Add smoothing to remapping of SSS restoring

[xylar]

This requires a major reorganization of the step to be more like remap_topography after its recent reorganization.

Remapping is performed with mbtempest. The public function remap_sss() has been removed. (It was not mentioned in the docs.)

This merge also adds a remapping step to the utility that generates the SSS field from WOA23. The data is remapped to a cubed-sphere grid at ne300 (~10 km) resolution. Remapping from the cubed sphere to MPAS meshes will be much more performant than remapping from the WOA original lat-lon grid, particularly when there is smoothing near the poles as we do here.

The smoothing goes linearly from zero at 70 N to 1000 km at 90 N.

Checklist

• User's Guide has been updated
• Developer's Guide has been updated
• Documentation has been built locally and changes look as expected
• Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

[xylar]

Testing

I ran the remap_sea_surface_salinity step for the IcoswISC30E3r5 mesh. The step took about 3 minutes on 4 Chrysalis nodes (without using the cubed-sphere grid it had not completed after 30 minutes). With a maximum smoothing length scale of 1000 km (the current default), the result is this field:

Compare with a similar plot without smoothing:

Results are in:

/lcrc/group/e3sm/ac.xylar/compass_1.7/chrysalis/test_20250114/icosr5_sss_restoring/ocean/global_ocean/files_for_e3sm/remap_sea_surface_salinity_restoring

[xylar]

Here is what the smoothing field looks like:

The units are meters.

[xylar]

@vanroekel, this is now ready for your review. Feel free to review the code (by inspection) but the main review I would like is whether you are good with the outcome.

[xylar]

@maltrud, @milenaveneziani, @alicebarthel and @irenavankova, you all commented on #872 so I wanted to make sure you're aware of this follow-up work.

[milenaveneziani]

It looks good to me @xylar, thanks for working on this.
If we wanted, could we apply this tool to smooth the WOA data that we use in MPAS-Analysis? This would eliminate the weird patching that we usually see in the bias plot.

[xylar]

If we wanted, could we apply this tool to smooth the WOA data that we use in MPAS-Analysis? This would eliminate the weird patching that we usually see in the bias plot.

It's something we could think about. It relies on conservative remapping with MOAB and that performs really badly with latitude-longitude grids (because there are so many tiny, overlapping cells near the poles). I had to switch to putting the SSS fields on a cubed-sphere grid here to make the remapping/smoothing to the MPAS mesh performant. I'm not sure how feasible it is to take a similar approach in MPAS-Analysis.

[milenaveneziani]

I see. Thanks for clarifying that part, I had missed it.

[xylar]

I had missed it.

I see I forgot to add that to the PR description! My mistake. It's there now.

[xylar]

If we wanted, could we apply this tool to smooth the WOA data that we use in MPAS-Analysis? This would eliminate the weird patching that we usually see in the bias plot.

Thinking about this further, I would feel weird about smoothing an "observational" dataset presented in MPAS-Analysis. I think it would be more honest to just mask out the parts we don't trust than to manipulate them.

[vanroekel]

Approving by visual inspection of code and examination of resulting smoothed field.

Thanks @xylar this looks great to me

[xylar]

Thanks @vanroekel!

[alicebarthel]

Thanks @xylar , this looks great!