Using CORDEX data when grid doesn't align to expected domain grid

[mo-acottrel]

Hello,
I'm working with CORDEX data but am getting an error for some cordex models, I think due to the grid of the CORDEX cube not matching the expected grid for the domain.

For example, with these cubes:
2023-05-11 16:15:57,772 UTC [84600] ERROR esmvalcore.preprocessor:353 Failed to run preprocessor function 'fix_metadata' on the data
[<iris 'Cube' of precipitation_flux / (kg m-2 s-1) (time: 1826; grid_latitude: 167; grid_longitude: 143)>,
<iris 'Cube' of precipitation_flux / (kg m-2 s-1) (time: 1826; grid_latitude: 167; grid_longitude: 143)>]
loaded from original input file(s)
[LocalFile('/project/champ/data/cordex/output/SAM-44/MPI-CSC/MPI-M-MPI-ESM-LR/rcp85/r1i1p1/REMO2009/v1/day/pr/v20160902/pr_SAM-44_MPI-M-MPI-ESM-LR_rcp85_r1i1p1_MPI-CSC-REMO2009_v1_day_20410101-20451231.nc'),
LocalFile('/project/champ/data/cordex/output/SAM-44/MPI-CSC/MPI-M-MPI-ESM-LR/rcp85/r1i1p1/REMO2009/v1/day/pr/v20160902/pr_SAM-44_MPI-M-MPI-ESM-LR_rcp85_r1i1p1_MPI-CSC-REMO2009_v1_day_20460101-20501231.nc')]

Then the following error is generated:
File "/opt/scitools/conda/environments/esmvaltool-2.8.0/lib/python3.10/site-packages/esmvalcore/cmor/fix.py", line 149, in fix_metadata
cube_list = fix.fix_metadata(cube_list)
File "/opt/scitools/conda/environments/esmvaltool-2.8.0/lib/python3.10/site-packages/esmvalcore/cmor/_fixes/cordex/cordex_fixes.py", line 193, in fix_metadata
self._fix_rotated_coords(cube, domain, domain_info)
File "/opt/scitools/conda/environments/esmvaltool-2.8.0/lib/python3.10/site-packages/esmvalcore/cmor/_fixes/cordex/cordex_fixes.py", line 144, in _fix_rotated_coords
self._check_grid_differences(old_coord, new_coord)
File "/opt/scitools/conda/environments/esmvaltool-2.8.0/lib/python3.10/site-packages/esmvalcore/cmor/_fixes/cordex/cordex_fixes.py", line 114, in _check_grid_differences
diff = np.max(np.abs(old_coord.points - new_coord.points))
ValueError: operands could not be broadcast together with shapes (143,) (146,)

main_log_debug.txt

Is there currently a way to use such CORDEX data in ESMValTool? Thank you!

[ehogan]

Hi @mo-acottrel 👋 and welcome to ESMValTool! 😁 I know @nhsavage did some work on regridding for CORDEX at the last workshop; @nhsavage, is this an issue you have experienced previously?

[nhsavage]

There are many many issues with CORDEX grids. Some are limitations of Iris (whcih is the area I was trying to work on in the workshop as we had the AVD team working with us). The problems with the metadata of the grids are more complicated. To try and improve the CORDEX grid metadata these fixers have been written, but it looks liek something isn't right. I will take a closer look but probably we need to approach the authors of these fixer