Obs Team:

No major changes

Still working on ingest and EMC is making progress on data conversion by Emily lieu

Verifying conversions and running them through variational (Sam and Lindsay)


OBSPROC: 

  • Quarterly review next week. Slides/bullets for quarterly reports are being collected from OBS core and in-kinds.

  • OBS core and in-kinds meeting was hosted yesterday (Wednesday 7/6/2023)

  • UFO NCEP Bending Angle Model (NBAM) operator was updated to match the latest NOAA GDAS v16.3. 

  • OBS team is setting up JEDI-skylab for the RO Modeling Experiment (ROMEX), which will incorporate 4 times operational RO data counts. This will be an opportunity to assess the performance of both JEDI and RO data assimilation.

  • Ongoing discussion: A bias correction sprint will be planned to incorporate VarBC for data types beyond radiance or channel-based data types, e.g., aircraft BC

  • scan position is described as an integer for the naming convention (JEDI-doc) but required as a float by UFO-CRTM. During yesterday's meeting, this issue was discussed and the decision was to have UFO handle the conversion. The IODA converters/naming convention will continue to use it as an integer.

  • Ongoing work on cloudy radiance using symmetric obs errors


ALGO:

Algorithms:

  • Merges in saber affecting BUMP users: https://github.com/orgs/JCSDA-internal/discussions/51 (merges next Monday and next Thursday)
  • 4DEnVar development: Yannick introduced 4D data structures for the cost function classes. The structures can handle 4D (with time), as well as handling different ensemble members. This allows to have different time slots (for 4DEnVar and weak-constraint 4DVar) either on different MPI communicators or on the same MPI communicator and would allow for use of time interpolation for H(x) within 4DEnVar. Next step: passing 4D Increments to the error covariances. Next step: correctly using 4D increments in the error covariances (to be summarized here: https://github.com/JCSDA-internal/oops/issues/2251). Currently the oops-level “Ensemble” covariance handles 4D aspects of 4DEnVar correctly, but saber covariances don’t and model-interface defined static covariances don’t. We’ll introduce 4D aspects in both places, with the following assumptions:
    • 3D localization is the same at every time step (already in assumption in Ensemble covariances in oops)
    • 3D static B (when using hybrid) is the same at every time step
    • Model-interface defined covariances are static covariances (i.e. do not introduce ensemble covariance with localization inside)
  • Cost functions: removal of variable change between model and control variables call in oops. Affects all 4DVar (both strong- and weak-constraint) users: the variable change would need to be called from the Model class within the model interface. Announcement to follow when the developme
  • Coupled development: introduced Increment class implementation for generically coupled models.
  • UFS with FGAT: updates with develop and instructions on building and running Skylab-like experiment (work on branches outside of develop branches currently)
  • HTLM: Removed unnecessary setting up of Increments in the code. Brought hybrid TLM training and use into same class and remove duplicate simplified linear model 
  • VADER: Had a discussion based on comments from Marek and Mayeul about the difficulties of using VADER for linear variable changes in the B. To make it easier, we are anticipating introducing separate setTrajectory method for the VADER recipes (that can be different than the call to the non-linear variable change) and ensuring it’s possible to have different trajectory and increment variables. https://github.com/JCSDA-internal/vader/issues/184
  • Progress on the document for atlas <-> jedi interfaces


Infrastructure:

Skylab v5 release
- Release branches are being merged into develop today (7/6/2023). Major change is that CRTMv3 (develop) is now built as part of the bundles.

spack-stack
- The spack-stack-1.4.1 release is being rolled out. This release is primarily for the UFS community, we won’t transition to it. But we can use it for testing the ioda-converters PR that needs bufr@12.
- New spack-stack installation on NRL’s Nautilus platform ready for testing. Uses Intel MKL instead of openblas and fixes problem with the MPI library on Nautilus.
- GMAO merged first set of libraries required for building GEOS into spack-stack. A second PR is anticipated, after which we should be able to build GEOS in spack-stack.

FMS
- FMS branch based on latest develop that includes the changes needed by SOCA is ready for testing. Once released it will unblock a lot of work on unifying FMS, FV3 and UFS weather model branches.

MPAS
- Fix for MPAS with external ESMF is available (set of three PRs). This needs to be reviewed and merged to be able to use the JEDI CI containers for MPAS, FV3 and UFS.

Observation ingest suite in EWOK
- Need to find a way to maintain the NCAR obs2ioda converters and make them part of the JCSDA ioda-converters (work with MMM)



Ben johnson

Still making finishing touches to coeffs for noaa 21, provided update to EMC

Fix polarization basis hopefully, testing now

Also added satellite id updates in files

Ensuring crtm v3 working correctly technical failures have mostly been resolved may be open mp accuracy issues will test to make sure they are the same and look for edge cases.

CRTM 3 major changes over next several months especially radar definitely need more assistance on the surface emissivity side for land and sea ice.

If anyone has experience with land emissivity and want to help please reach out.




  • No labels