Update global ocean init mode for meshes with ice-shelf cavities #6375

xylar · 2024-04-29T15:58:02Z

This merge updates MPAS-Ocean init mode for global_ocean in several ways that are critical for initialization that is compatible with planned coupling to MALI.

The fields landIcePressure, landIceDraft and ssh are now read from a topography file. This means that landIcePressure is no longer computed based on the density at the ocean surface in init mode. Similarly, ssh is not required to be the same as landIceDraft.

Support is removed for reading land-ice topography fields from a dataset on a lat-lon grid. We no longer use this capability and maintaining it is not worth the effort.

The land-ice topography data set is now required to have landIceFloatingFracObserved as a field, and this is used to set landIceFloatingFraction. Previously, landIceFracObserved was being used to compute both landIceFraction and landIceFloatingFraction (assuming no grounded ice in the domain).

We no longer set landIceFraction and landIceFloatingFraction to zero when landIceMask and landIceFloatingMask, respectively, are zero. This is in order to allow melting in cells that are less than 50% land ice in order to better support conservation between ice sheet and ocean.

In addition, we use sshAdjustmentMask instead of landIceMask as the starting point for smoothing mask used to produce the Haney-number vertical coordinate. This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability.

xylar · 2024-04-29T15:58:26Z

This has been moved from E3SM-Ocean-Discussion#90, which includes relevant discussion.

xylar · 2024-04-29T16:17:11Z

@mark-petersen, you are the best person I could think of to sanity check changes to init mode. Let me know if you can do that.

xylar · 2024-04-30T18:46:16Z

@mark-petersen, this is a high priority for the FAnSSIE project. Can you review when you can?

xylar · 2024-04-30T18:46:58Z

Sorry, I forgot I already pinged you...

xylar · 2024-05-14T15:01:39Z

@sbrus89, @mark-petersen is away and I need someone to review this that knows about MPAS-Ocean's global ocean init mode. You're the best candidate. Would you be up for reviewing this? If not, remove yourself as a reviewer and I'll find someone else.

sbrus89 · 2024-05-15T15:01:02Z

@xylar, sure I will take a look.

rljacob · 2024-05-30T17:45:57Z

@sbrus89 please review.

xylar · 2024-06-05T09:58:58Z

@mark-petersen, @cbegeman and @sbrus89, a reminder that I would appreciate a review of this PR. This would allow me to make important changes to how meshes with ice-shelf cavities are initialized in MPAS-Dev/compass#813. This, in turn, is needed for me to make progress on coupling to MALI.

We do not use this mode anymore and it is too difficult to maintain.

This merge uses the `landIceDraftObserved`, `landIcePressureObserved` and `landIceFloatingFracObserved` to initialize the corresponding MPAS-Ocean variables. Previously, the `landIcePressure` was being computed in a subroutine and the `landIceFloatingFraction` was required to be identical to `landIceFraction`. This merge also removes the unused variable `landIceGroundedFracObserved`

Instead, it will be read in from the topography file directly, meaning it can be updated (e.g. from ssh adjustment).

this is a more accurate description for how it is uses, since it can determine where we modify either ssh or landIcePressure, depending on the approach.

This could allow us to choose to have melting bleed into cells that are partly open ocean (e.g. for conservation when coupling to MALI).

xylar · 2024-06-10T14:04:23Z

components/mpas-ocean/src/mode_init/mpas_ocn_init_global_ocean.F

+             ! don't mask these because we may want to include melt that bleeds
+             ! into the open ocean
+             landIceFraction(iCell) = landIceFracObserved(iCell)
+             landIceFloatingFraction(iCell) = landIceFloatingFracObserved(iCell)


In case anyone has been reviewing this already, I just added the commit to change this. I think we don't want to hard-code this masking here because we want some flexibility about how to allow fluxes to bleed into cells that are partly open ocean and partly ice. This will be important for conservation when coupling with MALI (without requiring rerouting to cells the ocean decides are under land ice, which I think we want to avoid). I don't see an clear harm in at least supporting this possibility.

This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability.

cbegeman · 2024-06-11T21:19:01Z

In addition, we use sshAdjustmentMask instead of landIceMask as the starting point for smoothign mask used to produce the Haney-number vertical coordinate. This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability.

