• Beckmann, A., , and H. Goosse, 2003: A parameterization of ice shelf-ocean interaction for climate models. Ocean Modell., 5 , 157170.

  • Bitz, C. M., , M. M. Holland, , E. C. Hunke, , and R. E. Moritz, 2005: Maintenance of the sea ice edge. J. Climate, 18 , 29032921.

  • Bromwich, D. H., , and R. L. Fogt, 2004: Strong trends in the skill of the ERA-40 and NCEP/NCAR reanalyses in the high and middle latitudes of the Southern Hemisphere, 1958–2001. J. Climate, 17 , 46034619.

    • Search Google Scholar
    • Export Citation
  • Canuto, V. M., , A. Howard, , P. Hogan, , Y. Cheng, , M. S. Dubovikov, , and L. M. Montenegro, 2004: Modeling ocean deep convection. Ocean Modell., 7 , 7595.

    • Search Google Scholar
    • Export Citation
  • Comiso, J. C., 1995: SSM/I sea ice concentrations using the bootstrap algorithm. NASA Reference Publication 1380, 49 pp.

  • Comiso, J. C., , D. J. Cavalieri, , C. L. Parkinson, , and P. Gloersen, 1997: Passive microwave algorithms for sea ice concentration—A comparison of two techniques. Remote Sens. Environ., 60 , 357384.

    • Search Google Scholar
    • Export Citation
  • Connolley, W. M., , and S. A. Harangozo, 2001: A comparison of five numerical weather prediction analysis climatologies in southern high latitudes. J. Climate, 14 , 3044.

    • Search Google Scholar
    • Export Citation
  • Drijfhout, S. S., , C. Heinze, , M. Latif, , and E. Maier-Reimer, 1996: Mean circulation and internal variability in an ocean primitive equation model. J. Phys. Oceanogr., 26 , 559580.

    • Search Google Scholar
    • Export Citation
  • Ganachaud, A., , and C. Wunsch, 2000: Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408 , 453456.

    • Search Google Scholar
    • Export Citation
  • Goosse, H., , and T. Fichefet, 1999: Importance of ice-ocean interactions for the global ocean circulation: A model study. J. Geophys. Res., 104 , 2333723355.

    • Search Google Scholar
    • Export Citation
  • Goosse, H., , E. Deleersnijder, , T. Fichefet, , and M. H. England, 1999: Sensitivity of a global coupled ocean-sea ice model to the parameterization of vertical mixing. J. Geophys. Res., 104 , 1368113695.

    • Search Google Scholar
    • Export Citation
  • Gordon, A. L., 1998: Western Weddell Sea thermohaline stratification. Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margin, S. S. Jacobs and R. F. Weiss, Eds., Antarctic Research Series, Vol. 75, Amer. Geophys. Union, 215–240.

    • Search Google Scholar
    • Export Citation
  • Gordon, A. L., , and B. A. Huber, 1990: Southern Ocean winter mixed layer. J. Geophys. Res., 95 , 1165511672.

  • Harms, S., , E. Fahrbach, , and V. H. Strass, 2001: Sea ice transports in the Weddell Sea. J. Geophys. Res., 106 , 90579073.

  • Hellmer, H. H., 2004: Impact of Antarctic ice shelf basal melting on sea ice and deep ocean properties. Geophys. Res. Lett., 31 .L10307, doi:10.1029/2004GL019506.

    • Search Google Scholar
    • Export Citation
  • Holland, D., , S. S. Jacobs, , and A. Jenkins, 2003: Modelling the ocean circulation beneath the Ross ice shelf. Antarct. Sci., 15 , 1323.

    • Search Google Scholar
    • Export Citation
  • Holland, M. M., , and M. N. Raphael, 2006: Twentieth century simulation of the southern hemisphere climate in coupled models. Part II: Sea ice conditions and variability. Climate Dyn., 26 , 229245.

    • Search Google Scholar
    • Export Citation
  • Jacobs, S. S., , and J. C. Comiso, 1989: Sea ice and oceanic processes on the Ross Sea continental shelf. J. Geophys. Res., 94 , 1819518211.

    • Search Google Scholar
    • Export Citation
  • Jenkins, A., , and D. Holland, 2002: A model study of ocean circulation beneath Filchner-Ronne Ice Shelf, Antarctica: Implications for bottom water formation. Geophys. Res. Lett., 29 .1193, doi:10.1029/2001GL014589.

    • Search Google Scholar
    • Export Citation
  • Jungclaus, J. H., , H. Haak, , M. Latif, , and U. Mikolajewicz, 2005: Arctic–North Atlantic interactions and multidecadal variability of the meridional overturning circulation. J. Climate, 18 , 40134031.

    • Search Google Scholar
    • Export Citation
  • Kaleschke, L., , C. Lüpkes, , T. Vihma, , J. Haarpaitner, , A. Bochert, , J. Hartmann, , and G. Heygster, 2001: SSM/I sea ice remote sensing for mesoscale ocean-atmosphere interaction analysis. Can. J. Remote Sens., 27 , 526537.

    • Search Google Scholar
    • Export Citation
  • Kim, S-J., , and A. Stössel, 2001: Impact of plume convection on global thermohaline properties and circulation. J. Phys. Oceanogr., 31 , 656674.

    • Search Google Scholar
    • Export Citation
  • King, J. C., , J. Turner, , G. J. Marshall, , W. M. Connolley, , and T. A. Lachlan-Cope, 2003: Antarctic Peninsula climate variability and its causes as revealed by analysis of instrumental records. Antarctic Research Series, Vol. 79 , Amer. Geophys. Union. 1730.

    • Search Google Scholar
    • Export Citation
  • Koentopp, M., , O. Eisen, , C. Kottmeier, , L. Padman, , and P. Lemke, 2005: Influence of tides on sea ice in the Weddell Sea: Investigations with a high-resolution dynamic-thermodynamic sea ice model. J. Geophys. Res., 110 .C02014, doi:10.1029/2004JC002405.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and Coauthors, 2004: Reconstructing, monitoring, and predicting multidecadal-scale changes in the North Atlantic thermohaline circulation with sea surface temperature. J. Climate, 17 , 16051614.

    • Search Google Scholar
    • Export Citation
  • Legutke, S., , and E. Maier-Reimer, 1999: Climatology of the HOPE-G global ocean general circulation model. German Climate Computer Center Tech. Rep. 21, Hamburg, Germany, 90 pp.

  • Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp.

  • Lindsay, R. W., , and J. Zhang, 2005: The thinning of Arctic sea ice, 1988–2003: Have we passed a tipping point? J. Climate, 18 , 48794894.

    • Search Google Scholar
    • Export Citation
  • Lindsay, R. W., , and J. Zhang, 2006: Assimilation of ice concentration in an ice–ocean model. J. Atmos. Oceanic Technol., 23 , 742749.

    • Search Google Scholar
    • Export Citation
  • Markus, T., , and D. J. Cavalieri, 2000: An enhancement of the NASA Team sea ice algorithm. IEEE Trans. Geosci. Remote Sens., 38 , 13871398.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , and F. Schott, 1999: Open-ocean convection: Observations, theory,and models. Rev. Geophys., 37 , 164.

  • Marsland, S., , and J-O. Wolff, 2001: On the sensitivity of Southern Ocean sea ice to the surface fresh-water flux: A model study. J. Geophys. Res., 106 , 27232741.

    • Search Google Scholar
    • Export Citation
  • Marsland, S., , H. Haak, , J. H. Jungclaus, , M. Latif, , and F. Roeske, 2003: The Max-Planck-Institute global ocean/sea ice model with orthogonal curvilinear coordinates. Ocean Modell., 5 , 91127.

    • Search Google Scholar
    • Export Citation
  • Marsland, S., , N. L. Bindoff, , G. D. Williams, , and W. F. Budd, 2004: Modeling water mass formation in the Mertz Glacier polynya and Adelie depression, East Antarctica. J. Geophys. Res., 109 .C11003, doi:10.1029/2004JC002441.

    • Search Google Scholar
    • Export Citation
  • Martinson, D. G., 1990: Evolution of the Southern Ocean winter mixed layer and sea ice: Open ocean deepwater formation and ventilation. J. Geophys. Res., 95 , 1164111654.

    • Search Google Scholar
    • Export Citation
  • Martinson, D. G., , and R. A. Iannuzzi, 1998: Antarctic ocean-ice interaction: Implications from ocean bulk property distributions in the Weddell Gyre. Antarctic Sea Ice: Physical Processes, Interactions and Variability, M. O. Jeffries, Ed., Antarctic Research Series, Vol. 74, Amer. Geophys. Union, 243–271.

    • Search Google Scholar
    • Export Citation
  • Massom, R. A., and Coauthors, 2001: Snow on Antarctic sea ice. Rev. Geophys., 39 , 413445.

