xarray change coordinate to variable? · python-questions

Stream: python-questions

Topic: xarray change coordinate to variable?

Stephen Yeager (Mar 18 2021 at 18:08):

I want to change 'time' from a coordinate to a Data variable in an xarray Dataset. How do I do it?

Anderson Banihirwe (Mar 18 2021 at 18:21):

There's a .reset_coords(): https://xarray.pydata.org/en/stable/generated/xarray.Dataset.reset_coords.html method on both dataset and datarray objects.

In [3]: ds
Out[3]:
<xarray.Dataset>
Dimensions:  (time: 36, x: 275, y: 205)
Coordinates:
  * time     (time) object 1980-09-16 12:00:00 ... 1983-08-17 00:00:00
    xc       (y, x) float64 ...
    yc       (y, x) float64 ...
Dimensions without coordinates: x, y
Data variables:
    Tair     (time, y, x) float64 ...

Invoking this method on a non-dimension coordinate (those without the *) works:

In [10]: ds.reset_coords(["xc"])
Out[10]:
<xarray.Dataset>
Dimensions:  (time: 36, x: 275, y: 205)
Coordinates:
  * time     (time) object 1980-09-16 12:00:00 ... 1983-08-17 00:00:00
    yc       (y, x) float64 ...
Dimensions without coordinates: x, y
Data variables:
    Tair     (time, y, x) float64 ...
    xc       (y, x) float64 ...

However, if your coordinate is a dimension coordinate, xarray doesn't allow resetting this coordinate to a regular variable:

In [11]: ds.reset_coords(["time"])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-11-53667df9faf0> in <module>
----> 1 ds.reset_coords(["time"])

~/opt/miniconda/envs/playground/lib/python3.9/site-packages/xarray/core/dataset.py in reset_coords(self, names, drop)
   1586             bad_coords = set(names) & set(self.dims)
   1587             if bad_coords:
-> 1588                 raise ValueError(
   1589                     "cannot remove index coordinates with reset_coords: %s" % bad_coords
   1590                 )

ValueError: cannot remove index coordinates with reset_coords: {'time'}

Anderson Banihirwe (Mar 18 2021 at 18:29):

I'm curious... what do you hope to accomplish once you've set the time coordinate to a regular variable?

Stephen Yeager (Mar 18 2021 at 18:45):

Thanks! That worked. This is for the SMYLE-analysis repo, for which there is a need for initialized prediction Datasets that are dimensioned as follows: (Y,L,M,...) where Y is start time, L is lead time, and M is ensemble member. It's useful to have the original "time" coordinate converted to a data variable that acquires dimensions (Y,L). You can have a look after I upload the finished notebook.

Matt Long (Mar 18 2021 at 18:47):

I think we prototyped an intake-esm approach to reading DPLE datasets

Deepak Cherian (Mar 18 2021 at 19:02):

One place to check these kinds of questions: https://xarray.pydata.org/en/latest/howdoi.html

Stephen Yeager (Mar 18 2021 at 21:25):

I think we prototyped an intake-esm approach to reading DPLE datasets

This will be good to revisit for SMYLE-analysis. For now, I'm finding it helpful to learn how to do the necessary data manipulations.

Last updated: Jan 30 2022 at 12:01 UTC