Is it possible to change a coordinate in an netCDF loaded with xarray?

Question

I have a netCDF which is loaded in xarray with a dimension named bands (it was originally an import via rioxarray of ENVI data), but actually, I want to be able to parse the data by time. I have tried this:

And indeed is created a "time" coordinate in the dataset, but is kind of useless when I try to parse the data as such:

ds.sel(time="1982-01-31")['leafleaf_area_index_GLASS'].where(ds.leaf_area_index_GLASS > 20).plot()

Any suggestions on how can I make this time coordinate to work? I'm not sure if there is an import option with xarray in which I could map directly the band into a datetime coordinate.

Answer 1

You created a new coordinate but it also created a new dimension instead of using the "band" dimension. @Bert Coever's suggestion should work. Also note that every operation creates a new dataset, therefore one needs to assign it back to the dataset:

ds= ds.assign_coords({"band" : pd.Series(pd.date_range ... ...)})
ds= ds.rename_dims({"band" : "time"})

Answer 2

Something like this should work:

time = pd.date_range("1982-01-01", periods=408, frequ="M")
ds.assign_coordinates(band=("band",time)).rename(band="time")

The way it works is that you should specify to xarray what is the dimension to this new coordinate. Then you rename the coordinate and dimension to time.

Here below I give you my reproducible code snippet for my example

import xarray as xr
import pandas as pd
import numpy as np

n = 408
ds = xr.Dataset(dict(param=("band", np.random.randn(n))), coords=dict(band=np.arange(n)))

time = pd.date_range("1982-01-01", periods=n, freq="M")
ds = ds.assign_coords(band=("band",time)).rename(band="time")

ds

Which returns the following: