JEDI Weekly Meeting

Thursday, August 14⋅9:00 – 10:00 am

General notes (Yannick Tremolet):

  • No announcements, only reports today.


OBS (Benjamin Ruston):

  • IODA converter:
    • Re-alignment of obs2ioda this specialized converter handles many NCEP BUFR types (and Himawari AHI) conversion to IODA. There were separate JCSDA and MMM branches this brings them together, first main commit with this alignment is done.  Cleanup of this finalized today.  Will now refactor make structure to make it consistent with pattern used in JEDI.
    • Update from Clara Draper on GHCN converter, realizing a massive speed-up 90x, making it less noisy
  • UFO:
    • RTTOV interface modifications for reconstructed radiances
    • https://github.com/JCSDA-internal/ufo/pull/3754
    • - had a meeting with UKMO (Stefano and Mike Cooke) along with JCSDA (BenR) and NASA (Bryan K), JCSDA and NASA will prepare CRTM equivalent.  UKMO worked to compartmentalize all changes into small section will require some interface specificity in RTTOV and CRTM, but actions like file I/O (reading external files like cmatrix) can be done using single routine.
    • https://github.com/JCSDA-internal/ufo/pull/3657
    • Observation operator that reduces the spatial density of requested GeoVaLs.  This allow high density observations to re-use the same GeoVaLs.
    • Small PR from UKMO regarding land DA: soilMoistureContentSurface, encouraging to see progress for this regime.   https://github.com/JCSDA-internal/ufo/pull/3761/files


CRTM (no attendance)

JEDI:

  • INFRA (Evan Parker):
    • SOCA integration
    • Eric Lingerfelt is back from furlough, currently working on the r2d2-api
  • ALGO (Christian Sampson):
    • Draft PR for continuous DA (sliding window)
    • Getting parameters out of OOPS
    • Correlated-R
    • Multi-scale B


Space-Weather (no attendance)

EMC (Daniel Holdaway):

  • Working towards GFSv17. Real time due to start next week.
  • oops PR that updates the base classes that are used for ensembles of states/increments to support more of the distributed ensemble capabilities (both in time and across ensemble members)? https://github.com/JCSDA-internal/oops/pull/2970
  • Several PRs for coupled DA.
  • Open PRs for the removal of Git LFS. Realized there is an issue with the way CMake does find_package for these data repos so looking into that.
  • GSIbec and MGBF regional capabilities are getting closer to PRs.
  • Working on GNSSRO with help from Ricardo.
  • Hybrid TLM work wrapped up by Lapenta Intern (Joey Rotondo, U.Washington)


GMAO (Ricardo Todling):

  • Aug 15, 2025 is the last day of Ron Gelaro at GMAO
  • Working on GNSSRO to match GSI increments
    • Found a bug and fixed it, now the increments look much more alike
    • JEDI still has considerably larger increments wrt GSI
    • Still working with NBAM (to match GSI)
  • ABI adjustments are coming in
  • A few changes to CrIS
  • Marine: no updates (Doruk was out of office)
  • Yannick's PR in OOPS (adding timers) helped profile the time spent on parts that were somewhat hidden (e.g., GSIBEC)


OAR (Ming Hu):

  • Testing surface observations (e.g., METAR)
    • Working on matching the terrain of the model and the stations
  • GETKF is working well in the tests
  • MGBF is part of the effort to be tested
  • 3DVar (ok), 3DEnVar (ok), but Hyb-3DEnVar not as good as expected, needs investigating (results are closer to 3DVar)
  • Working on a single analysis process, instead of one with radar-only, and another with all the other observations
  • Plan to share results soon


MetOffice (Michael Cooke):

  • Local volume solvers affected by Jo's work
  • Correlations for obs locations and on grid
  • Optimizing the code
    • OpenMP for GetGeovals (mostly affects radiances, looks good)
  • Added recipes for VADER


NRL

  • Sarah King:
    • No notice of memory increase
    • Working on the 4DVar system
    • A little behind on the obs usage side
  • Dom Heinzeller:
    • NRL is continuing efforts to spack-ify JEDI components (make them spack-installable so that they can be loaded as modules)
    • Mystery around CRTM v2 code for Intel oneAPI resolved; there are two tags with the same name but different code in the public and internal JCSDA repositories; according to Ben J, the public repo is the one to be used and the internal repo should be ignored (question from Dom: archive or delete?) 
    • Spack-stack memory increase from 1.6.0 to 1.9.1 on Derecho with GNU in mpas-jedi experiments: investigation ongoing, reverting all package version updates so that 1.9.1 uses the same versions as 1.6.0 did not fix the issue


MMM (Jake Liu):

  • Made a PR in UFO a couple of weeks ago to upgrade 'rttovcpp' operator from using rttov12 to using rttov14. I plan to meet with UKMO to see their plan and how we should proceed.
  • Benjamin Menetrier provided a solution to speed up in responding my request in https://github.com/JCSDA-internal/mpas-jedi/issues/1082. Significant speed up and memory use (at least 20%) by skipping re-initialization of BUMP and I/O from the second outer loop. It would be good if Benjamin's changes can be merged with PRs to open.


ROMS (Hernan Arango):

  • Question about averaging observations over time bins (perhaps a thinning filter?!)
  • No labels