Datasets failing on get_bbox

[ellesmith88]

Some datasets for c3s-cordex are failing on get_bbox:

catalog-maker/catalog_maker/catalog.py

Lines 49 to 67 in 4ba705a

	def get_bbox(ds):
	lat = get_coord_by_type(ds, "latitude", ignore_aux_coords=False)
	lon = get_coord_by_type(ds, "longitude", ignore_aux_coords=False)
	min_y, max_y = get_coord_info(lat)
	min_x, max_x = get_coord_info(lon)
	if min_y < -90 or max_y > 90:
	raise Exception(
	f"Latitude is not within expected bounds. The minimum and maximum are {min_y}, {max_y}"
	)
	if min_x < -360 or max_x > 360:
	raise Exception(
	f"Longitude is not within expected bounds. The minimum and maximum are {min_x}, {max_x}"
	)
	bbox = f"{min_x:.2f}, {min_y:.2f}, {max_x:.2f}, {max_y:.2f}"
	return bbox

with the error AttributeError: 'NoneType' object has no attribute 'values'

An example file is /gws/nopw/j04/cp4cds1_vol1/data/c3s-cordex/output/EUR-11/CNRM/CNRM-CERFACS-CNRM-CM5/rcp45/r1i1p1/CNRM-ALADIN63/v2/day/sfcWind/v20190828/sfcWind_EUR-11_CNRM-CERFACS-CNRM-CM5_rcp45_r1i1p1_CNRM-ALADIN63_v2_day_20360101-20401231.nc

Looking into these files I see that lat/lon are data variables rather than coordinates (with, for example, x and y and time only as coordinates), where as previously all datasets have had latitude and longitude as coordinates.

So our function get_coord_by_type (https://github.com/roocs/roocs-utils/blob/152f33650a44fe3466ffbc8dbd99ebb9f8de783e/roocs_utils/xarray_utils/xarray_utils.py#L235-L260) doesn't find them.

I can fix this by using ds.cf['latitude'] and ds.cf['longitude'] - when I use these I can generate the bbox.

However, we use get_coord_by_type in several places, e.g. in clisops, so this would need to be updated throughout, otherwise these datasets would fail elsewhere.

@agstephens So, I can include the datasets in the catalog with their correct bbox and fix the issue in the other packages. Maybe the best way would be to update the get_coord_by_type function to use cf-xarray?

[agstephens]

Hi @ellesmith88: I think that I am in favour of updating get_coord_by_type to use ds.cf[...] as long as it is doing the right thing. Please can you review how the cf_xarray library is managing to find these coordinates. Please report your findings here and we can decide how to proceed. Thanks

[ellesmith88]

• ds.cf returns

Coordinates:
- CF Axes: * X: ['x']
           * Y: ['y']
           * T: ['time']
             Z: n/a

- CF Coordinates: * time: ['time']
                    longitude, latitude, vertical: n/a

- Cell Measures:   area, volume: n/a

- Standard Names: * projection_x_coordinate: ['x']
                  * projection_y_coordinate: ['y']
                  * time: ['time']

Data Variables:
- Cell Measures:   area, volume: n/a

- Standard Names:   latitude: ['lat']
                    longitude: ['lon']
                    wind_speed: ['sfcWind']

• Data variables are identified here https://github.com/xarray-contrib/cf-xarray/blob/31ceff2e71db7646062c74a3488a8045bd81f5c5/cf_xarray/accessor.py#L1280-L1282
  simply by using 'ds.data_vars'
• The standard name and variable names are identified within the method make_text_section
• ds.cf['latitude'] and ds.cf['longitude'] then use Dataset.__getitem__ : https://github.com/xarray-contrib/cf-xarray/blob/31ceff2e71db7646062c74a3488a8045bd81f5c5/cf_xarray/accessor.py#L1843
• which uses _getitem https://github.com/xarray-contrib/cf-xarray/blob/31ceff2e71db7646062c74a3488a8045bd81f5c5/cf_xarray/accessor.py#L603
• This is where I can't fully work out whats going on, but I think it just uses the names found for the above output and indexes the dataset.