Add recipe for sea ice area and extents in southern polar region

[flicj191]

Description

New recipe for sea ice area trends and extents around Antarctica. Using code from @anton-seaice for a COSIMA (Consortium for Ocean-Sea Ice Modelling in Australia) 'cookbook' example.

• Closes New recipe for sea ice area trends and extents #3606
• Link to documentation:

---

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.

• 🛠 This pull request has a descriptive title
• 🛠 Code is written according to the code quality guidelines
• 🛠 Documentation is available
• 🛠 Tests run successfully
• 🛠 The list of authors is up to date
• 🛠 Any changed dependencies have been added or removed correctly
• 🛠 All checks below this pull request were successful

New or updated recipe/diagnostic

• 🧪 Recipe runs successfully
• 🧪 Recipe is well documented
• 🧪 Figure(s) and data look as expected from literature
• 🛠 Provenance information has been added

To help with the number of pull requests:

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

[flicj191]

@rbeucher see draft pr for sea ice recipe

[flicj191]

this recipe could possibly be moved to examples ?

[flicj191]

@headmetal @rhaegar325 can you also have a look at this when you get a chance?

[bouweandela]

@flicj191 Could you please have a look at the checklist and make sure you've addressed all the items as best as you can? Our convention is that the author of the pull request fills in the checklist before asking for a review. If anything is unclear don't hesitate to ask.

[flicj191]

Codacy issues were too many local variables and a docstring that matches PEP257 from what I can see. Are these acceptable? @ESMValGroup/esmvaltool-developmentteam

[bouweandela]

Yes, a few issues should be fine.

[bouweandela]

the input data file type is a mystery to me - why use xarray here?

The choice for xarray here seems mostly a personal preference and that is absolutely fine!

However, re-implementing existing preprocessor functions in the diagnostic is something that should preferably be avoided.

[valeriupredoi]

The choice for xarray here seems mostly a personal preference and that is absolutely fine!

In general, I tend to agree with this, but this may lead to a number of issues:

• we could end up, in the future, with a constant move away from Iris, with other diagnostic developers relying more and more on xarray; so far only four diagnostics use it, so let's not get too impatient, but I'm starting to notice xarray's much higher popularity vs Iris
• this means using ESMValCore preprocessors will get harder and harder to be used in diags, due to the incompatibility between xarray and Iris IOs

That's why I asked Felicity, if possible, to stick with the oldschool Iris way 🍺

[rbeucher]

I agree with @valeriupredoi and @bouweandela that the preprocessors should handle the heavy lifting. Since they're based on IRIS, we're somewhat limited at the moment. Supporting XArray in the future would be great, though I understand it's a complex task.

[bouweandela]

@rbeucher Please open a GitHub issue if you encounter such limitations. We are applying for funding for future work at the moment, so if we are aware of shortcomings in the preprocessor, we can plan to solve these issues in future projects.

[rbeucher]

Sure, I will. We haven't worked on this recently. Hopefully soon.

[alistairsellar]

Hi @flicj191 if you are still looking for a science reviewer I could ask one of the sea-ice team at the Met Office. I would think that @anton-seaice's review would suffice for science review since he knows what the outputs should look like. But let me know if you think it really needs another science reviewer.

[flicj191]

Hi @flicj191 if you are still looking for a science reviewer I could ask one of the sea-ice team at the Met Office. I would think that @anton-seaice's review would suffice for science review since he knows what the outputs should look like. But let me know if you think it really needs another science reviewer.

Thanks @alistairsellar ! I will do a bit more work on this to test some regridders then will get @anton-seaice to have a look at the outputs so sounds good for that to be the science review. Cheers

[alistairsellar]

Thanks @alistairsellar ! I will do a bit more work on this to test some regridders then will get @anton-seaice to have a look at the outputs so sounds good for that to be the science review. Cheers

Sounds good @flicj191 !

[flicj191]

Hi @bouweandela @valeriupredoi I have restructured the recipe to use the preprocessors ready for re-review

Checking some different regridding preprocessors, This is a plot from previous diagnostic script

with these packages on the main branch I believe:

INFO    [3335387] ESMValCore: 2.11.0.dev8+gb47f7c2ca
INFO    [3335387] ESMValTool: 2.10.0.dev68+gf16c7eab7.d20231122

then with this preprocessor:

  map_extents: &map
    climate_statistics:
      operator: mean
      period: monthly
    extract_region: 
      start_longitude: 0
      end_longitude: 360
      start_latitude: -90
      end_latitude: -50
    regrid:
      target_grid: 0.5x0.5
      scheme: linear

  map_esmpy:
    <<: *map
    regrid:
      target_grid: 0.5x0.5
      scheme:
        reference: esmf_regrid.schemes:ESMFBilinear

then the target_grid to observations seems to do something strange:

  map_target:
    <<: *map
    regrid:
      target_grid: NSIDC-G02202-sh
      scheme: linear

  map_esmpy_target:
    <<: *map
    regrid:
      target_grid: NSIDC-G02202-sh
      scheme:
        reference: esmf_regrid.schemes:ESMFBilinear
        mdtol: 0.7

I'll have a look at the preprocessed files when I can.
@anton-seaice, not sure if there's a specific regridding scheme you would need?

[anton-seaice]

I expect bilinear interpretation is a good choice. The monthly averages of sea ice concentration should be fairly 'smooth' in space, so biliear is appropriate.