General Update  (Yannick)

  • No special announcement

JEDI Infra (Eric)

Workflows

  •  Added a new script to Jedi-tools that will build only the JEDI workflow applications needed for the JEDI Skylab Environment (r2d2, ewok, simobs, skylab, and solo). See jedi-tools/buildscripts for more information. This pairs well with the new option in Skylab to to build jedi during the experiment

Environments

  • Have a Rocky8 AMI instance with spack-stack environments of gnu and intel. Writing documentation for it and should be snapshot soon for others to test for development work.
  • Clang is having issues building spack-stack 1.8, mostly related to distros having a variety of versions (Debian based being older, and RedHat based newer)
  • Target date for spack-stack-1.9.0 is mid December
  • Looking into upgrading the following packages to their latest versions for spack-stack-1.9.0
  • NOAA-EMC: bufr-query@0.0.4 ECMWF: eckit@1.28.2, fckit@0.13.2, ectrans@0.39.0, atlas@1.5.0 testing jedi-bundle with these versions

Data

  • Finishing up a new authentication scheme using API KEYS for the new R2D2 client server system. This will remove the requirement for an AWS account to operate R2D2 and add more security to the server. 
  • Finishing up R2D2 test systems on MSU and NCCS for the new system.

Question(s)

[Dom] : mentioned something about shapely update
[Yannick]: mentioned something about skipping spack stack 1.8 and moving to spack stack 1.9

JEDI Model Interface (Francois H.)

Steve V

  •  Working to understand and debug some linear variable change issues 

Steve H

  • PR merged that transfers hardcoded configuration to YAML configuration in the ODB reader 
  • One PR merged, and one PR under review, that support the usage and validation of OBS Convention variable names 
  • Two PRs under review to enable using the new IODA io pool based reader in skylab 
  • Work in progress for BUFR parallel IO capability, obs data frame (in-memory) and obs iterator capability 
  • Many thanks to our partner in-kinds for their contributions to IODA

Liam

  • Started investigation into making use of the GPU offload version of ECTrans. 
  • Set-up ROCm on a mini PC and tried to validate the approach on that platform - so far unsuccessful.
  • Worked on GPU offload of Atlas interpolation - prep work for PRs after getting consensus for Atlas dev team.

Liam and Francois H.

  • Set-up an NVHPC JEDI environment and source build of ECTrans with NVHPC Toolchain - looking to fckit errors.

Observation (Hui)

Project review meeting

  • Ben and Hui attended NOAA commercial program end of year project review. We have two projects from the commercial program: one is for ocean surface wind and the other is for RO experiment project management. Ben gave a presentation for the ocean surface wind, which was well received.  

UFO:

  • PR #3522: Mike cook addressed the zero obs on an MPI task for several variable transforms (Merged)

IODA-CONVERTERS:

  • PR #1566: Patches to the build system to properly support importing ioda-converters. This PR makes a few small fixes to the ioda-converters build system to allow it to be found using find_package(iodaconv) by dependent package. (Merged)
  • PR #1575: MODIS AOD converter updated by Shi-Wei. Using epoch dateTime and Adding date_range for selecting data within the time window. (Open, needs review)
  • PR #1573: Update pandora converter with site classification information. Mariam. Removes Vertice information from this converter, as it is a total column observation, and Vertices information is unnecessary. (Open, needs review). 
  • PR #1576: converter for ground based TEC from tenet network by Ben R. (Open, needs review)

IODA:

  • PR #1348: Add names of ionosonde variables. Update units for TEC. (Open, needs review) 

Space Weather: 

  • IODA naming convention: Discussed conventions specific to space weather, as outlined in IODA #1348.
  • Hofx generation for ionosonde observations: Have been fixing the geoval generator with NRL. Created a UFO test and will draft a ctest for the Hofx generation (though still pending on the geoval file). 
  • Model variable convention: Recognized the need for model variable convention. Creating a chart for the existing variables. Will share with the JEDI team and NRL team for inputs.

CRTM 

  • No update

Compo (Maryam)

  • Continue working on the 4DEnVar experiment with geos-cf. Goal: find the optimal ensemble size
  • Spend some time fixing a bug in the experiment, which was putting too much NO2 in the stratosphere. The issue was due to the wrong no2/no partitioning and we addressed that.
  • Bugs in SIMOBS’s compo plots. Fixed with the help of the obs team.
  • Bugfix in AirNow converter
  • AQClassification is added to the Pandora data. Pandora ingest suite is ready and works as expected (Thanks Ashley)
  • AirNow and Pandora observations along with tempo, troopomi for Aug 2023 are ready to be uploaded to R2D2 (r2d2-ingest issues are submitted)

Question(s) 

[Ron]: Is COMPO using the Milan Node

[Maryam]: yes, moved from cascade to Milan and it used to be better at the beginning when less people were using it. Also, It is not just the queue that’s slow but also the filesystem 

[Ron]: We are aware NCCS has told them what’s wrong.

