Clip WRF data using shapefile

[gabrielops84]

Hello fellows! How is it going ?

I'd like to clip precipitation data from WRF output file. Usually I deal with datasets in WGS84 datum where lat lon are 1D arrays coordintes of dataset. Then I apply the following commands using geopandas, xarray and rioxarray:

ds = ds.rio.set_spatial_dims(x_dim=lon_str, y_dim=lat_str, inplace=True)
ds = ds.rio.write_crs('epsg:4326', inplace=True)
ds = ds.rio.clip(geo.geometry.apply(mapping), 'epsg:4326', drop=True)

Where geo is a gemetry shapefile and lat_str and lon_str are labeled names of lat long coordinates and dimensions of dataset like this one:

I'd like to do it on my WRF output file:

But XLAT and XLONG are 3D arrays on lambert conformal conic projection:
'+proj=lcc +lat_0=-23.6999893188477 +lon_0=-43.5 +lat_1=-0.5 +lat_2=-45 +x_0=0 +y_0=0 +R=6370000 +units=m +no_defs'

I'd like to reproject data to WGS84 datum and set latitude and longitude as spatial dims to clip space over time. My WRF file has irregular grid and I think I need to interpolate to a regular grid. I've been searching about projections using pyproj and interpolate using salem. I've found some solutions to plot data but nothing about clip data correctly.

rioxarray.clip method works with 'x' and 'y' as dimension labeled 1D arrays. Can you guys help ? Thanks!

[jthielen]

The following is not tested, but instead what I would guess to be the case based on the rioxarray docs. Feel free to try it out and report back with any issues!

By using xwrf's postprocess, CF-compliant x/y coordinate and grid mapping information will be added to your dataset. rioxarray then automatically interprets that information internally, including the differences in projections/datums between xarray data and the clipping geometry. So, your WRF data example should be able to look like:

ds_wrf_processed = ds_wrf.xwrf.postprocess()
ds_wrf_clipped = ds_wrf_processed.rio.clip(geo.geometry.apply(mapping), 'epsg:4326', drop=True)

[gabrielops84]

@jthielen Thanks for the advice! But it didn't work. I still have 2D for x and y dimensions and rioxarray.clip works with 1D latitude/y longitude/x arrays.
I worked around applying interpolation using xesmf.Regridder.

First I created a zero values Dataset ds_b with 1D lat/lon arrays from shapefile bounds;
Then I used ds_b grid to interpolate my WRF Dataset:

I started from here:

And got this result (bilinear method):

So I could clip my dataset using wgs84 datum:

I'll check the results and this warning after use the regridder object. At first looking it seems ok. If you have some tip let me know!