def test_xr_rm_trend_2d_dataarray(two_dim_da):
two_dim_da_dt = xr_rm_trend(two_dim_da, 'col')
assert two_dim_da_dt.shape == (4, 5)
# should have all values near 0 after detrending because it's linear
# but because of floating point precision, may not be 0
assert (two_dim_da_dt <= 1e-5).sum() == 20
def test_xr_rm_trend_3d_dataset_dim_order(multi_dim_ds):
multi_dim_ds = multi_dim_ds.transpose('lon', 'time', 'lat')
multi_dim_ds_dt = xr_rm_trend(multi_dim_ds)
# ensure the dims are back in its original state
assert list(multi_dim_ds_dt['air'].dims) == ['lon', 'time', 'lat']
assert list(multi_dim_ds_dt['airx2'].dims) == ['lon', 'time', 'lat']
def test_xr_rm_trend_3d_dataset(multi_dim_ds):
multi_dim_ds_dt = xr_rm_trend(multi_dim_ds)
# originally values were ~270 K, after detrending between -50 and 50
assert multi_dim_ds_dt['air'].shape == (2920, 25, 53)
assert multi_dim_ds_dt['airx2'].shape == (2920, 25, 53)
assert float(multi_dim_ds_dt['air'].min()) > -50
assert float(multi_dim_ds_dt['air'].max()) < 50
def test_xr_rm_trend_1d_dataarray(two_dim_da):
one_dim_da = two_dim_da.isel(row=0)
one_dim_da_dt = xr_rm_trend(one_dim_da, 'col')
assert one_dim_da_dt.shape == (5, )
# should have all values near 0 after detrending because it's linear
# but because of floating point precision, may not be 0
assert (one_dim_da_dt <= 1e-5).sum() == 5
def test_xr_rm_trend_2d_dataarray_interp_nan(two_dim_da):
two_dim_da[2, :] = np.nan
two_dim_da_dt = xr_rm_trend(two_dim_da, 'col')
assert two_dim_da_dt.shape == (4, 5)
# should have 15 values (5 NaNs) near 0 after detrending because it's
# linear. But because of floating point precision, it may not be 0.
assert (two_dim_da_dt <= 1e-5).sum() == 15
assert np.isnan(two_dim_da_dt[2]).all() # should be replaced with nan
def test_xr_rm_trend_1d_dataarray_interp_nan(two_dim_da):
one_dim_da = two_dim_da.isel(row=0)
one_dim_da[:3] = np.nan
one_dim_da_dt = xr_rm_trend(one_dim_da, 'col')
assert one_dim_da_dt.shape == (5, )
assert np.isnan(one_dim_da_dt[:3]).all() # should be replaced with nan
# since it's bfill, the linear trend no longer holds
assert (one_dim_da_dt <= 1e-5).sum() <= 20
def test_xr_rm_trend_missing_dim():
with pytest.raises(KeyError) as excinfo:
xr_rm_trend(xr.DataArray([0, 1, 2]), dim='non_existent')
assert "Input dim, 'non_existent'" in excinfo.value.message
