Report of the CSM Polar Climate Working Group Meeting

18-19 January, 2000, NCAR Mesa Lab

 

 

C. Bitz presented results from a simulation performed with CCM3 running at T42/18 resolution, and forced by an 18-year time series of SST and sea ice extent, prescribed from observations (courtesy of Reynolds). In the central Arctic, the CCM simulated a sea level pressure field (SLP) lacking the climatological mean cyclone seen in summertime observations, and in winter the simulated anticyclone was displaced toward the North Pole from its observed position. The result is that the simulated SLP gradient is in many places almost perpendicular to the observed SLP gradient. Simulated SLP contours extending from the center of the Icelandic low are more zonally oriented than observed, which would tend to force wind driven sea ice towards the Greenland coast.

 

Bitz also reported the results of running a developer's version of the new CSM sea ice model forced by output from the CCM3 T42/18 experiment described above. The map of simulated mean arctic ice thickness has a very large maximum offshore of Siberia, with a secondary maximum centered approximately on the "pole of inaccessibility" north of the Beaufort Sea. This pattern contrasts sharply with observations, which show the maximum mean ice thickness off the Greenland/Ellesmere north coasts. Similar problems are reported with the GFDL model (pers. comm. M. Winton).

 

Bitz also presented climatological time series of SLP from selected arctic sites (Beaufort Sea, Icelandic Low, Aleutian Low) simulated by CCM and by many other models participating in the AMIP1 experiments (this is joint work with Fyfe at CCC). The CCM SLP in the Beaufort is too high in summer, too low in winter. In the Icelandic Low area, CCM SLP is too low year round, and lags the observations by about 1 month. The best of the AMIP1 model SLP time series appeared to be those of the BMRC and UGAMP models. (NOTE: analysis performed after this meeting indicates BMRC and UGAMP also do not capture the distribution of mean winter and summer SLP in the central Arctic).

 

Action Item, re: Arctic SLP

 

PCWG will analyze T85 simulation by the PCM group (when available) to assess resolution effects on arctic SLP; PCWG will design and conduct experiments with CCM with increased (over T42) resolution; PCWg will analyze results from regional (mesoscale) models. Participants - Bettge, Bitz, Briegleb, Lynch, Moritz.

 

M. Holland presented results of simulations performed with the UVic global model forced by NCEP winds, and by CSM winds, for the period 1958-1996. The sea ice model employed EVP dynamics and a multiple category thickness distribution. When forced by the CSM winds, the simulated sea ice displays the problems similar to those presented by C. Bitz above: too thick off the Siberian coast, and too much exported ice going east of Spits Bergen instead of through Fram Strait. These problems will propagate through the simulated climate system through errors in downward short-wave radiation, ice growth rates, fresh water fluxes, sensible and latent heat fluxes, etc.

 

During the discussion, A. Proshutinsky indicated that cyclonic atmospheric flow at the surface is needed to "flush" sea ice from the Arctic Basin, and does not appear to occur in CCM and CSM simulations. In general, the magnitude of atmospheric variability is too low in the 300-year CSM run. B. Briegleb noted that problems with the atmospheric circulation also affect simulations of Antarctic sea ice in similar ways. M. Holland also pointed out that with filtering at the North Pole and the Bering Strait and Canadian Archipelago closed, no one has seriously analyzed the NCOM Arctic ocean simulation. A. Proshutinsky noted the development of a new Arctic Ocean Model Intercomparison Project. The first task will be to organize data sets, targeting for a 50-year data set, including sea level, ice thickness and extent. The goal is to understand how well model reproduce the Arctic Ocean climate. An organizational meeting will be held at the Polar Science Center in summer, 2000.

 

Action Item, re: Sea ice model evaluation

 

PCWG will compare sea ice simulated with the new CSM model forced by (a) NCEP winds and (b) CCM and CSM winds. The analysis will proceed in two steps, first verifying that the model produces good results with good forcing, is stable And understood, and second assessing the likely problems that arise in coupling to the atmosphere. Participants - Briegleb, Bitz, Holland.

 

QUESTION: aren't csm winds = ccm winds?

 

Action Item, re: Data Sets

 

PCWG will establish forcing and verification data sets for simulations with the new CSM ice model, to include the following variables: air temperature, humidity, wind velocity, incident radiation fluxes, ocean currents, heat flux from the ocean, ice velocity, ice thickness, and thickness distribution. It was noted that a new re-analysis will be coming out from ECMWF and that should be looked at as a candidate for forcing data. Participants - Briegleb, Moritz.

 

Progress on the new CSM Sea Ice Model:

 

a. Development of the active-ice-only (AIO) framework is in progress at NCAR. Data sets are needed to help distinguish problems of model performance from problems with the forcing data. A good mixed layer model may be needed in the absence of the full ocean model (Briegleb).

 

b. CICE2.0 is available on the web from LANL, and is being readied for LANL Ocean/Ice spinups (Hunke).

 

c. The current CSM surface albedo scheme over sea ice is not too bad, based on analysis of the 40-year coupled integration. It may be a bit too low in spring, and a bit higher than typically observed in summer. (Briegleb). J. Curry indicates CICE and CSM albedo parameterizations should be compared with other data sets including SHEBA, and with other models. A relevant paper is posted on the GCSS arctic cloud web page.

 

Action Item Evaluate albedo parameterizations. .

 

D. B. Lipscomb presented a new numerical scheme for advecting the ice thickness distribution in thickness space, called "Remapping". g(h) is represented as polynomial functions of h on each interval. 1-D experiments with this scheme show it converges with fewer thickness categories than the delta-function scheme, simulates a different response of mean thickness to a warming scenario, and is not a significant additional computing burden.

 

C. Bitz showed preliminary results from the PCM implementation of the new ice model, run by T. Craig (EVP dynamics, 5-category delta-function g(h), MPDATA advection, 2/3 degree resolution). The model was forced by climatological winds and air temperatures. The model exhibited some large variation in velocity on small spatial scale (1 grid point, checker boarding), which varies with the model time step, and may be due to strength variations associated with the variable thickness distribution. The model does not exhibit the checker boarding when upstream advection is used in place of MPDATA, probably because of numerical diffusion. Bitz, Craig and Hunke will look into this.

 

C. Deser presented results on the CCM3 response to prescribed sea ice anomalies based on observations that ice has been advancing in the Labrador Sea and retreating in the Greenland Sea over the last 40 years. Sea ice variability was prescribed using EOF's. The CCM3 response to variability in the first EOF of sea ice coverage exhibited a pattern similar to the Arctic Oscillation pattern, with amplitude 60m in geopotential height at 500 mb. Analysis indicates that the CCM3 response tends to be "restoring", i.e. the atmospheric response tends to damp out the ice anomaly. A further experiment in which just the SST was perturbed resulted in a much smaller response.

 

Bette Otto-Bliesner presented an overview of the Paleo CSM work. The Paleo group is particularly interested in pre-industrial scenarios involving solar variability, volcanic forcing and the last glacial maximum. In an uncoupled mode they are looking at sea ice variations over the past 10,000 years. Giff Miller has new proxy data on sea ice variations during this interval.