Converting WRF field to an iris dataset fails

[kelcyno]

What happened?

When using Xarray's to iris:

field_iris=field_in.to_iris()

the following error is thrown: ValueError: [UT_SUCCESS] Failed to parse unit

A work around is to drop 'XTIME' and 'Times' from field_in. This allows to_iris to pick up on the cf-compliant 'Time'.

Minimal Complete Verifiable Example

vars = ['COMDBZ']
files = sorted(glob.("wrf2dout_d02_2021*.nc"))
def preprocess(ds):
    return ds[vars]

ds = xr.open_mfdataset(
    files,
    engine="xwrf",
    parallel=True,
    concat_dim="Time",
    combine="nested",
    preprocess=preprocess,
    chunks={'Time': 1},decode_times=False
)
#Time originally has an integer, while Times is a datetime.  
#I need the datetime
ds['Time'] = ds['Times'] 
comdbz = ds['COMDBZ']
comdbz = comdbz.drop('XTIME')
comdbz = comdbz.drop('Times')
field_iris=comdbz.to_iris()
data_time=field_iris.slices_over('Time')

Relevant log output

No response

Version

main

Which Python version are you using?

3.8

Anything else we need to know?

This isn't a direct bug so much as a way to use wrf data with the Xarray to_iris functionality.

[lpilz]

I think this is the same topic we are discussing in #31 and #11. At the moment it seems the preferred option is to keep XTIME as a coordinate but have it as np.timedelta64 datatype

[kmpaul]

This is definitely a question of how best to import WRF data into Iris, which I don't really think is in xWRF's remit. But perhaps we can come up with a good solution nonetheless.

What preprocessing is usually required to import WRF data into Iris? What happens if XTIME is not a coordinate, but still in the dataset?

My experience with Iris has been that it expects very specific metadata and coordinates in the dataset. Xarray is more general. Perhaps this is something that needs to be improved upstream in Xarray's to_iris() method?