May 2001 Report for the Project

Study of the Atlantic thermohaline circulation in a highly efficient global ocean-atmosphere model

Principal Investigators: S. Rahmstorf and A. Ganopolski
Postdoctoral Researcher: M. A. Morales Maqueda
Potsdam Institute for Climate Impact Research, PO Box 601203, 14412 Potsdam

1 Introduction: Objectives and outline of the research
As expounded in the project proposal and in our June 2000 report, the thrust of this piece of research is the clarification of the role played by the oceanic overturning circulation in shaping the global climate and its variability on multi-year to millennial time scales. Our workhorse for this study is the coupled atmosphere-ocean-sea ice model CLIMBER 3, whose preliminary alpha version is already operational.

The specific objectives of the project are as follows.

  1. To complete the development phase of the CLIMBER 3. tex2html_wrap_inline106 model.

  2. While the atmospheric component of the model (POTSDAM; Petoukhov et al., 2000, Ganopolski et al., 2000a) has been in service for a number of years, the ocean (MOM3) and sea-ice (ISIS) modules have undergone significant and indispensable modifications in the course of the last year and a half (see June 2000 report and below). In addition, an efficient and physically-based coupler has been constructed as a necessary interface between the model components. The coupled model has now reached its fine-tuning stage.

  3. To study the large-scale modes of circulation of the modern ocean with the aim of helping elucidating the stability of the overturning circulation, the mechanisms whereby it is sustained, and its impact on today's global climate.
    Previous modelling work on this subject has been carried out with either computationally inexpensive, but extremely simplistic, representations of the atmospheric dynamics and thermodynamics or with highly elaborate, and hence very slow, atmosphere models (cf. references in Rahmstorf and Ganopolski, 1999). In contrast, the atmosphere module of CLIMBER 3. tex2html_wrap_inline108 provides an excellent compromise between physical realism and computational efficiency. It has succeeded in simulating most large-scale mean climatic fields with an accuracy similar to that of Atmospheric General Circulation Models (AGCMs), while its computing requirement is 100 to 1000 times lower. Seasonal-equilibrium integrations with the coupled model and with the ocen-sea ice component alone are about to be started. Once the experiments are completed, the modelled present-day ocean circulation, temperature and salinity distribution, and sea-ice fields will be compared against existing data sets both for validation/tuning purposes and in order to further tune the model components.

  4. To analyse the different equilibrium states of the system and their stability.
    The sensitivity of the meridional overturning to the boundary conditions at the air-ocean interface has been the subject of various studies (Bryan, 1986; Döscher et al., 1994; Rahmstorf and Willebrand, 1995). Simulations with stand-alone ocean models using diverse parameterisations of the surface boundary conditions for freshwater, heat, and wind stress are limited in that they can but poorly mimic, if at all, the interactions between atmosphere, ocean, and sea ice, notably in regard to the reaction of the atmosphere to large changes in ocean circulation. This shortcoming is not present in coupled climate models, which have, at least in principle, the potential to account for all relevant feedbacks, and whose only boundary conditions are the physical ones applied at the top of the atmosphere, namely heat input from the Sun and heat release from the Earth.

  5. To investigate the "natural" variability of the model and its relation to the observed climate variability.
  6. The model CLIMBER 2, forerunner of CLIMBER 3. tex2html_wrap_inline110 , has displayed a high level of skill in simulating the long-term variability of the climate system on orbital time scales (Ganopolski et al., 2000b). It has also been used to investigate possible mechanisms for abrupt changes in climate, such as Dansgaard-Oeschger and Heinrich events (Ganopolski and Rahmstorf, 2001), showing that relatively small variations of the thermohaline forcing in the North Atlantic might be sufficient to significantly alter the state of the Atlantic overturning circulation. Less attention has been paid so far to the investigation of shorter-term variability (multi-year, decadal, and centennial) with this model. In our experiments with CLIMBER 3. tex2html_wrap_inline112 we shall focus on this latter variability window. Ocean and coupled ocean-sea ice-atmosphere modes will be examined (e.g., North Atlantic Oscillation, Great Ice and Salinity Anomaly). The possibility of exciting variability modes through the introduction of stochastic terms in the atmosphere model will also be considered.

