[MITgcm-devel] Re: corner RAZ for cube sphere

[Jean-Michel Campin]

Hi Dimitri,

> Salut Jean-Michel, I want to try out your corner RAZ modification for
> the hi-res cube to see what happens.  Is the modification already
> checked-in the repository.  If not can you tell me which routine I need
> to modify.
> 
> Cheers, D.

It's not really in the repository but the experiment aim.5l_cs
use a modified RAZ for the shapiro filter, which play the role of
dissipation: 
in aim.5l_cs/code , S/R shap_filt_relvort3.F , lines 66 to 78:
     66 C     Special stuff for Cubed Sphere
     67       IF (useCubedSphereExchange) THEN
     68          AZcorner = 0.75 _d 0
     69          I=1
     70          J=1
     71 c        vort3(I,J)=
     72 c    &     +recip_rAz(I,J,bi,bj)*(
     73          vort3(I,J)=
     74      &     +recip_rA(I,J,bi,bj)/AZcorner*(
     75      &      vFld(I,J)*dyc(I,J,bi,bj)
     76      &     -uFld(I,J)*dxc(I,J,bi,bj)
     77      &     +uFld(I,J-1)*dxc(I,J-1,bi,bj)
     78      &     )

It means that instead of using recip_rAz at the corner, the
shapiro filter uses 4/3*recip_rA.
If I remember correctly, 3/4 * rA is larger than rAz at the 
corner and smaller than any other rAz arround (at least with 
the c32x32 grid) and give a smoother area variation arround 
the corner.
I choose 3/4 simply because at the corner there is only 3 
contributions to the vorticity instead of 4 usually.

I think before implementing a modification like this,
it would be interesting to check that with your grid:
rAz(corner) < 3/4 rAc(next_to_the_corner) < rAz(elsewhere).

Cheers,

Jean-Michel