Reproject aifs to a regular lat/lon grid

[sailgrib]

I am trying to reproject the reduced-gaussian aifs grib file to a regulat 0.25° lat/lon grid.
I have tried 3 approaches, one using nearest neighbor and 2 others that should be better but both failed.

The first one using nearest neighbor :
cdo remapnn,target_grid_world_025 input_file output_file
That works but is it really appropriate?

The second one with bilenear interpolation:
cdo remapbil,target_grid_world_025.txt input_file output_file
=> cdo remapbil: Bilinear weights from gaussian_reduced (542080) to lonlat (1440x721) grid
=> cdo remapbil (Abort): Can't do bilinear interpolation if the source grid is not a regular 2D grid!

The third one using first-order conservative remapping:
cdo remapycon,target_grid_world_025.txt input_file output_file
=> cdo remapcon (Abort): Source grid cell corner coordinates missing!

Any idea as to how I should correctly reproject the aifs to a regular lat/lon grid?

[igorROIK]

I've tried several ways to plot ECMWF AIFS data in Python, but without success. Could you help me or provide a script using this data? Thanks.

[sailgrib]

Sorry, I cannot help with this. I have not tried to plot ECMWF AIFS. i am using the data as lat/lon grib files.

[sol1105]

I had the same problem and found this issue. Meanwhile I have found the solution, so adding it here:

cdo -f nc4 remapbil,global_0.25 -setgridtype,regular in.grib out.nc

The explanation can be found here: https://code.mpimet.mpg.de/boards/1/topics/11546
In short: cdo only supports remapnn (nearest neighbour) and remapdis (inverse distance weighting) for unstructured grids, so the data must first be transformed to a regular grid. remapcon generally needs the grid cell boundaries / vertices defined to be able to calculate the area of the grid cells.