On the influence of adequate Weddell Sea characteristics in a large-scale global ocean circulation model
Global ocean circulation models usually lack an adequate consideration of high-latitude processes due to a limited model domain or insufficient resolution. Without the processes in key areas of the global thermohaline circulation, the characteristics and flow of deep and bottom waters cannot be modeled realistically. In this study, a high-resolution (~20 km) ocean model focused on the Weddell Sea sector of the Southern Ocean is combined with a low-resolution (2° × 2°) global ocean model applying the state estimation technique. Temperature, salinity, and velocity data on two Weddell Sea sections from the regional model are used as constraints for the large-scale model in addition to satellite altimetry and sea-surface temperatures. The differences between the model with additional constraints and without document that the Weddell Sea circulation exerts significant influence on the course of the Antarctic Circumpolar Current with consequences for Southern Ocean water mass characteristics and the spreading of deep and bottom waters in the South Atlantic. Furthermore, a warming trend in the period 1993–2001 was found in the Weddell Sea and adjacent basins in agreement with float measurements in the upper Southern Ocean. Teleconnections to the North Atlantic are suggested but need further studies to demonstrate their statistical significance.
KeywordsState estimation Adjoint method Weddell Sea Global ocean
Flow and water mass characteristics of the WSDW crossing the ridge are subject to seasonal and interannual variability (Robertson et al. 2002; Schröder et al. 2002) caused by local winds, sea ice forcing acting in the inner Weddell Sea, and/or changes within and at the eastern extension of the Weddell gyre (Schröder and Fahrbach 1999). Water mass property anomalies have been observed in the eastern Scotia Sea (Meredith et al. 2001) and southern Drake Passage (Rubython et al. 2001), the western entrance to the Scotia Sea. Here, changes in deep water characteristics have a more direct impact on the ACC which experiences an increase in volume transport and in density as it passes the Scotia Sea. In particular, LCDW which contributes ~20% to the total ACC volume transport (Sloyan and Rintoul 2001) is cooled and freshened. However, it is still subject to discussion whether cross-isopycnal mixing of LCDW and WSDW is responsible for these changes or whether this occurs by isopycnal flow from the northwestern Weddell toward the ACC (Naveira Garabato et al. 2002a). A possible source for the latter would be the Weddell–Scotia confluence, a well mixed band stretching eastward from the Antarctic Peninsula along the crest of the SSR (Whitworth et al. 1994). The study indicates that environmental conditions in Bransfield Strait and on the adjacent continental shelf have a strong influence on ACC dynamics.
Only in recent years the interaction between the ACC and waters leaving the Weddell Sea became a subject of intensive field work, e.g., as part of the international Deep Ocean VEntilation Through Antarctic Intermediate Layers (DOVETAIL) program (Muench and Hellmer 2002). However, numerical studies still have to be designed, as even regional models are rather coarse, to resolve the topographic details of the SSR with passages only a few kilometers wide. To study the influence of Weddell Sea waters on the dynamics of the ACC and their escape from the Southern Ocean, we apply a global circulation model which has been optimized for state estimation over periods of 10 years and more (Wenzel et al. 2001; Wenzel and Schröter 2002). This model is constrained to observations of sea-surface elevation and temperature between 1993 and 2001. Although its performance is, in general, acceptable, the ACC is too wide and a strong warming trend is observed near Antarctica. To further study this area and to investigate its influence on larger scales, we prescribe averaged monthly means from two sections (top to bottom) of a high-resolution (20 km) model for the Weddell Sea as additional “soft” constraints for a global model using the adjoint method (Wenzel et al. 2001). The model results of the Weddell Sea are not only in agreement with observations (e.g., Naveira Garabato et al. 2002b) in terms of water mass characteristics and gyre transport but provide information about passage transports and their interannual variability (Schodlok et al. 2002). We, therefore, expect that the general circulation of the global model improves mostly in the vicinity of the additional data.
