CRTM (Ben J.):

  • v2.4.1 release imminent, pending some last minute documentation updates. This marks the final version of the CRTM v2.x series.
  • v3.0.0 will merge right on top of it, and upon first release June 30, 2022, will also be "UFO Ready".
  • Met with Jim Jung and Nick Nalli on Nick's recent updates to snow/ice surface emissivity modeling, already demonstrating improved bias results in GSI limited testing for IR channels.  v3.0.0 will include this updated LUT.

General OBS Team (Francois V.):

  • We are establishing a 3D-Var testbed for verifying various components, what to target will be iterated with the ET.  Among the first items are to examine and exercise varBC and correlated error within the UFO.

OBS1 (Francois V.):

  • Working on cycling experiments: (1) added and continue adding more obs types (RO, radiosonde, satwinds...) (2) added new obs error for RO (COSMIC-2 project). Trial different interpolation schemes to adopt the obs errors for different lat and levels (3) checked convergence of cycling experiments
  • Plan UFO clean up together with OBS3. Plan to focus on one obs type first as an example then spread out to the rest of obs families. This will be the Q1 focus (Ryan has more details)

OBS2 (Francois V.):

  • We are discussing with NRL to share code for diagnostics and graphics.

OBS3 (Ryan H.):

We had a pair of discussions this week with the Algorithms group on model and observation variable interfaces and names. In the Algorithms meeting, we generally discussed the JEDI Observation naming conventions and how they would interact with model variables. In yesterday's Observation Infrastructure meeting, we considered examples of where this directly affects UFO and discussed how to resolve.

First, we considered UFO issue 1777, which discusses GeoVaLs and ObsSpace variable name interactions using the Identity ObsOperator. Essentially, in simulateObs, we get a list of variable names from the  ObsVector, which gets its names from the ObsSpace, and then ask GeoVaLs to retrieve those variables. The observation variable name must match the name used in the model interface. As we are rolling out improved observation variable names, the names do not match, and we need to work around this. We discussed modifying GeoVaLs to take a set of variable name mappings that we can provide from YAML, and this was seen as a good path to explore.

A similar name mismatch issue may also occur when using ObsFilters and functions. We are looking into making a few changes to IODA to allow name aliases.

Thanks to everyone for a good discussion. Afterwards, we had a small bit of general discussion. There was a general request to ObsOperator developers that they use ReadTheDocs to document the variables needed to run their operators. This is particularly needed for CRTM.

Hui mentioned a UFO cleanup. We actually didn't get into it much this week, but we want to clean up our test data and YAMLs this quarter.


COMPO (Jerome B.):

  • Working on setting GFS-EWOK aerosol and populating AOD with MODIS
  • Looking into running GEOS-Chem on AWS


SOCA (Travis S.):

  • Hamideh working on ioda converters for PACE ocean color
  • Kriti / Travis preparing for regional ocean ensemble DA experiments
  • soca has a superobbing tool, we need to discuss more where this should live and how we should make it generic.... UFO?


LAND (Andy F.):

  • Collaborating with RAL to run an experiment with wrf-hydro. Currently wrf-hydro is broken.
  • Started planning for a soil moisture DA experiment with PSL(?)


JEDI1 (Steve H.):

  • IODA writer improvement has been merged. This improvement will greatly reduce the number of output files (https://github.com/JCSDA-internal/ioda/pull/658). This feature will create one file per MPI task in the io pool which will greatly reduce the number of output files. Eg, you have 5000 tasks total working on the DA flow and 5 tasks in the io pool, you now get 5 output files instead of 5000. Instructions for new (optional) YAML control are documented in the jedi-models discussion group: https://github.com/orgs/JCSDA-internal/teams/jedi-models/discussions/68. Next step is to consolidate the files that the pool write into a single file: - The plan is to use the hdf5 parallel I/O feature for this - This will need to be designed with the Met Office to make sure that the method used by the ODB writer can coexist with the hdf5 parallel I/O method. The final step is to instantiate more than one io pool so that multiple obs spaces can be written out in parallel
  • Eric and Dom are working to set up EWOK/R2D2 using Spack on S4 and Discover. 
  • Eric, Maryam, and Yannick are working on implementing a cloud-based CI/CD for EWOK/R2D2
    • Using GitHub Actions and AWS
  • NRL is catching up with the develop branches of JEDI repos, porting to IODA v2 format files is almost done. There are issues with knowing which metadata to supply for CRTM – Ben J. will help with this issue. 
  • There are a number of diagnostics tools (RadGrams, Hovmoller diagrams, sensor diags, etc). We need to collect what the different centers expect for diagnostics. This will be helpful for EWOK development. 
  • Ryan and Dom are working on establishing HPC accounts for Patrick N.

JEDI2,3 (Anna):

JEDI4 (Yannick):

  • Currently, working on creating C192 experiment and later move to full-resolution experiment.




  • No labels