No general announcements.

INFRA:

  • Planning AOP25 and finishing touches some AOP24 issues
  • Nvidia spack-stack config and AMI
  • Large progress on r2d2 automated tests.
  • OMPS and METAR added to observation ingest suite
  • Automation scripts in jedi-tools now uses spack-stack 1.9
    • Build scripts minimize git-lfs usage too via s3 bucket cache, please use them if applicable to your work

Interfaces & Optimization:

  • Interface change: adding Increment::sqrt method at the oops level, models must add an implementation; merging Monday Apr 7th
  • Vader LVC improvements have been reviewed and will be merged today
  • Continuing work porting surface operators in UFO from Fortran to C++
  • Refactoring of the IODA reader to populate the OSDF ObsSpace memory backend underway: obs filtering is complete, file I/O and MPI redistribution are under design
  • Preparing to run a demo of an lfric-jedi 3dvar with certain SABER components being offloaded to GPU

ALGO:

  • Working on bringing weatherbench scores to ewok. Work with INFRA to find a solution when the python packages from google are incompatible with ours ): it looks like we'll investigate calling singularity or docker from ewok tasks.
  • Debugging cause of bad increments in high-res multi-scale B experiments
  • Officially adding a hack from the OBS team into EWOK so that forecast lat-lon files get written for computing scores

OBS:

  • Isaac Moradi from NASA added a filter to interpolate and average observations to GeoVals levels. User inputs environmental model levels, extends ObsSpace, and averages observations within the level. (ufo#3588, jedi-docs#876)
  • There is an ongoing merging of the surface operators being worked on right now. JCSDA worked with NOAA GSL to add a computation of geopotential to geometric height conversion. (ufo#3631) This helps to complete ufo#3622, where Sujie Pan is adding NOAA surface temperature correction methods.  There is a plan to have consolidated corrections, pressure, and temperature, and lastly will work on moisture of surface variables.
  • Emily Liu NOAA fixed one of the bias correction predictors (CosineOfLatitudeTimesOrbitNode), the code was incorrectly using the sensor azimuth angle instead of the satellite ascending flag. (ufo#3649, jedi-docs#895)
  • Mike Cooke from UKMO added the final piece to do ObsBias correction by record, this enables writing the covariances: ufo#3652. This has approvals and is seeking more input from the users, hoping it will move forward soon.
  • Observation operator that reduces the spatial density of requested GeoVaLs, for very dense observations this is allowing multiple observations to be assigned the same GeoVaL. (ufo#3657, ufo#3653)

CRTM

  • Work on surface emissivity jacobian, will need to coordinate PRs because this will change reference values
  • Work toward final version of netcdf tarball release of CRTM data

COMPO

  • Updating IODA converts for AOD (pace, viirs, aeronet) and adding ingest suites
  • Shih-Wei and Jerome start looking into dust, biomass burning, volcano, and polluted cases for level 1 DA work.
  • Maryam and Jerome working on COMPO's paper1
  • Maryam and Jerome on review panels

SOCA

  • Hamideh leaving JCSDA, will lead to pause of oasim developments
  • Kriti working on HTLM for the ocean, after Christian fixed issues; hoping to show HTLM better than identity!

GMAO

  • Working to add converter for (did not hear?) to IODA format
  • Looking into issues with large Jo from some obs spaces (land and marine surface fields)
  • Testing new interpolators around GSIBEC giving 3-4 faster runtime; enables running at operational resolution where the develop code doesn't even run; will also enable turning on halos and GSI TLNMC

UKMO

  • UFO PR adding new operator for spatially-dense operators to reduce number of geovals
  • IODA PR for parallel reading of odb files
  • Final trials of JEDI obs processing, should be operational by end of year

EMC

  • Working on transition to operations requirements (static libs, no forks, reproducibility when changing number of tasks)

ROMS

  • Working to accelerate 4dvar by combination of lower-res inner loop geometry and single-precision inner-loop math; but can we do this in JEDI?
  • PIO working well to accelerate the IO

No significant updates from USN or MMM this week.


  • No labels