2 State of advancement of the project
A description of the configuration of the ocean-sea ice model MOM3/ISIS can be found in the past June 2000 report for this project (Rahmstorf et al., 2000). At the time, uncoupled integrations with three different resolutions (5 tex2html_wrap_inline114tex2html_wrap_inline116 , 4 tex2html_wrap_inline118tex2html_wrap_inline120 , and 2 tex2html_wrap_inline122tex2html_wrap_inline124 ) and using the surface forcing of Fichefet and Morales Maqueda (1997) and Goose and Fichefet (1999) were being carried out. The outcome of these experiments revealed a number of shortcomings in the MOM3 ocean model used. Specifically, the ocean volume turned out not to be appropriately conserved, noisy and unstable patterns were prominent in the model's free surface and vertically integrated currents, and the global overturning circulation was poorly represented: deep water formation in the North Atlantic was weak and very shallow, while excessive deep and bottom waters were being created in the Southern Ocean. In addition, the sea-ice field around Antarctica was unable to reach a thermodynamic equilibrium, as not enough oceanic sensible heat flux was provided at the base of the ice to counter the strong cooling at the air-ice interface. These results prompted us to develop a number of improvements to the ocean component.

2.1 Further development of MOM3
The Geophysical Fluid Dynamics Laboratory (GFDL) released MOM3 into the public domain last year. However, the model was still in the prototype phase and aspects of its numerics and physics presented a number of serious problems that either were the direct cause or contributed to the difficulties encountered in our simulations. Consequently, parts of the MOM3 algorithmics had to be revised. Here we shall give only a summary of the most important modifications we have introduced.

  1. Nonlinear free-surface scheme. The nonlinear free-surface formulation of MOM contains a conspicuous checkerboard null mode that is inherently associated to the B-grid on which the model equations are cast. Since variations in the free surface control the changes in volume of the top grid cell of the model, this mode can impact on other fields, thus distorting crucial processes such as the local ocean circulation and convective activity, or the formation of sea ice. To put a remedy to this problem, we have tested several filtering techniques, such as Fourier, Laplacian, and Shapiro filters, all of which smooth out the checkerboard mode, but have a tendency to introduce noise in the surface vertical component of the velocity. We have found, however, that an upstream correction to the continuity equation for the free surface elevation is the most efficient and least disruptive procedure for filtering the mode. We are in the process of documenting this procedure, which we have adopted, and expect that it will be introduced in future MOM releases.

  2. Another, and more serious problem arising from the use of a nonlinear free-surface scheme relates to the lack of tracer-conservation that comes about when, in order to save computing time, tracer and current equations are integrated with unequal time steps. Long-term integrations with MOM are very resource consuming. A technique to circumvent this difficulty is the stretching of tracer time steps relative to ocean-dynamics time steps, which accelerates the integration by up to a factor of 10. The price to pay is a distortion of the ocean dynamics and transport processes. But integrations incorporating seasonal variability have shown that many important properties of the world ocean can be reasonably represented with this approach. However, in a nonlinear free surface scheme the effects of the undulating surface height are incorporated not only into the ocean dynamics but also into the tracer equations, which now explicitly take into account the presence of fresh water forcing. Thus, the model is fully conservative only if tracers and surface height are integrated synchronously, which would lead to an prohibitively expensive integration if tracers had to be advected-diffused with time steps appropriate for the free surface (seconds to minutes). We have developed an algorithm that guarantees volume and tracer conservation but still allows for stretched tracer time steps using a nonlinear free surface. The corner-stone of the technique is the splitting of the barotropic current and free surface time steps in such a way that barotropic currents run synchronously with the rest of the ocean dynamics while the free surface evolves with the tracer time step. The algorithm has been implemented in MOM3. As ilustrated in Fig. 1, results from two seasonal-cycle simulations, one employing synchronous dynamics and tracers, the other using the new technique, reveal that most ocean large-scale features, including depth and strength of the overturning, and tracer distributions, can be realistically simulated with the method (Griesel and Morales Maqueda, 2001).

