Spatial Correlation and Regression Analysis with xarray

[Gyuknj]

I am trying to use xarray for spatial correlation and regression analysis of two fields. Below is the script that have been given me error:

import numpy as np
import pandas as pd
import xarray as xr
import cartopy.crs as ccrs
import cartopy.feature as cfeature

import matplotlib.pyplot as plt
import matplotlib.patheffects as PathEffects

import os
from datetime import datetime

# == netcdf file name and location"
fnc_sst = 'sst.mnmean.nc'
fnc_wnd = 'uwnd10m.mon.mean.nc'
fnc_stress = 'uflx.mon.mean.nc'

# Load datasets
ds_sst = xr.open_dataset(fnc_sst)
ds_wnd = xr.open_dataset(fnc_wnd)
ds_stress = xr.open_dataset(fnc_stress)

# Compute climatological mean and anomaly of sst data
clm = ds_sst.sel(time=slice('1982-01-01','2022-12-01')).groupby('time.month').mean(dim='time')
anm = (ds_sst.groupby('time.month') - clm)

# Compute climatological mean and anomaly of surface wind data
clm_wnd = ds_wnd.sel(time=slice('1982-01-01','2022-12-01')).groupby('time.month').mean(dim='time')
anm_wnd = (ds_wnd.groupby('time.month') - clm_wnd)

# Compute Correlation and Regression between sst anomaly and surface wind anomaly
cor = xr.corr(anm, anm_wnd, dim="time")
reg = xr.cov(anm, anm_wnd, dim="time")/anm.var(dim='time',skipna=True).values

Below is the error I have been getting each time I tried running the "xr.corr" or "xr.cov":

TypeError: Only xr.DataArray is supported.Given [<class 'xarray.core.dataset.Dataset'>, <class 'xarray.core.dataset.Dataset'>].

[pierre-manchon]

Hello Gyuknj,
imo the error message speaks for itself.
Both xr.corr and xr.cov needs two xr.DataArrays as inputs while you try to give them xr.Datasets. In fact, we can see that anm and anm_wnd are read with xr.open_dataset then simply modified with sel and groupby which outputs objects of type xr.Dataset.
I believe what you want is to access the underlying DataArray of your modified values. This can be done very easily by using attribute style access (think ds.property or ds['property']) where property would be one of the keys in ds.data_vars or ds.coords (see)
Pierre