Revert GM buoyancy gradient calculation to v1 form by vanroekel · Pull Request #687 · MPAS-Dev/MPAS-Model

vanroekel · 2020-09-08T15:33:16Z

E3SM testing suggests that the new formulation of buoyancy gradient calculation that leverages infrastructure from Redi is at least partly responsible for excessive heat uptake in v2 alpha simulations. This PR will revert those changes back to the v1 formulation.

xylar · 2020-09-08T15:51:46Z

@darincomeau is seeing too little sea ice in the northern hemisphere in his SORRM runs. Could this be related?

milenaveneziani · 2020-09-08T16:24:58Z

@darincomeau: what hash are you using for the SORRM runs?

jonbob · 2020-09-08T16:28:00Z

test merge into E3SM passes:
PET_Ln9_PS.ne30_oECv3_ICG.A_WCYCL1850S.anvil_intel
PEM_Ln9_PS.ne30_oECv3_ICG.A_WCYCL1850S.anvil_intel

waiting on:
ERS.T62_oEC60to30v3wLI.GMPAS-DIB-IAF-ISMF.anvil_intel

milenaveneziani · 2020-09-08T16:28:41Z

@vanroekel: is this the specific MPAS-Model PR that is involved in the new GM buoyancy gradient calculation?:
#633
so, anything before that uses the v1 formulation?

vanroekel · 2020-09-08T16:37:13Z

@xylar and @darincomeau it is not clear if the v2 buoyancy gradient would result in too little sea ice. It is possible for sure as upper ocean stratification is pretty strongly affected by the choice.

@milenaveneziani it is before that where it was introduced - #445 , but it was improved in the PR you mention

darincomeau · 2020-09-08T16:57:45Z

@darincomeau: what hash are you using for the SORRM runs?

@milenaveneziani the hash is b9f3341f2, and the branch is here: https://github.com/E3SM-Project/E3SM/tree/darincomeau/SORRMr2-clubbv2. The branch was split off master on May 29th I believe.

milenaveneziani · 2020-09-08T17:01:09Z

@darincomeau: so #445 (merged in Apr) was there but not #633.

jonbob · 2020-09-08T17:01:59Z

test merge into E3SM also passed:
ERS.T62_oEC60to30v3wLI.GMPAS-DIB-IAF-ISMF.anvil_intel

milenaveneziani · 2020-09-08T17:18:16Z

so, @darincomeau, @xylar, I agree with @vanroekel that it is hard to say. I think best will be to restart the SORRM runs after v2 ocean is finalized, since so many changes are going in (and out) of v2 beta ocean. But maybe by then we'll have a final SORRM mesh to focus on?

darincomeau · 2020-09-08T18:13:40Z

Yeah these runs used GM_closure=N2_dependent, and redi was off, so SORRM reruns are definitely warranted.

xylar · 2020-09-08T18:29:37Z

Okay, happy to discuss next steps for SORRM but this PR isn’t the place.

mark-petersen · 2020-09-09T21:59:08Z

src/core_ocean/shared/mpas_ocn_gm.F

Some of these new arrays need nVertLevels+1. I can fix that.

okay, thanks! I could also take a look and fix as well if you'd like

vanroekel · 2020-09-10T04:38:32Z

an update, I've tested this for 20+ years in a B-case and the results are positive. Biases are reduced pretty substantially in terms of heat uptake and MLD bias. @mark-petersen if you could give take a look I'd appreciate it. This one is high priority for WC.

mark-petersen · 2020-09-10T18:15:59Z

src/core_ocean/shared/mpas_ocn_gm.F

@vanroekel in the current PR, dzdxEdge and drhoDzTopOfEdge are allocated and set to zero but never used. I can take them out, but can you double check that not using them was intentional, rather than a copy and paste error?

@mark-petersen this was a copy-paste error Thanks for catching that!

mark-petersen · 2020-09-10T19:59:20Z

rebased, fixed index out of bounds errors. Now passes nightly regression with gnu debug.

mark-petersen · 2020-09-10T20:32:32Z

@mattdturner This PR changes several loops to an older formulation from last year. Could you look at the OMP pragmas around the revised loops for this PR? We need to make sure it retains the newer format since #513.

mattdturner · 2020-09-10T21:02:32Z

There were some missing !$omp parallel regions around some of the loops. I added them, and the thread test PASSes on Cori w/ intel + debug.

Approved based on this test, and the other testing by @mark-petersen

mark-petersen · 2020-09-10T22:48:18Z

Thanks @mattdturner. I also ran a test with gnu 6 debug, passes all.

Revert GM buoyancy gradient calculation to v1 form #687 E3SM testing suggests that the new formulation of buoyancy gradient calculation that leverages infrastructure from Redi is at least partly responsible for excessive heat uptake in v2 alpha simulations. This PR will revert those changes back to the v1 formulation.

Fixes limiting on redi k33 #684 Fixes surface buoyancy forcing calculation #690 Revert GM buoyancy gradient calculation to v1 form #687

mark-petersen

Reviewed code, passes stand-alone tests. Testing in E3SM now, but @vanroekel has tested as well.

Fixes limiting on redi k33 #684 Fixes surface buoyancy forcing calculation #690 Revert GM buoyancy gradient calculation to v1 form #687

This PR brings in a new mpas-source submodule with changes only to the ocean core, plus scripts updates corresponding to Registry changes in the ocean. It includes three changes to improve ocean heat uptake: * Fixes limiting on redi k33 (MPAS-Dev/MPAS-Model/pull/684); * Fixes surface buoyancy forcing calculation (MPAS-Dev/MPAS-Model/pull/690); and * Revert GM buoyancy gradient calculation to v1 form (MPAS-Dev/MPAS-Model/pull/687). [NML] [non-BFB]

Update mpas-source: GM/Redi and surface buoyancy fixes This PR brings in a new mpas-source submodule with changes only to the ocean core, plus scripts updates corresponding to Registry changes in the ocean. It includes three changes to improve ocean heat uptake: * Fixes limiting on redi k33 (MPAS-Dev/MPAS-Model/pull/684); * Fixes surface buoyancy forcing calculation (MPAS-Dev/MPAS-Model/pull/690); and * Revert GM buoyancy gradient calculation to v1 form (MPAS-Dev/MPAS-Model/pull/687). [NML] [non-BFB]

…an/develop Revert GM buoyancy gradient calculation to v1 form MPAS-Dev#687 E3SM testing suggests that the new formulation of buoyancy gradient calculation that leverages infrastructure from Redi is at least partly responsible for excessive heat uptake in v2 alpha simulations. This PR will revert those changes back to the v1 formulation.

vanroekel added Ocean in progress labels Sep 8, 2020

vanroekel requested a review from mark-petersen September 8, 2020 15:33

vanroekel marked this pull request as draft September 8, 2020 15:33

mark-petersen self-assigned this Sep 9, 2020

mark-petersen reviewed Sep 9, 2020

View reviewed changes

vanroekel removed the in progress label Sep 10, 2020

vanroekel marked this pull request as ready for review September 10, 2020 04:39

mark-petersen reviewed Sep 10, 2020

View reviewed changes

vanroekel and others added 3 commits September 10, 2020 13:31

reverts GM buoyancy grad calculation to v1

6cfdc68

Fixes to the reversion

b06a186

Change allocate dims to nVerLevels+1, remove unused variables

1ca85f2

mark-petersen force-pushed the ocean/v1GMbuoyancyGradient branch from a2f4ce5 to 1ca85f2 Compare September 10, 2020 19:56

mark-petersen requested a review from mattdturner September 10, 2020 20:33

Add missing OMP parallel regions

8396205

mattdturner approved these changes Sep 10, 2020

View reviewed changes

mark-petersen added a commit that referenced this pull request Sep 10, 2020

Merge branch 'ocean/develop' into e3sm/develop

65699ce

Fixes limiting on redi k33 #684 Fixes surface buoyancy forcing calculation #690 Revert GM buoyancy gradient calculation to v1 form #687

mark-petersen mentioned this pull request Sep 10, 2020

Update mpas-source: GM/Redi and surface buoyancy fixes E3SM-Project/E3SM#3825

Merged

mark-petersen self-requested a review September 10, 2020 23:24

mark-petersen approved these changes Sep 10, 2020

View reviewed changes

mark-petersen merged commit 1f7abcc into MPAS-Dev:ocean/develop Sep 11, 2020

mark-petersen added a commit that referenced this pull request Sep 11, 2020

Merge branch 'ocean/develop' into e3sm/develop

7e32f8e

Fixes limiting on redi k33 #684 Fixes surface buoyancy forcing calculation #690 Revert GM buoyancy gradient calculation to v1 form #687

matthewhoffman added in progress Ocean labels Mar 17, 2021

Conversation

vanroekel commented Sep 8, 2020

Uh oh!

xylar commented Sep 8, 2020

Uh oh!

milenaveneziani commented Sep 8, 2020

Uh oh!

jonbob commented Sep 8, 2020

Uh oh!

milenaveneziani commented Sep 8, 2020

Uh oh!

vanroekel commented Sep 8, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

darincomeau commented Sep 8, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

milenaveneziani commented Sep 8, 2020

Uh oh!

jonbob commented Sep 8, 2020

Uh oh!

milenaveneziani commented Sep 8, 2020

Uh oh!

darincomeau commented Sep 8, 2020

Uh oh!

xylar commented Sep 8, 2020

Uh oh!

mark-petersen Sep 9, 2020

Choose a reason for hiding this comment

Uh oh!

vanroekel Sep 10, 2020

Choose a reason for hiding this comment

Uh oh!

vanroekel commented Sep 10, 2020

Uh oh!

mark-petersen Sep 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vanroekel Sep 10, 2020

Choose a reason for hiding this comment

Uh oh!

mark-petersen commented Sep 10, 2020

Uh oh!

mark-petersen commented Sep 10, 2020

Uh oh!

mattdturner commented Sep 10, 2020

Uh oh!

mark-petersen commented Sep 10, 2020

Uh oh!

mark-petersen left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

vanroekel commented Sep 8, 2020 •

edited

Loading

darincomeau commented Sep 8, 2020 •

edited

Loading

mark-petersen Sep 10, 2020 •

edited

Loading