[an error occurred while processing this directive] [an error occurred while processing this directive] Report of the CCSM Polar Climate Working Group



Report of the CCSM Polar Climate Working Group Meeting


Fifth Annual Community Climate System Model Workshop

Co-Chairs: Dick Moritz and Elizabeth Hunke

The Village at Breckenridge, 28 June 2000



The CCSM Polar Climate Working Group (PCWG) met on Wednesday, 28 June 2000, in Breckenridge, Colorado, as part of the Fifth Annual CCSM Workshop. A list of attendees is at the end of this report. The working group meeting was preceded by presentations by D. Moritz and C. Bitz in the CCSM plenary session on Tuesday. These presentations brought the community up to date on the status of the CCSM sea ice model and studies by members of the PCWG using the coupled CCSM ice-ocean model and other CCSM components.



I. Studies of Sea Ice Related to CCSM


A. The Current CCSM Sea Ice Model (CSIM)


Previous PCWG meetings outlined goals for upgrading the CSIM, which now largely have been achieved. As a result of efforts by many individuals, the CSIM currently includes [workforce noted in brackets]:



Regarding other model issues discussed in the September, 1999 and January, 2000 PCWG meetings:


Regarding sensitivity studies proposed in the September, 2000 meeting:



B. The "Active-Ice-Only" Framework


A primary objective established at the September, 1999 PCWG ice modeling meeting was to develop an "active-ice-only" framework for testing the ice model parameterizations and sensitivities. Bruce Briegleb developed this framework at NCAR, which consists of the CSIM, an ocean slab mixed layer, "dummy" models for the atmosphere and land, and a coupler similar to that used in the full CCSM. Development of the framework also included compilation of data sets suitable for forcing the ice model. Briegleb has collected such data and adjusted them to better match reliable observations in the polar regions. The active-ice-only framework has been used for two studies that were reported on during the June, 2000 PCWG meeting, one concerned with the ice model (Briegleb) and the other concerned with ice/upper ocean feedbacks (Holland).


Bruce Briegleb compared the results from the current CSIM (on the x3' "Greenland" displaced pole grid) with a previous CSM1 ice model run (on a rotated spherical grid). In the first comparison, parameters in the new CSIM were selected to mimic the physics of the old CSM1. For example, the ITD resolved only one category of ice, and the MPDATA advection scheme was run in upwind mode. Surface forcing data sets were derived from the CSM1 run. This comparison provided an opportunity to compare simulations using essentially the same physics, but with one integration based on new code, a new computer (the IBM Blackforest), and a new grid (POP). Although the new model produced increased ice area in both hemispheres, Briegleb concluded from the comparison that the new model code was running properly on the IBM and on the POP grid. The new model physics were then turned back on, and numerical integrations performed to test them, forced by both CCM/CSM outputs and NCEP/ISCCP data sets. Initial results show that the summer ice area coverage is low compared to observations in the Arctic, and the range of the annual cycle in total ice volume is larger than observed. In the Antarctic, summer minimum ice volume compares well with estimates based on satellite and isolated thickness data, while the winter maximum is high. Meridional transport of sea ice around Antarctica showed large differences between the different model runs.


C. Linear Remapping ITD Scheme


During Wednesday's PCWG meeting, Bill Lipscomb presented preliminary comparisons between the remapping and delta-function ITD schemes using his column model. Results indicate that the remapping scheme better resolves the thin end of the thickness distribution and that transitions between categories are smoother throughout the seasonal cycle. In the delta-function scheme, all of the ice must be transferred when the category thickness reaches a boundary through thermodynamic growth or melt. In the linear remapping scheme, the distribution of ice between two category boundaries is assumed to be linear (instead of a delta function), and only a portion of the ice is transferred from one category to another due to thermodynamic growth or melt, resulting in a much smoother distribution with all categories participating. This scheme may also relieve the B-grid checkerboard instability that appears with the delta-function ITD (the reason that the current CSIM is most frequently integrated using the upwind advection option instead of second-order MPDATA). We envision that both schemes will be available as part of CSIM, although discussions among PCWG members should decide which is recommended to the SSC as part of the standard model.


II. Studies of the Polar Atmosphere Using CCSM


A. Uma Bhatt presented initial results from a study of natural variability in the CSM1 coupled model. In particular, the CSM1 results exhibit an observed dipole in sea ice extent between the Sea of Okhotsk and the Bering Sea, evident in the second EOF. Bhatt proposes to use the model output for forcing the atmospheric model alone to investigate the dipole's contribution to Northern Hemisphere atmospheric variability.


B. In their presentations during Tuesday's plenary session, both D. Moritz and C. Bitz indicated the importance of improving the simulation of the polar atmosphere by CCSM and CCM. Output from new CCM model runs with improved parameterizations is needed for analysis by the PCWG to aid in diagnosing and solving the problems observed in CSM1 simulations. Bitz demonstrated conclusively that the poor ice distribution simulated by CSM1 is related to modeled surface pressure/winds, rather than a problem internal to the sea ice model--an identical simulation using NCEP reanalysis winds yields an ice distribution more in line with observations, with thicker ice near Greenland and Canada instead of on the Siberian side of the Arctic basin. Moritz noted that in addition to poor polar surface pressure simulations, cloud and radiation biases present in the CCSM atmosphere model may be contributing to problems with coupling in the polar regions. Moritz, Bitz, and Boville have just begun a project to examine how horizontal resolution of atmospheric dynamics, orography, and surface heating affect the polar atmospheric circulation simulated by CCM.


III. Coupled Model Simulations



A. During the ocean working group meeting Wednesday morning, G. Danabasoglu reported on coupled ice-ocean simulations using the most recent versions of the CCSM sea ice and ocean models that he and B. Briegleb have been running. Danabasoglu focused his talk on bugs that cause imbalances in the ocean tracer budgets, but he also noted that the sea surface temperature is highly sensitive to the method used to determine the amount of new ice growth in the ocean. This problem of balancing the surface heat budget apparently originates with the advection scheme used in the ocean model. In the PCWG meeting Wednesday afternoon, Briegleb summed up the sea ice results: simulated Arctic ice is too thin, while the Antarctic simulation is "pretty good."


B. Tony Craig reported that the NCAR PCM group is running the current CCSM model in fully coupled mode (ocean-ice-atmosphere) on the POP <2/3 degree> displaced pole grid, but using a different coupler than that used by the CSM group. They are forging a path through bugs and coupling issues contributing to an unexpectedly strong ice-atmosphere feedback and will be able to provide helpful insight when the rest of us begin to attack the fully coupled problem.


C. Marika Holland reported on studies of the sensitivity of CSIM variability to the ice-ocean feedback mechanism associated with the effect of solar radiation absorbed in leads on the heat flux from the ocean to the ice. Four integrations were performed, two with 5-category ITD and two with 1-category ITD. One of each pair of runs was performed with the ocean/ice heat flux feedback turned off, i.e., the ocean-to-ice heat flux was prescribed, and did not respond to variations in the solar flux absorbed in the leads. The experiments were forced with atmospheric data from the years 1948-1998. The 1-category ITD produces lower ice concentration in summer than the 5-cateogry run. More than 80% of the simulated ice concentration anomaly variance is due to the feedback mechanism, as well as 60% of the anomaly variance in mean ice thickness. From these results it was suggested that this feedback may be too strong in the new model.


IV. Discussion Notes


J. Curry asked whether lateral melt was playing a major role in feedbacks in the AIO simulations. B. Briegleb indicated it was not the major contributor to the melt.


P. Gent noted that differences in the sea ice simulations performed in the AIO mode and the coupled ocean/ice mode were larger in the Antarctic than in the Arctic. He suggests PCWG look at the sensitivity of the model results to uncertainties in the forcing functions used, and compare that to the sensitivity of results to changing model physics and parameters.


W. Washington asked whether the new albedo parameterization in CSIM makes a large difference. B. Briegleb replied that there are differences, but they are modest.


E. Hunke suggested diagnostic analysis of the simulated fraction of snow ice in the Antarctic sea ice zone, because this ice type is significant there.


J. Curry and E. Hunke suggested lists of items to pursue in further CSIM development including: spectral radiative transfer through ice; snow distribution and morphology, especially during the transition seasons; keel ablation; rafting versus ridging processes; snow ice formation; pancake and frazil ice; drag coefficients; and roughness length effects.


C. Bitz expressed some concern that we are testing CSIM in the AIO and coupled ice/ocean mode but not coupled to the atmosphere, except the early work by the PCM group.

V. Future Plans


1. Assemble improved forcing and verification data sets for use in the AIO experimental framework. A candidate is the most recent re-analyses done by ECMWF.


2. Assess the effects of horizontal resolution on the CCM-simulated atmospheric circulation over the Arctic.


3. Perform sensitivity studies in the AIO framework to compare the effects of changing model physics and parameters to the effects of uncertainties in the forcing functions.


4. Six-month CSIM outlook: The PCWG expects to deliver to the SSC, and release to the community, a well-documented, well-tested CSIM by the end of calendar 2000. This model will include a plastic rheology with an elliptical yield curve; a multiple-category ice thickness distribution; enhanced thermodynamics including multiple-level internal ice temperatures in each ice thickness category; and enhanced surface albedo parameterization, running on the POP grid for compatibility with the CCSM ocean model, compatible with the CCSM coupler* and running on the parallel IBM machines at NCAR.


*Either the current coupler or a new coupler if the latter can be developed, debugged, tested, and provided to the PCWG by September, 2000.


The CSIM is basically complete, as defined by PCWG model goals decided in previous meetings. Besides further model testing to work out remaining bugs and coupling issues, the primary job left to do before releasing the code to the public involves cleaning up the code (for example, removing and combining redundant routines), and updating the code documentation. Members of the PCWG have agreed to make the code clean-up a priority [Hunke, Bitz, Craig]. Documentation will proceed from the existing draft progress report (by Briegleb and Schramm, Moritz will assist here) and the existing on-line documentation of CICE (Hunke). The clean-up crew will also keep their eyes open for possible efficiency improvements.


Many ideas for improving CSIM parameterizations were floated during the PCWG meeting on Wednesday (see section 4, following). While most of these will probably not make it into CSIM in the next six months, at least two are possible:


a) Linear remapping ITD scheme of Lipscomb (2000). This scheme has been implemented in CICE at LANL, and comparisons between it and the delta-function scheme currently standard in CSIM are being performed.


b) improved algorithm for spectral radiative transfer through the ice. "This should be easy," as they say, and J. Curry believes it will make a considerable difference.


VI. Longer-term Outlook:

Other possible improvements to CSIM are being explored by various members of the PCWG at many different institutions. The following list is in no particular order. [J. Curry spearheaded this discussion.]



VI. List of Attendees