missval for zero sea ice

[chrisdane]

Hi

Currently a sea ice concentration of zero is saved as a missval (see the changing number of missvals per time step):

cd /scratch/a/a270092/pycmor_hr/cli37_extraatm_512g_filecache_vec/cmorized/core_seaice/cmorized
cdo infon siconc_tavg-u-hxy-u_mon_glb_gn_AWI-ESM-3_picontrol_r1i1p1f1_158701-158712.nc
Warning (cdf_set_var): Inconsistent variable definition for lat_bnds!
Warning (cdf_set_var): Inconsistent variable definition for lon_bnds!
    -1 :       Date     Time   Level Gridsize    Miss :     Minimum        Mean     Maximum : Parameter name
     1 : 1587-01-16 12:00:00       0  3146761 2174108 :  6.5424e-07      70.862      100.00 : siconc        
     2 : 1587-02-15 00:00:00       0  3146761 2231377 :  4.4944e-07      73.981      100.00 : siconc        
     3 : 1587-03-16 12:00:00       0  3146761 2198433 :  4.9542e-07      73.465      100.00 : siconc        
     4 : 1587-04-16 00:00:00       0  3146761 2166216 :  4.4839e-07      73.823      100.00 : siconc        
     5 : 1587-05-16 12:00:00       0  3146761 2139637 :  5.9624e-07      73.694      100.00 : siconc        
     6 : 1587-06-16 00:00:00       0  3146761 2149237 :  5.7427e-07      72.582      100.00 : siconc        
     7 : 1587-07-16 12:00:00       0  3146761 2188952 :  7.0793e-07      70.421      100.00 : siconc        
     8 : 1587-08-16 12:00:00       0  3146761 2270375 :  4.9355e-07      69.503      100.00 : siconc        
     9 : 1587-09-16 00:00:00       0  3146761 2319188 :  5.7495e-07      71.116      100.00 : siconc        
    10 : 1587-10-16 12:00:00       0  3146761 2230990 :  4.5295e-07      71.999      100.00 : siconc        
    11 : 1587-11-16 00:00:00       0  3146761 2141784 :  5.0039e-07      73.725      100.00 : siconc        
    12 : 1587-12-16 12:00:00       0  3146761 2106925 :  5.8550e-07      72.354      100.00 : siconc    

I can't really give a reference but I think it's more correct to set a zero sea ice concentration to zero and not to missval.

All sea ice variables are affected by this.

just for reference: FESOM/fesom2#799 (comment)

Cheers,
Chris

[taylor13]

Yes, I'm almost positive that sea ice concentration (and the other area fractions in CMIP7) should be defined everywhere, even over land, so there should be no missing values. (Some of the "descriptions" in the data request imply that.). The reason for defining these fractions everywhere is that simplifies conservative regridding and area averaging algorithms.

[nwieters]

Hi, I made a fix for this. It is/will be in branch feat/cmip7-awiesm3-veg-hr.
For data on native fesom unstructured grid fillValues in sea-ice and other related variables will be set to zero. In the fesom version I use for the CMIP7 experiments, Fesom writes zero sea-ice as fillValues. These output has only ocean cells. So no land at all.
For regridded output that we do along the native grid output, it will use a land-ocean mask and set fillValues over ocean to zero but leaves values over land as fillValues.