Figure: Upper panel: amplitude of the seasonal cycle (winter-minus-summer mean) of the barotropic transport (in millions of m tex2html_wrap_inline126 /s) from a seasonal-equilibrium integration of MOM with equal time steps for dynamics and tracers. Lower panel: difference in the amplitude of the seasonal cycle of the barotropic transport between a seasonal-equilibrium integration of MOM with time steps of 1 hour and 1 day for dynamics and tracers, respectively, and the simulation depicted in the upper panel. Note that, excluding the North Pacific area, differences between the two simulations are well below 10%.
K Profile Parameterisation (KPP) for vertical-mixing. The KPP parameterisation of Large et al. (1994) was implemented in MOM3, but contained programing errors that led to the creation of negative vertical viscosities and diffusivities. These errors have now been removed. The parameterisation leads to a considerable improvement of the oceanic sensible heat flux at the base of the ice, notably at the sea-ice front in Southern Ocean.
Bottom-Boundary Layer (BBL) and near-bottom mixing schemes. The BBL scheme of MOM has been complemented with the improvements recommended by Nakano and Suginohara (2000). Additionally, the method of enhancement of the vertical diffusivity over rough bathymetry of Hasumi and Suginohara (1999) has been tested and adopted in our MOM version.
Isopycnal Gent-McWilliams stirring. We have detected a weak near-surface instability in the model's overturning field that is created by the use of the skew-formulation of the Gent-McWilliams isoneutral mixing scheme (Griffies, 1998). We are in the process of solving this problem in collaboration with the GFDL.

As announced in our previous report, and in conjunction with the model development outlined above, we have installed in our computers a number of oceanographic data sets (Levitus (1982), Da Silva et al. (1994), and WOCE/AIMS) and visualisation software (Ocean Data View, ODV; see that will be used in the analysis of our model simulations of present-day climatic conditions. For this phase of the project, we shall put our emphasis in the simulations with the uncoupled ocean-sea ice model in resolution 2 tex2html_wrap_inline128tex2html_wrap_inline130 and the coupled atmosphere-ocean-sea ice in resolution 5 tex2html_wrap_inline132 5bsp;tex2html_wrap_inline134 .

2.2 Further development of ISIS
All the modifications to the thermodynamic-dynamic sea-ice model ISIS mentioned in Rahmstorf et al. (2000) have been incorporated in the model. Within the frame of our collaboration with T. Fichefet and H. Goosse (UCL), one of the authors (MAMM) has introduced in the model the sea-ice dynamics parameterisation of Zhang and Hibler (1979), which speeds up model integrations by circa a factor 10. As part of the testing procedure of ISIS, we have embarked in an investigation of the possible role that glacial sea-ice expansion in the Southern Ocean might have played in the lowering of atmospheric CO tex2html_wrap_inline136 during the Last Glacial Maximum (LGM). This work is certainly not without links with the overall subject of this project, and it can be summarised as follows (for details, see Morales Maqueda and Rahmstorf, 2001).

Recently, Stephens and Keeling (2000) have put forward an appealing theory for explaining the decrease in glacial atmospheric CO tex2html_wrap_inline138 inferred from ice-core records. They argue that a compact sea-ice cover extending southward of tex2html_wrap_inline140 55 tex2html_wrap_inline142 S would have trapped large amounts of CO tex2html_wrap_inline144 beneath the sea surface, thus accounting for the lower atmospheric concentrations. An atmosphere-ocean box model in which sea-ice area is prescribed allows them to simulate tex2html_wrap_inline146 80% of the CO tex2html_wrap_inline148 drawdown. However, glacial CO tex2html_wrap_inline150 levels can be attained in their model only when the fraction of ice-covered area southward of the Antarctic Polar Front (APF) rises to 99-100%. To assess whether the existence of such a compact ice cover is feasible in a LGM environment, we have carried out several integrations of the annual cycle of the sea-ice cover in the Southern Hemisphere under boundary conditions representative of today's and LGM's climates. To this end, we have employed ISIS coupled to the mixed-layer model described in Fichefet and Morales Maqueda (1997). Our best guess LGM scenario indicates that the ice-area fraction southward the APF was probably well below tex2html_wrap_inline152 92% at the time (Fig. 2), the persistence of relatively large amounts of open water (leads) within the ice pack stemming from the alliance of three dynamical effects that tend to counter lead closure: shearing deformation and ice divergence, which promote lead opening, and wind-induced variations in the collection thickness of new ice, which decelerates lead closure. These simulations show as well that the Southern Ocean mixed layer might have been up to tex2html_wrap_inline154 90 m deeper during the LGM. If this was the case, the stronger oceanic vertical mixing would have presumably supplied additional deep-ocean CO tex2html_wrap_inline156 to the surface, thus partly compensating for the increase in ice-area fractions. By inserting our simulated ice-coverage values into the ice area-versus-CO tex2html_wrap_inline158 concentration curve of Stephens and Keeling (2000), we estimate that entrapment of CO tex2html_wrap_inline160 by sea ice could account for not more than 15-50% of the glacial CO tex2html_wrap_inline162 decrease. Although it is implausible that Antarctic sea-ice expansion alone can fully explain the atmospheric CO tex2html_wrap_inline164 depletion during the LGM, our results do not discard however sea ice as a potentially important control.


