Infra:
Evan
- cylc project: a majority of our skylab suites work running cylc as the backend
- Spack stack release plan:
o Feature branch release plan today.
o One week for testing.
o All sites supported
Dom:
This is for INFRA / spack-stack v2 timeline
- Feature freeze Dec 4, roll out of first release candidate Dec 5
- Acceptance testing / bug fixes until Dec 12, then roll out of release until Dec 19 with
final announcement of spack-stack-2.0.0 on Dec 19
- Not all HPCs will be supported in spack-stack-2.0.0 - a follow-up release 2.1.0 is
expected for Jan/Feb 2026 to add support for these sites
Interfaces:
Steve H.:
- Testing and bug fixing has resulted in the skylab gfs-3dvar-c24 experiment
producing identical results for AMSU-A N19 assimilation when using either the
current ObsGroup container or the new OSDF container in the ObsSpace.
Partner contributions:
o Lewis Napper (Met Office) repaired the unit conversion capability in the ODB
reader: https://github.com/jcsda-internal/ioda/pull/1573
o Samuel Trehan (NOAA) provided a fix that eliminates spurious error message
that intermittently occurred when closing HDF5 objects: https://github.com/jcsda-
internal/ioda/pull/1571
o Samantha Pullen (Met Office) added a new ODB reader test for IMS Snow
Cover: https://github.com/jcsda-internal/ioda/pull/1566
Raji:
- PR in review has been merged [Cloud fraction warning removed, Difference Check
filter works with channels, Bug fix for WRFDA surface pressure operator], This sprint
task - adding where check in Specific Humidity variable transform is in review now.
Working on adding multi-threading feature in vertical interpolator operator.
Steve V.:
- UFO PRs:
o Two UFO maintenance PR's I worked on got merged, and two more are in
review.
o One of the UFO PRs in review adds new Parameters to
the DifferenceCheck and BoundsCheck filters, which allow you to specify that
you don't want the max/min endpoints of the range to be included in the range.
(e.g. the difference between the range [1,5], which includes the values 1 and 5,
and (1,5), which excludes the end values.) Prior behavior was to always include
the endpoints, and that is still the default behavior, so no yaml adaptation is
required, but the new capability should soon be available.
- IODA OSDF Distribution:
o One PR merged that advances the implementation of the new design
o Another PR in progress
o Getting close to being able to assemble all the tested parts into the whole
process.
- Liam has been on PTO last two weeks. Francois H. will be working very limited
hours for the next four weeks.
ALGO:
Christian:
- Did some benchmarking with the HTLM saw strange results on an HPC doing a bit
of lower res testing on AWS
- Found another potential optimization for the updateIncTL and updateIncAD functions
by hoisting the make_view and dxFSet lookups outside the for_each_index deepest
nested loop which looks promising giving a big speed boost, testing this and will
make a pr when ready.
- CDA is continuing
Nate:
- Saber PRs: merged more backend capabilities for the the BiFourier spectral
covariance blocks for regional systems.
- Upcoming Saber merges: New feature (from UKMO) allowing you to set an unequal
MPI split for hybrid B components; Adding filter method to NICAS block (Benjamin
Menetrier)
- Opened collection of PRs that refactor the ObsErrors in OOPS/UFO. Now the
default ObsError will be the UFO diagonal R, which is functionally equivalent to the
old OOPS diagonal ObsError which is being removed. No more explicit instantiation
of the ObsError factories. Will announce a merge date soon after some wrinkles in
lfric-jedi are ironed out.
Obs team:
- IODA converter:
o Will bring MEPO support R up-to-date #3735
- UFO:
o Clean-up to allow empty ObsSpace on single processor (for various reasons) in
`groupObservationsByCategoryVariable`. #3911
o Bring Difference check filter up to date with ability to specify range of channels
rather than individually (see jedi-docs) #3869
o Updated SfcCorrected operator to not require all surface variables (such as
specific humidity for sfc P) https://github.com/JCSDA-internal/ufo/pull/3830 –
Ricardo please let us know if you find any issues
- Pyiri PR /UFO
o Generic structured grid interpolator #80
o Interfaces to work with UFO #82
o Had to adjust tolerance after etkf precision PR (https://github.com/JCSDA-
internal/oops/pull/2962) was merged
o HofX crests with ionosonde (issues were raised with further investigation)
o Vertical integration operator: refactor the code so that any weighting can be used
in the future for the vertical integration without changing the main operator; end
point interpolation is also allowed in case the vertical interval is too large for
using the nearest model levels
CRTM/BenJ
- Working on surface emissivity utilities for CRTM and the associated C tests,
coefficient functionality etc; found some bugs
- Enable the ability for the current infrared snow surface emissivity model; to act like
the ice surface model. This was a request from U. of Wisconsin.
- Working on a converter to convert that to appropriate format. During the process it
also highlighted some existing issues with the way we interface with surface tables.
- Still working on the cloud coefficient tables
- have a delivery from STAR; adding the liquid optical properties into it.
- Cheng is working on tests in UFO; going to perform a comparison between
tomorrow.io and STAR.
- Cheng is integrating CRTM AOD with WRF, as part of her collaboration with MMM.
EMC/D. Holdaway:
- GFSv17 Retros Have Started. Four retro experiments covering different periods.
- Comparing ROPP for GNSS-RO with NBAM
- SDL with Process Perts
- Marine LETKF being reactivated and worked on.
- Working on optimization of FMS reads in FV3-JEDI.
- Added qoption 2 to the GSIbec background error model.
- Working on MTG-IRS in UFO.
- Need to tag JEDI soon
GMAO/Ricardo
- GETKF issue: the ensemble member mean is supposed to select the observations,
but the members end not seeing that selection properly and get misaligned.
- VARBC not working; trying to reset to static BC, which worked. Not sure why VARBC
was coded in GETKF. Soothing questionable in the codes.
- Continuing the work on skin temperature for IASI.
- Looking at the changes associated with pseudo Q2 and GSI back; can use GSI back
to use any variables of choice; can literally put chemical constitutes.
Tom: VARBC shouldn’t work for GETKF.
Michael Cooke
- Working on 4DVAR trial; benchmark on new hPC
- Speeding up the B matrix; running highest resolution; targeting from 50 min to 20min
- Thoughts about JEDI documentation; wondering if Yannick has updates on that
Yannick: will have follow up discussion after new year
David Simone: re-enforce Doc PR with code PR
Sara King
- Working on 4dvar, weren’t seeing performance improvement over 3dvar; think it is
related to how we are treating vertical coordinates. Forecast model in in geometric
height and assimilation is in pressure; adding adjustment for that
- Getting that finished soon for the higher altitude version of Neptune
- Looking at LETKF; ran into issues with the model
- Asking the plan for putting together an optimization team
Dom: the hdf5 chunking speed up 300 times
MMM /Chris Syner
- Junmei was running hofx for monitoring DPR from GPM.
- Finding weird hanging problem for unexplained reasons.
- Working on MPAS Ensemble recentering
NOAA GSL/Hu Ming
- GSL is working on building a MPAS+JEDI real-time hourly cycling system covering
CONUS 3km grid. The system will include both deterministic and ensemble analysis
with conventional observations.
- We are also actively developing functions for radar reflectivity, couple surface
analysis, multiscale analysis.
Hernan Arango
- Like Dan mentioned, we need to figure out how to do fork in front of the public
repositories.
- Github repository provides wiki so we can put all information about the interface,
how to use tutorials.
- Having a separate documentation is a must.
- Want to assimilate satellite sea surface observations; need to deal with biases and
have bias term in the cost function; do not know how to initialize these vectors in
JEDI.