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

[Martin Losch]

Hi again,

I have another question for Jinlun. I am now quite confident, that I have found all (serious) problems in the metric terms. However, my code still blows up. Could the discretization of \eta be a problem? On Z-points ("bottom left" corner of C-grid cell, on a B-grid this is where the velocities are), I average eta like this:
        DO j=1,sNy+1
         DO i=1,sNx+1
          etaMeanZ (I,J,bi,bj) =
     &         (           ETA (I,J  ,bi,bj)  + ETA (I-1,J  ,bi,bj)
     &         +           ETA (I,J-1,bi,bj)  + ETA (I-1,J-1,bi,bj) )
     &         / MAX(1.D0,maskC(I,J,  k,bi,bj)+maskC(I-1,J,  k,bi,bj)
     &         +          maskC(I,J-1,k,bi,bj)+maskC(I-1,J-1,k,bi,bj) )
         ENDDO
        ENDDO
In the B-grid code I find this:
        DO j=1-Oly+1,sNy+Oly
         DO i=1-Olx+1,sNx+Olx
          ETAMEAN(I,J,bi,bj) =QUART*(
     &          ETA(I,J-1,bi,bj) + ETA(I-1,J-1,bi,bj)
     &         +ETA(I,J  ,bi,bj) + ETA(I-1,J  ,bi,bj))
          ZETAMEAN(I,J,bi,bj)=QUART*(
     &          ZETA(I,J-1,bi,bj) + ZETA(I-1,J-1,bi,bj)
     &         +ZETA(I,J  ,bi,bj) + ZETA(I-1,J  ,bi,bj))
         ENDDO
        ENDDO
This means that \eta/\zeta on the boundaries are only half as large as in my C-grid discretization (however, these values may actually never be used because of the no-slip BCs on the B-grid). So I am wondering, what the "correct" value of \eta on the boundary is. I observe, that code with
        DO j=1,sNy+1
         DO i=1,sNx+1
          etaMeanZ (I,J,bi,bj) = 0.25 _d 0 *
     &         (           ETA (I,J  ,bi,bj)  + ETA (I-1,J  ,bi,bj)
     &         +           ETA (I,J-1,bi,bj)  + ETA (I-1,J-1,bi,bj) )
         ENDDO
        ENDDO
tends to be stable (at least in the shorter test that I made). Jinlun, what's your opinion on this?

Martin

----- Original Message -----
From: Martin Losch <Martin.Losch at awi.de>
Date: Friday, March 6, 2009 8:38
Subject: Re: [MITgcm-devel] (not so) funny things happen in seaice_lsr and pickups
To: MITgcm-devel at mitgcm.org

> Hi Jean-Michel,
> 
> I'll create a SEAICE_GRID.h, there are actually quite a few 
> fields in  
> SEAICE.h, that are really grid parameters (e.g., masks). These 
> are  
> then initialized in seaice_init_varia.F which is not quite 
> consistent  
> either. I'll do something about that once the C-LSR with metric 
> terms  
> proves to be stable (it's still running, but I need at least 
> the  
> weekend before I can say more)
> 
> Martin
> 
> On Mar 5, 2009, at 10:44 PM, Jean-Michel Campin wrote:
> 
> > Hi Martin,
> >
> > I guess it depends if we want
> > tanPhiAtZ and tanPhiAtC arrays to always be defined
> > (in this case, they will sit in GRID.h) or if we want
> > to define those 2 2-D arrays only with pkg/seaice,
> > and in this case I will prefer to put them elsewhere,
> > so that when GRID.h is included without PACKAGES_CONFIG.h
> > we don't have an incomplete file. Why not keeping those
> > 2 simple names tanPhiAtZ/C, but in a new header file
> > SEAICE_GRID.h ?  This way, if, in the future, they need
> > to be move back to GRID.h, it will not be a big deal.
> > And they could be set in ini_spherical_polar_grid.F
> > before the grid-rotation.
> >
> > Just a comment: regarding the the grid-rotation calls, them
> > could have been put in ini_grid.F (and then used also with
> > curvilinear grids, when x & y are lon & lat), no ?
> >
> > Jean-Michel
> >
> > On Thu, Mar 05, 2009 at 12:16:31PM -0800, Dimitris Menemenlis wrote:
> >> I will let JM answer this.
> >>
> >> On Mar 5, 2009, at 11:17 AM, Martin Losch wrote:
> >>
> >>> However, first I'd like to get a comment on how to add these new
> >>> fields tanPhiAtZ and tanPhiAtC (see a previous email in 
> this  
> >>> thread).
> >>> Currently I have implemented a version that is "private" to
> >>> pkg/seaice, because these fields would not be used 
> elsewhere. Or  
> >>> is it
> >>> better to define these fields in GRID.h where they would actually
> >>> belong?
> >>
> >> _______________________________________________
> >> 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
> 
> _______________________________________________
> 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/20090308/d1afb676/attachment.htm>