interpolation of an unstructured grid with known cell areas
Added by Paolo Davini over 1 year ago
Hi there,
I am trying to produce an interpolation use the remap command family of an unstructured global grid of which:
1. I don't have information on the grid-corners
2. I know the area of each grid cell.
I tried to feed the data with the undocumented operator ` cdo setgridarea,area.nc myfile.nc ` but it seems that the information is not retained, and can only be used chained on the ` fldmean ` calls.
When trying to compute the remapping weights on my unstructured grid, the command is obviously complaining that the grid corner are missing and CDO cannot evaluate the weights. However, it should be possible to do it since we know the areas - although they are on a different file. Is there a way to feed them to the remap command so that it avoids the grid corner search?
Of course I can provide a (not) working sample data.
Many thanks
Paolo
Many thanks
Replies (4)
RE: interpolation of an unstructured grid with known cell areas - Added by Paolo Davini over 1 year ago
One important information that I realize is missing: I am trying to use conservative remapping, nearest neighbour should work out of the box...
RE: interpolation of an unstructured grid with known cell areas - Added by Ralf Mueller over 1 year ago
hi Paolo!
Please upload he input data. that would help a lot, I hope.
cheers
ralf
RE: interpolation of an unstructured grid with known cell areas - Added by Ralf Mueller over 1 year ago
hi again!
Thought about it again. The grid area is not enough to perform conservative remapping. Without the corners of each source and target cell it is not possible to compute their fractional area overlay. Hence the interpolation weight cannot be computed, sry.
BTW: depending on the grid bi-linear, distance weighted or others are possible, too.
cheers
ralf
RE: interpolation of an unstructured grid with known cell areas - Added by Paolo Davini over 1 year ago
Thanks Ralf, you are absolutely right!!!
I need the corners, there is no way to avoid it. I will try to get those corners from the data producers...