[an error occurred while processing this directive] [an error occurred while processing this directive]
Report of the CSM Ocean Model Working Group
Peter Gent and Michael Spall, Co-Chairs
23 June 1998
The Village at Breckenridge
The CSM Ocean Model Working Group met on 23 June 1998 at the Third Annual CSM Workshop in Breckenridge, Colorado. The discussions covered a wide range of subjects and themes. Most of the presentations reported on progress in improving or adding physics packages to the ocean model. The status and testing of several related but distinct climate models were also presented. Two applications of existing models to particular regions or science issues were summarized. There was a lengthy discussion on priorities and options for the next generation of the CSM ocean model. This range of subjects marks a distinct difference from the annual summer meeting of 1997, where the focus was on deficiencies in the ocean model climatology. Recommendations in 1997 targeted physical parameterizations that were needed to address these shortcomings. The focus of the 1998 meeting was on future directions and the improvements resulting from implementations recommended at the 1997 meeting. The working group felt that the physics options that are currently being implemented are of high priority and that these efforts should continue.
The status of improvements and additions to the model physics were discussed by five speakers. The parameterization of bottom boundary layer physics and entrainment near oceanic overflows was identified in the 1997 meeting as an area of high priority, and work along this line has been pursued over the past year. Matthew Hecht (NCAR) discussed the implementation of the Beckmann/Doescher bottom boundary layer parameterization into the x3' NCAR Ocean Model (NCOM). The results show a significant improvement in the temperature, salinity, and density of the North Atlantic Deep Waters compared to the standard mixing parameterization but indicate that there are still deficiencies in the water mass characteristics downstream of the overflows. This is at least in part due to problems in the source water properties in the Greenland-Iceland-Norwegian (GIN) Seas and near Antarctica. Song (NASA/Jet Propulsion Lab, JPL) and Chao (NASA/JPL) presented results from several idealized overflow calculations in which a new bottom boundary layer scheme was tested and compared with previous approaches and with a high-resolution reference calculation. General problems relating to the parameterization of bottom boundary layers include: conservation issues, truncation errors, and lack of an exact solution with which to evaluate parameterization schemes.
An upper ocean version of NCOM to be used for the study of variability on seasonal-to-interannual timescales is being developed by Jim McWilliams (NCAR) and Gokhan Danabasoglu (NCAR). The primary advantages of this approach are: rapid equilibration, numerical efficiency, option of terrain following coordinates with reduced truncation errors related to steep and tall topography, and increased resolution of the bottom boundary layer near sill overflows. The primary disadvantage is that the role of the abyssal ocean for variability on these timescales is unknown. The basic code is developed and testing is underway.
Bill Large presented a review of the fresh water boundary conditions in NCOM. The surface freshwater flux is represented as a virtual salt flux. There is no net mass flux through the surface as there is for precipitation or evaporation into the real ocean. In the model, the salinity of the uppermost model grid cell is changed by an amount that is equivalent to the change that would be experienced had freshwater been introduced through the surface. A freshwater flux boundary condition will be implemented in future versions of NCOM in which the present rigid-lid approximation is replaced with a free surface. Freshwater run-off from land masses is now included in the model. This addition was recommended at the CSM Ocean Model Working Group meeting in 1997. The approach first divides the continental land masses into an arbitrary number of drainage basins. The net precipitation over each of these regions is then introduced into the nearby ocean grid points through a virtual salt flux. The approach is general, so that more sophisticated land run-off models can be readily incorporated. The approaches used to treat sea ice formation and melt and to parameterize exchanges with unresolved marginal seas were also reviewed.
Esther Brady (NCAR) discussed the use of two sub-models to parameterize the mass and tracer fluxes between the North Atlantic and the Mediterranean Sea. A source model uses the atmospheric forcing and the ambient waters of the North Atlantic to calculate the mass flux and tracer characteristics of the source waters at the sill of the Strait of Gibraltar. An entrainment model is then used to determine the water mass characteristics of the overflow waters after they are modified by strong mixing with the ambient North Atlantic waters just downstream of the sill. The general circulation model (GCM) is forced at the location of the Strait of Gibraltar with the baroclinic mass and tracer flux produced by the entrainment model. The models have been implemented and testing is underway. Preliminary results with the x3 model produce a salt tongue that is too shallow and too salty. Future calculations will include a double diffusive parameterization in the GCM, which is expected to redistribute the salt in the vertical.
Applications of two non-eddy resolving global ocean models to address specific science issues were discussed. Bert Semtner (Naval Postgraduate School) presented results from a 2/3 degree global model based on the Parallel Ocean Program (POP) code with a coupled viscous-plastic sea ice model. The focus was on the mean temperature, salinity, circulation, and convection in the Ross and Weddell Seas. General agreement with observations was found, both in terms of the water mass properties and the general circulation characteristics. Analogies were drawn with the deep convection and water mass formation sites in the North Atlantic.
The upper ocean heat budget and sea surface temperature in the subtropical North Atlantic simulated with the x3 NCOM was compared to mooring data at five locations for the period June, 1991 to June, 1993 by Michael Spall (Woods Hole Oceanographic Institution). Close agreement was found between the model and observations. The mean difference over all five mooring locations for the two-year period was 0.2 degrees C with a standard deviation of 0.5 degrees C. The mixed layer depth also compared well with a mean difference of -4 m and a standard deviation of 15 m. The shortwave flux used to force NCOM is approximately 10% lower than estimates based on measurements made at the moorings. Additional calibration of the surface flux data and evaluation of the upper ocean heat budget are underway.
The large-scale circulation of Mediterranean Water and Labrador Sea Water (LSW) is being evaluated by Michael Spall. There appear to be inconsistencies in the advective pathways of LSW in the x2 NCOM when compared to observational estimates. This is potentially important for the low-frequency propagation of climate anomalies generated in the subpolar North Atlantic. The dynamics of the mid-depth circulation, inter-gyre exchange, and the possible importance of double diffusion are being investigated using simple analytic models and NCOM.
Progress with several related global modeling efforts was summarized. Phil Jones (Los Alamos National Laboratory) discussed the spin-up of a POP-based code with land run-off and a coupled sea ice model. The focus of the discussion was on a warming trend between 200 m and 400 m depth. The cause of the warming is unknown but may be related to a surface flux imbalance and downward mixing. Todd Ringler (Colorado State University, CSU) presented results from several spin-up runs of the POP code with the European Centre for Medium-range Weather Forecasts (ECMWF) forcing and various subgridscale mixing parameterizations. The ocean model will eventually be coupled to the CSU atmospheric GCM. The most apparent discrepancy in the ocean spin-up cases is a warming trend at 1000 m to 2000 m depth. The warming at mid-depths found in these two calculations is potentially important as plans (discussed further below) are being made to transition NCOM to a POP-based code.
Phil Duffy (Lawrence Livermore National Laboratory) discussed results from a global ocean model (Geophysical Fluid Dynamics Laboratory's (GFDL) Modular Ocean Model version 1 (MOM1) plus physical and numerical upgrades) coupled to the Oberhuber ice model, with attention on the tracer distributions in the Southern Ocean. The control run with standard vertical mixing in regions of ice formation (brine rejection) results in chlorofluorocarbons (CFCs) being mixed too deep in the Southern Ocean. A modified scheme that mixes salt to moderate depths, but does not mix temperature or momentum, gives CFC distributions that compare much closer with the World Ocean Circulation Experiment (WOCE) data. The distribution of anthropogenic carbon dioxide in the model compares reasonably well with observations in the Indian Ocean, although the amplitude is 10 to 20% too high.
Wei Chang discussed results from runs using the Miami Isopycnal Ocean Model and NCAR's Community Climate Model version 3 (CCM3) in coupled and uncoupled modes. The fully coupled calculations use no flux corrections. Initial evaluation has focussed on comparisons of hydrography with the Levitus climatology and evaluation of the surface fluxes. Meridional heat flux and meridional overturning circulation in the ocean are also being diagnosed.
Recent changes in the computing environment at NCAR and the need for improved numerics make this a good time to consider transitioning NCOM to a new base code. Frank Bryan (NCAR) led a discussion of the advantages and disadvantages of several approaches. Efficiency on a wide variety of computing platforms was viewed as a high priority. This will facilitate use of the code over a wider user community and hopefully extend its lifetime into future generations of computing platforms. The general needs of the user community are also of concern (i.e., ease of use, physics options, portability, efficiency). The new code must be able to reproduce the current level of physics and numerics in NCOM and preferably improve on other aspects. The effort required for the transition was weighed against the potential benefits.
The four primary options are: continue with the current NCOM, transition to a code based on the GFDL MOM3, transition to a POP-based code, or develop a new code from scratch. It was felt that the POP code is more portable to a variety of computing platforms than either the NCOM or MOM3 codes. The ability to rotate the model grid in POP would help to address several problems associated with convergent meridians at the North Pole in NCOM 1.4. Finally, efforts are already underway to transition physical parameterizations developed for NCOM to the POP code. The working group recommends that the transition of NCOM to a POP-based algorithm begin immediately, and that an opportunity for the user community to provide feedback be provided at a mid-winter meeting of the CSM Ocean Model Working Group.
A transition work plan is to be developed that will include: implementation of NCOM 1.4 physics and diagnostics packages; implementation of the next generation of physical parameterizations (including a bottom boundary layer model, the Mediterranean outflow parameterization, and the upper ocean model); and partitioning of tasks between NCAR and other interested groups. The short-term goal for the code transition is to produce a benchmark calculation with the POP-based code for comparison with NCOM 1.4. It was viewed as essential that the new code not compromise the results of the most recent NCOM. The goal is to complete these benchmark comparisons before a mid-winter CSM Ocean Model Working Group meeting to obtain feedback from the user community.