  • McPhee, M. G., , C. Kottmeier, , and J. H. Morison, 1999: Ocean heat flux in the central Weddell Sea during winter. J. Phys. Oceanogr., 29 , 11661179.

    • Search Google Scholar
    • Export Citation
  • Ogura, T., , A. Abe-Ouchi, , and H. Hasumi, 2004: Effects of sea ice dynamics on the Antarctic sea ice distribution in a coupled ocean atmosphere model. J. Geophys. Res., 109 .C04025, doi:10.1029/2003JC002022.

    • Search Google Scholar
    • Export Citation
  • Orsi, A. H., , and T. Whitworth III, 2005: Southern Ocean. Vol. 1, Hydrographic Atlas of the World Ocean Circulation Experiment (WOCE), International WOCE Project Office, 223 pp.

    • Search Google Scholar
    • Export Citation
  • Paluszkiewicz, T., , and R. D. Romea, 1997: A one-dimensional plume model for the parameterisation of oceanic deep convection. Dyn. Atmos. Oceans, 26 , 95130.

    • Search Google Scholar
    • Export Citation
  • Powell, D. C., , T. Markus, , and A. Stössel, 2005: Effects of snow depth forcing on Southern Ocean sea ice simulations. J. Geophys. Res., 110 .C06001, doi:10.1029/2003JC002212.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 996 pp.

    • Search Google Scholar
    • Export Citation
  • Renfrew, I. A., , J. C. King, , and T. Markus, 2002: Coastal polynyas in the southern Weddell Sea: Variability of the surface energy budget. J. Geophys. Res., 107 .3063, doi:10.1029/2000JC000720.

    • Search Google Scholar
    • Export Citation
  • Schodlok, M. P., , H. H. Hellmer, , G. Rohardt, , and E. Fahrbach, 2006: Weddell Sea iceberg drift: Five years of observations. J. Geophys. Res., 111 .C06018, doi:10.1029/2004JC002661.

    • Search Google Scholar
    • Export Citation
  • Sloyan, B. M., , and S. R. Rintoul, 2001: The Southern Ocean limb of the global deep overturning circulation. J. Phys. Oceanogr., 31 , 143173.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , and T. Markus, 2004: Using satellite-derived ice concentration to represent Antarctic coastal polynyas in ocean climate models. J. Geophys. Res., 109 .C02014, doi:10.1029/2003JC001779.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , S-J. Kim, , and S. S. Drijfhout, 1998: The impact of Southern Ocean sea ice in a global ocean model. J. Phys. Oceanogr., 28 , 19992018.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , K. Yang, , and S-J. Kim, 2002: On the role of sea ice and convection in a global ocean model. J. Phys. Oceanogr., 32 , 11941208.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , M. M. Stössel, , and J-T. Kim, 2007: High-resolution sea ice in long-term global ocean GCM integrations. Ocean Modell., 16 , 206223.

    • Search Google Scholar
    • Export Citation
  • Strass, V. H., , and E. Fahrbach, 1998: Temporal and regional variation of sea ice draft and coverage in the Weddell Sea obtained from upward looking sonars. Antarctic Sea Ice: Physical Processes, Interactions and Variability, M. O. Jeffries, Ed., Antarctic Research Series, Vol. 74, Amer. Geophys. Union, 123–139.

    • Search Google Scholar
    • Export Citation
  • Timmermann, R., , and A. Beckmann, 2004: Parameterization of vertical mixing in the Weddell Sea. Ocean Modell., 6 , 83100.

  • Timmermann, R., , A. Worby, , H. Goosse, , and T. Fichefet, 2004: Utilizing the ASPeCt sea ice thickness data set to evaluate a global coupled sea ice–ocean model. J. Geophys. Res., 109 .C07017, doi:10.1029/2003JC002242.

    • Search Google Scholar
    • Export Citation
  • Timmermann, R., , H. Goosse, , G. Madec, , T. Fichefet, , C. Ethe, , and V. Duliere, 2005: On the representation of high latitude processes in the ORCA-LIM global coupled sea-ice-ocean model. Ocean Modell., 8 , 175201.

    • Search Google Scholar
    • Export Citation
  • Vihma, T., , J. Uotila, , B. Cheng, , and J. Launiainen, 2002: Surface heat budget over the Weddell Sea: Buoy results and model comparisons. J. Geophys. Res., 107 .3013, doi:10.1029/2000JC000372.

    • Search Google Scholar
    • Export Citation
  • Whitworth, T., , A. H. Orsi, , S-J. Kim, , W. D. Nowlin, , and R. A. Locarnini, 1998: Water masses and mixing near the Antarctic slope front. Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margin, S. S. Jacobs and R. F. Weiss, Eds., Antarctic Research Series, Vol. 75, Amer. Geophys. Union, 1–27.

    • Search Google Scholar
    • Export Citation
  • Wolff, J-O., , E. Maier-Reimer, , and S. Legutke, 1997: The Hamburg Ocean Primitive Equation model HOPE. German Climate Computer Center Tech. Rep. 31, Hamburg, Germany, 98 pp.

  • View in gallery

    Seasonal cycles of ice area (1012 m2) as simulated with the indicated experiments (see Table 1) and as derived from passive-microwave data using NT2.

  • View in gallery

    (left) February and (right) September snapshots of ice concentration as simulated with experiment (top) RO, (middle) RN, and (bottom) as derived from satellite passive-microwave (year 1992) data using the NT2 algorithm. The isoline interval between 90% and 100% is 2%, whereas that between 20% and 90% is 10%.

  • View in gallery

    Early December snapshots of ice concentration as simulated with experiments (top) RN, (middle) SM, and (bottom) as derived from passive-microwave data using NT2. Otherwise as in Fig. 2.

  • View in gallery

    September variance of ice concentration, as simulated with experiment (top) RO, (middle) RN, and (bottom) as derived from passive-microwave data using NT2. Here, 1 = 100% ice concentration, contour interval is 0.02, minimum shown is 0.02, and maximum (red) is 0.24.

  • View in gallery

    (left) February and (right) September snapshots of ice thickness in the Weddell Sea as simulated with experiment (top) SM and (bottom) SMI. The isoline interval is 0.4 m.

  • View in gallery

    Atlantic meridional overturning circulation as simulated with experiment (top left) RO, (top right) RN, (bottom left) SD, and (bottom right) SM. The isoline interval is 3 Sv.

  • View in gallery

    Meridional sections along 40°W (averaged between 50° and 30°W) of potential temperature as simulated with experiments (top left) RO, (top right) RN, (middle left) SD, (middle right) SM, (bottom left) SMI, and (bottom right) as compiled from measurements by Levitus (1982). The isoline interval is 0.4°C, while the color scale interval is 1.2°C.

  • View in gallery

    Meridional sections of salinity. The isoline interval is 0.04 g kg−1, while the color-scale interval is 0.12 g kg−1. Otherwise as in Fig. 7.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 10 10 3
PDF Downloads 1 1 1

Employing Satellite-Derived Sea Ice Concentration to Constrain Upper-Ocean Temperature in a Global Ocean GCM

View More View Less
  • 1 Department of Oceanography, Texas A&M University, College Station, Texas
© Get Permissions
Full access

Abstract

The quality of Southern Ocean sea ice simulations in a global ocean general circulation model (GCM) depends decisively on the simulated upper-ocean temperature. This is confirmed by assimilating satellite-derived sea ice concentration to constrain the upper-layer temperature of a sea ice–ocean GCM. The resolution of the model’s sea ice component is about 22 km and thus comparable to the pixel resolution of the satellite data. The ocean component is coarse resolution to afford long-term integrations for investigations of the deep-ocean equilibrium response. Besides improving the sea ice simulation considerably, the simulations with constrained upper-ocean temperature yield much more realistic global deep-ocean properties, in particular when combined with glacial freshwater input. Both outcomes are relatively insensitive to the passive-microwave algorithm used to retrieve the ice concentration being assimilated. The sensitivity of the long-term global deep-ocean properties and circulation to the possible freshwater input from ice shelves and to the parameterization of vertical mixing in the Southern Ocean is reevaluated under the new constraint.

Corresponding author address: Dr. Achim Stössel, 3146 TAMU, College Station, TX 77843. Email: astoessel@ocean.tamu.edu

Abstract

The quality of Southern Ocean sea ice simulations in a global ocean general circulation model (GCM) depends decisively on the simulated upper-ocean temperature. This is confirmed by assimilating satellite-derived sea ice concentration to constrain the upper-layer temperature of a sea ice–ocean GCM. The resolution of the model’s sea ice component is about 22 km and thus comparable to the pixel resolution of the satellite data. The ocean component is coarse resolution to afford long-term integrations for investigations of the deep-ocean equilibrium response. Besides improving the sea ice simulation considerably, the simulations with constrained upper-ocean temperature yield much more realistic global deep-ocean properties, in particular when combined with glacial freshwater input. Both outcomes are relatively insensitive to the passive-microwave algorithm used to retrieve the ice concentration being assimilated. The sensitivity of the long-term global deep-ocean properties and circulation to the possible freshwater input from ice shelves and to the parameterization of vertical mixing in the Southern Ocean is reevaluated under the new constraint.

Corresponding author address: Dr. Achim Stössel, 3146 TAMU, College Station, TX 77843. Email: astoessel@ocean.tamu.edu

1. Introduction

It is a major challenge for sea ice–ocean general circulation models (GCMs) to arrive at a reasonable simulation of Southern Ocean sea ice simultaneously with long-term global deep-ocean properties and circulation. This applies to coupled atmosphere–ice–ocean GCMs (e.g., Holland and Raphael 2006; Bitz et al. 2005; Ogura et al. 2004; Jungclaus et al. 2005) as much as to ice–ocean GCMs that are forced by atmospheric variables (e.g., Goosse and Fichefet 1999; Timmermann et al. 2005; Marsland et al. 2004; Stössel et al. 2002).

This challenge is to some extent due to the lack of adequate atmospheric variables for the Southern Ocean. Sea ice is highly sensitive to winds including their variability and precipitation. Winds determine the lead fraction within the sea ice pack and the turbulent heat fluxes, both affecting cooling and brine release in fall and winter, and melting and seasonal stratification in spring and summer. Precipitation has an indirect impact on sea ice by affecting the oceanic stratification, and thus the vertical oceanic heat flux that affects sea ice (e.g., Marsland and Wolff 2001). Through this interaction with sea ice, these two atmospheric variables strongly affect the rates of deep- and bottom-water formation, and therefore the long-term deep-ocean properties.

For a sea ice–ocean GCM that is not coupled to a GCM of the atmosphere, (re)analyses from numerical weather prediction centers are considered the most reliable source of present-day atmospheric forcing. In data-sparse areas such as the Southern Ocean, these products are mainly results of an atmosphere GCM. Critical features, such as the zonal temperature gradient across the Antarctic Peninsula, and local katabatic and barrier winds arising from the steep and cold Antarctic terrain, are only barely captured (e.g., King et al. 2003; Connolley and Harangozo 2001; Renfrew et al. 2002; Bromwich and Fogt 2004; Vihma et al. 2002). Precipitation including its distribution in the form of snow on sea ice has even more uncertainty (e.g., Marsland and Wolff 2001; Massom et al. 2001; Powell et al. 2005).

In a fully coupled model, while allowing for feedbacks between atmosphere, ocean, and sea ice, the atmospheric circulation is and can only be worse than the (re)analyses. Even if the atmosphere GCM used for the fully coupled simulation was as high-resolution and sophisticated as the one used to produce the analyses, it would not be constrained by observational data assimilated into the model, and surface boundary conditions provided by the ice–ocean GCM would certainly be less realistic than what is used for the operational analyses or the reanalyses. So by design, one cannot expect to achieve reasonably realistic long-term deep-ocean properties in such models as long as these properties depend critically on the wind and precipitation fields over the deep- and bottom-water formation regions.

Other aspects pertaining to the specific conditions around Antarctica that affect the long-term global water-mass distribution and circulation are the ice shelves and their role as a source of freshwater (e.g., Hellmer 2004; Schodlok et al. 2006), and as a shield from cold high-latitude air masses. How the glacial freshwater input might affect the oceanic stratification and thus the wintertime sea ice cover has been stressed (e.g., Marsland and Wolff 2001), and ocean–ice shelf interactions have been studied (e.g., Jenkins and Holland 2002; Holland et al. 2003). Beckmann and Goosse (2003) suggested an ad hoc parameterization of the freshwater flux forcing of ice shelves for global ocean GCMs. Stössel et al. (2007, hereafter SSK) indicated that ice shelf melt can potentially play a critical role in the long-term global deep-ocean properties. They also stressed the importance of capturing the geographic location of the Antarctic coastline and the ice shelves at high resolution.

The above-mentioned uncertainties in the atmospheric forcing and glacial meltwater are readily transferred to the oceanic conditions, which in turn affect the quality of Southern Ocean sea ice simulations. Furthermore, the oceanic boundary condition for sea ice (e.g., in the case of a sea ice–ocean GCM) generally suffers from critical physical processes being poorly resolved in the ocean component. To determine deep- and bottom-water formation in global ocean GCMs, the physical process by which these water masses are getting formed need to be captured. While near-boundary convection in association with coastal polynyas bears the uncertainties discussed above, ocean GCMs tend to form bottom water mainly through open-ocean convection (as discussed, e.g., in Kim and Stössel 2001). Plume convection occurring on horizontal scales of about 500 m (e.g., Marshall and Schott 1999) is still far from being resolved in global ocean GCMs. So the long-term deep-ocean properties and circulation in such models will to a large extent depend on the way convection is parameterized (e.g., Goosse et al. 1999; Stössel et al. 2002). Since the high-latitude Southern Ocean is generally weakly stratified (e.g., Martinson 1990) and features a large body of relatively warm water [Circumpolar Deep Water (CDW)] at intermediate depth (e.g., Gordon 1998), any failure in simulating vertical mixing correctly will result in an erroneous sea ice cover (e.g., Timmermann and Beckmann 2004) and thus in erroneous surface buoyancy fluxes.

In view of all these first-order uncertainties, an attempt has been made in this study to change the forcing of the upper-ocean layer of a global sea ice–ocean GCM by rendering its temperature more consistent with the instantaneous satellite-derived sea ice concentration. The coarse-grid ocean GCM is coupled to a sea ice model of much higher resolution, in particular one that is comparable to the pixel resolution of the satellite passive-microwave Scanning Multichannel Microwave Radiometer (SMMR) and Special Sensor Microwave Imager (SSM/I) data (about 22 km) (SSK). Assuming the ice concentration retrieved from this data reflects the real world, this approach allows for direct evaluation of the sea ice simulation. SSK identified a series of discrepancies between the simulated and the satellite-derived Southern Ocean sea ice concentration, and investigated under which conditions these discrepancies could be minimized. It was found that reasonable long-term deep-ocean properties could not be maintained together with reasonable Southern Ocean sea ice.

The philosophy in this study is to constrain the upper-ocean temperature of the ocean model to achieve more realistic sea ice simulations (by providing for a more realistic ocean boundary condition for the sea ice model), and therefore more realistic high-latitude upper-ocean boundary conditions in this sensitive region of deep-ocean ventilation. As will be seen, this measure does not only improve the sea ice simulation (as we would expect with a more realistic upper-ocean temperature), but also the long-term deep-ocean properties and circulation. The paper is organized as follows: section 2 describes the model, the assimilation procedure, and the experimental setup. Section 3 discusses the outcome as it affects the Southern Ocean sea ice and the global deep-ocean circulation. Section 4 provides a discussion and summary.

2. Model and experiments

a. The model

The model and its setup are identical to that described in SSK. It constitutes a global sea ice–ocean GCM, originally based on Drijfhout et al. (1996) and Wolff et al. (1997), and is also referred to as the Hamburg Ocean Primitive Equation (HOPE) model. It has been used extensively in its original spherical coordinate version (e.g., Latif et al. 2004; Marsland and Wolff 2001; Legutke and Maier-Reimer 1999; Stössel et al. 1998, 2002; Kim and Stössel 2001), and has been converted into curvilinear form to accommodate higher resolution in certain regions of interest [the so-called Max Planck Institute Ocean Model (MPI-OM1) version; Marsland et al. 2003, 2004; Jungclaus et al. 2005]. For technical descriptions of both versions see Wolff et al. (1997) and Marsland et al. (2003), respectively.

Following SSK, the version used in this study contains a Southern Ocean sea ice component that features a resolution 9 times higher (about 22 km) in both principal horizontal directions than the ocean component. Since coastal polynyas are largely determined by katabatic winds and thus the slope of the coastal terrain, the actual geographic location of the coast- or ice shelf line is critical. Therefore, the coastal boundary of the sea ice component does not follow that of the coarse ocean grid, but rather one that has been determined on the fine grid based on the satellite-derived demarcation line between ocean and land or ice shelf (see SSK). This measure requires a fraction of a coarse ocean grid cell to be covered by a rigid lid. This lid has been specified to be 10-m-thick, snow-covered sea ice to effectively suppress any heat exchange between atmosphere and ocean. The lid serves two purposes: it shields the ocean from the cold atmosphere following the geographic location of the high-resolution grid (rather than the coarse-resolution grid of the ocean), and it allows for a source of (glacial) meltwater. The coastal lids can thus be considered a crude representation of ice shelves, crude in the sense that potential meltwater will be released into the upper model layer, and that the rigid lid area does not match the actual ice shelf area. The effect of such (“glacial”) ice melt has been investigated in SSK, and will be further examined in this study. For further details on the model setup, including its forcing, see SSK.

b. The assimilation procedure

The assimilation of the satellite-derived sea ice concentration is done in an unconventional manner by affecting the temperature of the uppermost layer (50 m plus sea surface elevation minus ice draft) of the ocean GCM, rather than directly affecting the simulated sea ice concentration. This can be viewed as an additional forcing of the upper-ocean temperature where sea ice is observed. The modification reads as follows:
i1520-0442-21-17-4498-e1
and
i1520-0442-21-17-4498-e2
where T1 is the temperature of the uppermost layer of the ocean model, Tf is the ocean’s freezing-point temperature, Nsat is the satellite-derived ice concentration, and subscripts n and o mean new and old, respectively. This simple modification forces the uppermost ocean layer to attain a freezing-point temperature where and when the satellite-derived ice concentration is 100%, with a linear increase of the influence of the simulated upper-ocean temperature with decreasing satellite-derived ice concentration. Where the satellite-derived data show no sea ice and T1 is within 1° of the freezing point, the upper ocean is rigorously specified to be 1° warmer than the freezing point.

Modifications of this type violate the conservation of energy within the ocean model, which is, however, true for any ocean model that is forced by surface heat fluxes. Besides providing more realistic ocean boundary conditions for the sea ice model, this approach also improves the high-latitude upper boundary conditions for the global ocean model, not just in terms of sea surface temperature, but also in terms of the freshwater flux associated with freezing and melting of sea ice.

The upper-ocean temperature modification has only been applied on the fine Southern Ocean sea ice grid (running from 50° to 80°S). Thus, depending on the coarse-grid ice concentration, a large part of the coarse ocean grid cell may not be affected. Following SSK, this is accounted for when the fine-grid temperature is averaged over a coarse grid cell to arrive at the coarse-grid temperature necessary for integrating the (coarse) ocean model forward in time.

The satellite-derived ice concentration primarily used in this study has been retrieved from SSM/I passive-microwave data using the National Aeronautics and Space Administration (NASA) Team 2 (NT2) algorithm (Markus and Cavalieri 2000). In determining the surface buoyancy flux, the atmospheric synoptic-scale variability of the sea ice pack is crucial. To capture this variability as accurately as possible, the ice concentration used to constrain the upper layer of the ocean GCM consists of daily real-time NT2 data of 1 yr (1992). For the model’s integration toward equilibrium, these data are employed with an annual repeat cycle (in concert with the annual repeat cycle of the atmospheric forcing). The features to be analyzed are the overall characteristics of the sea ice pack and their long-term impact on the ocean. In terms of leads arising from synoptic-scale variability, passive-microwave ice concentration only yields an open-water fraction within a 20 km × 20 km pixel, the accuracy of which may to a large extent depend on the employed algorithm (see, e.g., Markus and Cavalieri 2000). The effect of using sea ice concentration that has been retrieved with an alternative algorithm will be investigated.

c. The experimental setup

The reference experiments (RO and RN) are to investigate the effect of the upper-ocean temperature being modified by the satellite-derived ice concentration. Two sensitivity experiments (SD and SM) are conducted to examine the effect of vertical mixing in the Southern Ocean and the potential effect of ice shelf melting, respectively, in the new setting with constrained upper-ocean temperature. Finally, the sensitivity to the particular assimilated ice concentration product is being investigated (SMI). All results shown reflect conditions after which the ocean GCM has reached a state of equilibrium, which requires model integrations on the order of 1000 yr for each experiment.

Table 1 summarizes the experimental setup. RO stands for the old reference case (experiment I in SSK). In this experiment, the upper-ocean temperature is not being modified other than by the atmospheric forcing and the interaction with the simulated sea ice. The vertical eddy diffusion is 10−4 m2 s−1 everywhere (Drijfhout et al. 1996), and there is no ice shelf melting allowed. Experiment RN is the new reference case, and differs from RO only through the modification described in section 2b. The sensitivity experiments start out from RN. In SD, the vertical diffusion in the Southern Ocean south of 50°S has been increased by one order of magnitude (as in experiment III in SSK). In SM, ice shelf melting is allowed south of 70°S at a rate that corresponds to the departure of the underlying ocean’s temperature from the freezing point. In the final experiment (SMI), the ice concentration assimilated into SM is based on the Bootstrap algorithm (Comiso et al. 1997; Comiso 1995) rather than the NT2 algorithm.

3. Results

a. Southern Ocean sea ice

Figure 1 shows the seasonal change of ice area, with the full solid line representing the NT2 product. Without upper-ocean temperature constraint (RO), the ice advance as well as the ice retreat occur clearly too early in the season. There is also no indication of a slower advance than retreat, as reflected in the satellite-derived product. During the season of ice growth and expansion, all other experiments follow NT2 very closely, suggesting that upper-ocean temperature is decisive in controlling the ice advance. Note that this is also the season during which convection plays the largest role.

The upper-ocean temperature also affects the ice retreat (i.e., all constrained experiments yield a better result than RO). However, the retreat is still too early by 1–2 months. This can be largely corrected for (or tuned) by reducing the departure of T1,n from the freezing point in Eq. (2), that is, by rendering the melting of ice by the upper-ocean temperature (where the satellite-derived product indicates no sea ice) less stringent. This measure, however, leads to too fast an ice advance in fall. A much more realistic ice retreat is achieved when the atmospheric temperature is constrained by the satellite-derived sea ice concentration in spring, suggesting that during fall and winter, the Southern Ocean sea ice is mainly determined by the ocean, while in spring and in summer it is mainly controlled by the atmosphere, including heating of the ocean by solar radiation. This is expected, since with surface cooling and brine rejection, the water column is becoming less stable in fall and winter, and thus the sea ice more coupled to the ocean. At the same time, the air column over the surface will become more stable, and the sea ice thus less coupled to the atmosphere. In spring and summer, the reverse is expected to occur, respectively.

Otherwise, Fig. 1 also reveals an earlier retreat in SM (and SMI), which will be discussed below. The choice of the satellite product is relatively insensitive, the slightly smaller ice area in SMI being qualitatively consistent with the differences between the Bootstrap and the NT2 data (Markus and Cavalieri 2000). The effect of a stronger vertical eddy diffusion, which generally leads to a reduced preconditioning for convection, does not have much of an influence on the sea ice cover when the upper-ocean temperature is constrained (SD thus being very similar to RN in that regard), in contrast to the corresponding unconstrained case (see SSK). So while reduced convection diminishes oceanic heat being brought to the surface to affect sea ice, the warmer upper-ocean temperature associated with enhanced convection is not allowed in the constrained case, thus rendering the influence of the weaker convection on sea ice rather negligible.

Figure 2 shows snapshots of ice concentration as simulated in RO (top panel), RN (middle panel), and as derived from passive-microwave data using the NT2 algorithm (bottom panel), for the end of February (left panel) and beginning of September (right panel). The chosen dates are the approximate times of minimum and maximum seasonal sea ice cover. The top right and bottom right panels correspond to the top left and bottom right panels in SSK’s Fig. 1. Validating against the NT2 product (bottom panel), there is a striking improvement from RO to RN, indicating that the upper-ocean temperature plays a decisive role in controlling the quality of the Southern Ocean sea ice simulation. Since the ocean model is formulated on the coarse grid, and the isolines are designed for fine-grid sea ice variables, one can readily identify where the ocean affects the sea ice cover. The winter panel of RO shows distinct open-ocean polynyas that are clearly due to open-ocean convection, as seen from the coarse-grid pattern. The ice edge is also largely determined by the upper-ocean temperature, as is again readily apparent from its coarse-grid pattern. This is expected with a coarse-grid ocean underneath, and as such is not unrealistic within the given resolution of the ocean model. The overall positioning of the ice edge, however, is unrealistic in RO, and this is mainly due to unrealistic upper-ocean temperatures as confirmed in RN, where these temperatures are modified as a function of the satellite-derived ice concentration. The relatively good agreement between RN and NT2 (as opposed to RO versus NT2) suggests that the ocean component does poorly in this region, while the sea ice component delivers reasonable results if provided with reasonable upper-ocean temperatures. Noticeable biases occur only in the northwestern Weddell Sea in summertime and are due to the fact that for the coarse ocean grid, the Antarctic Peninsula does not extend as far north as for the fine sea ice grid shown in the figures (see SSK), and to the zonal air temperature gradient across the Antarctic Peninsula being too weak (e.g., King et al. 2003). This deficiency is particularly apparent in the ice thickness field (see Fig. 5). Furthermore, assuming that NT2 captures the lead fraction (i.e., the open-water part) correctly, in comparison the sea ice model tends to underestimate the lead fraction, which may be due to a missing source of variability. Besides daily wind variability, which is contained in the wind forcing (though perhaps poorly), a (another) missing candidate is the tidal current (e.g., Koentopp et al. 2005; SSK). Also, between 0° and 25°E, the simulated sea ice does not extend as far north as the satellite data suggest, which is probably due to deficiencies in the atmospheric forcing. In any case, considering that only the ocean boundary condition for the sea ice model has been modified, the agreement of RN with the satellite data is remarkable.

