[MITgcm-devel] verification/seaice_obcs

[Dimitris Menemenlis]

> Not sure where the problem is.
> e.g., using obcs:
> uwind(i=Nx) has 2 center neighbours XC(Nx-1) & XC(Nx) which should  
> receive
> valid values (and unaffected by the exchanges) from exf
> and it should not be a problem to interpolate the wind to this  
> location.

The funny edge winds are caused by vector rotation in exf_set_uv.F,  
which requires valid xG and yG values at Nx+1 and Ny+1.

> In addition, we are not solving the momentum eq at this location
> since it's either at the OBC or further outside the OBC, although
> it's probably needed for KPP(i=Nx-1) and other stuff like that.

OK.

> and if not using obcs, we can agree that the domain needs to be
> closed at both end when the grid is different at both end:
> e.g., N-S edges, we would add 2 walls at j=1 & j=Ny,
> like we do with Lat-Lon grid when tricky terms
> (e.g. some side drag with fancy Horiz. viscosity) are used.

Won't the same issues, horizontal viscosity, etc., also apply to the  
open boundary problem when the domain is open at both ends and the  
grid is different?  How far away from the edges do the open boundary  
conditions need to be?

> Please give us more details on what is wrong.

I still have not managed to get seaice_obcs to satisfy tst_2+2.  That  
is why I am on a fishing expedition for possible problems.

D.