In trcdia.F90, kindic is a local integer of routine trc_dia. It is not initialized and it is passed as second argument to the routines

• trcdit_wr
• trcdii_wr
• trcdib_wr

In these subroutines, kindic is for INTENT(in). Therefore, the value should come from trc_dia itself.

kindic is used in the subroutines, a negative value forcing a call to histclo. And if this happens spuriously, the program ends up later with an IOIPSL error (histvar_seq), whose origin looks rather unclear to the user ;).

I am surprised to see those lines in blk_ice_core :

...
 649                ! Long  Wave (lw)
 650                ! iovino
 651                IF( ff(ji,jj) .GT. 0._wp ) THEN
 652                   z_qlw(ji,jj,jl) = ( 0.95 * sf(jp_qlw)%fnow(ji,jj,1) - Stef * pst(ji,jj,jl) * zst3 ) * tmask(ji,jj,1)
 653                ELSE
 654                   z_qlw(ji,jj,jl) = 0.95 * ( sf(jp_qlw)%fnow(ji,jj,1) - Stef * pst(ji,jj,jl) * zst3 ) * tmask(ji,jj,1)
 655                ENDIF
...

• Special treatment for northern hemisphere looks like an ad-hoc fix for a forcing set.
• In many cases (eg DRAKKAR forcing set ), tuning is done on the forcing files, or eventually in customized nemo for a given configuration
• I think that this kind of ad-hoc fix should NOT be in the reference version, unless I am missing a good rationale for that.

Cheers,

Jean-Marc

Context

ABL version implemented during 2019 MP is not restartable

Analysis

problem with ABL time swap indexes + air density calculated at the wrong place

Fix

apply bugfixes already tested in dev_r12563_ASINTER-06_ABL_improvement branch (see r12770, r12799 and r12808 in ticket #2419) to the trunk