Allowing for ice shelf melting in the limits of the current model design leads to a substantial difference in ice extent and concentration in spring, as illustrated in Fig. 3, an early December snapshot of ice concentration. There is much less ice in SM than in RN, the latter clearly delivering more realistic results when verified against NT2. A closer examination reveals that large regions that are ice free in SM are actually occupied by ice of a lower concentration (around 90%) in NT2, while the same regions show very high concentration (>96%) in RN. The reason for the difference between RN and SM is due to the additional meltwater making the water column more stable along most of the Antarctic coast- or ice shelf line (consistent with, e.g., Hellmer 2004). The additional meltwater and the associated reduction of near-boundary convection render the water column along the coastal regions overall warmer and fresher, thus in particular less dense. This, in turn, enhances the westward coastal current, strengthening the adjacent cyclonic gyre circulation (e.g., in the eastern Weddell Sea). The associated enhanced doming of isopycnals leads to a larger upper-layer salinity farther offshore in SM versus RN. The upper-layer temperature, on the other hand, is largely constrained by the satellite-derived ice concentration, and thus is mostly still at or close to the freezing point in spring. This type of preconditioning offshore gives rise to rapid shallow overturning in early December, and thus to a much more rapid decrease in ice extent and thickness in SM than in RN (see also Fig. 1). This suggests that besides making the water column more stable inshore, thereby increasing ice thickness and reducing sea ice formation (see, e.g., Hellmer 2004; Beckmann and Goosse 2003), ice shelf melt also has an indirect impact on sea ice offshore by influencing the strength of the gyre circulation.

Figure 4 shows estimates of daily variability of sea ice in September, which according to the NT2 data, is mostly contained along the ice edge. RO and RN reveal a huge difference. In RO, the variability is dominated by oceanic heat flux due to convection occurring sporadically in space and time, as clearly indicated by the coarse-grid pattern. The satellite-derived variance of ice concentration shows a rather coherent pattern along the ice edge as would be expected from a wavelike motion following the atmospheric synoptic scale as determined by the daily wind field. RN shows a very similar pattern. From this we can conclude that the upper-ocean temperature determines the position of the ice edge, but we cannot conclude that it is the variability of the upper ocean’s temperature that determines the variability of the ice edge, since the former has been constrained by the satellite-derived ice edge, the variability of which is more likely due to the synoptic variability of the wind field.

Figure 5 shows the simulated ice thickness in the Weddell Sea for the approximate times of minimum (left) and maximum (right) seasonal ice extent for SM (top panel) and SMI (bottom panel). The ice pack shows a general tendency toward thicker ice in the western Weddell Sea, in particular in summertime. This simulated pattern is much more realistic than in SSK, where the upper-ocean temperature was not constrained. In most parts, the ice thickness is comparable to what has been observed (e.g., Harms et al. 2001; Strass and Fahrbach 1998; Timmermann et al. 2004). However, within about 10° of the peninsula, the ice thickness is clearly underestimated, while it is overestimated in the central part of the Weddell Gyre, in particular in SM.

While the ice concentration dataset based on the Bootstrap algorithm (Comiso et al. 1997) shows regionally significantly lower ice concentrations than the one based on the NT2 algorithm (Markus and Cavalieri 2000), the impact of this difference is in general relatively small when used to constrain the upper-ocean temperature as described in section 2b. This alternative dataset yields nevertheless a noticeably thinner ice cover, as revealed from Fig. 5. This follows directly from Eq. (1); that is, with more open water the modified upper-ocean temperature should become warmer for the same T1,o, thus generally leading to thinner ice.

b. Global impact

A selection of standard control values of the global circulation is shown in Table 2. The Antarctic sea ice volume is much larger in all constrained simulations than in the unconstrained one (RO). This can also be deduced from the upper panels of Figs. 2 and 4, indicating excessive open-ocean convection in RO that ultimately leads to a thinner mean ice thickness. For the long-term mean Antarctic Circumpolar Current (ACC), the range is relatively small, between 110 (RN) and 95 Sv (SD; 1 Sv ≡ 106 m3 s−1). The latter can be attributed to the fact that the stronger vertical eddy diffusion in the Southern Ocean reduces the preconditioning for convection south of the ACC, and thus reduces the overall convective potential energy release in that region. This, in turn, reduces the meridional density gradient across the ACC.

The strongest formation of Antarctic Bottom Water (AABW) occurs in RN, which is reflected in the coldest and saltiest and thus densest global-mean deep-ocean water masses, and the weakest outflow of North Atlantic Deep Water (NADW). A weak (strong) NADW outflow occurs when the AABW overturning cell in the Atlantic occupies a large (small) part of the water column, as seen in Fig. 6. In RN, the line of zero streamfunction that separates the southward flow of NADW from that of AABW at 30°S lies at a depth of about 2300 m. The associated AABW intrusion is about 5 Sv. This relatively large AABW cell reduces the outflow of NADW, and thus the Atlantic meridional overturning cell. SD has clearly the weakest AABW intrusion, from the bottom to about 3200 m at 30°S occupying a relatively small part of the water column.

The difference between RO and RN, in particular the stronger rate of AABW formation in RN, is solely due to keeping the upper-ocean temperature at a level that corresponds closely to the satellite-derived ice concentration. As a consequence, the simulated sea ice is much more realistic in RN than it is in RO (see Fig. 2). The rate of AABW formation is nevertheless too strong in RN, as implied from the global-mean deep ocean being too cold (by about 0.5°C) and too salty (by about 0.04), and by the outflow of NADW being too weak [by about 6–11 Sv, referring to, e.g., Ganachaud and Wunsch (2000) and Sloyan and Rintoul (2001)]. A larger vertical eddy diffusivity in the Southern Ocean (SD) reduces convection there, clearly leading to warmer and fresher global deep-ocean conditions and a much stronger outflow of NADW. The global deep ocean, however, is in SD biased in the opposite direction of RN (i.e., too warm and far too fresh, consistent with SSK). It should be noted that enhancing vertical diffusion was one way to prevent excessive open-ocean convection and to maintain a continuous sea ice cover without constraining the upper-ocean temperature.

Keeping the vertical diffusivity at its original level and adding ice shelf melt instead (SM) also yields a reduction of the rate of AABW formation versus RN, in fact (serendipitously) to a degree that the global deep-ocean properties and the overturning circulation look rather realistic. The reduction of AABW formation is consistent with the reduction of near-boundary convection as a result of meltwater input along the coast (as described in section 3a for Fig. 3).

Constraining the upper-ocean temperature by the alternative satellite-derived product (SMI) does seemingly not lead to a large difference in the long-term deep-ocean properties and circulation, although a slight deep-ocean warming and freshening is noticeable. With the assimilated lower-than-NT2 Bootstrap ice concentration (in particular along the coast; see Markus and Cavalieri 2000), the upper-ocean temperature is getting less tied to the freezing point, thus generally becoming warmer. This warming contributes to a stabilization of the water column, and thus to less convection and less bottom-water formation. This outcome is to some extent counterintuitive, as one might expect a lower ice concentration to enhance heat loss and ice formation and thus convection and bottom-water formation. This would indeed be the case if the two different satellite-derived products were directly assimilated to modify the simulated ice concentration, as in Stössel and Markus (2004).

In trying to estimate the linkage between southern high-latitude processes and long-term deep-ocean properties and circulation, the water-mass distribution in the Southern Ocean is also of concern. One way to investigate this is to look at meridional sections of temperature and salinity spanning the region from the Antarctic continent all the way across the ACC up to the subtropical gyres. Figures 7 and 8 show such meridional sections for the western Weddell Sea–west Atlantic sector (centered at 40°W) for all experiments, and as compiled by Levitus (1982). A critical component in the water-mass distribution is the southward penetration of the relatively warm and salty CDW up to the Antarctic continental margin (e.g., Whitworth et al. 1998; Gordon 1998). This feature is clearly indicated in the Levitus climatology, and even more pronounced in individual sections (e.g., Orsi and Whitworth 2005, section A23). All experiments indicate some southward and upward migration of CDW, while the absolute CDW properties are in most experiments deviating considerably from the in situ measurement-derived climatology. SD shows far too low salinities, which are also reflected in its global deep-ocean properties (Table 2). Reasonable intermediate- and deep-ocean salinities are reproduced with RN and SM, while the water is too cold along the Antarctic continental margin in RN. SMI captures the Antarctic Intermediate Water (AAIW) tongue somewhat better than SM, but is overall unrealistically fresher than SM. To summarize, SM is an improvement over RN, and RN is a major improvement over RO. RN, and in particular SM, are reasonable given the coarse resolution of the ocean model, and given the crude parameterization of subgrid-scale vertical mixing.

