Update gcycle to be able to read re-gridded GSI increment land files.#1010
Conversation
|
@GeorgeGayno-NOAA Submitting this as a draft for now, while I do the tests. I'm not sure which are the unit tests and which are the consistency tests? I'm currently running rt.sh, with all the tests commented out except global_cycle. |
The regression tests are in The unit tests may be run by some slight modifications to the You may also need to update |
|
All 5 consistency checks passed on hera. |
|
@GeorgeGayno-NOAA How do I confirm that the ctests passed. The first pass through, they didn't and I fixed it. build_all.sh compiled, but I can't find a report of the test results. |
When I test your branch on Hera, I see one unit test failing: The test results will be in ./build/Testing/Temporary/LastTest.log. The test fails because of a missing file: I can take a look. |
|
@GeorgeGayno-NOAA Is that from standard out? I think I'm not actually running the tests. Let me follow up on that. |
That is the output from the |
Just came here to say I found the ctest line! I'm running it now. I'll sort out the failing test. |
|
I fixed it. The reg_tests and ctests pass on hera, and gnu compilation works on hera. Converting from a draft. Yuan Xue and Tseganeh Gichamo would be good reviewers. |
GeorgeGayno-NOAA
left a comment
There was a problem hiding this comment.
Do we need another regression test for a GSI case on the native grid?
i) write out the ensemble mean reanalysis, and apply it to the deterministic member. ii) replace UFS_UTILS regridding of Gaussian soil analysis to model grid with an external ESMF-based program (with appropriate land masking) Requires: sorc/gsi_enkf.fd develop after Tue Nov 12. Latest works. sorc/gdas.cd/sorc/da-utils PR: NOAA-EMC/DA-utils#5 sorc/ufs_utils.fd from PR: ufs-community/UFS_UTILS#1010
|
Overall, looks good to me. Thanks! |
|
@ClaraDraper-NOAA - make sure you have merged the latest updates from 'develop'. |
There was a problem hiding this comment.
Looks good to me.
@ClaraDraper-NOAA I am guessing the "gaussian_to_fv3_interp" function will soon be replaced by the ESMF re-gridding code you just wrote?
To expand on my comment, there are two regression tests for applying land increments: |
|
Is logical |
done. |
Yes. Please do. |
This PR effectively adds the option to do either. However, the ESMF regridding is more appropriate - so perhaps in the future we can just delete the gaussian_to_fv3_interp as you suggest. |
I just added a new test, and have checked it passes on hera. I added the new test to the drivers for the other machines, except s4 which looks a bit different to the others. The additional files that you need to add to the COMIN are: Examples of the baseline files are here: |
@GeorgeGayno-NOAA Yes, do_lndinc is redundant, since it can be turned on by the do_*_inc options. However, my preference would be to keep it (together with the DO_SFCCYCLE and DONST) to make it clear which actions (adding land increments, updating land model parameters, and/or updating NST) cycle if being called for. It's up to you though - if you'd like me to remove it, let me know. We could either completely remove it, or remove it from the namelists, and set it internally. |
|
@GeorgeGayno-NOAA I think I've addressed everything, let me know if there's anything else I can do. |
Ok. We can keep it. But I would update the script and code prologs, so the general user knows how to set each variable. For example, I am guessing if 'do_soi_inc' is set to true, then 'do_lndinc' must also be set to true. |
| # DO_SOI_INC_GSI Call routine to update soil states with gsi(gaussian) increment files | ||
| # DO_SNO_INC_JEDI Call routine to update snow states with jedi increment files | ||
| # DO_SOI_INC_JEDI Call routine to update soil states with jedi increment files | ||
| # DO_SOI_INC Call routine to update soil states |
There was a problem hiding this comment.
I would add more detail on how to set each logical.
There was a problem hiding this comment.
I've updated the comments, and added some logic into the global_driver.sh to set DO_LNDINC to true if either of DO_SNO_INC or DO_SOI_INC it true.
Great. I will host these files in the official location on each machine. |
Don't worry about s4. That is not an officially supported machine. Only WCOSS2, Jet, Hera, Hercules and Orion are supported.
|
|
I placed the new baseline files on Hera: and the new input files here: I will push the data to the other machines after we verify everything is working. |
|
@GeorgeGayno-NOAA Apologies for the poor PR behaviour. I just pushed a minor change to a variable name in the scripts (INTERP_LNDINC to GCYCLE_INTERP_LNDINC), since this makes the function clearer in global_workflow. The tests pass on hera with this change. |
|
I found a bug in |
I fixed this. |
|
@GeorgeGayno-NOAA I've addressed everything. |
|
The The input files are here: |
I updated to files to fix this. The new ones are here: /scratch2/BMC/gsienkf/Clara.Draper/gerrit-hera/PRprep/NEWTEST_UFS_UTILS/UFS_UTILS_COMIN $ ncdump -c soil_xainc.006 |
|
@ClaraDraper-NOAA - 'develop' was updated on Friday. Please merge these changes to your branch. They should not conflict with your PR. |
done. |
DESCRIPTION OF CHANGES:
Generalized the reading in of land increments to be:
Native model grid files may come from JEDI, or be from GSI and been regridded using a separate program. The read routine will detect which.
Also, changed the number of soil levels to be updated from 3 to 2. This will change the test output for all tests adding soil increments.
TESTS CONDUCTED:
If there are changes to the build or source code, the tests below must be conducted. Contact a repository manager if you need assistance.
global_cyclecode had no warnings.global_cycleconsistency tests locally on all Tier 1 machines. Done using 4d6403f. The new test ran to completion with no errors. The other tests all passed as expected.Describe any additional tests performed.
DEPENDENCIES:
None.
DOCUMENTATION:
N/A
ISSUE:
Fixes #1008