Add notebook to validate spatial averaging

[tomvothecoder]

Validation Checklist

1. API usage (and plots where helpful)
2. Performance metrics (use timeit)
3. Bugs, improvements
4. Questions

Compare below APIs against CDAT equivalent.

• ds.xcdat.spatial_average() or ds.spatial.avg()

[tomvothecoder]

Hi @lee1043, I noticed you already started work on this issue so I assigned you to it. Steve and I can review the PR whenever you're ready to open one.

[lee1043]

Hi @tomvothecoder, yes I started playing with it in this notebook, which eventually can be used for demo. I'd be happy to take lead on that. I will probably write a separate (but similar) notebook that is dedicated for the validation.

I have some questions.

1. I think I understand that the spatial average should be conducted on dataset (as in cell [19] of the notebook), not data array (which I hoped for doing such as in cell [17] and [18] of the notebook). I was hopping for the later case because the data array is a generated one (land masking applied to the given data array of the dataset). Maybe a workaround to be define a new dataset by using the new data array, then apply spatial average to the new dataset?

2. I think cdms assigns weighting for spatial average as a default while xcdat does not. Should we consider providing default weightings in xcdat for spatial averaging?

[pochedls]

1. I think I understand that the spatial average should be conducted on dataset (as in cell [19] of the notebook), not data array (which I hoped for doing such as in cell [17] and [18] of the notebook). I was hopping for the later case because the data array is a generated one (land masking applied to the given data array of the dataset). Maybe a workaround to be define a new dataset by using the new data array, then apply spatial average to the new dataset?

Yes - you can create a new masked dataset and then call the spatial averager.

2. I think cdms assigns weighting for spatial average as a default while xcdat does not. Should we consider providing default weightings in xcdat for spatial averaging?

xcdat - weights data (by area) by default [Note - I now see the confusion; if you do not pass in a weighting matrix (i.e., default None), then it will calculate the spatial weights for you].

[tomvothecoder]

I think cdms assigns weighting for spatial average as a default while xcdat does not. Should we consider providing default weightings in xcdat for spatial averaging?

xcdat - weights data (by area) by default [Note - I now see the confusion; if you do not pass in a weighting matrix (i.e., default None), then it will calculate the spatial weights for you].

To avoid confusion, we can either rename the method param from weights to custom_weights (or some other clearer name), or set the default value to weights="generate" instead of None.

I'm lean towards custom_weights=None because it is clearer that the weights are user-specified (with a tradeoff in verbosity). I'm open to other opinions though.

[tomvothecoder]

To avoid confusion, we can either rename the method param from weights to custom_weights (or some other clearer name), or set the default value to weights="generate" instead of None.

Issue opened to address this: xCDAT/xcdat#147

[tomvothecoder]

This issue is from almost 3 years ago and does not need to be addressed anymore.