4. Discussion and summary

This study demonstrates how the long-term deep-ocean properties and circulation in a global ocean GCM depend on the upper-ocean boundary conditions in the Southern Ocean sea ice–covered regions. One major finding of this study is that a sophisticated and relatively high-resolution sea ice component itself will not lead to satisfactory results without reasonable upper-ocean boundary conditions. These boundary conditions have been improved by forcing the ocean model’s upper-layer temperature in the Southern Ocean to closely comply with satellite-derived sea ice concentration. The assimilated ice concentration does not affect the simulated ice concentration or thickness directly, thus enabling the sea ice model to respond freely to the given atmospheric and oceanic boundary conditions in each time step. This avoids the problem of constraining ice thickness, which would become inevitable if observed ice concentration were assimilated to directly affect the simulated ice concentration (e.g., Lindsay and Zhang 2005, 2006; Stössel and Markus 2004). One may view this strategy as an alternative way to improve the upper-ocean boundary conditions in ocean GCM studies on long-term global deep-ocean properties in regions where these conditions are affected by sea ice, which itself is highly sensitive to the simulated heat flux due to ocean circulation and (vertical) mixing.

With this upper-ocean temperature constraint, the Southern Ocean sea ice simulations and thus the upper-ocean boundary conditions improve considerably compared to the unconstrained case. As a result, the long-term deep-ocean properties and circulation are much more realistic, now in concert with realistic sea ice. This result indicates that various misrepresented features (e.g., open-ocean polynyas, too weak AABW formation rates) in long-term ocean GCM integrations can be corrected by rendering the upper-ocean temperature in ice-covered regions more realistic. Some of the misrepresentations have to do with poor atmospheric forcing, but most are actually due to errors in the ocean model, that is, its upper-ocean circulation and, in particular, its vertical mixing in a region that is characterized by a relatively weak stratification and by relatively warm water masses at intermediate depths (e.g., Martinson 1990; McPhee et al. 1999). While improved sea ice simulations sound trivial considering that satellite-derived ice concentration against which the simulation is verified has been assimilated, it is noted that only the upper-ocean temperature has been affected by this “assimilation.” The sea ice simulations with the modified ocean boundary condition doing substantially better supports the notion that Southern Ocean sea ice is highly determined by the upper-ocean temperature. The latter, in turn, is strongly affected by the combined heat flux through the ice-free and ice-covered surface, and thus the atmospheric boundary condition. If the water column is rendered unstable, either through surface cooling or brine release, or warm subsurface anomalies, warmer water from deeper layers will affect the upper-ocean temperature, and hence sea ice. While the ocean stratification and advection of heat by ocean currents may thus play a substantial role in determining the Southern Ocean sea ice pack, the local atmospheric boundary conditions (and their direct impact on sea ice) can in turn decisively affect the ocean stratification. By assimilating “observed” sea ice concentration in the manner described here, the observed atmospheric (and oceanic) imprint on sea ice is transferred to the upper ocean of the model. Sea ice being highly dependent on the upper-ocean temperature explains the described improvements.

Sensitivity studies under the constrained conditions revealed that enhanced vertical diffusion in the southern high latitudes lead to worse regional and global water-mass distributions, even though it helped prevent excessive convection in the unconstrained experiments of SSK (where this measure created a realistically compact sea ice cover). Allowing for meltwater from pseudo–ice shelves leads to considerably more realistic deep-ocean properties due to a reduced rate of AABW formation. Furthermore, crucial control values of the global thermohaline circulation, such as the outflow of NADW, became more realistic. As pointed out in the literature (e.g., Hellmer 2004; Marsland and Wolff 2001), the effect of glacial ice melt is mainly to stabilize the water column, thus reducing convection at critical sites around Antarctica. A secondary effect seems to be an acceleration of the gyre circulation around Antarctica, which actually tends to enhance vertical mixing offshore in spring, and thus lead to a more rapid ice retreat than in the case without ice shelf melt. The rapid retreat, however, is an artifact of the upper ocean being kept at lower temperatures than what the ocean model would predict. Finally, employing satellite ice concentration derived from an alternative algorithm (Bootstrap rather than NT2) to constrain the upper-ocean temperature has a relatively small impact on the formation rate of AABW.

It is concluded that the satellite passive-microwave ice concentrations, which are meanwhile available over a continuous time frame of almost 30 years at a resolution of at least 20 km in space and about 1 day in time, can be used as a powerful alternative in forcing the upper ocean in regions affected by sea ice. To some extent, this is a step backward, since the ice–ocean model is now more constrained by observations. However, when it comes to investigating the sensitivity of the long-term deep-ocean properties, as in this study, a realistic representation of the high-latitude upper-ocean conditions is decisive for a reasonably realistic simulation of the long-term deep-ocean properties and circulation. The improved sea ice simulations under the new upper-ocean temperature constraint are essential for providing more realistic upper-ocean boundary conditions (in terms of surface heat and freshwater fluxes). Thus, while the upper-ocean temperature is constrained, the simulated sea ice still affects the surface fluxes as in the unconstrained case. From the sea ice point of view, the only difference in the constrained case is that the ocean boundary condition for the simulation of sea ice is more realistic. From the ocean point of view, the surface boundary conditions have improved where they are determined by sea ice. However, since the intrinsic upper-ocean dynamics are getting distorted in the constrained case, this approach is not suitable for studying atmosphere–ice–ocean interactions.

There is a concern with the quality of the assimilated data, since there are considerable uncertainties associated with the conversion of passive-microwave signals into ice concentration (e.g., Markus and Cavalieri 2000; Comiso 1995; Kaleschke et al. 2001). Nevertheless, these data provide continuous spatial and temporal coverage at relatively high resolution. This can be compared to atmospheric forcing data, in particular near-surface winds and temperature as well as precipitation and glacial meltwater. Measurements of these variables are very sparse in the Southern Ocean, and mostly confined to the summer season and to the Antarctic coastline. Halfway reliable gridpoint datasets are only those from (re)analyses (see section 1). In this region, however, these data mainly constitute output from atmosphere GCMs. They are also not (yet) available at a spatial resolution of 20 km.

When referring to “lead fraction” in conjunction with the effect of wind variability on sea ice concentration, one should keep in mind that the passive-microwave data just reveal an open-water (possibly including thin ice) fraction within a 20 km × 20 km pixel. One should also point out that modifying the upper-ocean temperature by satellite-derived sea ice concentration does not replace or make the atmospheric forcing (and the problems associated with it, as described in section 1) obsolete. In regions affected by sea ice, however, the total surface buoyancy fluxes per model grid cell are mostly determined by the ice concentration, in particular in wintertime [i.e., large cooling or brine release in the ice-free part (the lead fraction), and relatively little of such in the ice-covered part]. Rendering the ice concentration and extent more realistic (e.g., by an assimilation of the sort presented in this study) thus constitutes a substantial step toward improving the surface buoyancy fluxes, and thus the upper-ocean boundary conditions that ultimately drive deep- and bottom-water formation.

Most failures in simulating Southern Ocean sea ice resulted from excessive oceanic heat flux associated with excessive vertical mixing. The upper-ocean layer thickness being 50 m, one could argue that a higher vertical resolution could lead to more realistic results without or with a weaker constraint than introduced in this study. Looking at observed temperature and salinity profiles in the Southern Ocean (e.g., Jacobs and Comiso 1989; McPhee et al. 1999; Gordon and Huber 1990; Martinson and Iannuzzi 1998; Orsi and Whitworth 2005), one finds the upper 50 m typically well mixed, except in summer conditions when the mixed layer can be as shallow as 30 m. Since the latter coincides with ice-free conditions, this discrepancy is not critical. To some extent the constraint is correcting deficiencies in the external forcing, in particular when and where it adversely affects the stratification in the ocean. What is mainly corrected with the constraint, however, is the failure of the ocean model to properly simulate convection. As long as it is not feasible to operate global nonhydrostatic ocean GCMs for studies of long-term changes in deep-ocean properties and circulation with a resolution on the order of 100 m to resolve convective plumes, the critical vertical mixing processes will continue to depend on subgrid-scale parameterizations. Even though progress has been made in recent years (e.g., Goosse et al. 1999; Canuto et al. 2004; Timmermann and Beckmann 2004), the choice of the various empirical parameters involved is still largely a matter of trial and error rather than principal physics. The GCM used for this study employs the plume-convection parameterization of Paluszkiewicz and Romea (1997), which led to a series of improvements over results achieved with the conventional convective-adjustment parameterization. In particular, excessively large open-ocean polynyas in wintertime were eliminated, but replaced by the sporadic occurrence of (coarse) grid-size polynyas throughout the ice growth season. While this change in behavior contributed to improving integral effects, such as the asymmetry of the seasonal cycle of ice extent (Kim and Stössel 2001) and a shift toward brine rejection in coastal polynyas being the dominant contributor to AABW formation (Stössel et al. 2002), the sporadically occurring open-ocean polynyas cannot be verified.

