Versions Compared

Old Version 1

changes.mady.by.user Eric Lingerfelt

Saved on

compared with

New Version Current

changes.mady.by.user Eric Lingerfelt

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

JEDI weekly meeting notes


Andy/Land:
- Doubled checked the way of grouping observations for R2D2 after Jerome’s comment last meeting, and they are correct. Just using a different naming convention
- Using EWS to run EWOK on Orion and AWS
- Mainly working with NCAR RAL folks on putting together some basic diagnostics and doing some testing now that they have begun cycling experiments with WRF-Hydro-JEDI.

Maryam/COMPO:

- Able to run GFS w/aerosols under ewok, small H(x) experiment in Aug 2021 on Orion

Travis/Ocean:

- Getting soca working under ewok: currently have hofx done, working on 3dvar next. Hope to get this wrapped up by the end of the week
- Develop branch of soca is "broken", waiting on fixes to oops. The model-specific version of interpolators in oops is not handling halos correctly (see https://github.com/JCSDA-internal/oops/issues/1762)

Patrick,Chen/CRTM:

- Remaining coefficients for tropics constellation, 5/7 completed
- Surface emissivity coefficient table improvements (sea surface, snow surface)

Francois V.,Hui,Ryan/OBS:

- Code sprint this/next week: prepare UFO for JEDI release.
    - Improve the organization of ufo and in particular clean up the ctests. (have over 5000 lines of cmake code in ufo describing over 450 tests)
    - Reorganize the test directories and review all tests.
    - Remove duplicate tests and strip out repetitive sections of YAML that are tested elsewhere.
    - Classify each test into a test tier, move long-running tests, instrument tests and other non-unit tests into tier 2, and use the ufo_add_test macro to make test maintenance easier.
    - Once this is all done, test ufo in a broad range of environments.
- Test ufo in cycling experiments.
    - Adding obs types one after another, focusing on scientific performance, radiance bias correction, and observation errors.
    - Currently, have radiosonde, satwinds, aircraft, GNSSRO, ATMS, AMSU-A included.
- Also working on converters to ingest more data from different resources, which includes working with various providers to examine new missions/observations (e.g., commercial data, TROPICS proxy data, etc.).

Dom/Infrastructure:

- Infrastructure team planning out first 1.5 code sprints for release end of June
- spack-stack beta version planned for tomorrow/email will be sent Monday next week
- Heavy development on EWOK and get it to run through CI on Github self-hosted runners
- When final version of spack-stack is rolled out on all HPCs, in the CI system and for users
    - Need to plan for a hard switch from jedi-stack to spack-stack, since it requires changes in several repositories due to updates in library names etc.

Anna/Algorithms:

In terms of preparation for the release:

  • Two lots of observations are in R2D2

  • Ewok YAMLs are under review

  • Have to change long names in FV3-Jedi for soil moisture to match observation name, that PR is in too.

On the side, I’ve been trying to sort out the ATLAS interface changes from a few weeks ago with the WRF-Hydro interface

Jerome/COMPO:

- The goals for the skylab1 release are completed
- We can run H(x) with GFS-aerosols with VIIRS AOD using EWOK and R2D2

Travis/SOCA:

- Hamideh working on ocean color forward operator, OASIM, will be integrated into UFO in the next couple of months
- Kriti: ensemble DA experiments with regional ocean domain
- Travis: wrapping up skylab integration, I just need to gather all the data files, submit the required PRs, and coordinate with Eric to get the files where they belong.
- Also, we are still dealing with problems from the GetValues refactoring that was done several months ago, leaving our develop branch unusable for realistic workloads. I'm coordinating with Francois H to hopefully get a fix in OOPS very soon.

Patrick/CRTM:

Tropics is available now. Antaris did not make it to orbit. He already mentioned the UFO code sprint and the minor issues getting tests to run on Orion. Issue with kernel installation there. At CRTM monthly meeting, we will have invited speaker about cloud project from Europe.

Ben,Hui/OBS:

  • OBS team is testing CRTM2.4 for a potential release with UFO.
  • The preliminary results show differences with current UFO tests with CRTM2.3.
  • Currently OBS team is working on identifying the source of the differences.
  • One change from CRTM2.4 is that CRTM2.4 would simulate clear sky radiance once cloud_fraction is less than a minimum threshold, which is different from CRTM2.3. This would affect the sensors using the cloudy radiance. 
  • OBS team is communicating with CRTM as well as EMC team, who confirmed the differences.
  • There are also other changes/bug fixes from CRTM2.4 and their impacts are not yet clear on UFO tests.

Dom/Infrastructure:

- eckit/fckit/atlas issues tracked down to version mismatches when building atlas-0.29 in jedi-stack
- spack-stack release candidate 2 created 2022/06/23, expect final release next week
- Dom currently rolling out on target platforms: Discover (Intel,GNU), Orion (Intel, GNU), AWS Red Hat 8 / Ubuntu 20.04 (GNU), macOS (clang+gfortran)
- Includes everything to build jedi-fv3/mpas/um/ufs/-bundle, soca-bundle, ewok
- Figuring out a few last details for building certain packages
- Need to know if the JEDI release should provide default/reference versions for solo, r2d2, ewok or not (i.e. users always have to install themselves)
- NASA requested adding packages to spack-stack to build GEOS, use on Discover
- Great opportunity to get NASA software developers join spack-stack effort and get help with GEOS build
- Work on automated CI testing of EWOK using Github self-hosted runners
- Include testing feature branches of bundle and its components together automatically
- Not cloning whole bundle every time, just check for updates of existing clone and run
- Wrapping up R2D2 and EWOK work for release, new model for aerosol GFS backgrounds, added observations for land in R2D2, install Python simobs Python library- PRs streamlining interfaces between Geometry/State/Increment and atlas functionspace/fieldset are planned for merge on Monday June 13. All model interfaces would need to adapt, see details here: https://github.com/orgs/JCSDA-internal/teams/jedi-models/discussions/71. 
- PRs enabling assimilation when observed and simulated variables are planned for merge on Wednesday June 15. None or minimal changes are expected in yamls, but please test your model with this development. Details here: https://github.com/orgs/JCSDA-internal/teams/jedi-models/discussions/73
- Speed up and clean up fv3-jedi CI tests
    - Split YAML validation and run on single processor, run application w/o validating the YAML is faster than running both in the MPI app. Added option in ufo (oops)? to do this.
    - https://github.com/JCSDA-internal/oops/pull/1774
- Question from Hui: "We noticed some memory issues while running hyperspectral radiance data assimilation. This is also reported by NCAR for their MPAS-JEDI tests. In addition, we are also trying to sort out the issue related to the running clock time of GNSSRO operators, noticed by the JEDI team earlier."