindly provided by John Bullister. Data on Antarctic sea ice concentration have been obtained from the National Snow and Ice Data Center. We thank E. Wiebe, M. Eby, and M. Holland for helpful discussions and assistance, as well as two anonymous reviewers for their constructive suggestions. REFERENCES Bitz, C. M., M. M. Holland, A. J. Weaver, and M. Eby, 2001: Simulating the ice-thickness distribution in a coupled climate model. J. Geophys. Res., 106, 2441 2464. Caldeira, K., and P. B. Duffy, 1998: Sensitivity of simulated CFC- 11 distributions in a global ocean model to the treatment of salt rejected during sea-ice formation. Geophys. Res. Lett., 25, 1003 1006. Doney, S. C., and J. L. Bullister, 1992: A chlorofluorocarbon section in the eastern North Atlantic. Deep-Sea Res., 39, 1857 1883. Duffy, P., M. Eby, and A. Weaver, 2001: Climate model simulations of effects of increased atmospheric CO 2 and loss of sea ice on ocean salinity and tracer uptake. J. Climate, 14, 520 532. Dutay, J.-C., J. L. Bullister, S. C. Doney, and Coauthors, 2002: Evaluation of ocean model ventilation with CFC-11: Comparison of 13 global ocean models. Ocean Model., 4, 89 120. England, M. H., 1992: On the formation of Antarctic intermediate and bottom water in ocean general circulation models. J. Phys. Oceanogr., 22, 918 926.

DECEMBER 2002 SAENKO ET AL. 3395, and E. Maier-Reimer, 2001: Using chemical tracers to assess ocean models. Rev. Geophys., 39, 29 70., J. S. Godfrey, A. C. Hirst, and M. Tomczak, 1993: The mechanism for Antarctic Intermediate Water renewal in a world ocean model. J. Phys. Oceanogr., 23, 1553 1560., V. Garcon, and J.-F. Minster, 1994: Chlorofluorocarbon uptake in a world ocean model. 1. Sensitivity to the surface gas forcing. J. Geophys. Res., 99, 25 215 25 233. Fahrbach, E., S. Harms, G. Rohardt, M. Schroder, and R. A. Woodgate, 2001: Flow of bottom water in the northwestern Weddell Sea. J. Geophys. Res., 106, 2761 2778. Fichefet, T., and M. A. Morales Maqueda, 1997: Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. J. Geophys. Res., 102, 12 609 12 646. Gent, P. R., and J. C. McWilliams, 1990: Isopycnal mixing in ocean general circulation models. J. Phys. Oceanogr., 20, 150 155., W. G. Large, and F. O. Bryan, 2001: What sets the mean transport through Drake Passage? J. Geophys. Res., 106, 2693 2712. Gregory, J. M., 2000: Vertical heat transports in the ocean and their effect on time-dependent climate change. Climate Dyn., 16, 501 515. Hewitt, C. D., C. A. Senior, and J. F. B. Mitchell, 2001: The impact of dynamic sea-ice on the climatology and climate sensitivity of a GCM: A study of past, present, and future climates. Climate Dyn., 17, 655 668. Holland, M. M., C. M. Bitz, M. Eby, and A. J. Weaver, 2001a: The role of ice ocean interactions in the variability of the North Atlantic thermohaline circulation. J. Climate, 14, 656 675.,, and A. J. Weaver, 2001b: The influence of sea ice physics on simulations of climate change. J. Geophys. Res., 106, 19 639 19 655. Hunke, E. C., and J. K. Dukowicz, 1997: An elastic viscous plastic model for sea ice dynamics. J. Phys. Oceanogr., 27, 1849 1867. Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437 471. Kreyscher, M., M. Harder, P. Lemke, and G. M. Flato, 2000: Results of the Sea Ice Model Intercomparison Project: Evaluation of sea ice rheology schemes for use in climate simulations. J. Geophys. Res., 105, 11 299 11 320. Levitus, S., and T. P. Boyer, 1994: Temperature. Vol. 4, World Ocean Atlas 1994, NOAA Atlas NESDIS 4, 117 pp. McCartney, M. S., 1977: Subantarctic Mode Water. A Voyage of Discovery, M. V. Angel, Ed., Pergamon, 103 119. Orsi, A. H., W. D. Nowlin Jr., and T. Whitworth III, 1993: On the circulation and stratification of the Weddell Gyre. Deep-Sea Res., 40, 169 203., G. C. Johnson, and J. L. Bullister, 1999: Circulation, mixing, and production of Antarctic Bottom Water. Progress in Oceanography, Vol. 43, Pergamon, 55 109. Pacanowski, R., 1995: MOM 2 documentation, user s guide and reference manual. GFDL Ocean Group Tech. Rep. 3, GFDL, Princeton, NJ, 232 pp. Pollard, D., and S. L. Thompson, 1994: Sea-ice dynamics and CO 2 sensitivity in a global climate model. Atmos. Ocean, 32, 449 467. Ribbe, J., 1999: On wind-driven mid-latitude convection in ocean general circulation models. Tellus, 51A, 517 525., 2001: Intermediate water mass production controlled by Southern Hemisphere winds. Geophys. Res. Lett., 28, 535 538. Robitaille, D. Y., and A. J. Weaver, 1995: Validation of sub-gridscale mixing schemes using CFCs in a global ocean model. Geophys. Res. Lett., 22, 2917 2920. Saenko, O. A., and A. J. Weaver, 2001: Importance of wind-driven sea ice motion for the formation of Antarctic Intermediate Water. Geophys. Res. Lett., 28, 4147 4150., E. C. Wiebe, and A. J. Weaver, 2002: North Atlantic response to the above-normal export of sea ice from the Arctic. J. Geophys. Res., in press. Sarmiento, J. L., and C. Le Quere, 1996: Oceanic CO 2 uptake in a model of century-scale global warming. Science, 274, 1346 1350. Walsh, J. E., and C. M. Johnson, 1979: Analysis of Arctic sea ice fluctuations, 1953 77. J. Phys. Oceanogr., 9, 580 591. Weatherly, J. W., T. W. Bettge, and W. M. Washington, 1997: Simulation of sea ice in the NCAR Climate System Model. Ann. Glaciol., 25, 107 110. Weaver, A. J., and Coauthors, 2001: The UVic Earth System Climate Model: Model description, climatology and application to past, present and future climates. Atmos. Ocean, 39, 361 428. Weiss, R. F., J. L. Bullister, M. J. Warner, F. A. Van Woy, and P. K. Salameh, 1990: Ajax Expedition chlorofluorocarbon measurements. SIO Reference 90-6, Scripps Institution of Oceanography, University of California, San Diego, 190 pp.