Sea ice meeting, Exeter, 2005/09/08

Pics show Obs (Comiso, Bootstrap); HadCM3 control (aaxzc run at the Hadley Centre); HadGEM1 (adzbf); HiGEM (eacew, Macolm Roberts, run on the Earth Simulator; and sometimes xbxrb Warwick Norton, UK). Note that eacew is only 5 years (my data) though it has run 8.

Talk plan:

  1. Thanks: HiGEM team (esp MR); Hadley Folk.
  2. HiGEM: 1/3x1/3 ocean, 5/4x5/6 atmos. New orog/bathymetry. HiGEM runs: xbxrb (5y UK); eacew (8y ES). Run with p*=5k.
  3. General maps of ice, Sep/Mar and NH/SH
  4. Thickness
  5. Ice Velocity (well speed really)
  6. Crashes

Mean area

.run p1

Mean area is not too bad, except in NH summer when there is clearly too little ice. However, unlike early runs it is no longer disastrously bad even in NH summer, this now just looks like a slight enhancement of HadGEMs tendency.

Notice the semi-polynya in the Ross in September (present in HadGEM somewhat) and the minimum in the Weddell towards 0E, as in HadGEM, rather contrary to obs.

Notice also the "kick out" at 160 W, which is present in HadGEM, exaggerated in HiGEM, but on close inspection visible in Obs and Cm3 too. Probably related to the pattern of climatological lows.

Notice that the max ice conc is around the coast (in all the models), unlike in the Obs where a broad lower-conc zone is seen near the coast.

p2 shows:

Trends in thickness

James Harle at NOC noticed trends in the HiGEM ice thickness. I'll look at the Antarctic a bit. p3

These trends are in m/y.

Ice gets rather thick near the Western Weddell / East side of the Peninsula, past 10m in places. Regrowing the Larsen!?!

p3a

Pic shows ice thickness for one Sept (after 4 y for HiGEM, after 50 for HadGEM). Vectors are long term ice motion. Note that the pattern is similar to HadGEM, but HiGEM is doing it rather faster. Ice build-up could be consistent with divergence of ice motion, but may be thermo (don't have therm/dyn increments available).

Ice speed

p4

So... top two pics are ice speed for Emery/Fowler and Eis (http://imkhp7.physik.uni-karlsruhe.de/~eisatlas/). There is clearly a major difference! Against Fowler, HiGEM/HadGEM (and HadCM3, not shown) are clearly far too fast. Against Eis, they are merely somewhat too fast. HiGEM speed/pattern very similar to HadGEM. However (perhaps for the first time) the addition higher rez of HiGEM is producing information (which may or may not be correct...) in the tighter coastal flow and the streamers in the ACC type regions.

Crashes

HiGEM has shown a tendency to crash in the sea ice code, especially around the north pole. This week, MR has found that his 8-y run is currently stopped because of this. This is due to a combination of high speeds (30-40 cm/s) and small grid cells leading to CFL problems (at least it looks like that, though MRs calc show its getting past a literal CFL).

In the long term, getting rid of the polar singularity would be a good idea.

Scientifically, finding the cause of the high ice speeds and reducing them would be sensible (perhaps just arbitrarily limit to 20 cm/s for the while).

Computationally, adjusting the grid near the pole, or filtering, may be the answer.