Today we did a round-robin update from everyone.
College Park
- Nearing the middle of a two week code sprint
- May 7 - 18
- Will stop taking updates from the main repositories
- So that we can do the ObsSpace reorg, and other updates without impacting the code sprint
- Issues
- The Locations class, being in UFO, is conflicting with their interpolation needs
- Want the Locations class to be in the model to enable vertical interpolation before horizontal interpolation
- Want to add vertical coordinate to Locations
- Want to add drifter ID to Locations
- Yannick commented that drifter ID does not belong in Locations
- Yannick suggested that a new Locations subclass (in the model) could be created to accommodate the vertical before horizontal interpolation requirement
- Yannick commented that the reasons for placing Locations in UFO were:
- Keeps the class generic
- Specifics such as model grid can be avoided
- Communication is difficult if Locations is in the model
- Learned this from experience
- Just a warning that placing a Locations subclass in the model might be more work than expected
- Keeps the class generic
- Want the Locations class to be in the model to enable vertical interpolation before horizontal interpolation
- It was asked when will more (necessary) information be placed into the Locations class
- Currently only latitude, longitude and time are stored, and there needs to be more than this
- It was noted that this is on the todo list, and that the GPSRO work is the first to extend the Locations class
- Please feel free to contact the core team for help, especially if this is blocking progress
- There was some discussion about interpolation
- Yannick commented that interpolation was kept out of the model so that it could remain generic and be used by other systems outside UFO and models
- ESMF said that the interpolation code would work regardless of obs distribution, but we are not aware of a case that verifies or refutes this claim
- There was some discussion about the organization of UFO and obs operator functions
- Some questions and concerns were raised, but it is not clear at this point if there is anything blocking progress
- Please feel free to contact the core team or other developers for help with this, especially if there are issues blocking progress
- The Locations class, being in UFO, is conflicting with their interpolation needs
Met Office
- Working on getting an incremental to run for 3D Var and 3D EnVar
- Currently working on a solution to enable conversion of temperature to virtual temperature during interpolation
- No barriers exist, and they should have things working shortly
NRL
- No update
NCAR
- Populating ZenHub project boards
- Learning ODB, C++
- David Davies from the Met Office has made a lot of progress with using ODB in JEDI
- David would be a good contact for ODB matters (david.davies@metoffice.gov.uk)
- David Davies from the Met Office has made a lot of progress with using ODB in JEDI
- Working on WRF
- Can run WRF inside JEDI
- Running into issues with files, I/O
- It was noted that they wanted help with some questions they have, and this will be taken offline
- For radiance, it seems that the interface with CRTM may not be too complicated
- Benjamin offered assistance for any CRTM questions that come up (benjamin.t.johnson@noaa.gov)
- Comparing with WRF-DA
- Working on WRF
- Exploring interfaces with ufo→MPAS, ufo→CRTM (JJ)
- Encouraged to coordinate with MET Office and JCSDA (Benjamin Johnson)
- Encouraged to coordinate with MET Office and JCSDA (Benjamin Johnson)
General Discussion
- Chris Harrop reported progress with building and running JEDI outside the containers for HPC systems
- Singularity is typically not available on HPC systems
- ufo-bundle is working on Theia and Cheyenne
- Have a set of scripts that do the builds for Theia and Cheyenne
- Where should these scripts be checked in?
- New repo, existing repo, ?
- Where should these scripts be checked in?
- It was noted that besides the script sets, there is additional software on Theia and Cheyenne that the scripts utilized
- BJ asked whether the MPAS bundle has been implemented on Cheyenne and the answer is no - not yet - Chris has been focusing on the ufo bundle but much of the setup can also be used for MPAS
- Yannnick brought up that we have many bundles now, and we need to come up with a solution of how to organize and manage these
- This will take some work to get a good solution in place
- This will go on the todo list and be addressed later
- Francois mentioned an issue with Git-LFS
- A local copy of a data file got corrupted when it was checked out
- The remote copy (on GitHub) is okay
- Removing the local copy and having git-lfs check it out again fixed the problem
- Not clear what caused the problem
- If people start experiencing this more frequently, please let the core team know so they can address this
- Guillaume mentioned that they have set up the git-lfs filter for the marine obs data, and someone needs to do the same for the atmospheric obs data
- The git-lfs filter is the mechanism that specifies which files go into git-lfs
- Steve H (core team) will take care of setting up the atmospheric git-lfs filter
- It was emphasized again that all binary and netCDF files should go on git-lfs
- Mariusz mentioned an upcoming issue concerning FV3-Chem
- FV3-Chem will be used by NOAA for chemical (e.g., aerosol) prediction
- Currently implemented through fms but more work to be done
- The chemistry part is required to couple to FV3 via an interface called NUOPS
- The question is how JEDI will deal with running such model using NUOPC and to what extent NUOPC will limit assimilation
capabilities
- NCAR asked when the April 15, 2018 obs, geovals files will be ready
- Steve H (core team) will take care of this, and noted that he will be working on this next week
- The files should be available in a week or so, but there is some coordination that needs to be done with model initialization files