Interpolation of point forcing data (station-based) using wradlib

[shartgring]

Issue addressed

Fixes #312

Explanation

Consists of a part in wflow.py for reading and checking the forcing data and station data, and a new workflow in workflow/forcing.py where interpolation takes place.

Checklist

• Updated tests or added new tests
• Branch is up to date with main
• Tests & pre-commit hooks pass
• Updated documentation if needed

Additional Notes

Interpolation types not working in tests

• natural neighbor gives error messages and does not work with N = 3 (which was my initial test, it is now N = 8). Not sure why though
• Cressman interpolation returns a uniform field, so that one needs some extra checking, maybe by directly calling the MetPy function with some test data and seeing what the output looks like. I have only checked the test data which was also processed by hydromt.workflows.precip (the forcing map itself), so maybe the raw data coming out of workflows.forcing.spatial_interpolation needs checking.

01-04-2025: replaced metpy with wradlib

[shartgring]

The images from the visual inspection of the test data. The title per subplot is the mean value times 1000, and the figure title is the mean over the subplots.

[shartgring]

Allright! After some work the code looks good. Only thing to improve is how the gauge stations are loaded. That code is now copied from setup_gauges, but it can be cleaner and the implementation of the GeoDataset can be improved (linking it to precip_fn and not the stations).

The idea is to leave this PR as a draft as MetPy functionality is not working for natural neighbor and Cressman. It would make sense to switch to https://docs.wradlib.org/en/latest/. However, this will not work with the xarray requirement, and to get rid off the pin we need to upgrade to HydroMT version 1.0. Therefore, the PR will be a draft until we know more about the update to version 1.0.

[shartgring]

Replaced metpy with wradlib now that newer xarray versions are supported :)

Tests are passing and all interpolation methods work fine so looks like it is ready for review

[shartgring]

I have included an option to filter stations using the basin geometry and a buffer. This might make it easier to work with large datasets. Would it maybe be logical to include somewhere which stations have been used for interpolation? Ar a possible geom / geojson for visualization? I am doubting as they are not part of the model. Listing all the stations in the log is an alternative, but might end up with a long list

Great work! I think it would be very nice to have the meteo stations added as a geojson as well! They are not directly part of the model, but can play quite a big role indirectly, so I can imagine that quickly plotting these locations is very valuable.

I have included it 👍

[hboisgon]

Very nice improvements @shartgring ! This will definitely become a useful function I think.
I have a few comments mostly on documentation.
And to keep in with the docs, you should not forget to add this function to the hydromt_wflow docs itself :) There are two places for that: api.rst (both for the wflow function and the new forcing workflow) and then update the table for all setup methods for wflow in user_guide/wflow_model_setup.rst

[hboisgon]

I was maybe not clear but I did check in hydromt and the buffer argument for get_geodataset and get_geodataframe is in meters. So we should update the default value and the docs here.

[hboisgon]

Great thanks a lot for the updates! It's good to go now :)