Refactor robustness methods

[aulemahal]

Pull Request Checklist:

• This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes #xyz
• Tests for the changes have been added (for bug fixes / features)
  • (If applicable) Documentation has been added / updated (for bug fixes / features)
• CHANGES.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?

• Add ipcc-advanced as a test to ensembles.change_significance. This corresponds to the simpler alternative of approach C as described in the Cross-Chapter Box 1 of the IPCC Atlas (AR6, WG1).
• Add two examples in the "Ensembles" notebook to show how xclim can be used to get the IPCC-recommended hatching masks.
• Add a very simple detrend function.

Does this PR introduce a breaking change?

No, but it might, see below.

Other information:

As you can see in the examples, the IPCC decouples the "change significance" from the "sign agreement", while xclim does not. This make the second example more complex than expected.

I thought of :

• Changing the output of change_significance so that pos_frac is independent of change significance
• Adding a third output so we have: significant_change_frac, significant_positive_change_frac, positive_change_frac.
• Adding a keyword argument to switch between the two modes.
  First two are breaking changes while the thirds feels overly heavy for something this simple...

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[huard]

• Looks good.
• The nb text assumes the reader's familiarity with the topic and the IPCC approach. I suggest to take a more pedagogic approach.
• I don't see a test ; )

[aulemahal]

@huard After our discussions and some thinking, I rewrote the function completely.

I was not satisfied with the output of change_significance so I rewrote it as robustness_fractions. The former still works but it will warn the user that it is deprecated. Also added robustness_categories to help with making the actual map.

Finally, I simplified the notebook a lot. May be too much...

I'll add tests tomorrow.

[aulemahal]

@RondeauG , would you have time to look at the new functions ? And may be @mccrayc, if you have time.

I'm simply looking for a review of the general idea and usage, as shown in the example : https://xclim--1514.org.readthedocs.build/en/1514/notebooks/ensembles.html#Change-significance-and-model-agreement

[coveralls]

coverage: 90.435%. first build
when pulling 930c272 on ipcc-robustness
into 3406661 on master.

[RondeauG]

Quelques détails de syntaxe ou typo:

• surimposed --> superimposed
• change significance : whether most of the ensemble members project a statistically significant climate change signal, in comparison to their internal variability.
• There are be 5 different members in the ensemble

(models agree and all show significant change)

Je dirais plutôt enough show a significant change. L'IPCC utilise un seuil de X% (60?).

Reply via ReviewNB

[huard]

Ça me semble beaucoup plus clair comme ça. Pas testé en vrai, juste lu en diagonale.

[RondeauG]

This looks pretty good! robustness_categories will be very useful.