Figure: Wintertime sea-ice concentrations in the Southern Ocean from (a) SMMR observations, (b) a present day climate simulation, and (c) an LGM scenario where air temperatures were 10 tex2html_wrap_inline166 C colder than today and winds 50% stronger. Contour intervals are 0.15 (ice front), 0.90, and 0.95. The pink line delineates the mean path of the APF.

2.3 Coupled atmosphere-ocean-sea ice interface
The Coupled Atmosphere-Ocean-Sea ice interface (CAOS) is an important piece of CLIMBER 3. tex2html_wrap_inline168 , as it controls the exchange of both physical and numerical information between the model components. The central principle that has guided the design of this interface is that the coupling must guarantee strict conservation of energy, momentum, and substances in the system, provided, of course, that the individual components of the coupled model are themselves conservative. This goal has been achieved by adhering to the following three directives:

  1. The atmosphere, ocean and sea ice models must provide only state variables (e.g. air temperature, humidity, and winds, sea-surface temperature, salinities, and currents, and sea-ice surface temperature, thickness, and drift) to the coupler.
  2. The coupler must compute all fluxes at the boundaries between system components by using pertinent empirical or physically-based parameterisations, and these fluxes are the only information that the coupler is allowed transfer back to the models, after spatial and temporal interpolation if need there be.
  3. In no case must the coupler transform the values of the models' state variables.
Additionally, CAOS has the task to check the compatibility of atmosphere, ocean and sea-ice configurations before an integration is started. This checking procedure ensures, for example, that model grids are congruent, that all components use the same values of key geophysical and empirical constants (e.g. Earth radius, acceleration of gravity, Coriolis parameter, Stefan-Boltzman radiative constant), that any two model components share the same number of fluxes, and that these fluxes have the adequate units.

The coupler has for the most part been completed by now, and its performance is being assessed in a 2000-year integration on the coarse 5 tex2html_wrap_inline170tex2html_wrap_inline172 -resolution grid.

The tool employed to launch and control the integration of the coupler and the coupled model is the OASIS (Ocean-Atmosphere-Sea Ice-Soil) software. OASIS is an efficient and easy-to-use set of Fortran 77, Fortran 90 and C routines whose main tasks are the synchronisation of the models being coupled. Together with CLIMBER 3. tex2html_wrap_inline174 , OASIS has been installed in the IBM-SP parallel supercomputer recently arrived in PIK. We are in the process of adapting the coupled model so that it runs in parallel in this new machine with the aid of OASIS and using the message passing interface MPI1. With this aim in mind, we have started a collaboration between our group, the Data & Computation Department at PIK, and S. Frickenhaus from the Computer Centre at the Alfred Wegener Institut (Bremerhaven).

3 Situation of the Project in comparison with the initial objectives and time schedule
According the plan outlined in the project proposal (Rahmstorf and Ganopolski, 1999), we should at this stage have completed the test and model spinup phases of the project and started the analysis of the present-day climate simulations both with CLIMBER 3. tex2html_wrap_inline176 and with the uncoupled MOM3/ISIS submodel in the 5 tex2html_wrap_inline178tex2html_wrap_inline180 , 4 tex2html_wrap_inline182tex2html_wrap_inline184 , and 2 tex2html_wrap_inline186tex2html_wrap_inline188 resolutions. Unfortunately, due mainly to the numerous difficulties we have been facing with MOM3, these tasks have not yet been finalised. We are optimistic, however, that the integrations will be completed by the end of Summer 2001, excluding perhaps the fine, 2 tex2html_wrap_inline190tex2html_wrap_inline192 -resolution one, which will require further preparation before it is started, notably regarding the ocean's bathymetry, for which we shall use a compilation of sills of the global ocean prepared by S. R. Thompson and B. de Cuevas (see

Although we have run into a delay of slightly over half a year with respect to the original plan, we feel that the high level of expertise and experience gained so far during the process of tackling the problems posed by the different components of CLIMBER 3. tex2html_wrap_inline194 will largely pay in the last phase of the project, as familiarity with the numerics and the parameterisations of the codes will be instrumental to understanding and analysing model results. PIK has recently equipped itself with a new computer which is significantly faster than the one we had available till now.

