Add new SOwISC12to30E3r3 ocean and sea-ice mesh

[xylar]

Long name: SOwISC12to30kmL80E3SMv3r3

This version of the Southern Ocean Regionally Refined Mesh (SORRM) has resolution that is:

• 12 km around Antarctica
• 30 km elsewhere
  It is intended to be similar to the IcoswISC30E3r5 mesh except in the Southern Ocean and around Antarctica.

This is a proposed E3SM v3 (E3) mesh for polar (formerly cryosphere) simulations. This is revision 3 (r3) of the mesh, which includes many changes from revision 2 outlined in MPAS-Dev/compass#807. The mesh will be tagged as: https://github.com/MPAS-Dev/compass/releases/tag/mesh_SOwISC12to30E3r3 if it is accepted.

The mesh and the G-case results will be reviewed here:
https://acme-climate.atlassian.net/wiki/spaces/OO/pages/4441145345/Review+SOwISC12to30E3r3

A G- and B-case will begin shortly and analysis will be posted on the review page as soon as it is available.

[xylar]

@jonbob and @darincomeau, the mesh is ready for some sanity checking and then making mapping files.

[xylar]

@jonbob, apologies! @cbegeman asked me to make some changes to the mesh (to ensure a thicker minimum water-column thickness). I don't think those changes will affect the horizontal mesh but I am not sure that they won't. Even if the cell count with these changes happens to be the same, the cell order might be different.

So if you happen to have started making mapping files, I'm afraid you'll need to start over. If you haven't had a chance to start yet, that would be a relief!

[xylar]

@jonbob, okay, @cbegeman has signed off on the water-column thickness with the latest version of this mesh and I have just updated the files in /lcrc/group/e3sm/data/inputdata/. I think we're ready for mapping files from the new version of the mesh/initial condition.

[jonbob]

Are you all interested in any particular configurations? I've made mapping files for the following atm resolutions:
T62
TL319
ne30pg2
ne120pg2
and r05 plus r025 if you want to run trigrid. I also made runoff mapping files for:
rx1
r05
r025
JRA025

[darincomeau]

That looks complete to me - thanks @jonbob !

[xylar]

I believe we're most interested in B-cases with the r05 try grid. We will probably use the JRA grid, too. But it's really good to have all those options. Thanks @jonbob!

[darincomeau]

If we don't want to stage them all now, I agree with @xylar that ne30pg2_r05_SOwISC12to30E3r3 and TL319_SOwISC12to30E3r3 (with JRA025 runoff map) are the ones we'll want first for testing.

[jonbob]

I added support for G-case with T62 and TL319 configurations, as well as fully-coupled using ne30pg2 in both bi- and tri-grid. I've staged the necessary files on the local lcrc inputdata directory, and tested:

SMS.ne30pg2_r05_SOwISC12to30E3r3.WCYCL1850NS.chrysalis_intel
SMS.TL319_SOwISC12to30E3r3.GMPAS-JRA1p5.chrysalis_intel

So it's ready for your testing. Please note I didn't make spunup IC files from a G-case for use in B-cases

[xylar]

@jonbob, oh, dear. @darincomeau already found a problem with the horizontal mesh (a place where we need a critical passage to keep there from being a hole in the land ice). I have finally fixed the issue after 2 false starts. I am rebuilding all the SORRM r3 Compass files now. I will likely ping you tomorrow to make new mapping files (again!). I'm very sorry about that.

[jonbob]

@xylar -- no problem. Darin had warned me yesterday, so I didn't push more on it. And I always need the practice making mapping files!

[xylar]

And I always need the practice making mapping files!

Oh, I doubt that! I bet you could make them in your sleep at this point.

[xylar]

@darincomeau, if you happen to be working today, could you give the new version of the mesh a sanity check before @jonbob makes mapping files. I really want the 3rd time to be the charm here...

[darincomeau]

@xylar looks good to me! I checked around the coast again and didn't see any other issues.

[xylar]

Thanks @darincomeau!!

[jonbob]

I updated the mapping and domain files, and pushed the new file datestamps and mesh size to this PR. I have staged the files in the lcrc local inputdata directory for testing. I did successfully make the following cases:

• SMS_P640.TL319_SOwISC12to30E3r3.GMPAS-JRA1p5.chrysalis_intel
• SMS_P1280.ne30pg2_r05_SOwISC12to30E3r3.WCYCL1850NS.chrysalis_intel
  but they are still in the queue.

[xylar]

Thanks so much @jonbob! I just updated the inputdata files from Compass with the fix that I mentioned in the Polar call yesterday.

I'll rerun the 2 smoke tests you ran (or at least submitted) to make sure things are still fine, but I expect them to be.

