-
Notifications
You must be signed in to change notification settings - Fork 191
Enable reduced ensemble size for early cycle in marine DA #3628
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: develop
Are you sure you want to change the base?
Enable reduced ensemble size for early cycle in marine DA #3628
Conversation
@AndrewEichmann-NOAA |
@aerorahul The GDASApp PR is NOAA-EMC/GDASApp#1650 - If you think we're ready to test I can merge it - it will break bmat in GDASApp so there should be some coordination |
@aerorahul I should clarify - If the g-w PR is merged before the GDASApp PR, I expect it will not break, but the marine bmat won't rotate the background ensemble members that it copies from COMROOT. Merging the GDASApp PR without the g-w PR will make bmat fail. The GDASApp PR will be ready to test when the g-w PR is merged. |
Experiment C48mx500_3DVarAOWCDA FAILED on Hera in Build# 2 with error logs:
Follow link here to view the contents of the above file(s): (gfs_marinebmat.log) |
Experiment C48mx500_3DVarAOWCDA FAILED on Hera in Build# 2 in |
Experiment C96C48mx500_S2SW_cyc_gfs FAILED on Hera in Build# 2 with error logs:
Follow link here to view the contents of the above file(s): (enkfgfs_earc_tars_01.log) |
Experiment C96C48mx500_S2SW_cyc_gfs FAILED on Hera in Build# 2 in |
Got it @AndrewEichmann-NOAA. Did you run a test at some point with both PRs together? |
@CatherineThomas-NOAA Yes - it produced an |
I started an experiment to confirm things behave as expected |
@CatherineThomas-NOAA I ran it again with the updated branches in global-workflow and GDASApp and it produces 30 ensemble members, visible here: |
Thanks for rerunning that test @AndrewEichmann-NOAA. Looks right to me as well. |
Rerunning experiment on Hera: |
Experiment ran several cycles, then |
I see it failed twice both in 202103260000 and 202103261800. I see that this PR: #3651 passed hera but the recent PRs tend to be tested on gaea and wcoss2. I can try this on gaea since I think you are still getting access there -- let me know if this would be helpful. I would usually ask @CatherineThomas-NOAA about a met job as a first pass but she's on leave today. @KateFriedman-NOAA would you be familiar enough to point us in the right direction? |
@JessicaMeixner-NOAA @AndrewEichmann-NOAA taking a look at the metplus job failures... |
@AndrewEichmann-NOAA @JessicaMeixner-NOAA I see that @AndrewEichmann-NOAA's test is a modified configuration of the |
This experiment was a |
@AndrewEichmann-NOAA - while the offset is not tested, a CI test with early cycle turned on (only 2 ensemble members) is https://github.com/NOAA-EMC/global-workflow/blob/develop/dev/ci/cases/pr/C96C48mx500_S2SW_cyc_gfs.yaml |
@JessicaMeixner-NOAA Would it be ok to run this with 80/30 members? |
@aerorahul @KateFriedman-NOAA - @AndrewEichmann-NOAA has updated the PR description to clear up the confusions. For next steps, does @AndrewEichmann-NOAA need another CI test without modifications? Otherwise, I think this is ready for CI. If we'd like another review from @CatherineThomas-NOAA we'll need to wait for her return tomrrow. |
Please let me know if this is ready for CI. We just merged a number of PR's, most impactful for this is #3642. |
@aerorahul I have a C96C48mx500_S2SW_cyc_gfs run that failed overnight due to quota overruns. I restarted and then ran into problems with the repo not updating right. I'll rebuild with the current merge and start again. |
4eb24dd successfully completed |
@aerorahul Ready to roll for CI testing |
Description
Enables reduced ensemble size for early cycle in marine DA by adding math to marine DA job scripts to reassign the ensemble size and allow for rotating sampling the background of the full ensemble.
Partially resolves NOAA-EMC/GDASApp#1588 - once NOAA-EMC/GDASApp#1650 is merged, the marine DA will use
NMEM_ENS_GFS
instead ofNMEM_ENS
for the number of ensemble members for the early cycle.Type of change
Change characteristics
How has this been tested?
Multi-model day cycling + forecast test on Hera, expdir here
/scratch1/NCEPDEV/da/Andrew.Eichmann/fv3gfs/reduced-ens/global-workflow/sorc/gdas.cd/build/gdas/test/gw-ci/C48mx500_hybAOWCDA/EXPDIR/C48mx500_hybAOWCDA
COMROOT here
/scratch1/NCEPDEV/stmp4/Andrew.Eichmann/COMROOT/C48mx500_hybAOWCDA
with examination of enkfgfs ensemble and gfs_marinebmat.log to check for proper behavior.
Checklist