[an error occurred while processing this directive] [an error occurred while processing this directive]
The CCSM Polar Climate Working Group met on Wednesday, 27 June 2002, in Breckenridge, Colorado, as part of the Seventh Annual CCSM Workshop.
J. Schramm (NCAR) brought the working group up to date on changes to the sea ice model and documentation over the past year, including its release. CSIM4_6_15 was released on May 17, 2002 as the sea ice component of CCSM2.0. Two more releases are planned - in late summer 2002 a paleo (lower resolution) version of the model, which has a 10% lower albedo for sea ice; and in autumn 2002 a version with improved cloud physics and cleaned-up ice-ocean coupler. Some changes to the ice model source code in the past year include shared constants for consistency between components, modifications to history and restart file names, the addition of ProTex to the source code, and a reduction in the amount of information exchanged between ice model and coupler. Overall model performance has increased about 19%, while ice model performance has increased about 30%. Documentation of the ice model is available and includes a Scientific Guide, User's Guide and a Code Reference Guide. Both the source code and the documentation are available at: http://www.cesm.ucar.edu/models/ccsm2.0
The ice model has three configurations: "B" (fully coupled), "F" (data ocean model and thermo-only ice model which reads in prescribed ice concentration), and "M" (data atmosphere model, data land model, slab ocean model). At this time it is assumed the user will provide his/her own forcing data sets for the "M" configuration. What's provided is 1 year of NCEP atmosphere data in latm.
A control run using the fully coupled configuration of CCSM2.0 has presently been run out to year 830. There are some known problems with the history files output from the ice component of the control run. Information is not received from the coupler at every grid point due to packing of the buffer. Complete fields can be retrieved from the atmosphere or ocean history files. Some of the fields written to the history files are those sent to the coupler (divided by ice area), not those that affect the ice. Both of these problems have been documented in the User's Guide and the latter has been fixed in the source code and will not be a problem in later simulations.
The active ice only (M) configuration has been released with a single year (1985) of NCEP forcing. A GUI (Graphical User Interface) has also been released with CCSM2.0 that will aid considerably in setting up the scripts for any model configuration.
Bruce Briegleb followed with a detailed description of the polar climate and sea ice simulation of CCSM2. (Bruce presented polar atmosphere results at the Atmosphere Model Working Group meeting on Wednesday.) Compared to CSM1, CCSM2 has much thinner Arctic ice and extensive and thick Antarctic ice. Mean annual Arctic ice thickness is 1.4 m, while in the Antarctic 2.0 m. The polar sea ice simulation was shown to be dynamically and thermodynamically realistic, given some known biases in the simulated atmospheric and oceanic fields (such as incident radiation fluxes at the surface) that force the sea ice model. The CCSM2 Arctic Ocean simulates the halocline, which was completely absent in CSM1, despite a warm Atlantic underlayer at 200-300 m depth. Antarctic temperature and salinity profiles are close to observations. It is likely that biases in elements of the atmosphere model and ocean model simulations contribute significantly to the bias in the simulated ice extent and thickness. Overall, CCSM2 polar and sea ice simulation are an improvement in several ways over that of CSM1.
Marika Holland analyzed arctic climate variability simulated by CCSM2, including comparisons with NCEP Reanalysis (RA) data. The NCEP RA in winter shows a trend in surface air temperature with a large dipole pattern in the high latitude North Atlantic, related to the trend in the NAO. CCSM2 control simulation exhibits less low frequency NAO variability than in NCEP RA. The spatial patterns of CCSM2 surface temperature and precipitation anomalies regressed on the CCSM2 NAO index are qualitatively reasonable, as are the ice concentration and thickness anomalies in the central Arctic. But all model patterns have smaller amplitude than observed. A comment from the working group suggested that the model's inability to simulated the natural variability of the last 30 years might be partially explained by lack of anthropogenic forcing in the model.
