launch jenkins(debug)

The testsuite you launched for this PR has failed. You can inspect the results at this link.

Test results summary after Antarctica coastal glacier fix

The change intentionally modifies output in Antarctic coastal regions to remove unrealistic high-albedo artefacts.

All tests pass except for one:

• mpim/icon_r2b4 (GLOBE input) → FAIL in tolerance_check_cdo.py
  • Affected fields: ALB, ALNID, ALUVD, LU_CLASS_FRACTION
  • ~87–90 grid points differ out of 20480 (~0.43%)
  • Typical absolute differences: 14.9 – 22.2 (albedo fields), some small fraction differences (0.196–1.0)
  • The differences are expected and directly caused by the new Antarctica coastal correction:
    • Reduced excessive snow_ice fractions
    • Forced albedo = 0.15 on low-elevation (<100 m), low land-fraction (<50%) points south of 60°S

The differences are expected and directly caused by this new correction

! enforce that the lu_class_fractions sum up to 1  (existing code above)

IF (lu_class_fraction(i,j,k,ilu_snow_ice) > 0.01_wp .AND. lat_geo(i,j,k) < -60._wp &
     .AND. fr_land_lu(i,j,k) < 0.5_wp .AND. hh_topo(i,j,k) < 100._wp ) THEN
   count_frland_ice = count_frland_ice + 1
   lu_class_fraction(i,j,k,ilu_snow_ice) = 0.6_wp * fr_land_lu(i,j,k)
   lu_class_fraction(i,j,k,ilu_bare_soil) = 0.4_wp * fr_land_lu(i,j,k)
   alb_field_mom (i,j,k,1:12) = 0.15_wp
   alnid_field_mom (i,j,k,1:12) = 0.15_wp
   aluvd_field_mom (i,j,k,1:12) = 0.15_wp
ENDIF

Other tests unaffected (as expected):

• All COSMO-7/1, CCLM 12km, ICON-D2(+caching), Ecoclimap-SG, ECCI, Corine → green / bit-identical or within tolerance

Recommendation:
Since this is a deliberate improvement for physical realism (especially relevant for global/coarse grids), I suggest merge this PR afterwards

Hi @jhelmert,

Ok, perfect! Then I will update the references for the failing test and merge this, ok?

Also, quick question: As far as I understood, this fix was previously done in the ICON code. Was the fix now removed from ICON? And does everyone that depends on that fix know that it will now be part of EXTPAR (i.e. that they would need to update their external parameter files)?