SOCA (Hamideh and Kriti)

  • Hamideh has been working on changing the mom6 model variable names to match soca in skylab and I have a pending PR on skylab on this. 
  • She also has  a pending PR in the "coupling" repo to add oasim hofx there and one more  corresponding PR in fv3-jedi-data to add the modis-aqua ocean color radiance obs data for the oasim hofx test.
  • Travis has wrapped up the parametric stddev block he was working on
  • He is now working on SOCA performance evaluation tuning
  • Kriti has been working on testing the scientific and computational performance of 4dVar with HTLM with ¼ degree regional ocean domain. 
  • She found that the linearization errors with HTLM look are small compared to persistence but there are intermediate TLM perturbations that are large and may indicate a bug somewhere

ALGO (Clementine)

  • Christian has been working on continuous DA, they have mapped out some changes to IODA that are needed as well as changes to the TimeWindow class that are needed for shifting windows. They are currently implementing things for testing.
  • Nate has been working on skylab experiment that does a multi-scale localization on a 60km MPAS ensemble.  
  • Updated the last group of variables from the naming sprint to the CCPP standard. Skylab is updated with the changes, and there is an r2d2 ingest for updating the BUMP files
  • Clémentine has been working on jedi-geos (satellites obs + wrapping up variable name convention changes to the geos skylab experiment).
  • Currently running Ricardo's case (w/ static GSI B) and will now add an ensemble B component to the experiment using localization created with bump

GMAO ( Ron and Doruk)?

  • Reported on Ricardo’s work on atmospheric side and the issue with the 3d pressure field not assimilating properly and issue with linear model similar to the problem at EMC
  • Doruk mentioned that NCCS Discover has been slow and experiments are failing randomly similar to what Maryam mentioned. 

NOAA EMC (Dan)

  • Memory usage issue identified in marine runs. Undertaking profiling with MAP. Similar issues around scalability seen for fv3-jedi hurricane config.
  • D-Grid wind PR, working with Clementine to check things work for GEOS.
  • Improved IO for the fv3-jedi History. The related PR should lead to as much as 100x speed up in the IO
  • Continued testing of the GSI Recursive Filter for regional. Regional FV3-jedi is comparable to regional GSI for conventional aircraft observations.
  • Parallel processing of BUFR in IODA PR is ready. 
  • Nick’s fix to the Aircraft bias correction converter.
  • Fix for surface pressure corrected UFO for regional domain.

Question(s)

[Ron]: Is there some document or PR that Ricardo can refer to the improvements related to the IO for the FV3-JEDI  history?
[Dan]: There is a PR and Ricardo is added to that PR. Ricardo should test the feature branch in fv3-jedi that improves the speed of IO for History files. This is the mode of IO that Ricardo is using for his GSI vs JEDI evaluation runs.

UK MetOffice (Philip Underwood)

  • An update to the ODB reader has been merged into IODA at PR 1356 ( link sent in chat: https://github.com/JCSDA-internal/ioda/pull/1356).  This removes observation specific dependencies from the code. These are now controlled from the YAML configuration files.  This will make it easier for others to make use of the ODB reader and writer. 
  • The jopa-bundle and ops-um-bundle legacy Met Office bundles have been archived on JCSDA-internal, as users have all migrated to mo-bundle hosted by the Met Office 

M cube

  • No updates 

ROMS ( Heran)

  • Worked on the nonlinear trajectory time interpolation after Yannick’s PR in OOPS   that removes PostTimerParameters. 
  • Mentioned that model using ATLAS has been problematic for them because of the staggering of some of the variables. He proposed a solution to add the variables to state

Question(s)

[Yannick]: mentioned something about variable change

[Ron]: destagger in model as part of the state so no need to make change sto the variable change code. Do what Dan is doing/recommended

[Dan]: asked for clarification about the new method that was added to the TLM class.

[Yannick/Hernan]: The new method is there to allow time interpolation in situations when the TLM trajectory is saved at time steps that are larger than the time step of the TLM. The new method can be used to propagate the trajectory or use time interpolation. It’s the model’s choice. Nothing is done genetically for time interpolation.

  • No labels