[xylar]

Yep, both tests passed for me:

SMS_P640.TL319_SOwISC12to30E3r3.GMPAS-JRA1p5.chrysalis_intel
SMS_P1280.ne30pg2_r05_SOwISC12to30E3r3.WCYCL1850NS.chrysalis_intel

[darincomeau]

Ok great, I'll get the G-case going

[jonbob]

Just fyi. I did make mapping files to support ne120_r025_SOwISC12to30E3r3 as well. Please let me know if you want to add this configuration to the PR

[xylar]

@jonbob, if you've bothered to make the mapping files, I think it makes sense to add the configuration here as well.

[xylar]

@darincomeau, thanks for adding the B-case initial conditions!

Keep us posted on how the G-case goes.

[darincomeau]

Keep us posted on how the G-case goes.

Looks like it'll still be in the queue for awhile.

I also have a B-case sitting in the queue. It's a CRYO1850 compset, not CRYO1950, because there isn't an fsurdat file for 1950 conditions on the r05 grid (which we should probably ask for).

[xylar]

@darincomeau, do you know who to ask? We do probably want that.

[darincomeau]

I think Wuyin?

[cbegeman]

OK. Perhaps instead of applying a Dirichlet boundary condition on the sea ice model (SSH is identical at the boundary as for one grid point in, hence gradSSH=0), we could apply a Neumann boundary condition, where gradSSH is identical at the boundary as the unaffected cell inside the boundary. By boundary, I mean cells with nonzero landIcePressure. I think that would work for sea ice and is more justifiable in publications.

@proteanplanet, that sounds fine to me. If I understand, this would be the same as extrapolating gradSSH instead of just pressureAdjustedSSH. I don't know off the top of my head how we might accomplish this in practice since extrapolating a scalar is a ton easier than extrapolating a vector in MPAS.

@xylar I don't understand what you mean because pressureAdjustedSSH is not extrapolated as far as I can tell. What @proteanplanet's proposing seems similar to the idea of excluding the land ice boundary edges' gradSSH(iEdge) from the reconstruction. But were you thinking we would do something more manual and take gradSSH from nearby edges and extrapolate and rotate it onto a land ice boundary edge before the reconstruction?

[xylar]

@cbegeman, my idea of setting gradSSH to zero is equivalent to extrapolating SSH into the land-ice region as far as I can tell.

I don't think we have a plausible method for reconstruction gradSSH at cell centers from a subset of edges, so we would need to create a value for gradSSH at all edges (including those bounded by land ice on one side and maybe even including open-ocean edges bounded on both sides by some fraction of land ice). I don't think there are any tricks we can use with the existing, very local vector reconstruction that would take care of this without some form of extrapolation of pressureAdjustedSSH or gradSSH.

[xylar]

I discovered that I made a mistake in masking the ice sheet thickness in the data set that combines GEBCO and BedMachine. This data set is so large that I can't visualize it in ParaView, which has sometimes made sanity checking tricky. Even so, it should have been immediately apparent that the ice sheet thickness after remapping was not correct:

I believe this issue is likely propagating the strange results in SSH adjustment that we are seeing and then to the sea ice problem. My hope is that once I solve this issue, we can retest without any additional fiddling with landIceDraft, pressureAdjustedSSH or gradSSH.

[proteanplanet]

Yay. Nice find.

[xylar]

Here is the same field as above after fixing this in compass:

The blue areas have bedrock above sea level and are purposefully zeroed out.

[xylar]

I'm still seeing big changes between ssh (after adjustment) and landIceDraft (before adjustment)

[xylar]

Here is the ice sheet thickness using the latest approach, which doesn't allow it to exceed the flotation thickness:

The ice draft and pressure are similarly limited (as they are just scaled versions of the same field).

[xylar]

After SSH adjustment, everything now looks sensible, with smooth changes that are less than 4 m:

[darincomeau]

Mesh review tests underway:

G-case:
https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.dcomeau/20240829.GMPAS-JRA1p5-DIB-PISMF.TL319_SOwISC12to30E3r3.mesh-review-test.chrysalis/

B-case:
https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.dcomeau/E3SMv3_dev/v3.SORRME3r3.piControl/

[proteanplanet]

@xylar Would it help me for to review the mesh once more with my script? I can set it running this morning.

[xylar]

@proteanplanet, yes, I think it wouldn't hurt to rerun your script. I don't think much has changed but I also absolutely don't want to assume that.

[darincomeau]

Approving based on test simulations here:
https://acme-climate.atlassian.net/wiki/spaces/OO/pages/4441145345/Review+SOwISC12to30E3r3

[xylar]

@jonbob, let me know if this mesh is ready to go to E3SM from your perspective.

[darincomeau]

I think @proteanplanet was about ready to sign off on this. We'd like to get it in for the new tag if possible.

[rljacob]

Please have this on next by Monday if you want it in the 3.0.1 tag.

[xylar]

Closed in favor of E3SM-Project#6631

[darincomeau]

We'd like to get it in for the new tag if possible.

Rescinding this statement, not a priority for 3.0.1