The current approach can be held against the argument that fully coupled GCMs should be employed to guarantee realistic feedbacks between the atmosphere, ocean, and sea ice. While substantial progress has been made with such models (e.g., Randall et al. 2007; Holland and Raphael 2006; Bitz et al. 2005; Ogura et al. 2004; Jungclaus et al. 2005), they are unsuitable for studies of the long-term thermohaline circulation. The reason is essentially due to the failures described above: basically, the high-latitude near-surface conditions in such models are too unrealistic to properly simulate deep- and bottom-water formation rates and consequently the long-term global deep-ocean properties.

Acknowledgments

Thanks are due to Thorsten Markus for the NT2 satellite data, and to Marion Stössel for processing these data for the model and for substantial work on the graphics. I am grateful for the numerous valuable recommendations I received from the two anonymous reviewers (one of whom almost deserves coauthorship), and Lynne Talley, editor of JPO, for her recommendation of having this contribution being considered for this journal (rather than JPO). This research has been sponsored through NASA Grant NNG05GN87G and NSF Grant ATM-0333341.

REFERENCES

  • Beckmann, A., , and H. Goosse, 2003: A parameterization of ice shelf-ocean interaction for climate models. Ocean Modell., 5 , 157170.

  • Bitz, C. M., , M. M. Holland, , E. C. Hunke, , and R. E. Moritz, 2005: Maintenance of the sea ice edge. J. Climate, 18 , 29032921.

  • Bromwich, D. H., , and R. L. Fogt, 2004: Strong trends in the skill of the ERA-40 and NCEP/NCAR reanalyses in the high and middle latitudes of the Southern Hemisphere, 1958–2001. J. Climate, 17 , 46034619.

    • Search Google Scholar
    • Export Citation
  • Canuto, V. M., , A. Howard, , P. Hogan, , Y. Cheng, , M. S. Dubovikov, , and L. M. Montenegro, 2004: Modeling ocean deep convection. Ocean Modell., 7 , 7595.

    • Search Google Scholar
    • Export Citation
  • Comiso, J. C., 1995: SSM/I sea ice concentrations using the bootstrap algorithm. NASA Reference Publication 1380, 49 pp.

  • Comiso, J. C., , D. J. Cavalieri, , C. L. Parkinson, , and P. Gloersen, 1997: Passive microwave algorithms for sea ice concentration—A comparison of two techniques. Remote Sens. Environ., 60 , 357384.

    • Search Google Scholar
    • Export Citation
  • Connolley, W. M., , and S. A. Harangozo, 2001: A comparison of five numerical weather prediction analysis climatologies in southern high latitudes. J. Climate, 14 , 3044.

    • Search Google Scholar
    • Export Citation
  • Drijfhout, S. S., , C. Heinze, , M. Latif, , and E. Maier-Reimer, 1996: Mean circulation and internal variability in an ocean primitive equation model. J. Phys. Oceanogr., 26 , 559580.

    • Search Google Scholar
    • Export Citation
  • Ganachaud, A., , and C. Wunsch, 2000: Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408 , 453456.

    • Search Google Scholar
    • Export Citation
  • Goosse, H., , and T. Fichefet, 1999: Importance of ice-ocean interactions for the global ocean circulation: A model study. J. Geophys. Res., 104 , 2333723355.

    • Search Google Scholar
    • Export Citation
  • Goosse, H., , E. Deleersnijder, , T. Fichefet, , and M. H. England, 1999: Sensitivity of a global coupled ocean-sea ice model to the parameterization of vertical mixing. J. Geophys. Res., 104 , 1368113695.

    • Search Google Scholar
    • Export Citation
  • Gordon, A. L., 1998: Western Weddell Sea thermohaline stratification. Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margin, S. S. Jacobs and R. F. Weiss, Eds., Antarctic Research Series, Vol. 75, Amer. Geophys. Union, 215–240.

    • Search Google Scholar
    • Export Citation
  • Gordon, A. L., , and B. A. Huber, 1990: Southern Ocean winter mixed layer. J. Geophys. Res., 95 , 1165511672.

  • Harms, S., , E. Fahrbach, , and V. H. Strass, 2001: Sea ice transports in the Weddell Sea. J. Geophys. Res., 106 , 90579073.

  • Hellmer, H. H., 2004: Impact of Antarctic ice shelf basal melting on sea ice and deep ocean properties. Geophys. Res. Lett., 31 .L10307, doi:10.1029/2004GL019506.

    • Search Google Scholar
    • Export Citation
  • Holland, D., , S. S. Jacobs, , and A. Jenkins, 2003: Modelling the ocean circulation beneath the Ross ice shelf. Antarct. Sci., 15 , 1323.

    • Search Google Scholar
    • Export Citation
  • Holland, M. M., , and M. N. Raphael, 2006: Twentieth century simulation of the southern hemisphere climate in coupled models. Part II: Sea ice conditions and variability. Climate Dyn., 26 , 229245.

    • Search Google Scholar
    • Export Citation
  • Jacobs, S. S., , and J. C. Comiso, 1989: Sea ice and oceanic processes on the Ross Sea continental shelf. J. Geophys. Res., 94 , 1819518211.

    • Search Google Scholar
    • Export Citation
  • Jenkins, A., , and D. Holland, 2002: A model study of ocean circulation beneath Filchner-Ronne Ice Shelf, Antarctica: Implications for bottom water formation. Geophys. Res. Lett., 29 .1193, doi:10.1029/2001GL014589.

    • Search Google Scholar
    • Export Citation
  • Jungclaus, J. H., , H. Haak, , M. Latif, , and U. Mikolajewicz, 2005: Arctic–North Atlantic interactions and multidecadal variability of the meridional overturning circulation. J. Climate, 18 , 40134031.

    • Search Google Scholar
    • Export Citation
  • Kaleschke, L., , C. Lüpkes, , T. Vihma, , J. Haarpaitner, , A. Bochert, , J. Hartmann, , and G. Heygster, 2001: SSM/I sea ice remote sensing for mesoscale ocean-atmosphere interaction analysis. Can. J. Remote Sens., 27 , 526537.

    • Search Google Scholar
    • Export Citation
  • Kim, S-J., , and A. Stössel, 2001: Impact of plume convection on global thermohaline properties and circulation. J. Phys. Oceanogr., 31 , 656674.

    • Search Google Scholar
    • Export Citation
  • King, J. C., , J. Turner, , G. J. Marshall, , W. M. Connolley, , and T. A. Lachlan-Cope, 2003: Antarctic Peninsula climate variability and its causes as revealed by analysis of instrumental records. Antarctic Research Series, Vol. 79 , Amer. Geophys. Union. 1730.

    • Search Google Scholar
    • Export Citation
  • Koentopp, M., , O. Eisen, , C. Kottmeier, , L. Padman, , and P. Lemke, 2005: Influence of tides on sea ice in the Weddell Sea: Investigations with a high-resolution dynamic-thermodynamic sea ice model. J. Geophys. Res., 110 .C02014, doi:10.1029/2004JC002405.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and Coauthors, 2004: Reconstructing, monitoring, and predicting multidecadal-scale changes in the North Atlantic thermohaline circulation with sea surface temperature. J. Climate, 17 , 16051614.

    • Search Google Scholar
    • Export Citation
  • Legutke, S., , and E. Maier-Reimer, 1999: Climatology of the HOPE-G global ocean general circulation model. German Climate Computer Center Tech. Rep. 21, Hamburg, Germany, 90 pp.

  • Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp.

  • Lindsay, R. W., , and J. Zhang, 2005: The thinning of Arctic sea ice, 1988–2003: Have we passed a tipping point? J. Climate, 18 , 48794894.

    • Search Google Scholar
    • Export Citation
  • Lindsay, R. W., , and J. Zhang, 2006: Assimilation of ice concentration in an ice–ocean model. J. Atmos. Oceanic Technol., 23 , 742749.

    • Search Google Scholar
    • Export Citation
  • Markus, T., , and D. J. Cavalieri, 2000: An enhancement of the NASA Team sea ice algorithm. IEEE Trans. Geosci. Remote Sens., 38 , 13871398.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., , and F. Schott, 1999: Open-ocean convection: Observations, theory,and models. Rev. Geophys., 37 , 164.

  • Marsland, S., , and J-O. Wolff, 2001: On the sensitivity of Southern Ocean sea ice to the surface fresh-water flux: A model study. J. Geophys. Res., 106 , 27232741.