Can you explain when this arose? Were these locations where landIceDraft > 0 and landIceMask = 0 ones where landIceFraction was less than the threshold?

cbegeman · 2024-06-11T22:21:46Z

@xylar I tested this PR@e0c0c0755b with MPAS-Dev/compass#813 against compass/main@1588d78416 and E3SM/master@470a42bdcb. I'm not sure whether this would change with your latest commit but I was seeing maxDeltaSSH of 0.3m with this PR and maxDeltaSSH of 0.02m with the baseline (landIcePressure is the adjusted variable) for the IcoswISC240 test at the last ssh adjustment iteration. Is this similar to what you've been seeing?

xylar · 2024-06-12T06:46:51Z

In addition, we use sshAdjustmentMask instead of landIceMask as the starting point for smoothing mask used to produce the Haney-number vertical coordinate. This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability.

Can you explain when this arose? Were these locations where landIceDraft > 0 and landIceMask = 0 ones where landIceFraction was less than the threshold?

Yes, there are many such locations around the peninsula. They may have been avoided or the topography was different when grounded ice being masked out of the interpolation process but now that I am including it (as we will be doing with MALI), I am seeing places where not only is the landIcePressure nonzero far from places where landIceFraction > 0.5 but also the Haney number in some of these locations is huge if we use a z-star coordinate so we need to include these regions in the haney-number coordinate as well.

For the new SORRM mesh (MPAS-Dev/compass#807):

Here is the landIceMask (where landIceFraction > 0.5):

Here is the sshAdjustmentMask (where landIcePressure > 0):

Here is the smoothing mask before (seeded from landIceMask):

Here is the resulting Haney Number:

Here is the smoothing mask after the change (seeded from sshAdjustmentMask):

And the resulting Haney Number (now capped at 20, as required):

xylar · 2024-06-12T06:49:38Z

I'm not sure whether this would change with your latest commit but I was seeing maxDeltaSSH of 0.3m with this PR and maxDeltaSSH of 0.02m with the baseline (landIcePressure is the adjusted variable) for the IcoswISC240 test at the last ssh adjustment iteration. Is this similar to what you've been seeing?

I haven't looked at that. My plan is to bring this in in conjunction with MPAS-Dev/compass#813 so I wasn't trying to compare with the current compass main branch. But I would not be surprised if this changes maxDeltaSSH (both with main and with MPAS-Dev/compass#813).

xylar · 2024-06-12T06:50:03Z

@cbegeman, thanks very much for looking at this and for your testing.

cbegeman

Thanks for the additional explanation and figures in response to my comments. I think the changes here look good. Thanks for all your work on this!

cbegeman · 2024-06-12T14:24:35Z

I'm not sure whether this would change with your latest commit but I was seeing maxDeltaSSH of 0.3m with this PR and maxDeltaSSH of 0.02m with the baseline (landIcePressure is the adjusted variable) for the IcoswISC240 test at the last ssh adjustment iteration. Is this similar to what you've been seeing?

I haven't looked at that. My plan is to bring this in in conjunction with MPAS-Dev/compass#813 so I wasn't trying to compare with the current compass main branch. But I would not be surprised if this changes maxDeltaSSH (both with main and with MPAS-Dev/compass#813).

Agreed. It's hard to tell what changes are related to this PR and the new topography file. Since it doesn't crash, it seems to have been adjusted enough.

xylar · 2024-06-12T15:25:40Z

@cbegeman, thanks so much for reviewing!

rljacob · 2024-06-13T17:40:23Z

@sbrus89 and @mark-petersen need your reviews.

mark-petersen · 2024-06-20T18:32:01Z

I compiled this with gnu (debug and optimized) on chicoma and intel on chrysalis and tested with the nightly test suite.

There are machine-level diffs on the QU240 init, which are negligible.

pwd
/lustre/scratch5/mpeterse/runs/n/ocean_model_240620_266c9d69_ch_gfortran_openmp_SSHinit

grep 'l1:' -B 1 case_outputs/*init*log
temperature          Time index: 0
0:  l1: 4.44089209850063e-16  l2: 4.44089209850063e-16  linf: 4.44089209850063e-16
--
layerThickness       Time index: 0
0:  l1: 5.11590769747272e-13  l2: 4.58286294150329e-13  linf: 4.54747350886464e-13

Tested QUwISC240 using this branch and MPAS-Dev/compass#813 together, with gnu debug, and it is successful.

Next I will test IcoswISC240 and rerun the Arctic 3km mesh initialization I set up for the Aerosi project.

mark-petersen · 2024-06-20T18:50:38Z

Results using this branch and MPAS-Dev/compass#813 together, with gnu debug.

output:


ocean/global_ocean/QUwISC240/mesh
  * step: base_mesh
  * step: remap_topography
  * step: cull_mesh
  test execution:      SUCCESS
  test runtime:        01:33
ocean/global_ocean/QUwISC240/WOA23/init
  * step: 01_init_with_draft_from_constant_density
  * step: 02_ssh_from_surface_density
  * step: 03_init
  * step: 04_adjust_ssh
  * step: 05_init
  * step: 06_adjust_ssh
  * step: 07_init
  * step: 08_adjust_ssh
  * step: 09_init
  * step: 10_adjust_ssh
  * step: initial_state
  * step: remap_ice_shelf_melt
  test execution:      SUCCESS
  test runtime:        04:32
ocean/global_ocean/QUwISC240/WOA23/performance_test
  * step: prognostic_ice_shelf_melt
  * step: data_ice_shelf_melt
  test execution:      SUCCESS
  test runtime:        00:53
ocean/global_ocean/IcoswISC240/mesh
  * step: base_mesh
  * step: remap_topography
  * step: cull_mesh
  test execution:      SUCCESS
  test runtime:        01:31
ocean/global_ocean/IcoswISC240/WOA23/init
  * step: 01_init_with_draft_from_constant_density
  * step: 02_ssh_from_surface_density
  * step: 03_init
  * step: 04_adjust_ssh
  * step: 05_init
  * step: 06_adjust_ssh
  * step: 07_init
  * step: 08_adjust_ssh
  * step: 09_init
  * step: 10_adjust_ssh
  * step: initial_state
  * step: remap_ice_shelf_melt
  test execution:      SUCCESS
  test runtime:        04:21
ocean/global_ocean/IcoswISC240/WOA23/performance_test
  * step: prognostic_ice_shelf_melt
  * step: data_ice_shelf_melt
  test execution:      SUCCESS
  test runtime:        00:53
Test Runtimes:
01:33 PASS ocean_global_ocean_QUwISC240_mesh
04:32 PASS ocean_global_ocean_QUwISC240_WOA23_init
00:53 PASS ocean_global_ocean_QUwISC240_WOA23_performance_test
01:31 PASS ocean_global_ocean_IcoswISC240_mesh
04:21 PASS ocean_global_ocean_IcoswISC240_WOA23_init
00:53 PASS ocean_global_ocean_IcoswISC240_WOA23_performance_test
Total runtime 13:44

xylar · 2024-06-21T16:03:20Z

Thanks very much @mark-petersen for reviewing!

xylar · 2024-06-21T16:03:46Z

@mark-petersen, could you approve the PR if you're happy?

xylar · 2024-06-21T16:04:59Z

@sbrus89, will you have time to review this very soon? If not, I'll probably just go forward with the reviews I have.

sbrus89

I've looked this over and the changes look good to me. I'm approving based on testing from @mark-petersen and @cbegeman.

xylar · 2024-06-21T18:30:06Z

Thank you, @sbrus89!

mark-petersen · 2024-06-24T13:26:22Z

I'm testing with the new Arctic mesh for the Aerosi project, which is the RRS6to18 with a 3km Arctic. I'm having trouble with the init/adjust_ssh step, as the 04_adjust_ssh is unstable. I'm sure I just need to adjust my flags for this resolution, so I am trying that.

If these steps already ran:

01_init_with_draft_from_constant_density  02_ssh_from_surface_density  03_init

is it possible to just run the 04_adjust_ssh step, so I don't need to wait again? Thanks.

mark-petersen

I was able to run the IcoswISC240 through mesh generation, init, and performance test on chicoma with gnu. I set up and started running RRS6to18 and my new Arctic 3km mesh. I had some problems with that last one, but it is just my flag settings, which I can adjust myself.