Bill Lipscomb presented his recent efforts to develop an improved ridging scheme for the sea ice model. Most of the physics components in CSIM have been redesigned during the past few years. The ridging component, however, is based on schemes developed during the 1970s when few ice thickness measurements were available. An analysis of sonar-observed ice drafts in the Arctic and Antarctic shows that the tail of the ice thickness distribution (ITD), which consists mainly of ridged ice, is invariably exponential. Our current ridging scheme does not reproduce this shape; the ITD falls off too abruptly to the right of the thermodynamic mode, then too slowly over the range 5-20 m. If we replace the current ridging redistribution function with an exponential function, we obtain an ITD in much better agreement with observations. The exponential shape can be derived from statistical mechanics, in analogy with the Boltzmann distribution, if we think of ridging as a process that redistributes ice at random under the constraint of constant volume. Using ice draft data, we can derive an empirical relationship between the e-folding scale and the parameter G*, which determines the portion of the ITD that participates in ridging. The effects of a new ridging parameterization on long-term climate simulations are unknown and will be studied during the next year.
Alex Hall discussed the impact of ice surface albedo on simulations of the climate response to increasing greenhouse gases, using an ocean-atmosphere model developed at the Geophysical Fluid Dynamics Laboratory. GHG scenarios are run with (a) fully interactive sea ice and (b) sea ice with albedo fixed to follow the control simulation. Polar amplification of surface temperature change is a prominent feature of both simulations, but is larger in scenario (a). In both scenarios, the polar amplification is associated with thinning sea ice, which permits a larger heat flux from the ocean to the surface in the cold season. Except where the sea ice disappears completely, surface temperature in summer is fixed at the melting point. Tropical temperatures also respond in these scenarios, indicating some sensitivity of poleward atmospheric heat transport to the ice albedo feedback in the Arctic. Alex stressed the importance of studying polar-tropical interactions, and the relationships between changes in surface albedo and planetary albedo.
Richard Grotjahn described studies of biases in the SLP and other fields simulated by CCM3.6 in the Arctic. The SLP bias in the Arctic is part of a hemispheric bias pattern with extrema over Europe (east of Iceland) and over E.Siberia-Alaska. The biases are statistically significant, i.e. not due to sampling error arising from natural variability in the model. Observations were studied to determine anomaly patterns in climate variables that occur when observed arctic features (e.g. the Beaufort High) are shifted from their climatological positions in a manner similar to the bias of CCM climatology. Temperature showed strongest links at locations remote from the center of SLP bias. Partly this can be explained as temperature advection by winds associated with the anomalous SLP field; but some features could not be explained by advection. Fields of precipitation, eddy fluxes, and cloud amount in the N. Pacific storm track and N. and W. Europe were related to the shifts in the Beaufort high. Overall, the model fields of bias associated with biased arctic circulation features matched well with anomaly fields in the NCEP data associated with anomalies in the corresponding arctic circulation features.
W. Hibler, in his presentation titled "High frequency sea ice variability and ice mechanics," warned that the ice-ocean boundary layer may be improperly modeled in CCSM2. High resolution data indicate a 12-hour peak in deformation rates owing to inertial oscillations; the quadratic drag law currently used in CCSM2 was not designed for inertial time scales (although it works fine for time scales longer than a few inertial oscillations, as in the current mode of CCSM2 operation). He suggested that the ice dynamics model needs to be embedded within the ocean boundary layer in order to properly model the ice response to inertial motions (i.e., ridge formation) and oriented, linear features observed in satellite images of the ice pack. He plans to make the case that the CCSM simulation can be improved by including formulations that account for high frequency ice deformation and oriented fractures in the ice cover.
Following the presentation was a general discussion. It was noted that the PCWG CSL allocation was approved. E. Hunke raised the possibility that future models might consider a separate boundary layer under each category of sea ice, given results presented in a poster by M. Holland and W. Large. There was a general desire to create a test suite for the M version of the model that would include a community-available forcing data set.