[Mitgcm-support] Interpolation & Land

[mitgcm-support at dev.mitgcm.org]

Hello Alistair,

It's about topography & matlab interpolation from the cubic sphere 
grid to lat-lon grid (to make some graphes):

o using directly the matlab routine is not good because of land.

o putting NaN in land before doing the interpolation does not works 
so well because all the neighbor cells of one land point become NaN.

o what seems better is to multiply the field by the land mask,
then to interpolate, and finally to divide the results 
by the interpolated mask. 
This is based on the assumption that the interpolation is 
linear. Is it the case ? 

For u,v , I assume that they are =zero in land and interpolate directly
using uvcube2latlon.
Finally, I divide by the interpolated hFacC mask. I use the same mask 
as for tracer since the output lat-lon grid is identical.

Does it make sense ? 

See you,

Jean-Michel