[Mitgcm-support] Re: jGl, iGl

[mitgcm-support at dev.mitgcm.org]

Hello Chris,

>    I don't understand why we would ever have to multiply the
>    overlap by the total domain size e.g
>    OLx*Nx, OLy*Ny.
>    Can anybody explain? Lines are in ini_spherical_polar_grid.F.
>    Surprised the code works if its wrong, so I am puzzled.        

Yes, I can explain this since I put this modification to work
with small domain (where Oly > Ny, as for instance, zonal-average
case):
This has no effect on the acuracy of the evaluation of iGl(I,bi)
and jGl(J,bj) since we take mod(a+OLx*Nx,Nx) and mod(b+OLy*Ny,Ny).
But in case a or b is negative, then the FORTRAN function "mod"
does not return the matematical value of the "modulus" function,
and this is not good for your purpose.
This is why I add +OLx*Nx and +OLy*Ny to be sure that the 1rst 
argument of the mod function is positive.

See you,

Jean-Michel