What if one of the IDF's has different dimensions, e.g. smaller cell size? Be aware, regridding is necessary for a correct display of the UGRID data in the (iMOD6) viewer

Good point, I assumed the existing iMODFLOW model had all matching grids (as imod-python writes them), but this is not the case for many existing models.

In that case, you can regrid as follows:

...
#%% Open data
top = imod.idf.open(top_idf)
bot = imod.idf.open(bot_idf)
khv = imod.idf.open(khv_idf)

# Regrid, I assume khv has a different grid, and the grid of "top" is required.
# Choose a "like" grid to regrid to, this is a dummy DataArray to which "source" is regridded.
like = top
#Regrid, different methods can be chosen, you can even supply your own method, see docs. We here choose "mean".
khv_regridded = imod.prepare.Regridder(method="mean").regrid(khv, like)

# Continue rest
ds = xr.merge(layer_coord_to_var(da) for da in [top, bot, khv_regridded])
...

For more information, see our documentation on regridding.

If I use the command given above:

# Convert to URGRID ds_ugrid = imod.util.to_ugrid2d(ds)

... there is an error message: AttributeError: module 'imod.util' has no attribute 'to_ugrid2d'

--> Do I have the wrong iMOD version or is the iMOD.util not yet released?

I updated iMOD and my version (according to print(imod.__version__)) is 0.10.1.

Ah yes, we probably did not release this functionality yet into PyPI (pip install) or conda-forge (conda install).

As we are adding quite some functionality currently, my advice is to do a "development install". This requires git, which you can download here.

rem you might want to activate you conda environment, e.g. "imod"
rem conda activate imod
cd <path/to/folder/>
git clone https://gitlab.com/deltares/imod/imod-python.git
cd imod-python
pip install -e .

Whenever you want/need to update imod-python to the last version, cd into your imod-python folder and run:

git pull

For more information, see our installation guide

@betsyromero I see you created an example script for UGRID here, similar to this snippet, in the documentation, very good.

Perhaps it is good to make the following remarks there:

1. For the new iMOD 3D viewer as well as QGIS, currently every layer needs to be a seperate variable. Hence the call here to layer_coord_to_var
2. Due to a bug in MDAL, we have to encode the times as np.float64. This is of course only for when people convert data with a time coordinate to UGRID, like transient model output.