Versions Compared

Key

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

2024-08-08

JEDI Biweekly meeting

ALGO (Anna)

  • Setting up Skylab 3DVar with GEOS using GSI B (Clementine)
  • Continuous DA: implementing ObsSpace append functionality and using it to add observations to the outer loop iterations (Christian and Steve H)
  • ObsDataFrame integration with ObsSpace (Steve H and Phil)
  • Using spectral filter SABER block with mpas-jedi for multiscale covariances (Nate)
  • Simplifying code for variables and their metadata (including number of levels) and further planning for model variable naming convention (Steve V with Interfaces team)
  • Bitmap diagnostic flags option in UFO (Anna)
  • Planning improvements to the saber generic linear variable change block (calling vader) to make it more usable (Steve V and Anna).



Model Interface (Francois H)

  • Liam identified GCC-12 offload tools dependency issue, and created work around; worked on fine-tuning GPU offload environment set-up instructions; developed an GPU offload example in JEDI.
  • Francois H opened PRs using the generic interpolator for resolution changes in fv3jedi + mpasjedi. In initial testing, it appears this gives ~10% speed up in 3dvar test, and potentially more in 3denvar. Encourage partners to test this in their experiments.


Infrastructure (Ashley)

  • Updated observation ingest to use a S3 bucket instead of relying on Globus
  • Addressed a couple bug fixes in the ingest suite
  • There is a pending change for PyIRI-jedi that should let us add it to the jedi-bundle
  • Beta version is complete, end-to-end R2D2 V3 client and server system are under initial code review
  • Skylab is being updated to utilize the new R2D2 Python REST client
  • The creation of spack-stack-1.8.0 release candidates are targeted for Aug 19
  • Atlas upgrade to 0.38.1 will be included in spack-stack-1.8.0



Observation (Hui)

INGEST:

  • fix for the thickness predictor in the upgrader (not the converter) caught by Bill Campbell thanks Bill!
  • ionosonde decoder in collaboration with NRL thanks Dave! watch for review requests for IODA convention adaptation for the new observables
  • tech debt reduction. removal of deprecated code and ctest and files in ioda-converters thanks Youlong and Clara
  • a converter for data from the ICARTT campaign from 2004 was added by Maryam (if Jérôme doesn't mention it ;0)


OBS:

  • a PR for the UKMO contribution of a generic structure for a ground-based radar operator thank-you Chris Thomas, undergoing final reviews
  • the capability to write out a diagnostic bitmap was introduced piggy-backing off the CreateDiagnosticFlag filter. thank you to Anna Shlyaeva and to everyone for their prompt reviews


CRTM (Ben J)

  • Getting ready for CRTM 3.1.1 release, with several minor updates including a bug fix to surface emissivity/reflection calculation.
  • New experimental GMAO cloud coefficient tables are added to the fix folder.
  • New PolSIR instrument coefficients are released and under evaluation.
  • Multiple new features are under development for CRTM v3.2, including the updated TROPICS coefficients for polarization calculations, and CRTM calculations for air-craft level and ground-based sensors.



COMPO (Jerome):

  • Good progress on the composition cycling experiment. (Maryam)
  • Cross-team discussion on node allocation.


SOCA (Hamideh)

NASA

Ricardo

  • Comparison between JEDI/GSI, noticeable differences likely due to the adaptation of different observations, especially for IR/Ozone sensitive channels.
  • Comparing JEDI with GSI: convergence looks similar with many obs; but jedi “converges” faster than gsi consistently.
  • Skin temperature effects on the Jacobians.


Dorukhn Ardag – send an email requesting updates

  • Cycling marine DA in SWELL with FGAT using Static B
  • Updated our CI-Workflow to work with commit-hashes that allows us to pin a JEDI build for a certain date, in addition to our nightly develop tests.
  • Submitted a PR for GMAO/ODAS to IODA converter script update


NOAA (Ming Hu)

  • RRFS DA team from GSL and EMC are working on testing and evaluating the key JEDI functions, such as
    • Processing and filter observation data: bufr2ioda, ioda
    • Observers: ufo for surface, upper air, and radiance
    • Background and observation errors: static and ensemble BEC
    • Analysis methods: hybrid 3DEnVar, LETKF.
  • Several cases are added for RRFS in RDASApp repository:
    • MPAS JEDI 3DEnVar analysis over CONUS 12km grid with 30 MPAS ensemble forecasts
  • Syncing all source code in RDASApp submodules with JCSDA/jedi-bundle
  • Using single observation to test the ensemble BEC
    • Generate the bumploc files with mpasjedi_error_covariance_toolbox.x from current (Hash ee4e0ad on Mar 15, 2024) and synced Saber (Hash e93b14f on Jul 29, 2024)
    • found some difference in the pattern of the analysis increments when using current and synced BUMP. Working on finding the reason for the difference.



MMM (N/A)

NRL(N/A)

MetOffice (Phil Underwood)



Hernan G Arango: cross-team discussions