INFRA
- IODA parallel reader - first functional version ready for testing around next week
- R2D2 interface improvements, kwarg validation soon, can relocate r2d2-experiments-localhost
- spack-stack - jedi - hercules - fckit shared pointer issue
- boost regex issue need to revert oops PR
- CI pipeline - found bug in caching; if errors found (old CI passes, new CI fails) please notify jedi-infra through the usual channels and tag @eap in the PR
- EWOK ingest suites for observations is progressing nicely, more and more obs types are added
- Dom on leave for the next three weeks
- use jedi-infra-support slack channel, jcsda-internal discussions, or forum; please avoid contacting individuals in the team directly
ALGO
- Removal of the call to model-analysis variable change from the 4D-Var and 3D-FGAT cost functions, and removal of a method in the Model class: https://github.com/orgs/JCSDA-internal/discussions/54 (merge today)
- Additional test output for Dirac for the localization: https://github.com/orgs/JCSDA-internal/discussions/56 (merge Monday July 24th)
- Pass Configuration instead of Parameters through repo interfaces: https://github.com/orgs/JCSDA-internal/discussions/55
- GlobalAtlasInterpolator replaced by GlobalInterpolator; changes for users that use it in yamls (and possibly code), e.g. when using saber Interpolation block and when using the feature of writing out output to lat/lon in a generic way: https://github.com/orgs/JCSDA-internal/discussions/58
- Need yaml change: key to switch interpolator
- Ricardo: Why the default can't be what it is now?
- Anna: The default depends on the situation/application, and previously one or the other was used (hardcoded); now you can switch for either of them.
- Hernan: Problem with interpolator with models on staggered grid
- Yannick: Decision was made in the past that model passes all variables on the same grid to avoid the complexity of dealing with different staggered grids in different models
- Dan: Can't interpolate winds on staggered grids, because they are tangential to the grid - requires rotation (too much to ask from generic code)
- Yannick working on improvements on 4denvar, Anna using that information for cross-covariances
- Bug fixes in bump concerning sampling in limited area models, new nicas normalization otion
- Progress on MPAS-Skylab
- Finalization of jedi-atlas interface requirements document
COMPO
- geos-cf cycling PR soon
- adding tracers to linear model
- tempo converters (series of PRs to integrate tempo data - for proxy data for now)
- omps suite of converters and others -
- CCPP standard names PR for new names from composition
- Working on using metplus to get score cards at JCSDA
- EMC is adopting the column retrieval operator that we are using with COMPO instead of the outdated atmvertinterplay. Right now we are trying to make sure the results are matching or are close enough to what they currently have for total ozone columns with the OMPS instrument.
MARINE
- Planning meeting for soca for this quarter
- Kriti: Evaluating qc for in-situ salinity profiles within Skylab
OBS
Ingest
- Work with INFRA on ingest suite
- EMC working on converters for all their bufr dump converters; JCSDA developed their own converters (some originally coming from NCAR), verifying against EMC converters, sharing back with MMM
- IODA meeting on July 19 (see IODA issue 1058): requesting groups of variables in obs space that shouldn't go to the output file (because only used internally etc.)
- Greg's upcoming change for ioda-converters, UFO: sensor scan position must be integer in the file (is float now)
- Need to convert and re-ingest about 6000 files in R2D2
- Met Office confirmed that they are ready, will need to verify with other centers
- If no issues raised, may merge next week
UFO/OBSPROC
- Work on supporting different versions of CRTM within UFO develop, because partners use JEDI for their own verification but older versions of CRTM
- CI/develop tests must support two versions of CRTM (detects CRTM version), coming soon
- Variational BC for conventional (non-channel-dependent) data types (mostly aircraft for now)
- Planning on code sprint in the week of Sep 11, scope will be worked out this/next week
- Incorporating features from GMAO on GNSSRO operators; making sure that generic code is working with specific parameters from GMAO
- Proposal PR ready today or tomorrow
CRTM
- v3 occasional underflow values in tests; fixed by forcing initial allocation to be zero (array being passed in and out of several routines)
- resolves long-standing problem with intermittently failing tests
- going through rest of code to identify similar issues (has a slight impact on performance)
- Getting sea surface emissivity model over from v2 to v3
- Working on radar PR from Isaac (massive PR)
- Expect changes in test references for v3 as elements of 2.4.1 are brought over
- Land surface emissivity changes in v3 lead to colder temperatures at surface, taking a closer look at land surface emissivity model
- JCSDA land lead one appointed needs to take over this work
- Patrick Stegmann leaving JCSDA, will be available for some time to answer questions around coefficient generation
- NESDIS star taking over some of the coefficient generation, new position at JCSDA will be opened soon
EMC
- Bugs related to B matrix calculation in SOCA (linear variable changes instead of SABER blocks)
- Increment off by a factor of two, trying to replicate in fv3-jedi since can't find bug in SOCA
- Anna: UK Met Office may have had the same problem, they fixed it but how?
- Olly: Problem was entirely on Met Office, bug in their code
- Ricardo: GSI increments may be wrong/half? But Guillaume has a simple test that shows that increment in SOCA is wrong (Dirac test)
- Once GPSRO UFO code ready, please notify EMC so that they can test it
Met Office:
- Observation processing: four types of global obs left, at least three to be ported in September
- ODB working on all variables in the obs space
- DA method: hybrid B progressing well, hold vertical level information and interpolate between saber blocks
- faster/simpler spectral cov calibraiton
- HTLM oops PR for hybrid linear model restructuring, optimizing ..., allows training of ... ?
- simplify linear model time step (factor of hybrid lin model time step)
- Control pert PR: responded to comments in PR, once approved will implement operastor in model ... control
- Cycling 3dvar w/ limited obs types
GMAO
- Matching GSI observation operators with CRTM v2.4, halfway done
- Using on-premises CI system: nightly run to build whole JEDI system to test UFO operators
- EVA work: code sprint in a few weeks on documents etc.
- Working on cycling ... (Ricardo) - PR to come, including GSIbec update
- should have a spack-stack update for that
ROMS
- Interested in BUMP features, but documentation is a disaster and debugging is difficult
- Many switches, flags with meaningless names
- Problems with isotropic correlation training
- Writing coefficients to ??? fails, bump not leveraging netCDF metadata information
- Debugging is difficult because bump is using Python (?) preprocessor
- Recommendation is to run separate preprocessor step in cmake and load those
- Analyzing sea-surface salinity
- Analyzing critical older 3dvar algorithms and porting to 4dvar
NRL
- Dual-resolution work progressing, IODA v3 converters merged, cylc suite improvements, adding diagnostics, evaluating iasi and metopb, ongoing HPC issues that have slowed things down
MMM
- Experimenting with letkf in mpas-jedi. Crashes when assimilating GNSSRO data, works fine in 4denvar