Thanks for your work on this! I approve.

#6375) Update global ocean init mode for meshes with ice-shelf cavities This merge updates MPAS-Ocean init mode for global_ocean in several ways that are critical for initialization that is compatible with planned coupling to MALI: * The fields landIcePressure, landIceDraft and ssh are now read from a topography file. This means that landIcePressure is no longer computed based on the density at the ocean surface in init mode. Similarly, ssh is not required to be the same as landIceDraft. * Support is removed for reading land-ice topography fields from a dataset on a lat-lon grid. We no longer use this capability and maintaining it is not worth the effort. * The land-ice topography data set is now required to have landIceFloatingFracObserved as a field, and this is used to set landIceFloatingFraction. Previously, landIceFracObserved was being used to compute both landIceFraction and landIceFloatingFraction (assuming no grounded ice in the domain). * We no longer set landIceFraction and landIceFloatingFraction to zero when landIceMask and landIceFloatingMask, respectively, are zero. This is in order to allow melting in cells that are less than 50% land ice in order to better support conservation between ice sheet and ocean. * In addition, we use sshAdjustmentMask instead of landIceMask as the starting point for smoothing mask used to produce the Haney-number vertical coordinate. This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability. [BFB] mpas-ocean standalone feature

jonbob · 2024-06-24T17:05:57Z

Passes:

ERS_Ld5.T62_oQU120.CMPASO-NYF.chrysalis_intel
SMS_D_Ld1.TL319_IcoswISC30E3r5.GMPAS-JRA1p5-DIB-PISMF.chrysalis_intel.mpaso-jra_1958

merged to next

xylar · 2024-06-24T22:10:43Z

is it possible to just run the 04_adjust_ssh step, so I don't need to wait again? Thanks.

@mark-petersen, yes, you can run that step on its own by going into its directory (adjust_ssh/04_adjust_ssh or whatever it's called) and submitting the job_script.sh. You can skip several steps by editing WOA23/init/init.cfg so that the steps you want to skip are not in steps_to_run and then submitting job_script.sh in init.

jonbob · 2024-06-25T14:41:25Z

merged to master

xylar · 2024-06-29T15:30:44Z

Thank you so much @jonbob! And also thank you @cbegeman, @sbrus89 and @mark-petersen for the reviews!

Bring in E3SM-Project/E3SM#6375

xylar added mpas-ocean BFB PR leaves answers BFB MPAS-Ocean standalone Issues and features for standalone MPAS-Ocean code that dont impact E3SM. labels Apr 29, 2024

xylar requested review from mark-petersen and cbegeman April 29, 2024 15:58

xylar assigned jonbob Apr 29, 2024

xylar mentioned this pull request Apr 29, 2024

Update global ocean init mode for meshes with ice-shelf cavities E3SM-Ocean-Discussion/E3SM#90

Closed

xylar force-pushed the ocn/init-land-ice-pressure-from-topo branch 2 times, most recently from cb9ca7e to f17151a Compare May 8, 2024 22:10

xylar mentioned this pull request May 9, 2024

For initialization with ice-shelf cavities, adjust SSH not land ice pressure MPAS-Dev/compass#813

Merged

7 tasks

xylar removed the request for review from mark-petersen May 14, 2024 15:00

xylar requested a review from sbrus89 May 14, 2024 15:01

mark-petersen self-requested a review May 28, 2024 14:36

xylar added 6 commits June 10, 2024 02:47

Disable reading land-ice topography from lat-lon file

11ae9ce

We do not use this mode anymore and it is too difficult to maintain.

Read topo variable landIcePressureObserved

e8363b8

Compute density in global ocean init mode

f372331

Don't set ssh from landIceDraft in global ocean init

0569e25

Instead, it will be read in from the topography file directly, meaning it can be updated (e.g. from ssh adjustment).

