[MITgcm-devel] (not so) funny things happen in seaice_lsr and pickups

[Martin Losch]

Hi Jinlun,
I used the "canonical" ZMIN=4.e8 that I found in earlier versions of the B-grid code, but I am quite sure (from what I learned from the analysis of some "blow-ups"), that a much smaller value, maybe even O(1), would do, too. 

However, this would be just a fix for some extreme situtations and I have the ambition to solve this discretization problem cleanly.
In this context, could you do me a favor and have a look at the current B-grid code (pkg/seaice/lsr.F). It has changed a little from your original, but you still should be able to recognize this. I have a hard time associating the discretized terms with the ones in you 1997 paper, e.g. terms like (d/dx) [eta tanphi vice] appear (to me) to be discretized as [eta tanphi] d vice/dx. Could you check that, as these terms seem to be the problem in the C-grid (in terms of boundary conditions). I'd rather rewrite them as (d/dx) [eta tanphi vice]=eta tanphi] d vice/dx + vice (d/dx) [eta tanphi] before discretizing.

Martin

Martin

----- Original Message -----
From: Jinlun Zhang <zhang at apl.washington.edu>
Date: Tuesday, March 10, 2009 19:57
Subject: Re: [MITgcm-devel] (not so) funny things happen in seaice_lsr	and pickups
To: MITgcm-devel at mitgcm.org

> Martin, so a non-zero ZMIN helps to stabilize the system. What 
> value do 
> you use to make it stable?
> 
> Jinlun
> 
> Martin Losch wrote:
> > The question marathon continues ...
> > Jinlun, my observation is, that the metric terms cause all the 
> trouble 
> > (no problems so far in cubed sphere integration of any 
> resolution, nor 
> > on cartesian grid).  I have the suspicion that the 
> symmetry of the 
> > laplacian operator is broken, because the metric terms are 
> distributed 
> > unevenly between left and right hand side of the u and v 
> equations. In 
> > order to ensure the diagonal dominance of the coefficient 
> matrix, 
> > would it be possible to move all metric terms to the rhs? Just 
> > wondering ...
> >
> > BTW, the system seems to be stable when I have a non-zero ZMIN.
> >
> > Martin
> >
>    
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

Martin Losch
Alfred Wegener Institute 
Postfach 120161, 27515 Bremerhaven, Germany; 
Tel./Fax: ++49(0471)4831-1872/1797




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20090312/51b3d9a8/attachment.htm>