[MITgcm-devel] lsr/obcs

[Jean-Michel Campin]

Hi Martin,

I think the way it is now is right, it's doing what the comments say:
> C--     prevent tri-diagonal solver to modify OB values:
and units are not really an issue since the 4 other coeffs
(AU,CU,uRt1,uRt2) are all zeros.
I don't like to change the "ONE" to something that can become zero,
because I prefer finite ice-velocity (rather than 0./0.).

But there is a potential problem for the residual calculation:
Seems to me that it's OK if we don't try to use the
free-drift residual, but there will be a problem if trying to use
LSR_mixIniGuess=2 or 4.

Did not look at your latest changes in S/R SEAICE_RESIDUAL,
but might be useful to add some maskInC(i)*maskInC(i-1)
factor in this routine.

Cheers,
Jean-Michel

On Thu, Dec 20, 2012 at 08:39:36AM +0100, Martin Losch wrote:
> Hi Jean-Michel,
> 
> just so I don't misunderstand the code for obcs in seaice_lsr.F:
> #ifndef OBCS_UVICE_OLD
> […]
>             BU(I,J,bi,bj) = ONE
> […]
>             rhsU(I,J,bi,bj) = uIce(I,J,bi,bj)
> #endif
> 
> This is absolutely correct, but would be more consistent (in terms of units) if it were
> BU =  seaiceMassU(I,J,bi,bj)/SEAICE_deltaTdyn
> rhsU = seaiceMassU(I,J,bi,bj)/SEAICE_deltaTdyn * uIce (I,J,bi,bj)
> 
> Is that correct? I guess scaling these values by mass/dt only has an effect on the residual computations. Do you agree?
> 
> Martin
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel