CDO

hi Alexander!

your target grid file does not provide cell corners, hence CDO cannot

• compute the area of each cell (I know this is given as an extra variable)
• compute overlaps with the input cells to calculate the area fractions of all input cells for the specific target grid cell

CF-conformant files provide additional coordinate variables and attach those to the coordinated with the bounds attribute

CDO has problems with your grid descriptions:

cdo verifygrid grid_CMN-5x5.nc
Warning (cdfInqContents): Coordinates variable lat can't be assigned!
Warning (cdfInqContents): Coordinates variable lon can't be assigned!
Warning (cdfInqContents): Coordinates variable lat can't be assigned!
Warning (cdfInqContents): Coordinates variable lon can't be assigned!

cdo    verifygrid (Abort): Projection type >latitude_longitude< unsupported!

cdo verifygrid grid_NH-40KM.nc
Warning (cdf_set_var): Inconsistent variable definition for lon2D!
Warning (cdf_set_var): Inconsistent variable definition for lat2D!
cdo    verifygrid: Grid consists of 97969 (313x313) cells (type: curvilinear), of which
cdo    verifygrid (Warning): Grid cell corner coordinates missing!
cdo    verifygrid:      lon2D : -179.9981 to 179.9949 degrees
cdo    verifygrid:      lat2D : 18.93623 to 90 degrees
cdo    verifygrid: Processed 4 variables [0.01s 46MB].

Additionally:

• the projection in the input file does not seem to be needed
• the coordinates do not have useful units ("degrees" must be ether degrees_east or degrees_north)

some helpful commands to clean up the input

  ncatted -a coordinates,border,d,, grid_CMN-5x5.nc
  ncatted -a coordinates,area,d,, grid_CMN-5x5.nc
  ncatted -a units,lon,m,c,'degrees_east' grid_CMN-5x5.nc
  ncatted -a units,lat,m,c,'degrees_north' grid_CMN-5x5.nc

but still: without target cell corners conservative remapping will not be possible. other methods do work, e.g. bilinear (genbil).

you don't need to generat text files with the griddes operator. Instead use the plain netcdf files - CDO takes the first grid from those files as input

cdo genbil,grid_NH-40KM.nc grid_CMN-5x5.nc scrip_CMN-5x5_NH-40KM.nc

Dear Ralf,

Thanks for the clear response, that has helped me improve my process substantially. I still have a remaining doubt however. Following your suggestions, I have cleaned grid attributes and I can successfully generate eg bilinear weights with the more straightforward command:

cdo genbil,grid_NH-40KM.nc grid_CMN-5x5.nc scrip-con_CMN-5x5_NH-40KM.nc

And I can successfully reproduce the conservative mapping error as expected:

cdo gencon,grid_NH-40KM.nc grid_CMN-5x5.nc scrip-con_CMN-5x5_NH-40KM.nc

>>> cdo    gencon (Abort): Target grid cell corner coordinates missing!

The remaining doubt is that when I generate a simple grid description file on my own (not using cdo) as attached in .txt files, then conservative mapping is completely successful:

cdo gencon,grid_NH-40KM.txt -setgrid,grid_CMN-5x5.txt grid_CMN-5x5.nc scrip_CMN-5x5_NH-40KM.nc

It only fails when I use the simple gencon command you proposed, or when I specify grid description files generated by cdo. I wonder, what is happening here?

Thanks again!
Alex

Of course, that is to say I am happy that it succeeded, as it is convenient not to have to define the grid borders when they are already clear from the grid definition. But it is surprising...