Question about remapping vector variables (e.g., tx_sur, ty_sur)

[gyuseokyi]

Hi,

I am Gyuseok.
I have a question regarding the remapping of vector variables such as tx_sur and ty_sur.
I currently remap these variables using CDO as follows:

cdo genycon,r1440x720 -selvar,tx_sur -setgrid,dart_griddes_elems.nc tx_sur.fesom.1950.nc weights.tx_sur.r1440x720.nc
cdo -L remap,r1440x720,weights.tx_sur.r1440x720.nc -setgrid,dart_griddes_elems.nc tx_sur.fesom.1950.nc tx_sur.fesom.1950.r1440x720.nc

Is any additional vector rotation required after this remapping?

For example, in the POP model, the model-grid velocity components are rotated to geographic eastward and northward components using a transformation such as:

true_zonal_velocity = cos(ANGLE) * UVEL - sin(ANGLE) * VVEL
true_meridional_velocity = sin(ANGLE) * UVEL + cos(ANGLE) * VVEL

Is a similar rotation necessary for FESOM vector variables (e.g., tx_sur and ty_sur) before or after remapping?

Thank you for your help.

[patrickscholz]

Hi @gyuseokyi

So within the model all vector quantities are in the rotated coordinate frame in which FESOM2 operates. By default these vector variables are also written out in this rotated coordinates and need to be rotated back to geo-coordinates in the post-processing. However there is a flag in namelist.io: vec_autorotate = .true./.false., where the model does the rotation to geo-coordinates before it writes them out.
You can also chech in the attributes of your netcdf files in which which coordinate frame the data are written e.g (attribute, 'autorotate_back_to_geo', merge(1, 0, vec_autorotate))

What you did so far, is only an conservative interpolation to a regular grid, this doesnt help you with the rotation. The more correct way would be to do the rotation before the remapping ! You can look in pyfesom2 or tripyview how this is done in python. If you cant use python than you could also try to do the rotation after the remapping with the remapped rotated vector components but it is less precise.

[gyuseokyi]

Dear @patrickscholz

Thank you very much for your kind response.
It seems that I will need to redo the remapping using pyfesom2.

I encountered errors related to version differences in the following file:
~/miniconda3/envs/python3/lib/python3.12/site-packages/pyfesom2/pfinterp.py

The following part needs to be modified as shown in the attached screenshots:

tonodes(
    data_forint.astype("float32"),
    mesh
)
#     mesh.n2d,
#     mesh.voltri,
#     mesh.elem,
#     mesh.e2d,
#     mesh.lump2,
#     )

In addition, in:
~/miniconda3/envs/python3/lib/python3.12/site-packages/pyfesom2/regridding.py

I modified line 64 as follows:
distances, inds = tree.query(list(zip(xt, yt, zt)), k=k, workers=n_jobs) # n_jobs=n_jobs)

I am sharing these modifications in case they are useful for others who encounter similar version-related issues.