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