Skip to content
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.

Sign up for GitHub

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

Jump to bottom

FGAT Function Test with 3DEnVar #7

Open
XuLu-NOAA opened this issue Jun 12, 2024 · 2 comments
Open

FGAT Function Test with 3DEnVar #7

XuLu-NOAA opened this issue Jun 12, 2024 · 2 comments
Assignees
@XuLu-NOAA

Comments

@XuLu-NOAA
Copy link
Collaborator

XuLu-NOAA commented Jun 12, 2024 •
edited
Loading

This is to record the test of FGAT function with 3DEnVar in JEDI using HAFS.
(1) A 10-member GDAS ensemble is first generated through atm_init_ens onto HAFS grid.
(2) A single dropsonde T observation is generated.
(3) Modify the following parts of the yaml file from 3DEnVar single ob test for 3DFGAT:

  1. cost type: 3D-Var -> cost type: 3D-FGAT
  2. background:
    datetime: 2020-08-25T12:00:00Z
    -> background:
    datetime: 2020-08-25T09:00:00Z
  3. Add pseudo model configuration:
    model:
    name: PSEUDO
    tstep: PT3H
    filetype: fms restart
    filename is datetime templated: true
    datapath: Data/bkg
    filename_core: '%yyyy%mm%dd.%hh%MM%ss.fv_core.res.tile1.nc'
    filename_trcr: '%yyyy%mm%dd.%hh%MM%ss.fv_tracer.res.tile1.nc'
    filename_sfcd: '%yyyy%mm%dd.%hh%MM%ss.sfc_data.nc'
    filename_sfcw: '%yyyy%mm%dd.%hh%MM%ss.fv_srf_wnd.res.tile1.nc'
    filename_cplr: '%yyyy%mm%dd.%hh%MM%ss.coupler.res'
  4. Add time interpolation in observations/observers:
    get values:
    time interpolation: linear

The OMB differences suggest the FGAT is working, but the OMA looks suspicious:
image
@XuLu-NOAA XuLu-NOAA self-assigned this Jun 12, 2024
BinLiu-NOAA pushed a commit that referenced this issue Jun 18, 2024
@guillaumevernieres
Optionally run specified rocoto task as part of the ctests (#1121)
af7fb73 
Only tested with the coupled UFS and the marine DA tasks:
```
Test project /scratch1/NCEPDEV/stmp2/Guillaume.Vernieres/runs/prs/global-workflow/sorc/gdas.cd/build/gdas/test/gw-ci
  Test  #1: C48mx500_3DVarAOWCDA
  Test  #2: C96_atmaerosnowDA
  Test  #3: C96C48_ufs_hybatmDA
  Test  #4: C48mx500_3DVarAOWCDA_gdasfcst_202103241200
  Test  #5: C48mx500_3DVarAOWCDA_gdasprepoceanobs_202103241800
  Test  #6: C48mx500_3DVarAOWCDA_gdasocnanalprep_202103241800
  Test  #7: C48mx500_3DVarAOWCDA_gdasocnanalbmat_202103241800
  Test  #8: C48mx500_3DVarAOWCDA_gdasocnanalrun_202103241800
  Test  #9: C48mx500_3DVarAOWCDA_gdasocnanalchkpt_202103241800
  Test #10: C48mx500_3DVarAOWCDA_gdasocnanalpost_202103241800

Total Tests: 10
```
BinLiu-NOAA pushed a commit that referenced this issue Jun 18, 2024
@guillaumevernieres @CoryMartin-NOAA
Addin a GFSv17 ctest (#1130)
4598d91 
**What was done:**
- reorganization/code tidy to facilitate the addition of a GFSv17
prototype ctest
- exclude the new ctest by default, the user will have to run ```cmake
-DRUNGWCI=ON .``` on an old build to configure the new ctest

He're the current list:

```
 Test project /scratch1/NCEPDEV/stmp2/Guillaume.Vernieres/runs/prs/global-workflow/sorc/gdas.cd/build/gdas/test/gw-ci
  Test  #1: WCDA-3DVAR-C48mx500
  Test  #2: WCDA-3DVAR-C48mx500_gdasfcst_202103241200
  Test  #3: WCDA-3DVAR-C48mx500_gdasprepoceanobs_202103241800
  Test  #4: WCDA-3DVAR-C48mx500_gdasocnanalprep_202103241800
  Test  #5: WCDA-3DVAR-C48mx500_gdasocnanalbmat_202103241800
  Test  #6: WCDA-3DVAR-C48mx500_gdasocnanalrun_202103241800
  Test  #7: WCDA-3DVAR-C48mx500_gdasocnanalchkpt_202103241800
  Test  #8: WCDA-3DVAR-C48mx500_gdasocnanalpost_202103241800
  Test  #9: Aero-Snow-3DVAR-C96
  Test #10: Aero-Snow-3DVAR-C96_gdasfcst_202112201200
  Test #11: Atm-hyb-C96C48
  Test #12: Atm-hyb-C96C48_gdasfcst_202402231800
  Test #13: GFSv17-3DVAR-C384mx025
  Test #14: GFSv17-3DVAR-C384mx025_gdasfcst_202106300000
  Test #15: GFSv17-3DVAR-C384mx025_gdasprepoceanobs_202106300600
  Test #16: GFSv17-3DVAR-C384mx025_gdasocnanalprep_202106300600
  Test #17: GFSv17-3DVAR-C384mx025_gdasocnanalbmat_202106300600
  Test #18: GFSv17-3DVAR-C384mx025_gdasocnanalrun_202106300600
  Test #19: GFSv17-3DVAR-C384mx025_gdasocnanalchkpt_202106300600
  Test #20: GFSv17-3DVAR-C384mx025_gdasocnanalpost_202106300600
  Test #21: GFSv17-3DVAR-C384mx025_gdasocnanalvrfy_202106300600
  Test #22: GFSv17-3DVAR-C384mx025_gdasprep_202106300600
  Test #23: GFSv17-3DVAR-C384mx025_gdasanal_202106300600

Total Tests: 23
```

---------

Co-authored-by: Cory Martin <[email protected]>
@XuLu-NOAA
Copy link
Collaborator Author

Update the ensemble and background using the corrected halo file according to Issue #11. The OMA/OMB table is updated with JEDI 4DEnVar and GSI FGAT as a reference.
image
Still having concerns regarding the OMA in this FGAT setup. A help issue is submitted to https://github.com/JCSDA-internal/fv3-jedi/issues/1233

@TingLei-NOAA TingLei-NOAA mentioned this issue Jul 2, 2024
Open
@XuLu-NOAA
Copy link
Collaborator Author

After comparing the yaml file between mine and NOAA-EMC/RDASApp#100, it appears the FGAT concern is due to the "over fit" of using two inner loops. If I only do a single inner loop, the results appears to be more reasonable:
image

The issue appears to be resolved for now. But it remains a question why "overfit" would create a larger OMA than OMB. Also it might be questionable why the OMA changes could be much larger than the OMB changes in FGAT09 over FGAT03.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@XuLu-NOAA XuLu-NOAA

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@XuLu-NOAA