CDO

Moin,

I would like to do a conservative remapping using cdo. As the same remapping should be applied to a large number of files I would prefer using a pre-generated weight file to re-calculating the remapping weights for each file. The following command works totally fine for cdo version 2.5.0:

module load cdo/2.5.0-gcc-11.2.0
cdo -s remap,r360x180,weights.nc testfile.nc testfile_out.nc # works with v. 2.5.0

However, switching to a newer cdo version, the same command fails:

module swap cdo/2.5.3-gcc-11.2.0
cdo -s remap,r360x180,weights.nc testfile.nc testfile_out.nc
cdo    remap (Abort): Source grid cell corner coordinates missing! # doesn't work with v. 2.5.3

However, when not using a pre-generated weight file but calculating the weights on the fly everything works fine also with version 2.5.3:

cdo -s remapcon,r360x180 testfile.nc testfile_out.nc # works

Also, deleting a random variable prior to interpolation seems to do the job:

cdo -s remap,r360x180,weights.nc -delname,sp_10m testfile.nc testfile_out.nc   # works
cdo -s remap,r360x180,weights.nc -delname,t_g testfile.nc testfile_out.nc   # works as well
cdo -s remap,r360x180,weights.nc -delname,u_10m testfile.nc testfile_out.nc   # works as well

So it seems to be totally irrelevant which variable I delete. I wondered if the problem might be related to memory issues or so, so what I also tried is to artificially duplicate one variable under a different name:

cdo merge testfile.nc -setname,duplicate_of_sp_10m -selvar,sp_10m testfile.nc testfile_mod.nc
cdo -s remap,r360x180,weights.nc testfile_mod.nc testfile_out.nc  # doesn't work
cdo    remap (Abort): Source grid cell corner coordinates missing!
cdo -s remap,r360x180,weights.nc -delname,sp_10m testfile_mod.nc testfile_out.nc # works

This behaviour was reproducable on two machines (levante and the HPC system of DWD), in both cases version 2.5.0 was the last version that works. On the DWD system the test was done using version 2.5.2 instead of 2.5.3 (availability of installed versions slightly differed between the two machines). The problem seems to persist for any version newer than 2.5.2.

Any ideas what might cause this?

Best,
Ralf