[MITgcm-devel] are curvilinear grids ever periodic?

[Martin Losch]

Hi there,

the netcdf output of XG, YG (in grid.t*.nc) on curvilinear grids has (nx+1,ny+1) entries. The last row/column contain the values of the first, unless we use the cubed sphere (or llc). For periodic domains that makes sense, but do we ever think that this type of grid has periodic boundary conditions? I.e. the Arctic cap configurations that are carved out from the cubed sphere certainly don’t.

Does is make sense to (only in this case, i.e. usingCurvilinearGrid=.true. and .not.cubedSphereExchange) to compute the coordinates in i=Nx+1 (for all j) ,j=Ny+1 (for all i) by linear extrapolation? 
Or is this not general (and important!) enough to be part of the repository? Or if at all, introduce this with an extra flag (e.g. nonPeriodicCurviLinearGrid)?

Martin