We finally note that the work carried out so far has generated two scientific publications. The first one deals with the technical problem of the treatment of the nonlinear free surface in an ocean general circulation model (Griesel and Morales Maqueda, 2001). It was presented at this year's General Assembly of the European Geophysical Society, and the manuscript is in the process in being completed. The second one, which has already been submitted, reports on our research regarding potential links between Antarctic sea ice expansion and atmospheric CO tex2html_wrap_inline196 variations (Morales Maqueda and Rahmstorf, 2001).


Bryan, K., 1986. High-latitude salinity effects and interhemispheric thermohaline circulations. Nature 323, 301-304.
Da Silva, A. M., C. C. Young, and S. Levitus, 1994. Atlas of surface marine data 1994. Vol. 1: Algorithms and Procedures. US Dept. of Commerce. NOAA Atlas NESIDS 6, Washington D.C., USA, 83 pp.
Döscher, R., C. W. Böning, and P. Herrmann, 1994. Response of circulation and heat transport in the North Atlantic to changes in thermohaline forcing in northern latitudes: a model study. J. Phys. Oceanogr. 24, 2306-2320.
Fichefet, T., and M. A. Morales Maqueda, 1997. Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. J. Geophys. Res. 102, 12,609-12,646.
Hasumi, H., and N. Suginohara, 1999. Effects of locally enhanced diffusivity over rough bathymetry on the world ocean circulation. J. Geophys. Res. 104, 23,367-23,374.
Ganopolski, A., and S. Rahmstorf, 2001. Simulation of rapid glacial climate changes in a coupled climate model. Nature, 409, 153-158.
Ganopolski, A., V. Petoukhov, S. Rahmstorf, V Brockin, M. Claussen, and C. Kubatzki, 2000a. CLIMBER-2: A climate system model of intermediate complexity. Part II: Model sensitivity. Clim. Dyn., in press
Ganopolski, A., S. Rahmstorf, V. Petoukhov, and M. Claussen, 2000b. Simulation of modern and glacial climates with a coupled global model of intermediate complexity. Nature, 391, 350-356.
Goosse, H., and T. Fichefet, 1999. Importance of ice-ocean interactions for the global ocean circulation: A model study. J. Geophys. Res. 104, 23,337-23,355.
Griesel, A., and M. A. Morales Maqueda, 2001. A volume and tracer conserving method for Asynchronous integrations of dynamics and tracers in an ocean model using a nonlinear free surface. In preparation. To be submitted to Ocean Modelling.
Levitus, S., 1982. Climatological Atlas of the World Ocean. NOAA Prof. Pap. 13, US Gov. Printing Office, Washington D.C.
Large, W. C., J. C. McWilliams, and S. C. Doney, 1994. Ocean vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys. 32, 363-403.
Morales Maqueda, M. A., and S. Rahmstorf, 2001. Did Antarctic sea-ice expansion cause glaciol CO tex2html_wrap_inline198 decay? Submitted to Geophys. Res. Lett.
Nakano, H., and N. Suginohara, 2000. Reproduction of overflow/downslope-flow in a world ocean model by incorporating bottom boundary layer. Submitted to Journal of Physical Oceanography.
Petoukhov, V., A. Ganopolski, M. Claussen, A. Eliseev, C. Kubatzki, and S. Rahmstorf, 2000. CLIMBER-2: A climate system model of intermediate complexity. Part I: Model description and performance for present climate. Clim. Dyn., 16, 1-17.
Rahmstorf, S., and A. Ganopolski, 1999. CLIVAR Antrag. Untersuchung der thermohalinen Atlantikzirkulation in einem hocheffizienten globalen Ozean-Atmosphären-Modell.
Rahmstorf, S., J. Willebrand, 1995. The role of temperature feedback in stabilizing the thermohaline circulation. J. Phys. Oceanogr. 25, 787-805.
Rahmstorf, S., A. Ganopolski, and M. A. Morales Maqueda, 2000. June 2000 Report for the Project "Study of the Atlantic thermohaline circulation in a highly efficient global ocean-atmosphere model".
Stephens, B. B., and R. F. Keeling, 2000. The influence of Antarctic sea ice on glacial-interglacial CO tex2html_wrap_inline200 variations, Nature, 404, 171-174.
Zhang, J., and W. D. Hibler III, 1997. On an efficient numerical method for modeling sea ice dynamics. J. Geophys. Res. 102, 8691-8702.

Dr. Miguel Angel Morales Maqueda

Thu May 17 11:50:17 NFT 2001