    • Search Google Scholar
    • Export Citation
  • Marsland, S., , H. Haak, , J. H. Jungclaus, , M. Latif, , and F. Roeske, 2003: The Max-Planck-Institute global ocean/sea ice model with orthogonal curvilinear coordinates. Ocean Modell., 5 , 91127.

    • Search Google Scholar
    • Export Citation
  • Marsland, S., , N. L. Bindoff, , G. D. Williams, , and W. F. Budd, 2004: Modeling water mass formation in the Mertz Glacier polynya and Adelie depression, East Antarctica. J. Geophys. Res., 109 .C11003, doi:10.1029/2004JC002441.

    • Search Google Scholar
    • Export Citation
  • Martinson, D. G., 1990: Evolution of the Southern Ocean winter mixed layer and sea ice: Open ocean deepwater formation and ventilation. J. Geophys. Res., 95 , 1164111654.

    • Search Google Scholar
    • Export Citation
  • Martinson, D. G., , and R. A. Iannuzzi, 1998: Antarctic ocean-ice interaction: Implications from ocean bulk property distributions in the Weddell Gyre. Antarctic Sea Ice: Physical Processes, Interactions and Variability, M. O. Jeffries, Ed., Antarctic Research Series, Vol. 74, Amer. Geophys. Union, 243–271.

    • Search Google Scholar
    • Export Citation
  • Massom, R. A., and Coauthors, 2001: Snow on Antarctic sea ice. Rev. Geophys., 39 , 413445.

  • McPhee, M. G., , C. Kottmeier, , and J. H. Morison, 1999: Ocean heat flux in the central Weddell Sea during winter. J. Phys. Oceanogr., 29 , 11661179.

    • Search Google Scholar
    • Export Citation
  • Ogura, T., , A. Abe-Ouchi, , and H. Hasumi, 2004: Effects of sea ice dynamics on the Antarctic sea ice distribution in a coupled ocean atmosphere model. J. Geophys. Res., 109 .C04025, doi:10.1029/2003JC002022.

    • Search Google Scholar
    • Export Citation
  • Orsi, A. H., , and T. Whitworth III, 2005: Southern Ocean. Vol. 1, Hydrographic Atlas of the World Ocean Circulation Experiment (WOCE), International WOCE Project Office, 223 pp.

    • Search Google Scholar
    • Export Citation
  • Paluszkiewicz, T., , and R. D. Romea, 1997: A one-dimensional plume model for the parameterisation of oceanic deep convection. Dyn. Atmos. Oceans, 26 , 95130.

    • Search Google Scholar
    • Export Citation
  • Powell, D. C., , T. Markus, , and A. Stössel, 2005: Effects of snow depth forcing on Southern Ocean sea ice simulations. J. Geophys. Res., 110 .C06001, doi:10.1029/2003JC002212.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 996 pp.

    • Search Google Scholar
    • Export Citation
  • Renfrew, I. A., , J. C. King, , and T. Markus, 2002: Coastal polynyas in the southern Weddell Sea: Variability of the surface energy budget. J. Geophys. Res., 107 .3063, doi:10.1029/2000JC000720.

    • Search Google Scholar
    • Export Citation
  • Schodlok, M. P., , H. H. Hellmer, , G. Rohardt, , and E. Fahrbach, 2006: Weddell Sea iceberg drift: Five years of observations. J. Geophys. Res., 111 .C06018, doi:10.1029/2004JC002661.

    • Search Google Scholar
    • Export Citation
  • Sloyan, B. M., , and S. R. Rintoul, 2001: The Southern Ocean limb of the global deep overturning circulation. J. Phys. Oceanogr., 31 , 143173.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , and T. Markus, 2004: Using satellite-derived ice concentration to represent Antarctic coastal polynyas in ocean climate models. J. Geophys. Res., 109 .C02014, doi:10.1029/2003JC001779.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , S-J. Kim, , and S. S. Drijfhout, 1998: The impact of Southern Ocean sea ice in a global ocean model. J. Phys. Oceanogr., 28 , 19992018.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , K. Yang, , and S-J. Kim, 2002: On the role of sea ice and convection in a global ocean model. J. Phys. Oceanogr., 32 , 11941208.

    • Search Google Scholar
    • Export Citation
  • Stössel, A., , M. M. Stössel, , and J-T. Kim, 2007: High-resolution sea ice in long-term global ocean GCM integrations. Ocean Modell., 16 , 206223.

    • Search Google Scholar
    • Export Citation
  • Strass, V. H., , and E. Fahrbach, 1998: Temporal and regional variation of sea ice draft and coverage in the Weddell Sea obtained from upward looking sonars. Antarctic Sea Ice: Physical Processes, Interactions and Variability, M. O. Jeffries, Ed., Antarctic Research Series, Vol. 74, Amer. Geophys. Union, 123–139.

    • Search Google Scholar
    • Export Citation
  • Timmermann, R., , and A. Beckmann, 2004: Parameterization of vertical mixing in the Weddell Sea. Ocean Modell., 6 , 83100.

  • Timmermann, R., , A. Worby, , H. Goosse, , and T. Fichefet, 2004: Utilizing the ASPeCt sea ice thickness data set to evaluate a global coupled sea ice–ocean model. J. Geophys. Res., 109 .C07017, doi:10.1029/2003JC002242.

    • Search Google Scholar
    • Export Citation
  • Timmermann, R., , H. Goosse, , G. Madec, , T. Fichefet, , C. Ethe, , and V. Duliere, 2005: On the representation of high latitude processes in the ORCA-LIM global coupled sea-ice-ocean model. Ocean Modell., 8 , 175201.

    • Search Google Scholar
    • Export Citation
  • Vihma, T., , J. Uotila, , B. Cheng, , and J. Launiainen, 2002: Surface heat budget over the Weddell Sea: Buoy results and model comparisons. J. Geophys. Res., 107 .3013, doi:10.1029/2000JC000372.

    • Search Google Scholar
    • Export Citation
  • Whitworth, T., , A. H. Orsi, , S-J. Kim, , W. D. Nowlin, , and R. A. Locarnini, 1998: Water masses and mixing near the Antarctic slope front. Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margin, S. S. Jacobs and R. F. Weiss, Eds., Antarctic Research Series, Vol. 75, Amer. Geophys. Union, 1–27.

    • Search Google Scholar
    • Export Citation
  • Wolff, J-O., , E. Maier-Reimer, , and S. Legutke, 1997: The Hamburg Ocean Primitive Equation model HOPE. German Climate Computer Center Tech. Rep. 31, Hamburg, Germany, 98 pp.

Fig. 1.
Fig. 1.

Seasonal cycles of ice area (1012 m2) as simulated with the indicated experiments (see Table 1) and as derived from passive-microwave data using NT2.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 2.
Fig. 2.

(left) February and (right) September snapshots of ice concentration as simulated with experiment (top) RO, (middle) RN, and (bottom) as derived from satellite passive-microwave (year 1992) data using the NT2 algorithm. The isoline interval between 90% and 100% is 2%, whereas that between 20% and 90% is 10%.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 3.
Fig. 3.

Early December snapshots of ice concentration as simulated with experiments (top) RN, (middle) SM, and (bottom) as derived from passive-microwave data using NT2. Otherwise as in Fig. 2.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 4.
Fig. 4.

September variance of ice concentration, as simulated with experiment (top) RO, (middle) RN, and (bottom) as derived from passive-microwave data using NT2. Here, 1 = 100% ice concentration, contour interval is 0.02, minimum shown is 0.02, and maximum (red) is 0.24.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 5.
Fig. 5.

(left) February and (right) September snapshots of ice thickness in the Weddell Sea as simulated with experiment (top) SM and (bottom) SMI. The isoline interval is 0.4 m.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 6.
Fig. 6.

Atlantic meridional overturning circulation as simulated with experiment (top left) RO, (top right) RN, (bottom left) SD, and (bottom right) SM. The isoline interval is 3 Sv.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 7.
Fig. 7.

Meridional sections along 40°W (averaged between 50° and 30°W) of potential temperature as simulated with experiments (top left) RO, (top right) RN, (middle left) SD, (middle right) SM, (bottom left) SMI, and (bottom right) as compiled from measurements by Levitus (1982). The isoline interval is 0.4°C, while the color scale interval is 1.2°C.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Fig. 8.
Fig. 8.

Meridional sections of salinity. The isoline interval is 0.04 g kg−1, while the color-scale interval is 0.12 g kg−1. Otherwise as in Fig. 7.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2256.1

Table 1.

The model experiments. See main text for details.

Table 1.
Table 2.

Annual equilibrium values of selected integrated or averaged quantities from the individual experiments. NADW: NADW outflow across 30°S (Sv); ACC: Drake Passage throughflow (Sv); AASI: Antarctic sea ice volume (1012 m3); θ4: global-mean potential temperature at 4000-m depth (°C); and S4: global-mean salinity at 4000-m depth (g kg−1).

Table 2.
Save