SOwISC12to60E2r02 mesh: coarser Southern Ocean, high res Arctic and North Atlantic

[xylar]

The "background" mesh before regional refinement now has 45-km resolution in southern mid latitudes and 25-km resolution near the south pole, before the 12-km refinement is applied around Antarctica. Resolution is also enhanced in the Arctic, with 15-km resolution around Greenland, tapering off to 30-km resolution at the Arctic margins and in the northern part of the Gulf Stream in the North Atlantic.

[xylar]

This has been done at @maltrud's request (will he ever be satisfied with the SO grid?!).

@maltrud's cartoon:

New mesh with coarser southern hemisphere

The requested resolution (passed to JIGSAW) at these longitudes:

The map of resolution passed to JIGSAW:

Drake Passage:

Old mesh for comparsion

We'll want to discuss if the change in resolution at Drake Passage is a problem, so that we should move the transition a little farther north.

[xylar]

@mark-petersen has also mentioned wanting to add a region of enhanced resolution (30 km?) in the North Atlantic, based on discussions with FESOM developers at AWI. This change would probably go along with a tightening of the higher resolution region around the equator, so that the total number of cells wasn't significantly different. This PR might be a good place to add these changes as well.

[xylar]

Current

nCells = 739904 ;
nEdges = 2234478 ;
nVertices = 1494156 ;

Coarser

nCells = 654592 ;
nEdges = 1977921 ;
nVertices = 1322919 ;

The sharper transition in resolution means lower resolution in the Drake Passage. I'm trying another iteration where the transition is moved 300 km farther north.

[kristin-hoch]

I think this looks reasonable.

[xylar]

@maltrud and anyone else, the data for this is available on LANL IC at:
/lustre/scratch4/turquoise/xylar/test_SOwISC_coarser_SH/ocean/global_ocean/SO60to10wISC/init/

[xylar]

Tighter 30-km resolution around the equator

To further reduce the number of cells, @mark-petersen has suggested reducing the high-res region around the equator by a factor of 2 (from 30 degrees latitude wide to 15 degrees wide). Below are results from that mesh.

The resolution of the proposed mesh:

For comparison, here is the current SOwISC (before this PR) at the same angle:

nCells = 609380 ;
nEdges = 1841373 ;
nVertices = 1231609 ;

Getting closer to a 20% reduction in total cells compared with the current mesh.

[xylar]

Another variant: 12 km finest resolution, 1600 km transition

Previous was 10 km finest, 1100 km transition

@maltrud's request:

What I came up with:

nCells = 389042 ;
nEdges = 1178412 ;
nVertices = 789030 ;

Current version from ocean/develop for comparison:

[xylar]

Higher resolution in the North Atlantic

As requested at the last E3SM Cryosphere meeting, @maltrud came up with the following request for enhanced resolution:

"Maybe something like a linear increase in grid scale from 15 km to ~30 km away from the coast of Greenland, but enclosed by the circle drawn in black."

Here's what I came up with as a first iteration:

nCells = 420112 ;
nEdges = 1274050 ;
nVertices = 853561 ;

Number of cells is only about 30k more than the previous version (a nearly negligible change)!

For comparison, the "current" mesh (ocean/develop) with the same colorbar:

Here are the usual plots of approximate resolution vs. lat (and sometimes lon):

Here are the shapes that define the contour of ~35 km resolution:
https://github.com/xylar/MPAS-Model/blob/coarser_SOwISC/testing_and_setup/compass/ocean/global_ocean/SO60to10wISC/init/north_mid_res_region.geojson

and of 15 km resolution (Greenland):
https://github.com/xylar/MPAS-Model/blob/coarser_SOwISC/testing_and_setup/compass/ocean/global_ocean/SO60to10wISC/init/greenland.geojson

These are very easy to edit (e.g. on http://geojson.io) so feel free to suggest changes.

[xylar]

I notice that the new mesh has the Black Sea and the "current" one doesn't. I wonder if we should put in a critical passage to make sure we do have the Black Sea or a critical land bridge to make sure we don't.

[milenaveneziani]

Thanks @xylar!
I think the transition from 20 to 30 km (current transition zone for variable GM) is covering about half of the subpolar gyre at the moment. My suggestion is to move the center of the circle more toward the center of the subpolar gyre, and also make the transition at lower latitudes (toward the GS extension) sharper, so that we don't transition GM right in the middle of the GS extension either.
I personally wouldn't worry about the Black Sea.

[xylar]

I think the transition from 20 to 30 km (current transition zone for variable GM) is covering about half of the subpolar gyre at the moment. My suggestion is to move the center of the circle more toward the center of the subpolar gyre, and also make the transition at lower latitudes (toward the GS extension) sharper, so that we don't transition GM right in the middle of the GS extension either.

These sound like reasonable suggestions but I don't speak Arctic or North Atlantic. ;-) What and where is the GS?

If you want to modify the shape(s) as you see fit, I'm happy to rerun.

I personally wouldn't worry about the Black Sea.

Let it randomly appear in different meshes? Or take it out?

[mark-petersen]

I think Black and Caspian Seas are in the same boat :). We should decide if they should both be in or both be out, and then force them in or out. If they don't influence the global ocean, it seems like they should both be out.

[xylar]

I think Black and Caspian Seas are in the same boat :). We should decide if they should both be in or both be out, and then force them in or out. If they don't influence the global ocean, it seems like they should both be out.

There's probably no danger of getting the Caspian Sea even at high res because we don't resolve the connection and it gets culled. I can add a critical land bridge so the same happens with the Black Sea if there's general agreement.

[milenaveneziani]

GS=Gulf Stream.
Let me see if I can render in graphics what I am thinking about.

[xylar]

GS=Gulf Stream.

Ah, of course!

Let me see if I can render in graphics what I am thinking about.

That'd be great!

[milenaveneziani]

Here is one cartoon, but imagine that the 20-30 km transition wraps around the high-res circle.
Please also wait for @maltrud to weigh in before changing anything.

[jonbob]

As for the Black and Caspian Seas, we do need to be consistent because it impacts other e3sm components. We had some trouble awhile back when MOSART assumed they were ocean and tried to add runoff to those cells -- and all the other components considered them inland seas.

[darincomeau]

minor point - we didn't use 'L64' in the r01 version of this mesh; should it be included here? I just want to make sure I'm consistent with the agreed upon naming convention.

Also, where'd we end up with the horizontal cell count?

[xylar]

minor point - we didn't use 'L64' in the r01 version of this mesh; should it be included here? I just want to make sure I'm consistent with the agreed upon naming convention.

Good point, the "short name" is without that. I'll edit.

Also, where'd we end up with the horizontal cell count?

I believe #518 (comment) is still accurate or close enough, so

nCells = 425809

[maltrud]

@mark-petersen this is just for mpas, right? this won't be propagated to e3sm, correct?

[xylar]

Right, the COMPASS changes might propagate to E3SM but that isn't the same as the mesh being available for use in E3SM. That part won't happen just based on this PR.

[maltrud]

thanks for the clarification, @xylar

[mark-petersen]

@xylar since it's the end of the day here, I'll let you rebase and make any adjustments. You can merge with the button if you like.

[xylar]

Latest

Here's the latest resolution plot. I'm not going to re-run spin-up until this gets merged to ocean/develop but will run through initital_state to make sure no problems emerge.

[mark-petersen]

PS: A mesh generation timing test took 21:58 on Chrysalis on one node.