Accomplishments and Plans of the Oceanography Section,

Climate and Global Dynamics Division, NCAR, 1998-2003

The mission of the Oceanography Section (OS) is understanding the large-scale ocean circulation and the dynamics of climate through studies of the important processes in the ocean and sea-ice, in air-sea-ice interactions, and in coupled systems. The section also upgrades and maintains the ocean, sea-ice and ocean biogeochemistry components of the CCSM, through participation in the CCSM Ocean, Polar Climate and Biogeochemistry Working Groups.

1) Studies Using the CSM-1 Ocean Component and Coupled System.

Significant progress was made in the areas of equatorial ocean current strength, and North Atlantic gyre structure. The ingredient necessary to achieve realistic equatorial currents was found to be a lateral eddy viscosity of order 1000 m^2/s acting on the meridional shear of zonal momentum. To achieve this at resolutions as coarse as three degrees, the horizontal viscosity was reformulated to be anisotropic. Thus, the along stream viscosity can be large enough to control numerical noise, and spatially variable, so that large viscosity near western boundaries can be used to resolve boundary currents. This viscosity is a major reason for improved spatial and temporal patterns of ENSO-like variability in the tropical Pacific, see Large et al (2001). Also, in the two degree version of the CSM ocean component, this anisotropic viscosity improved the southward penetration of the subpolar gyre off the east coast of the United States, and the Gulf Stream separation.

A major effort was expended to construct the best possible ocean forcing over the forty years 1958–1997. It is based on near surface winds, temperature and humidity from the six-hourly NCEP/NCAR reanalysis, monthly satellite radiation (1983–1991) and precipitation (1979–1997). Using these fields, a forty year hindcast of ocean variability was performed, which reproduces many aspects of the observed seasonal cycle and interannual variability. It is clearly superior to some highly smoothed analyses of historical hydrographic data, such as the Levitus et al World Ocean Atlas of 1994, in terms of representing the recent interannual variability of the real ocean.

A series of ocean tracer and carbon cycle calculations have been completed using the CSM global ocean model. The tracer work provides valuable information on the ventilation rates of the ocean model on timescales from several years out to centuries. Simulations of deep water chlorofluorocarbons demonstrated that the formation of Antarctic Bottom Water in the case using the usual forcing was weak. Significant improvements were found by modifying the surface salinity/freshwater forcing in the Ross and Weddell Seas; see Doney and Hecht (2001). The carbon simulations are used to study the physical and biological factors governing the ocean inorganic carbon system, as well as links with the atmospheric and terrestrial biosphere. Doney and Lindsay have submitted results from the tracer and biogeochemical runs to the international Ocean Carbon Model Intercomparison Project. This project has a standard set of tracer simulations, and a number of new tracer simulations have been generated in the CSM ocean component over the last two years: natural equilibrium radiocarbon and abiotic carbon; equilibrium biotic carbon; anthropogenic perturbation radiocarbon and anthropogenic carbon; and chlorofluorocarbons.

Gent et al (2001) asks the question "What sets the mean transport through Drake Passage"? The paper is an analysis of twelve experiments using the CSM ocean component alone, coupled to a sea-ice model, and in fully coupled CSM mode. These experiments have a very wide range of strengths of the Antarctic Circumpolar Current and transport through Drake Passage. The conclusion is that the transport is set mostly by the zonal wind stress, or meridional Ekman transport, and by the strength of the thermohaline circulation off the Antarctic shelf. This conclusion disagrees with previous hypotheses for what governs Drake Passage transport. It is shown that the transport is definitely not set by the curl of the wind stress at the latitude of Cape Horn, or by the square root of the average zonal wind stress. Both these previous theories totally ignore any effects from the thermohaline circulation.

The North Atlantic Ocean thermohaline circulation was studied in three experiments using the fully coupled CSM. They are a control integration for 1870 conditions, and particular emission scenarios for the 20th and 21st centuries. Gent (2001) shows that the strength of the thermohaline circulation does not change significantly over the 21st century. This result contrasts with several recent studies done at other climate centers, that have projected a significant reduction over the 21st century. The reason for the difference is that the Northwest Atlantic becomes warmer and more saline in the CSM. These changes combine to make little change to the surface ocean density in this region, and hence to the rate of deep water formation.

Danabasoglu and McWilliams (2000) proposed and assessed principles for the design of an upper-ocean model, suitable for studies of large-scale oceanic variability over periods of a few months to many years. Its essential simplification compared to a conventional full-depth model is the specification of an abyssal climatology for material properties. The upper-ocean model for the general circulation is constructed based on the CSM ocean component, and its solutions are compared to those of an equilibrium run of the full-depth model. The two model solutions agree well in both the mean state and short-term climate fluctuations. Therefore, the upper-ocean model is an efficient tool for studies of coupled climate dynamics, sensitivity to model parameters and forcing fields, and for hypothesis testing about the role of the abyssal ocean.

2) Development of the Ocean and Sea-Ice Components of CCSM-2.

