Aggregation functions and other minor fixes

[RondeauG]

Pull Request Checklist:

• pre-commit hooks are installed/active in my local clone ($ pre-commit install)
• This PR addresses an already opened issue (for bug fixes / features)
  • This PR is a prerequisite for Implement ensemble selection #4
• (If applicable) Documentation has been added / updated (for bug fixes / features)
• If a merge request has been made in parallel to this PR in xscen-notebooks, it is merged and the submodules have been updated.
• HISTORY.rst has been updated (with summary of main changes)
  • Link to issue (:issue:number) and pull request (:pull:number) has been added

What kind of change does this PR introduce?

Main changes

• Major API changes from another branch.
• New file: aggregate.py
  • aggregate.climatological_mean computes rolling means over a specified window.
  • aggregate.deltas computes deltas.
  • aggregate.spatial_mean computes a spatial means using either .mean(), .interp() or xesmf.SpatialAverager()
• compute_indicators now supports disjointed time periods.

Minor changes

Does this PR introduce a breaking change?

• Both shapely and geopandas would be requirements for xscen. Are we fine with that?

Other information:

[juliettelavoie]

Cool!
Are you planning on adding a section to GettingStarted ?
I will try a few tests of your functions on my data.

[RondeauG]

Cool! Are you planning on adding a section to GettingStarted ? I will try a few tests of your functions on my data.

I was thinking of making a specific Notebook with aggregate and reduce, focusing on their use for ensemble selection. I was waiting until everything is ready, but I could start working on it if you want to.

[juliettelavoie]

I agree that reduce should be in its own notebook. But, I feel like climatological mean and deltas have their place in Getting Started between indicators and ensemble.

[juliettelavoie]

For compute_deltas, the time axis is missing the horizon of reference. Is this on purpose ? It will always be 0 (or 1), but I would be encline to keep it.

1. It is better to have regular axis without holes.
2. We need it for ccdp.
3. It is a good validation of what the data is.

[RondeauG]

For compute_deltas, the time axis is missing the horizon of reference. Is this on purpose ? It will always be 0 (or 1), but I would be encline to keep it.

1. It is better to have regular axis without holes.
2. We need it for ccdp.
3. It is a good validation of what the data is.

It was on purpose since I didn't see the point of keeping it, but I can change that since you have a use case.

[RondeauG]

In the code that I copied for compute_deltas, the variables are always renamed. For example, growing_degree_days becomes growing_degree_days_delta_1981_2010. What do we think of that? Should we make it an option?

[juliettelavoie]

In the code that I copied for compute_deltas, the variables are always renamed. For example, growing_degree_days becomes growing_degree_days_delta_1981_2010. What do we think of that? Should we make it an option?

Making it optional is okay with me. The variable was always renamed because it was always in the same dataset has the 30y mean (no delta). But, in xscen, the mean and the deltas are separate.

[juliettelavoie]

In the code that I copied for compute_deltas, the variables are always renamed. For example, growing_degree_days becomes growing_degree_days_delta_1981_2010. What do we think of that? Should we make it an option?

Making it optional is okay with me. The variable was always renamed because it was always in the same dataset has the 30y mean (no delta). But, in xscen, the mean and the deltas are separate.

[RondeauG]

This PR should be OK for a new round of reviews. I also updated the notebooks.

[aulemahal]

Je n'ai pas testé les fonctions, mais ça me semble beau!

[juliettelavoie]

looks good!