Rename modifyLandIcePressureMask to sshAdjustmentMask

2389f6f

this is a more accurate description for how it is uses, since it can determine where we modify either ssh or landIcePressure, depending on the approach.

xylar force-pushed the ocn/init-land-ice-pressure-from-topo branch from f17151a to 2389f6f Compare June 10, 2024 07:47

Leave the landIce(Floating)Fraction unmasked

e0c0c07

This could allow us to choose to have melting bleed into cells that are partly open ocean (e.g. for conservation when coupling to MALI).

xylar commented Jun 10, 2024

View reviewed changes

Use sshAdjustmentMask instead of landIceMask for rx1 smoothing mask

266c9d6

This is needed because otherwise locations with nonzero ice draft get missed and the Haney number may be too high for model stability.

cbegeman approved these changes Jun 12, 2024

View reviewed changes

sbrus89 approved these changes Jun 21, 2024

View reviewed changes

mark-petersen approved these changes Jun 24, 2024

View reviewed changes

jonbob merged commit 31f771c into E3SM-Project:master Jun 25, 2024
13 checks passed

xylar deleted the ocn/init-land-ice-pressure-from-topo branch June 29, 2024 15:30

xylar added a commit to xylar/compass that referenced this pull request Jun 30, 2024

Update E3SM-Project submodle

1af70f8

Bring in E3SM-Project/E3SM#6375

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update global ocean init mode for meshes with ice-shelf cavities #6375

Update global ocean init mode for meshes with ice-shelf cavities #6375

xylar commented Apr 29, 2024 •

edited

Loading

xylar commented Apr 29, 2024

xylar commented Apr 29, 2024

xylar commented Apr 30, 2024

xylar commented Apr 30, 2024

xylar commented May 14, 2024

sbrus89 commented May 15, 2024

rljacob commented May 30, 2024

xylar commented Jun 5, 2024

xylar Jun 10, 2024

cbegeman commented Jun 11, 2024

cbegeman commented Jun 11, 2024

xylar commented Jun 12, 2024 •

edited

Loading

xylar commented Jun 12, 2024

xylar commented Jun 12, 2024

cbegeman left a comment

cbegeman commented Jun 12, 2024

xylar commented Jun 12, 2024

rljacob commented Jun 13, 2024

mark-petersen commented Jun 20, 2024

mark-petersen commented Jun 20, 2024

xylar commented Jun 21, 2024

xylar commented Jun 21, 2024 •

edited

Loading

xylar commented Jun 21, 2024

sbrus89 left a comment

xylar commented Jun 21, 2024

mark-petersen commented Jun 24, 2024

mark-petersen left a comment

jonbob commented Jun 24, 2024

xylar commented Jun 24, 2024

jonbob commented Jun 25, 2024

xylar commented Jun 29, 2024

Update global ocean init mode for meshes with ice-shelf cavities #6375

Update global ocean init mode for meshes with ice-shelf cavities #6375

Conversation

xylar commented Apr 29, 2024 • edited Loading

xylar commented Apr 29, 2024

xylar commented Apr 29, 2024

xylar commented Apr 30, 2024

xylar commented Apr 30, 2024

xylar commented May 14, 2024

sbrus89 commented May 15, 2024

rljacob commented May 30, 2024

xylar commented Jun 5, 2024

xylar Jun 10, 2024

Choose a reason for hiding this comment

cbegeman commented Jun 11, 2024

cbegeman commented Jun 11, 2024

xylar commented Jun 12, 2024 • edited Loading

xylar commented Jun 12, 2024

xylar commented Jun 12, 2024

cbegeman left a comment

Choose a reason for hiding this comment

cbegeman commented Jun 12, 2024

xylar commented Jun 12, 2024

rljacob commented Jun 13, 2024

mark-petersen commented Jun 20, 2024

mark-petersen commented Jun 20, 2024

xylar commented Jun 21, 2024

xylar commented Jun 21, 2024 • edited Loading

xylar commented Jun 21, 2024

sbrus89 left a comment

Choose a reason for hiding this comment

xylar commented Jun 21, 2024

mark-petersen commented Jun 24, 2024

mark-petersen left a comment

Choose a reason for hiding this comment

jonbob commented Jun 24, 2024

xylar commented Jun 24, 2024

jonbob commented Jun 25, 2024

xylar commented Jun 29, 2024

xylar commented Apr 29, 2024 •

edited

Loading

xylar commented Jun 12, 2024 •

edited

Loading

xylar commented Jun 21, 2024 •

edited

Loading