[MITgcm-devel] upcoming changes in seaice_growth.F

[Jean-Michel Campin]

Dear Brothers and Sisters,

I have a different vision on this issue.
First, I have to say, I don't believe in the EVOLUTION (branch).
The Seaice package was given to us with a A22 parameter.
It was there in the CREATION branch (also referred as "LEGACY" branch
by few heretics like Gael), and everyone was able to contemplate
the happy seaice growth.
Then A22 got removed, and things started to go bad, as Dimitris
mentioned. But I have a different interpretation: It's because it
got hot like hell that all the seaice started to go away and
disappear under the areMin threshold.
There was some early warnings in the sermon of Reverend TAF,
but, we have to recognize, it's not always easy to get the true
meaning of TAF's messages, and even the most fervent disciple
Dr Patrick H. admitted the problem other day (private communication).
Therefore, I propose to start a petition to restore A22 in it's
original function, in the spirit of the genuine Seaice package.

In Peace,
Jean-Michel

On Fri, May 27, 2011 at 09:34:31AM -0400, Gael Forget wrote:
> Hi Dimitris,
> 
> we are not going to remove areaMin because it is needed in the LEGACY branch.
> 
> rev.87 is when I added the merged branch (which I refer to as EVOLUTION) including
> the set of pathological cases that I thought required some treatment. The one you are referring 
> to (area=max(area,areaMin) if ice is present) I included based on Martin's comment
> C     The default of A22 = 0.15 is a common threshold for defining
> C     the ice edge. This ice concentration usually does not occur
> C     due to thermodynamics but due to advection.
> 
> After I completed this phase of the merging in October, I sent a detailed email to the devel list, entitled
> 	[MITgcm-devel] wrapping up seaice_growth.F  october 2010 revision
> where I was referring to this pathological cases part specifically, and suggesting comparison runs. 
> I don't recall getting much feedback. So, to be on the safe side, SEAICE_GROWTH_LEGACY has 
> remained the default since then. I am still hoping to get more feedback...
> 
> Have you ever tried the EVOLUTION branch? Did you notice any big difference compared 
> with LEGACY branch results? (I didn't at the time)
> 
> Cheers,
> Gael
> 
> On May 27, 2011, at 4:30 AM, Menemenlis, Dimitris (3248) wrote:
> 
> > Gael and Ian, I am also a little worried about suggested removal of areaMin, mainly because I don't remember why areamin is there in the first place.
> > 
> > Digging back in CVS tree, areaMin (formerly under more informative name of A22) has been around since beginning of time, i.e., since first check-in of pkg/seaice on branch release1 some 8.5 years ago.  That is, it originates from the original Jinlun/Hibler code.
> > 
> > First a clarification, areamin (or A22) was never supposed to be used as an actual minimum on AREA.  Up to seaice_growth version 1.86 areamin was only used to regularize thickness for seaice_solve4temp computations.  Attached figure shows histogram of AREA for the cube92 integration carried out with checkpoint 62h (revision 1.64 of seaice_growth).  There is no 0.15 threshold.  Starting in seaice_growth version 1.87, areamin is used to modify AREA directly.
> > 
> > Gael can you explain the 1.86 to 1.87 modification.  Did you actually mean to establish a lower threshold of .15 for AREA or is this threshold somehow removed later in routine?
> > 
> > Second, Patrick noted that the areamin regularization causes trouble with adjoint back in revision 1.2 of seaice_growth:
> > "cph
> > (all
> > these
> > initialisations
> > involving
> > AREA
> > are
> > nasty
> > "non-linear")"
> > and I added a cryptic comment in revision 1.6 that says:
> > "transfering all regularization of local ice thickness to seaice_growth
> > as a first step towards possibly getting rid of A22 altogether".
> > 
> > 
> > So it seems that removing the areamin regularization was something we meant to do for a long time but I forget why it never got done.
> > 
> > Ian do you know what is impact of removing the areamin regularization on seaice_solve4temp?
> > 
> > Thanks
> > 
> > Dimitris Menemenlis
> > 
> > On May 25, 2011, at 2:04 PM, Pierre Rampal wrote:
> > 
> >>> With regard to 1.1) Ian argued at the time that we don't need
> >>> areaMin (old A22) to be a run time parameter. Opinions?
> >> If I understand well, if areaMin is no longer a run time parameter, it will be impossible to change his value from data.seaice, is it right? If it is, I'm not sure that it is a good idea as it would be better to look first at the sensitivity of models results to a change in this value. To do that with convenience, it seems to me important to be able to change its value from data.seaice. Indeed, I would expect some important changes in air-sea fluxes in coarse resolution setup by changing this value. As you mention below, areaMin is currently set to 0.15, which is somehow very large. So, before having tested the effect of changing it from 0.15 to 10^-5  or any other value, I would keep it as a run time variable.
> > 
> > _______________________________________________
> > MITgcm-devel mailing list
> > MITgcm-devel at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 

> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel