Add Langmuir mixing parameterization via CVMix #134

qingli411 · 2019-01-16T22:51:21Z

This merge adds Langmuir mixing parameterization via CVMix

Change the CVMix repository to use the 'theory-wave' approximation of Li et al., 2017 to estimate the Langmuir enhancement factor from 10-meter wind, surface friction velocity and boundary layer depth
New entries in mpaso namelist allowing different options of Langmuir mixing parameterization

MPAS-Analysis:

G-case without Langmuir turbulence parameterization

config_cvmix_kpp_langmuir_mixing_opt = 'NONE'
config_cvmix_kpp_langmuir_entrainment_opt = 'NONE'

G-case with Langmuir (LWF16)

config_cvmix_kpp_langmuir_mixing_opt = 'LWF16'
config_cvmix_kpp_langmuir_entrainment_opt = 'LWF16'

G-case with Langmuir (LF17)

config_cvmix_kpp_langmuir_mixing_opt = 'LWF16'
config_cvmix_kpp_langmuir_entrainment_opt = 'LF17'

B-case with Langmuir (LF17, 40 years)

config_cvmix_kpp_langmuir_mixing_opt = 'LWF16'
config_cvmix_kpp_langmuir_entrainment_opt = 'LF17'

1. Change the CVMix repository to use the 'theory-wave' approximation of Li et al., 2017 to estimate the Langmuir enhancement factor from 10-meter wind, surface friction velocity and boundary layer depth 2. New entries in mpaso namelist allowing different options of Langmuir mixing parameterization

vanroekel

@qingli411 this looks good. I have just two requests. Let me know if you have any questions

vanroekel · 2019-01-25T15:29:43Z

src/core_ocean/shared/mpas_ocn_vmix_cvmix.F

@@ -872,6 +917,9 @@ subroutine ocn_vmix_cvmix_init(domain,err)!{{{
      call mpas_pool_get_config(ocnConfigs, 'config_cvmix_convective_diffusion', config_cvmix_convective_diffusion)
      call mpas_pool_get_config(ocnConfigs, 'config_cvmix_convective_viscosity', config_cvmix_convective_viscosity)
      call mpas_pool_get_config(ocnConfigs, 'config_cvmix_kpp_use_enhanced_diff', config_cvmix_kpp_use_enhanced_diff)
+      call mpas_pool_get_config(ocnConfigs, 'config_cvmix_kpp_langmuir_mixing_opt', config_cvmix_kpp_langmuir_mixing_opt)
+      call mpas_pool_get_config(ocnConfigs, 'config_cvmix_kpp_langmuir_entrainment_opt', &
+                                config_cvmix_kpp_langmuir_entrainment_opt)


please add a check below that user input values are acceptable values of your two character strings. You can follow L1022 and following for an example check.

@vanroekel Added.

vanroekel · 2019-01-25T15:31:00Z

src/core_ocean/shared/mpas_ocn_vmix_cvmix.F

+            else
+               langmuirEnhancementFactor = 1.0_RKIND
+            end if
+


please move this down after L651, you may miss some boundary layer changes in your computation

@vanroekel Done.

vanroekel · 2019-01-25T15:36:04Z

@mark-petersen two questions on this PR

this should be BFB with ocean/develop as this is off by default and will be climate changing if on. How do you want to test this for merging to MPAS-Model? @qingli411 has run a core cycle with these changes and 40 years of a B-case. Results are as expected.
is it okay to merge this even though we have to point to @qingli411 personal cvmix repo? He has issued a PR into cvmix for the changes. Also asking @jonbob

1. Check the values of parameters 'config_cvmix_kpp_langmuir_mixing_opt' and 'config_cvmix_kpp_langmuir_entrainment_opt' 2. Move the lines that update Langmuir enhancement factor after the update of boundary layer depth

mark-petersen · 2019-01-25T18:29:05Z

@qingli411 and @vanroekel thanks for your work here.

Please conduct a bfb comparison in MPAS-Ocean that exercises your code options, for two tests:

exact restart
comparison across different partition numbers

that will catch any issues with initialization, restart fields, and halo updates. I know the odds of those mistakes are low on a new cvmix option, but these are easy tests to do. You can use six hour tests at QU240, which are in COMPASS.

cd testing_and_setup/compass/
./list_testcases.py |grep restart
      113: -o ocean -c global_ocean_V1 -r QU240 -t restart_test
./list_testcases.py |grep blocks
      115: -o ocean -c global_ocean_V1 -r QU240 -t se_blocks_test
./setup_testcase.py -f general.config.ocean_turq --work_dir DIR -n 115

etc. Or you can just use your standard test right now, but write a restart file at the end and use ncdiff for comparison.

Also run your standard test with both a gnu and intel compiler in debug mode, which will catch things you don't see without debug.

mark-petersen · 2019-01-25T19:01:17Z

Also good to run a performance test to catch anything major. Run your standard test with standard cvmix and your new option. Compare the vertical mixing timers at the end of the ocean log file. We should know if time is much larger. This is the time to catch major performance changes, like loops nested in the wrong order.

mark-petersen · 2019-01-25T19:14:20Z

src/core_ocean/get_cvmix.sh

+CVMIX_GIT_HTTP_ADDRESS=https://github.com/qingli411/CVMix-src.git
+CVMIX_GIT_SSH_ADDRESS=git@github.com:qingli411/CVMix-src.git
+CVMIX_SVN_ADDRESS=https://github.com/qingli411/CVMix-src/tags
+CVMIX_WEB_ADDRESS=https://github.com/qingli411/CVMix-src/archive


We can't use @qingli's repo here, as then all mainline E3SM simulations would depend on a nonstandard CVMix tag. Worse, we need to be able to go back to any previous E3SM/master commit and run it in future years. That means all submodules also need to be permanent.

The solution is to simply not make the above changes. We can still merge this PR. In order to run with @qingli411's non-default option, one would need to make the above change with your local E3SM repo copy.

After Qing's changes are merged into CVMIX, we make an MPAS PR to update the above CVMIX_TAG line.

qingli411 · 2019-04-08T20:22:59Z

@vanroekel and @mark-petersen, here are some updates on this PR:

My changes to CVMix have been merged into CVMix with tag v0.94b-beta. I have updated the git repo of CVMix in this PR.
I've conducted two tests (six hour tests at QU240 in COMPASS) on Grizzly with target 'ifort-gcc': 1) exact restart and 2) comparison across different partition numbers, which are all passed.
I've also compared the vertical mixing timers in these two tests and the timers are not significantly different from each other (within 3%). But I haven't tried comparing the timers in a longer run.

qingli411 · 2019-04-09T16:05:52Z

I just checked the vertical mixing timers (vmix imp) in my G test cases listed above. It takes 6.9% and 13.8% longer than the standard CVMix for vertical mixing with the first (LWF16) and the second (LF17) options, respectively, in a 5-year run. This is closer to what I was expecting, as there are additional computations in the second option and it should take longer time than the first option.

vanroekel · 2019-04-09T16:21:57Z

@qingli411 could you also include the percentage increase in the overall ocean timestep timer? I can't remember how much of the timestep is CVMix.

qingli411 · 2019-04-09T16:37:36Z

@vanroekel For ocean timestep timer ('o:se timestep') it's 0.16% and 0.64% increase respectively for the two options.

vanroekel · 2019-05-08T13:17:48Z

@mark-petersen is there anything @qingli411 and I can do to move this PR along? It should be ready to go.

mark-petersen · 2019-05-08T13:21:12Z

I am merging in changes that are bfb on all E3SM tests. See order here:
#229 (comment)
Should this one go in now? I would put it at the end.

mark-petersen · 2019-05-08T13:21:55Z

@vanroekel OK, just saw your comment. Will test and merge.