2 Model description
We use the Hamburg Large Scale Geostrophic model (LSG, Maier-Reimer and Mikolajewicz 1991) to perform two assimilation experiments that differ in additional constraints for the Weddell Sea. This model was originally designed for climate studies with time scales of thousands of years (e.g., Maier-Reimer et al. 1993). In combination with its adjoint, however, it has also been used successfully for ocean state estimation (e.g., Wenzel et al. 2001; Wenzel and Schröter 2002; Staneva et al. 2002). The model version used in this paper has 2° × 2° horizontal resolution, 23 vertical layers (varying from 20-m thickness for the top layer to 750-m for the deepest ones), and the implicit formulation in time allowing for a time step of 10 days. The overall bottom topography of the model is derived from the Scripps 1°-topography (Gates and Nelson 1975) but modified/improved in the South Atlantic to better fit to the topography of the high-resolution Weddell Sea model (Schodlok et al. 2002). Compared to the original coding two improvements to the model were implemented. The model’s evolution of the sea-surface height now includes the effects due to steric expansion, and the description of the convective adjustment allows for a 50% entrainment in addition to the exchange of unstable layers. Although global models with resolutions less than 1° × 1° are operated routinely at climate modeling centers with superior computer facilities, this coarse version of a global model allows for fast sensitivity studies to various constraints on prevalent computers.
monthly sea-surface temperatures for the period 1993–2001 (Reynolds et al. 2002)
ten day gridded fields of sea-surface height anomalies as measured by the TOPEX/Poseidon altimetric mission for the period 1993–2001
the SHOM98.2 mean sea-surface height (CLS) referenced to the EIGEN-GRACE01S geoid (GfZ)
temperature and salinity of the Levitus climatology (WOA94), but with large error bars thus serving only as background information.
Since the SDR solution shows essential deficiencies in the Weddell Sea circulation (see below) an additional assimilation experiment (WEDEX) is performed in which a 9-year (1985–1993) average of monthly mean temperatures, salinities, and horizontal velocities is constrained on two sections running across the inner Weddell Sea from Kapp Norvegia to Joinville Island, and along the SSR crest from Joinville to the South Sandwich Islands (Fig. 1, Sects. I and II, respectively). The corresponding data is taken from the high-resolution model for the Weddell Sea BRIOS-1.1 (Schodlok et al. 2002). Although its simulation period (1985–1993) does not coincide with our period of assimilation (1993–2001), the use of the results is justified because of the seasonal cycle being the dominant signal.
2.2 Model technique
We look for a global ocean state that is as close to the observations as possible. To combine the model and the data we use the adjoint method which is a variational optimization method. It is based on the definition of a costfunction J that describes the misfit between the data and corresponding model counterparts in an Eulerian sense. J implicitly depends on a set of control parameters that have to be adjusted to achieve minimum cost. The model’s trajectory is totally determined in space and time by its initial conditions and the surface forcing fields (wind stress, air temperature, and freshwater flux). Therefore, it appears to be natural to use these as control parameters but with one exception: only the initial temperature, salinity, and sea surface height are subject to optimization. The initial velocities are not directly modified by the procedure. Instead, we allow them to adjust to the changed initial density structure during the first year of integration. The comparison between the data and the model, i.e., the evaluation of the costfunction, thus starts with year 2 of the simulation.
To reduce the number of control parameters of optimization, we use a Hilbert empirical orthogonal function (HEOF) decomposition for the surface forcing fields. The HEOFs are a special group of the complex empirical orthogonal functions (CEOF), described in detail by von Storch and Zwiers (1999, Chap.16). The major advantage of such a decomposition for representing the forcing is that only the amplitudes of variable patterns that occur naturally are allowed to be adjusted. No unrealistic bulls-eye like fluxes are possible to correct for local model errors keeping the optimized forcing fields as realistic as possible. The obvious disadvantage is that it may introduce spurious teleconnections. Here, we focus on the influence of the Weddell Sea data on the global solution. Therefore, both experiments, SDR and WEDEX, start from the same first guess for the control parameters. The model’s initial state is derived by a model spin-up with 1,000 years of perpetual 1950-forcing fields from the NCEP reanalysis followed by an integration forward to the actual start time of the experiments, again using the NCEP reanalysis as forcing. The first guess fields of the temporal mean and the trend of the forcing within the period of our interest as well as their HEOF amplitudes are taken as obtained in the basic decomposition of the corresponding forcing fields (see above).
3 Model results
the individual results (temperature and salinity) of the two simulations (SDR and WEDEX) and
the differences between the simulations (WEDEX–SDR).
3.1 Mean temperature and salinity distribution
3.2 Horizontal velocities and transports
In Drake Passage a slow-down of the ACC starts at ~1,100-m depth, extending all the way to the bottom. Since transport through Drake Passage is weakly constrained to a time-mean of 135 Sv, less deep flow is compensated by enhanced surface flow near the coast of South America. A phenomenon, beyond the scope of this paper but noteworthy, is the slow-down of the Labrador Sea gyre. Such interhemispheric connection between the formation sites for North Atlantic Deep Water (NADW) and AABW has already been reported by Brix and Gerdes (2003).
