Automatically fix small deviations in vertical levels

[bouweandela]

Description

This pull request aims to make working with vertical levels that have tiny errors in the coordinate values easier. It does this by:

1. trying to automatically fix the problem
2. setting the levels equal to those requested in extract_levels when cube slicing is used instead of interpolation

This will make it easier to compute multi-model statististics, which now need exactly matching coordinates.

See #956 (comment)

---

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

It is the responsibility of the author to make sure the pull request is ready to review. The icons indicate whether the item will be subject to the 🛠 Technical or 🧪 Scientific review.

• 🧪 The new functionality is relevant and scientifically sound
• 🛠 This pull request has a descriptive title and labels
• 🛠 Code is written according to the code quality guidelines
• 🧪 and 🛠 Documentation is available
• 🛠 Unit tests have been added
• 🛠 Changes are backward compatible
• 🛠 The list of authors is up to date
• 🛠 All checks below this pull request were successful

---

To help with the number pull requests:

• 🙏 We kindly ask you to review two other open pull requests in this repository

[valeriupredoi]

this is cool! maybe we should make the tolerances a user tunable parameter? I can see uses for in a configuration file

[bouweandela]

I'm not sure. Do you think the values that I've chosen now make sense?

[zklaus]

I think in this case it's better if we control the tolerances. Note also that it might be tempting to do a similar thing for horizontal coordinates, where e.g. in CORDEX very small deviations can occur. The difference is that there is no "right" value in CORDEX, this is decided by the model, whereas here we have an authoritative value from the CMIP tables. This is also what allows us to decide on tolerances, imho.

[bouweandela]

Note also that it might be tempting to do a similar thing for horizontal coordinates, where e.g. in CORDEX very small deviations can occur.

Indeed, this was implemented in #507 with the default tolerances. The default relative tolerance might be a bit lax (rtol 1e-5, atol 1e-8).

[bouweandela]

I think in this case it's better if we control the tolerances.

I could add that for extract_levels. What would you propose as the default? The current values?

[zklaus]

I think the current values look fine to me, but let's have another look for the next release.

[codecov]

Codecov Report

Merging #1177 (076243a) into main (43e6b16) will increase coverage by 0.27%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1177      +/-   ##
==========================================
+ Coverage   87.47%   87.74%   +0.27%     
==========================================
  Files         193      193              
  Lines        9649     9683      +34     
==========================================
+ Hits         8440     8496      +56     
+ Misses       1209     1187      -22     

Impacted Files	Coverage Δ
esmvalcore/_recipe.py	93.19% <100.00%> (+0.97%)	⬆️
esmvalcore/cmor/check.py	97.71% <100.00%> (+<0.01%)	⬆️
esmvalcore/experimental/recipe_info.py	79.59% <100.00%> (+14.37%)	⬆️
esmvalcore/experimental/recipe_output.py	77.51% <100.00%> (+3.36%)	⬆️
esmvalcore/preprocessor/_regrid.py	96.34% <100.00%> (+0.88%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c77474d...076243a. Read the comment docs.

[bouweandela]

The remaining Codacy issue is about too many arguments for the extract_levels preprocessor function, so unfortunately it cannot be solved.

[bouweandela]

@zklaus I added the option to tweak the parameters as requested. Could you have another look, please?

[valeriupredoi]

nice add-on @bouweandela - do you want to add a couple words about it in the documentation (other than the API documentation)?

[bouweandela]

Thank you for reviewing @valeriupredoi! @zklaus Do you have any further comments on this?