vanroekel · 2019-05-08T13:23:13Z

It is fine to wait till after the BFB PRs in #229. Since this is climate changing it should probably go on its own and on its own to E3SM.

mark-petersen

Merged into ocean/develop locally. Tested on grizzly, gnu/intel, opt/debug, nightly regression suite. Passes all. Merging in.

vanroekel self-assigned this Jan 16, 2019

vanroekel added feature Ocean labels Jan 16, 2019

vanroekel self-requested a review January 16, 2019 22:52

vanroekel suggested changes Jan 25, 2019

View reviewed changes

Add parameter check for Langmuir turbulence parameterization

68accd3

1. Check the values of parameters 'config_cvmix_kpp_langmuir_mixing_opt' and 'config_cvmix_kpp_langmuir_entrainment_opt' 2. Move the lines that update Langmuir enhancement factor after the update of boundary layer depth

mark-petersen reviewed Jan 25, 2019

View reviewed changes

Update git repo for CVMix

11c949d

vanroekel approved these changes May 8, 2019

View reviewed changes

vanroekel mentioned this pull request May 15, 2019

Update CVMix after recent release with python 3 support #250

Closed

xylar mentioned this pull request May 15, 2019

Update CVMix to 0.95-beta #251

Merged

mark-petersen mentioned this pull request May 29, 2019

Update mpas-source for Langmuir ocean vertical mixing parameterization E3SM-Project/E3SM#2952

Merged

mark-petersen self-requested a review May 30, 2019 14:50

mark-petersen approved these changes May 30, 2019

View reviewed changes

mark-petersen merged commit 11c949d into MPAS-Dev:ocean/develop May 30, 2019

matthewhoffman added feature Ocean labels Mar 17, 2021

qingli411 deleted the langmuir_mixing branch May 4, 2021 21:32

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Langmuir mixing parameterization via CVMix #134

Add Langmuir mixing parameterization via CVMix #134

qingli411 commented Jan 16, 2019 •

edited

vanroekel left a comment

vanroekel Jan 25, 2019

qingli411 Jan 25, 2019

vanroekel Jan 25, 2019

qingli411 Jan 25, 2019

vanroekel commented Jan 25, 2019

mark-petersen commented Jan 25, 2019 •

edited

mark-petersen commented Jan 25, 2019

mark-petersen Jan 25, 2019 •

edited

qingli411 commented Apr 8, 2019

qingli411 commented Apr 9, 2019

vanroekel commented Apr 9, 2019

qingli411 commented Apr 9, 2019

vanroekel commented May 8, 2019

mark-petersen commented May 8, 2019

mark-petersen commented May 8, 2019

vanroekel commented May 8, 2019

mark-petersen left a comment

Add Langmuir mixing parameterization via CVMix #134

Add Langmuir mixing parameterization via CVMix #134

Conversation

qingli411 commented Jan 16, 2019 • edited

vanroekel left a comment

Choose a reason for hiding this comment

vanroekel Jan 25, 2019

Choose a reason for hiding this comment

qingli411 Jan 25, 2019

Choose a reason for hiding this comment

vanroekel Jan 25, 2019

Choose a reason for hiding this comment

qingli411 Jan 25, 2019

Choose a reason for hiding this comment

vanroekel commented Jan 25, 2019

mark-petersen commented Jan 25, 2019 • edited

mark-petersen commented Jan 25, 2019

mark-petersen Jan 25, 2019 • edited

Choose a reason for hiding this comment

qingli411 commented Apr 8, 2019

qingli411 commented Apr 9, 2019

vanroekel commented Apr 9, 2019

qingli411 commented Apr 9, 2019

vanroekel commented May 8, 2019

mark-petersen commented May 8, 2019

mark-petersen commented May 8, 2019

vanroekel commented May 8, 2019

mark-petersen left a comment

Choose a reason for hiding this comment

qingli411 commented Jan 16, 2019 •

edited

mark-petersen commented Jan 25, 2019 •

edited

mark-petersen Jan 25, 2019 •

edited