ds.spatial.get_weights used to work nicely, but does not work anymore

[bonfils2]

ds = xc.open_dataset('/p/css03/esgf_publish/CMIP6/ScenarioMIP/IPSL/IPSL-CM6A-LR/ssp585/r2i1p1f1/Amon/tas/gr/v20191121/tas_Amon_IPSL-CM6A-LR_ssp585_r2i1p1f1_gr_201501-210012.nc',add_bounds=True,decode_times=True,center_times=True)
ds = ds.temporal.group_average("tas", freq="year", weighted=True)
ds.lat.attrs["axis"] = "Y"
ds.lon.attrs["axis"] = "X"
weights4 = ds.spatial.get_weights(axis=['Y', 'X'], lat_bounds=[-90,90],lon_bounds=[0,360]) #*m0x
ts_global = ds.spatial.average("tas", axis=["Y", "X"],weights=weights4)["tas"] #.load().data

---> 7 weights4 = ds.spatial.get_weights(axis=['Y', 'X'], lat_bounds=[-90,90],lon_bounds=[0,360]) #*m0x

File ~/my_envs/xcdatsimple/lib/python3.10/site-packages/xcdat/spatial.py:274, in SpatialAccessor.get_weights(self, axis, lat_bounds, lon_bounds, data_var)
272 axis_weights: AxisWeights = {}
273 for key in axis:
--> 274 d_bounds = self._dataset.bounds.get_bounds(axis=key, var_key=data_var)
276 if isinstance(d_bounds, xr.Dataset):
277 raise TypeError(
278 "Generating area weights requires a single bounds per "
279 f"axis, but the dataset has multiple bounds for the '{key}' axis "
280 f"{list(d_bounds.data_vars)}. Pass a data_var key "
281 "to reference a specific data variable's axis bounds."
282 )

File ~/my_envs/xcdatsimple/lib/python3.10/site-packages/xcdat/bounds.py:218, in BoundsAccessor.get_bounds(self, axis, var_key)
215 bounds_keys = []
217 if len(bounds_keys) == 0:
--> 218 raise KeyError(
219 f"No bounds data variables were found for the '{axis}' axis. Make sure "
220 "the dataset has bound data vars and their names match the 'bounds' "
221 "attributes found on their related time coordinate variables. "
222 "Alternatively, you can add bounds with xcdat.add_missing_bounds "
223 "or xcdat.add_bounds."
224 )
226 bounds: Union[xr.Dataset, xr.DataArray] = self._dataset[
227 bounds_keys if len(bounds_keys) > 1 else bounds_keys[0]
228 ].copy()
230 return bounds

KeyError: "No bounds data variables were found for the 'Y' axis. Make sure the dataset has bound data vars and their names match the 'bounds' attributes found on their related time coordinate variables. Alternatively, you can add bounds with xcdat.add_missing_bounds or xcdat.add_bounds."

[pochedls]

This code works for me in xcdat 0.5.0. Is this the version you're on (print xc.__version__)?

Also – do you know that you can pass the bounds in spatial averager and have it compute the weights on its own?

# note that add_bounds/decode_times are True by default (I dropped it)
ds = xc.open_dataset('/p/css03/esgf_publish/CMIP6/ScenarioMIP/IPSL/IPSL-CM6A-LR/ssp585/r2i1p1f1/Amon/tas/gr/v20191121/tas_Amon_IPSL-CM6A-LR_ssp585_r2i1p1f1_gr_201501-210012.nc', center_times=True)
# note that weighted is True by default (I dropped it)
ds = ds.temporal.group_average("tas", freq="year")
weights = ds.spatial.get_weights(axis=['Y', 'X'], lat_bounds=[-90,90],lon_bounds=[0,360])
ts_global_user_weights = ds.spatial.average("tas", axis=["Y", "X"],weights=weights)["tas"]

# or compute directly
# note that this is a global domain, so you don't actually need to specify the bounds
# in this case (but I do to show the syntax)
ts_global_direct = ds.spatial.average("tas", axis=['Y', 'X'], lat_bounds=(-90, 90), lon_bounds=(0, 360))["tas"]

print(np.allclose(ts_global_user_weights, ts_global_direct))

True

[bonfils2]

You are right @lee1043, you found the origin of the problem : applying an older version of xarray ("conda install -c conda-forge xarray=2023.3.0") works for me!

[bonfils2]

@pochedls @lee1043 thanks both for your help!!

[bonfils2]

I submitted a bug report : #458

[tomvothecoder]

Thanks @bonfils2! We can continue this discussion in #458 as needed.