Trend calculation in xarray

[kuchaale]

There are basically two convenient ways how to calculate trends in xarray:

• you vectorize stats.linregress function using xr.apply_ufunc:

from scipy import stats

x = np.arange(da.time.shape[0])

def new_linregress(y):
    # Wrapper around scipy linregress to use in apply_ufunc
    slope, intercept, r_value, p_value, std_err = stats.linregress(x, y)
    return np.array([slope, intercept, r_value, p_value, std_err]

stats = xr.apply_ufunc(new_linregress, da, 
                       input_core_dims=[['time']],
                       output_core_dims=[["parameter"]],
                       vectorize=True,
                       dask="parallelized",
                       output_dtypes=['float64'],
                       output_sizes={"parameter": 5},
                      )

• or wait a while since polyfit should be available in the next release (Implementation of polyfit and polyval pydata/xarray#3733, AttributeError: 'DataArray' object has no attribute 'polyfit' pydata/xarray#3902). However, it will not allow calculating things such as p-values.