def filter_2D(data, std, dim, dtype=None):
if astropy is None:
raise RuntimeError(
"Module `astropy` not found. Please install optional dependency with `conda install -c conda-forge astropy"
)
if dtype is None:
dtype = detect_dtype(data)
kernel = Gaussian2DKernel(std)
def smooth_raw(data):
raw_data = getattr(data, "values", data)
result = convolve_fft(raw_data, kernel, boundary="wrap")
result[np.isnan(raw_data)] = np.nan
return result
def smoother(data):
dims = dim
# this is different from the 1d case
return xr.apply_ufunc(
smooth_raw,
data,
vectorize=True,
dask="parallelized",
input_core_dims=[dims],
output_core_dims=[dims],
output_dtypes=[dtype],
)
return smoother(data)
def filter_2D(data, std, dim, dtype=None):
if dtype is None:
dtype = detect_dtype(data)
kernel = Gaussian2DKernel(std)
def smooth_raw(data):
raw_data = getattr(data, "values", data)
result = convolve_fft(raw_data, kernel, boundary="wrap")
result[np.isnan(raw_data)] = np.nan
return result
def smoother(data):
dims = dim
# this is different from the 1d case
return xr.apply_ufunc(
smooth_raw,
data,
vectorize=True,
dask="parallelized",
input_core_dims=[dims],
output_core_dims=[dims],
output_dtypes=[dtype],
)
return smoother(data)
def filter_1D(data, std, dim="time", dtype=None):
if dtype is None:
dtype = detect_dtype(data)
kernel = Gaussian1DKernel(std)
def smooth_raw(data):
raw_data = getattr(data, "values", data)
result = convolve_fft(raw_data, kernel, boundary="wrap")
result[np.isnan(raw_data)] = np.nan
return result
def temporal_smoother(data):
dims = [dim]
return xr.apply_ufunc(
smooth_raw,
data,
vectorize=True,
dask="parallelized",
input_core_dims=[dims],
output_core_dims=[dims],
output_dtypes=[dtype],
)
return temporal_smoother(data)
def xr_linregress(a, b, dim="time", convert_to_dataset=True, dtype=None, nanmask=False):
"""Applies scipy.stats.linregress over two xr.DataArrays or xr.Datasets.
Parameters
----------
a : {xr.DataArray}
Independent variable for linear regression. E.g. time.
b : {xr.DataArray, xr.Dataset}
Dependent variable.
dim : str
Dimension over which to perform linear regression.
Must be present in both `a` and `b` (the default is 'time').
convert_to_dataset : bool
Converts the output parameter to data_variables instead of an
additional dimension (the default is True).
dtype : dtype
Dtype for the output. If None, defaults to dtype of `b`, or `b`s
first data variable(the default is None).
nanmask: bool
optional masking of nans in the data to avoid nan output from
regression (the default is False)
Returns
-------
type(b)
Returns a dataarray containing the parameter values of
scipy.stats.linregress for each data_variable in `b`.
"""
if dtype is None:
dtype = detect_dtype(b)
stats = xr.apply_ufunc(
_linregress_ufunc,
a,
b,
nanmask,
input_core_dims=[[dim], [dim], []],
output_core_dims=[["parameter"]],
vectorize=True,
dask="parallelized",
output_dtypes=[dtype],
output_sizes={"parameter": 5},
)
stats["parameter"] = xr.DataArray(
["slope", "intercept", "r_value", "p_value", "std_err"], dims=["parameter"]
)
if convert_to_dataset:
# chug them all into a dataset
ds_stats = xr.Dataset()
for ff in stats.parameter.data:
ds_stats[ff] = stats.sel(parameter=ff)
out = ds_stats
else:
out = stats
return out