The OS continues to maintain and upgrade the ocean component of the CCSM. This work is done in very close cooperation with the ocean modeling group at the Los Alamos National Laboratory. Porting all the parameterizations in the CSM ocean component to the Parallel Ocean Program (POP) code is now completed, and 1 degree and 3 degree versions have been assembled for the CCSM-2. At both resolutions, it was decided to use the Gent and McWilliams eddy parameterization and the K-profile parameterization scheme for vertical mixing. This followed work which compared these to horizontal tracer eddy mixing and the Pacanowski and Philander vertical mixing scheme in a 1 degree version using POP. Ocean alone runs clearly showed that the new schemes do a much better job in maintaining realistic temperature and salinity profiles in the upper ocean. In fully coupled integrations, the new schemes result in a much reduced climate drift, and much better simulations of the areas and thick nesses of sea-ice, especially in the Arctic. This comparison of the best ocean physics to use at 1 degree resolution was a major factor that enabled the merger of the CSM and the Parallel Climate Model into the CCSM. The ocean component of the Parallel Climate Model had used 1 degree resolution with the older physics parameterizations.

The development of the new CCSM-2 sea-ice model has involved strong collaborations between NCAR, the Los Alamos National Laboratory, and the University of Washington. The primary dynamical improvement in the model is the inclusion of the elastic-viscous-plastic rheology to determine the force due to internal ice stress. This rheology uses an elliptical yield curve and allows the ice to resist both convergence and shear. Improvements have also been made to the thermodynamic parameterizations used in the new sea-ice model. The vertical heat conduction and storage is now solved using the formulation of Bitz and Lipscomb, which is an energy-conserving scheme that accounts for the effect of internal brine-pocket melting on surface ablation. To account for the high spatial variability that is present in the observed ice cover, the sub-gridscale ice thickness distribution of Bitz et al is used, which allows for five ice and one open water category within each model grid cell. An active-ice-only system has been developed by Briegleb for testing the new sea-ice model. This system includes the active ice model coupled to a slab ocean model, driven by atmospheric forcing and run through the CCSM coupled system.

M. Holland has performed sea-ice variability simulations from 1958-1998 using the new CCSM active-ice-only system, forced by the atmospheric fields described earlier. The strength of the feedback mechanisms and the influence of various sea-ice model parameterizations on these feedbacks has been evaluated in this context. It was found that ocean mixed layer feedbacks, particularly those associated with the albedo feedback mechanism, have a strong influence on the sea-ice variability, accounting for up to 60% of the summertime sea-ice concentration and thickness variance in the Central Arctic. Additionally, resolving the ice thickness distribution modifies the feedback mechanism's impact, due to its influence on the sea-ice strength and open water formation.

3) Some Other Section Projects.

Bryan and Hecht have analyzed a series of North Atlantic basin integrations at 0.4, 0.2, and 0.1 degrees horizontal resolution using the POP ocean model; see Smith et al (2000). There is a sharp regime transition between the simulations at 0.1 and 0.2 degrees, with the representation of both the mean flow and variability becoming both qualitatively and quantitatively much more accurate. True numerical convergence of the solutions has yet to be demonstrated, however. An analysis of the dynamics of eddy-mean-flow interaction in the simulated Gulf Stream was initiated, including a direct comparison against corresponding analyses using dense observations obtained during the Synoptic Eddies observational program. The geographical distributions and magnitudes of mean flow and eddy energy are realistic. However, discrepancies are apparent in the spatial distribution of eddy-mean flow energy conversions, and other measures of eddy dynamics. This may be indicative of remaining problems in simulating instability processes in the Gulf Stream, even using horizontal resolutions of order 10 km.

Marine biogeochemical processes as they relate to the global carbon cycle and climate system were studied by Doney. Three paths were pursued: ecosystem modeling and remote sensing; global ocean tracer and biogeochemical modeling; and observational data analysis. A global, mixed-layer marine ecosystem model has been developed, based on extensions of a simple, nitrogen-based ecosystem model for the Sargasso Sea. The mixed-layer model was used as a testbed for evaluating and improving biological parameterizations for such things as iron limitation, zooplankton grazing, nitrogen fixation and calcification. These ecosystem processes are thought to be critical factors in the ocean biogeochemistry and the potential future response of the marine carbon cycle to climate change, see Doney (1999). Satellite ocean color images, a proxy for surface phytoplankton distributions, play an important role in evaluating model ecosystem solutions. Doney and university collaborators have completed a statistical data analysis of the Sea-viewing Wide Field-of-view Sensor satellite ocean color images showing that the magnitude and spatial scales of variability in the biology are closely tied to those of the mesoscale physics.