3.3 Heat and salt content
3.4 Analyzed trends for 1993–2001
The heat content of the global ocean increases almost linearly for the 9-year period. Compared to the SDR solution it is, however, strongly reduced when Weddell Sea data is included in the estimation (not shown). Especially in the bottom layer the warming trend becomes negligible while the change in the upper ocean (< 512.5 m) for the period 1993–2001 closely resembles the trend of the sea-surface elevation as measured by altimetry (Wenzel and Schröter 2002).
4 Discussion and conclusions
Improved constraints in the southern model domain related to a better representation of the Weddell Sea hydrography naturally causes a more realistic situation in the adjacent South Atlantic. The major improvements are related to a redirection of the ACC in the southern Scotia Sea due to the westward displacement and strengthening of the cyclonic Weddell gyre. Consequences for the ACC transport, however, cannot be studied by this model because the time-mean transport through Drake Passage is weakly constrained to 135 Sv.
The westward shift of the gyre allows for the transfer of colder (but still warmer than observed) and relatively fresh water masses from the interior Weddell Sea in to the adjacent basins. The opposed distribution in the temporal mean salt content of the whole water column (Fig. 7) to the west of the Antarctic Peninsula indicates that flow in to the southeast Pacific via the slope front/coastal current related to the Antarctic Slope Front (Whitworth et al. 1998) occurs only sporadically. The structure might be related to the Antarctic Dipole manifested at the surface in the sea ice variability in the western Amundsen Sea and central Weddell gyre (Yuan and Martinson 2001). The changes in the layers below are caused by a shift of the area of enhanced deep convection from the southeast Pacific to the Weddell Sea (not shown). Here, convection is driven by more realistic surface fluxes causing a saltier (than in SDR) upper ocean (Fig. 7). Bottom currents tunneling the ACC as they pass Scotia Sea, Georgia and southern Argentine Basins carry Weddell Sea characteristics in to the circumpolar current and further north. Nevertheless, secondary centers of activity in the trends of heat and salt can be identified in Amundsen, Bellingshausen, and eastern Ross Seas.
The cooling and freshening in the eastern Ross Sea is caused by a more meridional flow of the ACC at the position of the eastern limb of the Ross Sea gyre. In the Southeast Pacific south of 60°S temperature and salinity maximum, corresponding to Upper and Lower Circumpolar Deep Water, are located within the depth range 500–1,000 m (http://www.awi-bremerhaven.de/GEO/eWOCE). Therefore, a slight change in flow path can result in differences between both simulations. Since bottom topography is identical, atmospheric forcing only slightly adjusted, and the Ross Sea gyre did not develop, the modification of the ACC flow in this area has to be related to a change in the density structure triggered by the, in general, warmer and saltier Amundsen and Bellingshausen Seas. The only source for such changes in water mass characteristics, however, is the ACC itself, indicating that Weddell Sea characteristics exert a significant influence on the circumpolar course of the current.
A tight connection between the Argentine Basin and the Weddell Sea coincides with earlier observations that the western South Atlantic responds within roughly a decade to disturbances in the Weddell Sea (e.g., Coles et al. 1996) thus serving as a conduit between high and low latitudes. The response at the formation sites of NADW might be an artifact due to model deficiencies (see model description) but seems plausible considering the spreading of AABW in the Atlantic Ocean (Emery and Meincke 1986) and the interplay between the NADW and AABW cells studied with a global coupled ice-ocean model (Brix and Gerdes 2003).
This study shows that the processes in the Weddell Sea significantly influence the hydrography of the Southern Ocean and South Atlantic. A similar result has to be expected for the South Pacific if the characteristics of the Ross Sea are adequately considered. The latter might also cause significant changes in the Indian Ocean as Ross Sea disturbances travel westward with the coastal current (Beckmann and Timmermann 2001). The connection between Weddell and Amundsen/Bellingshausen Seas via the coastal current, as proposed in earlier studies (e.g., Hellmer 2004), cannot be supported by this model certainly due to its insufficient resolution at the northern margin of the Antarctic Peninsula.
