Losing attributes of coords when using weights in ensemble stats

[juliettelavoie]

Setup Information

• Xclim version: '0.38.1-beta'
• Python version: 3.9
• Operating System: mac

Description

When I add the weights argument to ensemble statistics, I lose the attributes of my coordinates.

I think this is related to the fact that the weights dataset only has a realization coord.

Steps To Reproduce

with weights:

datasets = pcat.search(processing_level= 'delta-+1.5C').to_dataset_dict(
                        xarray_open_kwargs={'decode_timedelta':False})

weights = xs.ensembles.generate_weights(datasets=datasets)
ens = xc.ensembles.create_ensemble(datasets)
stats= xc.ensembles.ensemble_mean_std_max_min(ens, weights=weights)
print(stats.horizon.attrs)
>>> {}

without weights:

datasets = pcat.search(processing_level= 'delta-+1.5C').to_dataset_dict(
                        xarray_open_kwargs={'decode_timedelta':False})
ens = xc.ensembles.create_ensemble(datasets)
stats= xc.ensembles.ensemble_mean_std_max_min(ens)
print(stats.horizon.attrs)
>>> {'baseline': '1850-1900'}

Additional context

This is also an issue for ensemble_percentiles.

Contribution

• I would be willing/able to open a Pull Request to address this bug.

Code of Conduct

• I agree to follow this project's Code of Conduct