Holland et al (2001) examines the influence of simulated Arctic sea-ice variability on ice/ocean interactions and the thermohaline circulation. Under stochastic wind forcing of the ice cover, the thermohaline circulation responds with variability that is approximately 10% of the mean. This variability occurs predominantly on interdecadal timescales which are concentrated at approximately twenty years. It is forced by fluctuations in the export of ice from the Arctic into the northern North Atlantic and the subsequent variations in sea-ice melting that occur in this region. The ice melt stochastically forces the surface ocean and appears to excite a damped ocean-only mode of variability. The ice/ocean thermal coupling damps the thermohaline circulation variability, causing a 25% reduction in its standard deviation. A further study which examined how increasing atmospheric CO2 modifies these ice/ocean interactions and variability has been completed.

Recognizing that ocean model solutions are a function of both the model physics and the surface forcing, and that global surface wind vectors are becoming routinely observed from satellites, Milliff and Large have further processed these observations and used the result to force ocean models. This processing transforms the irregularly sampled satellite scatterometer data into regularly gridded wind fields. The observed statistics of wavelet coefficients are used to simulate wind components at the high resolution of 50 km globally every six hours. These wind fields have been produced from August 1996 through July 1997, covering the NCSAT satellite, and since the beginning of the QSCAT satellite data stream in August 1999. Milliff et al (2000) shows that using these winds is essential to producing model equatorial currents that are comparable to observations.

4) Future Work.

The OS will continue to develop and maintain the ocean, sea-ice and ocean biogeochemistry components of the CCSM. New developments for the ocean component will include spatial and temporal distributions of the vertical and isopycnal mixing coefficients, the use of partial bottom cells to obtain a better representation of topography, and the use of a bottom boundary layer scheme to improve the simulation of overflows. For the sea-ice component, this will include parameterizations of the ice/ocean turbulent heat exchange, lead and ridged ice processes, and surface melt ponds and their influence on the surface albedo. The OS will also participate in evaluating the equilibrium climate and various future climate scenario integrations using the fully coupled CCSM-2.

Doney will continue to study the coupled dynamics of ocean physics–biology–chemistry with a particular focus on the natural and anthropogenically perturbed carbon cycle. The tools for this research will include the CCSM-2 ocean component, incorporating recent and to be developed biogeochemical modules, satellite remote sensing, and in-situ data analysis. A significant fraction of the effort will be devoted to model development and evaluation.

M. Holland will continue to examine the role of the polar regions in climate change and variability. The influence of the North Atlantic Oscillation on recent Arctic changes will be assessed and the feedbacks in the Arctic system will be addressed. The influence of sea-ice on interdecadal variability in the system will also be investigated. A hierarchy of models will be used in these studies, from single-column ice/ocean coupled models to the fully coupled global CCSM-2. In addition, the influence of sea-ice on paleoclimates will be examined, because feedbacks associated with sea-ice are likely to be important for the maintenance and variability of perturbed climates.

The OS will continue to gather more observational data, and extend and improve the ocean hindcast, with the goal of improved understanding of mechanisms generating ocean variability. A high priority is to improve the ocean forcing in two ways. First, extend the period of consistent satellite radiative forcing through at least 2000, to match currently available NCEP reanalysis. Second, to improve the buoyancy forcing through open leads in the presence of sea-ice by utilizing satellite measurements of the ice concentration. This high latitude exchange is an important factor in the surface water mass transformation rates of the largest water masses of the world's oceans. The overall strategy is to form partnerships with observational individuals and institutions, thereby making model results available to aid in the interpretation of observations, and observations available to demonstrate where model fidelity suggests that known model physics pertains to the real ocean. Often this information exchange has been either nonexistent or one-way.

Large will design and perform experiments to test hypotheses about the role of the ocean in generating its own variability either locally or through remote ocean pathways, and about direct ocean forcing of atmospheric variability. This work requires the development of the capability to control the frequency of air-sea coupling in fully coupled model integrations. With such a tool, only specified regions need be fully coupled, while others see no forcing in prescribed frequency bands, such as ENSO and the NAO, while still retaining the seasonal and higher frequency coupling required for model stability.

Bryan will extend the series of "eddy-permitting" to "eddy-resolving" North Atlantic simulations described above. Only the 0.1 degree case has a poleward heat transport that agrees with observations to within their estimated uncertainty. But, experience with ocean models in the CCSM has shown that it is possible to realistically simulate poleward ocean heat transport at low resolution with adequate parameterization of the effects of mesoscale eddies. The results of these North Atlantic simulations indicate that eddy effects must still be parameterized in the "eddy-permitting" resolution regime. However, simulations at 0.4 degrees using the standard Gent-McWilliams eddy mixing parameterization have led to unsatisfactory results. While the heat transport increases to near observed levels, many of the desirable features of the high resolution simulation, such as tightness of frontal features and eddy energy levels, are lost. In anticipation that the ocean component of coupled climate models will move into the "eddy-permitting" regime, new efforts in refining the eddy parameterizations to make them effective in this regime will be necessary. This will require additional, basin to global scale simulations at 10 km and higher resolution, to serve as the basis to investigate the dynamics of eddies and their interaction with the large-scale general circulation. In addition, it will be necessary to investigate the convergence properties of the solutions at resolutions finer than 10 km, in order to prescribe confidence levels to the eddy statistics obtained in these "eddy-resolving" integrations.