The maximum trend (WEDEX) in water column heat content of 8 W m−2 for the 512–2,250 m depth range in the eastern Weddell Sea (Fig. 8—middle) corresponds to a warming of 0.035°C a−1. This is similar to the warming estimated for the upper Southern Ocean (700–1,100 m) by Gille (2002). However, the reduced trend in WEDEX compared to SDR indicates that the warming in the Southern Ocean predicted by global climate simulations (e.g., Manabe et al. 1991) could be less if southern high-latitude processes are considered in more detail.
As a next step, data from the Ross Sea should be used as additional constraints in the South Pacific while the ACC transport through Drake Passage should be allowed to develop unopposed. The latter would allow to study all aspects of the interaction between Weddell gyre and ACC, and both steps possibly have serious consequences for the modeled Southern Ocean and deep world ocean.
Thanks to S. Esselborn for providing the TOPEX/Poseidon altimeter data, to U. Mikolajewicz and two anonymous reviewers for careful reading and a constructive evaluation, and to the editors of Ocean Dynamics for a courageous handling of the manuscript.
- Assmann KM, Timmermann R (2005) Variability of dense water formation in the Ross Sea. Ocean Dyn (in press)Google Scholar
- Brix H, Gerdes R (2003) NADW and AABW: Their interaction and influence on the variability of the global ocean circulation. J Geophys Res 108. DOI 10.1029/2002JC001335Google Scholar
- Bryden HL, Imawaki S (2001) Ocean heat transport. In: Siedler G, Church J, Gould J (eds) Ocean Circulation and Climate, Academic Press. Intern Geophys Ser 77:455–474Google Scholar
- Carmack EC (1977) Water characteristics of the Southern Ocean south of the Polar Front. In: Angel M (ed) A voyage of discovery—George Deacon 70th Anniversary Volume, Pergamon Press, New York, pp 15–41Google Scholar
- CLS: SHOM98.2 mean sea surface http://www.cls.fr/html/oceano/projets/mss/cls_shom_en.html
- Emery WJ, Meincke J (1986) Global water masses: summary and review. Oceanol Acta 9:383–391Google Scholar
- Gates, WL, Nelson AB (1975) A new (revised) tabulation of the Scripps topography on a 1 degree global grid. Part II: Ocean depths. Tech Rep R-1277-1-ARPA, The Rand Corporation, Santa MonicaGoogle Scholar
- GfZ: EIGEN-GRACE01S geoid. http://op.gfz-potsdam.de/grace/index_GRACE.html
- Maier-Reimer E, Mikolajewicz U (1991) The Hamburg Large Scale Geostrophic Ocean General Circulation Model (Cycle 1). Tech Rep No. 2. Deutsches Klimarechenzentrum, HamburgGoogle Scholar
- Olbers D, Wübber C (1991) The role of wind and buoyancy forcing of the Antarctic Circumpolar Current. In: Latif M (ed) Strategies for future climate research, MPI, Hamburg, pp 161–192Google Scholar
- Staneva J, Wenzel M, Schröter J (2002) Oceanic state during 1993-1999 determined by 4D Var data assimilation. WOCE Int Newslett 42:11–13Google Scholar
- Storch H von, Zwiers FW (1999) Statistical Analysis in Climate Reseach. Cambridge University Press, CambridgeGoogle Scholar
- Timmermann R, Beckmann A, Hellmer HH (2002) Simulations of ice-ocean dynamics in the Weddell Sea 1. Model configuration and validation. J Geophys Res 107. DOI10.1029/2000JC000741Google Scholar
- Wenzel M, Schröter J (2002) Assimilation of TOPEX/POSEIDON data in a global ocean model: differences in 1995–1996. Phys Chem Earth 27:1433–1437Google Scholar
- Whitworth III T, Nowlin Jr WD, Locarnini RA, Smith SG (1994) Weddell Sea shelf water in the Bransfield Strait and Weddell-Scotia Confluence. Deep-Sea Res 41:629–641Google Scholar
- Whitworth III T, Orsi AH, Kim SJ, Nowlin Jr WD, Locarnini RA (1998) Water masses and mixing near the Antarctic slope front. In: Jacobs SS, Weiss R (eds) Ocean, ice, and atmosphere: interactions at the Antarctic continental margin. Antarc Res Ser 75:1–27Google Scholar
- Wüst G (1935) Schichtung und Zirkulation des atlantischen Ozeans. Das Bodenwasser und die Stratosphäre. Wiss. Ergeb. dtsch. atlant. Exped. Meteor 1925-27 6:1–288Google Scholar