Notes taken by Lindsey Mattson
- Liam visited Boulder, spent the week on onboarding and discussions
- met with UKMO and ECMWF to learn about atlas's GPU capabilities; we're currently exploring further to see if these capabilities will meet our needs
Work on allowing different outer loops to be run as separate executables is progressing with Tom Hill at the met office. This is aimed at both improved HTLM testing as well at continuous DA.
Christian is also working on adding multiple outer loops as well as the ability to use the general OOPS:PesudoModel in Skylab with ensembles and with the HTLM
Clémentine is working with NASA to build geos and jedi their way (the nasa way) and our way, to get some crucial exchange of data between the model and JEDI going (needed for any in-core applications without having to go through IO) --> very good progress on this last week.
We are also having constructive conversations on how to handle the propagation of an increment to all derived variables (not necessarily assimilated, but affected still example ps is assimilated, and delp should be updated too) (edited)
For IODA:
BUFR reader PRs have been merged introducing the capability to construct ObsSpace objects directly from BUFR files
A PR is under review that enables the IODA reader to ingest multiple input files for the construction of a single ObsSpace object
Good progress with implementation of the new ObsDataFrame memory container. Phil's PRs are under review and once merged will complete the initial implementation of the ObsDataFrame class. Phil and SteveH are making progress on the next step which is to create a translation layer to connect the ObsSpace API to the ObsDataFrame API. Once that is completed the following step is to test and profile the new ObsDataFrame container.
Finished & merged a big update to the B-Matrix training suite to make it more flexible
Opened a PR in MPAS-JEDI that adds a test using spectral blocks to do a single-scale ensemble localization
Starting to work on a 3denvar mpas experiment that uses the same single-scale spectral ensemble localization
Steve V.: Designing an enhancement to VADER Linear Variable Changes, which will allow distinct definition of Trajectory variables and increment variables
ingest:
cleanup of snow cover fraction converter completed,, finished with update to the README thanks to Youlong from NOAA
VIIRS L1B converter for albedo being finalized by Shih-wei capability being used for aerosol by COMPO team but obviously has other applications
CloudSat decoder from Isaac Moradi at NASA being integrated, initial prototype in a draft PR some consolidation and organization and clean-up being performed should be ready in couple of weeks
UFO:
Anna S. completed the addition of an action that can be performed on PRE filters only, this action will remove the observation from the ObsSpace and memory footprint , no GeoVals will be retrieved. Goes without saying any check that required GeoVal cannot use this action.
Fabio has a draft PR on handling zero locations with variable transforms that derive variables. Joanne Waller has given this a precursor review, we are making sure there's not already an existing way to handle this correctly before promoting from draft state.
UFO PR #3343 adds a limited area model domain mask to potential GeoVals retrieved for an observation location. This handles observations at the domain edge boundaries