JEDI Weekly Meeting

Thursday, August 28⋅9:00 – 10:00 am

JEDI:

Yannick: fieldSetOperation optimization broke SOCA for a while. Now it was fixed and PR was merged yesterday

ALGO: 

• From Nate: 
  
  • Finalized plan for initial prototype of diffusion-based correlated R (with Kriti)
  • Spinning up on Space Weather project, Need to add IODA and UFO as dependencies to PyIRI-JEDI, and update PyIRI-JEDI CI with this change
  • Finishing up hand-off of multi-scale B
• From Christian:
  • PR for sliding window continuous DA is ready
  • Beginning to add workflow in ewok to enable larger scale testing with with continuous DA
  • Removed parameters from the HTLM
  • Will begin adding some adaptive regularization to the HTLM

OBS (Benjamin Ruston):

• IODA converter:
  Draft PR from Raji, this build on the work contributed by MMM, and re-aligns the NCAR obs2ioda converter adding a c++ layer and some bugfixes.  Raji converted the make structure to match the convention used in JEDI, and removed dependency on an external yaml parser. https://github.com/JCSDA-internal/ioda-converters/pull/1695
• UFO:  
  • Replacing the SfcPcorrected with, SfcCorrected operator (UFO https://github.com/JCSDA-internal/ufo/pull/3779).  Please be aware, if you are using SfcPCorrected you would shift to this more general operator.
  • Reconstructed radiance PR for RTTOV https://github.com/JCSDA-internal/ufo/pull/3754, this allows for use of hyperspectral radiance reconstructed from PCs.  An equivalent effort for CRTM by Bryan K. is underway, meeting with Stefano and Mike Cooke occurred and moving along nicely.
  • NASA GMAO R. Todling, JCSDA, and NOAA EMC met to discuss the implementation of NBAM operator for dual use in the NOAA and NASA systems.  https://github.com/JCSDA-internal/ufo/pull/3778

NOAA: 

Daniel Holdaway mentioned HALF and BUMP memory issues.  Anna Shlyaeva mentioned the memory growth during MPI broadcast. These are under investigation

CRTM (Ben Johnson)

EMC noted an issue that under very rare circumstances, a negative radiance or inf/nan could be produced in CRTM.  It was noted that this occurs when graupel is present in the model data, and when using the GFDL CloudCoeff file in CRTM, and only occurs at 166 GHz (possibly higher).  

The root cause appears to be the truncation factors for delta fit are not normalized.    See issue https://github.com/JCSDA/CRTMv3/issues/241

A fix was provided to EMC which addressed the short term issue, but CRTM team will work on correcting the table and providing a comprehensive assessment of all of the cloud LUTs provided by PSU to ensure this problem does not occur in other tables.

GMAO (Ricardo Todling):

• Continue JEDI tests. Unfortunately, some tests doesn't generate reasonable results as they did in the past runs, e.g., skin temperature for IR
• Also tackled FV3 side issues passing pressure to NBAM operator. Now surface pressure is reasonable. Also wondered about the adjoint tests of this operator. Can't find adjoint tests for UFO operators (- Hernan recommended to perform single obs test). Also super-refrectivity QC doesn't generate the GMAO flavored results.

OAR/GSL (Ming Hu):

• working on improving MPAS-JEDI surface analysis with METAR observations: Run retrospective experiment shows positive impact from using 2m-T from the model as background to calculate surface T observation innovation; Investigate the impact and tuning terrain adjustment for surface T analysis.

• Run hourly GETKF ensemble data assimilation cycles with aircraft observations and radar reflectivity observations. Evaluate the impact of the ensemble DA on the ensemble forecast and the 3DEnVar.

• Test the GSIBec with a bug fixed version. Try to understand the impact of the static BEC and the hybrid analysis in MPAS-JEDI and compare the results to the similar test from GSI.

NRL

• Sarah King:
  • Working testing dual resolution, adding new obs, including satellites observing the mid-atmosphere. 
• Dom Heinzeller:
  • NRL has done limited testing with Intel oneAPI's ifx instead of ifort for the components of JEDI used in neptune-bundle; requires Intel oneAPI 2025
  • Continuing the debugging efforts on the memory growth discussed in detail in Executables built with spack-stack v 1.9.1 use significantly more memory · Issue #1588 · JCSDA/spack-stack 
• John Haiducek: 
  • Received help from Nate, Hui, Ben and Lindsey to get PyIRI integrated
  • PR from Nate to get PyIRI ctest added
  • Is searching a way to pass satellite position from obs space to model interface. (- Yannick recommended to look at the FOV implementation, which was implemented at both UFO and OPPS sides to get the FOV passed from UFO to model interface for radiance data)

MMM (Jake Liu):

We did some further development on rttovcpp+rttov14 interface in ufo. It now technically works Ok for microwave sensors like amsua and mhs with QC filters and VarBC. A few days cycling shows reasonable results.

ROMS (Hernan Arango):

• Working on 4D-Var, which is working well so far