Impact of l_fixed_area flag in idealized variable grid simulations

[JFLemieux73]

See also issues #786, #792 and #809.

Issue #792 shows that l_fixed_area=.true. eliminates the checkerboard pattern in aice, hice fields. However, gx3 simulations crash with this flag set to true (issue #809). I want to see if idealized runs with a variable grid crash or not with l_fixed_area=.true.

./cice.setup --case CvgridFA -g gbox80 -m ppp6 -e intel -p 1x1 -s boxforcee,boxclosed,vargrid,buildincremental

I then changed things in the namelist:
npt=14 (days)
ndte = 2400
elasticDamp = 0.1d0
deltaminEVP = 2e-9
capping_method = 'sum'
kridge = 1
ktransport = 1
atm_data_type = 'uniform_northeast'
ocn_data_type = 'calm'
bgc_data_type = 'default'
fe_data_type = 'default'
ice_data_type = 'uniform'
ice_data_conc = 'p8'
ice_data_dist = 'uniform'
scale_dxdy = .true.
dxscale = 1.01d0
dyscale = 1.01d0

The runs are 14 days long. I look at the results after 14 days.

[JFLemieux73]

I set l_fixed_area=.true. for both B and C grids. Here is the B grid solution (aice):

The C grid run did not crash. Here is the solution also after 14 days:

Looks good. Its time to go back to issue #809...

[JFLemieux73]

I redid the exp above but with the modified approach for computing the edge masks as described in issue #809. This is in my branch Fixed_area_remap with commit 86d3409.

Here is the B-grid solution with fixed_area=T:

And the C-grid solution with fixed_area=T:

[JFLemieux73]

The C grid noise has been eliminated with l_fixed_area=T (see also issue #809). The code does not crash anymore for gx simulations (see PR #833).