General update (Francois H.)

  • No special announcement

JEDI Infra (Evan)

  • JEDI build workflow is now in final stage for review, now you can select specific build directory. If anyone is interested in trying them please reach out to Asheley for information.
  • There has been updated on MPAS ingest suite on all platforms.
  • There is a couple of bug fixes for the ingest suite.
  • CI backend changes, if
  • R2D2 client server is in testing phase.

[Dom]: Spack-Stack 1.8 has been released. It is installed on Discover and parallel clusters. On S4, it is holding back until the system upgrade. For EMC people, there is a problem on Orion and Hercules. You will not be able to use 1.8 for JEDI until the Intel compiler libraries fixed.

[Francois H.]: Have a follow-up question which is you've described a problem with So Intel? Is the spec stack 18 available with new on these systems or not at all?

[Dom]: Yeah.

[Steve V.]: Are there any issues with Macos and is all the documentation updated for building on the Mac?

[Dom]: Steve submitted a PR to fix all that and I was specifically for Sonoma with Xcode 15. It is definitely not supported at the latest MacOS, Sequoia. I think it calls Xcode 16, so stay away from that.

[Evan]: Sonoma is the best platform for using Spack-Stack.

[Dom]: Add one more thing. We are targeting end of the year to go back to the 3-month cycle. Spack-Stack 1.9 will be released at the end of this year, every one feels free to update or skip one version.

JEDI Algo (Clementine)

  • Anna and Francois will be in a code sprint and working on preparing the model variable and naming code this week.
  • Anna also work on refactoring to move the loop over QC filters and UFO specific QC functionality from OOPS to UFO. No changes are expected, but this is still in review.
  • They also work on aircraft bias correction. It is expected to work on branches, when using a static bias correction. Variational bias correction for aircraft is in working progress.
  • Christian is working on improvements to how the observation appended between outer loops, to make it possible to apply QC filters and others. He is also working on shifting the end of DA window.
  • The HTLM generic pseudo model is now working in Skylab and close to merge.
  • Nate was helping with the model variable naming sprint. He is now in the Howard U. in DC presenting JEDI at the MPAS-JEDI tutorial.
  • I modified the feedback files last week you will see OMB and OMA and intermediate innovations are saved automatically and don’t need to specify it in the YAML file anymore. It is the beginning of effort to simplify the YAML files, we are trying to make this more consistent in the code and YAML files. I’m also working on GEOS DA within Skylab.

[Steve V.]

  • Next week we’ll have a code sprint for model variable naming. So VADER, SABER and UFO are changing some of the variable name strings to the CCPP naming standard. So that we have a clear and consistent usage of variable names in the generic code.
  • For scientists who may have their own experiment YAMLs, you may see that you have to change some of the variables in you YAMLs and we will distributing information about that.
  • For users working on UFO, validation from GeoVals files, the names of the model fields in GeoVals files will also be updated and will provide instructions on that as well.

[Hernan]: Are the new system on names are going to be hard coded or set dictionary in YAML file?

[Steve V.]: They will be hard coded, because they spread across multiple repos.

Hernan expressed his concern about it is too rigid and wonder if there is an easy to manage, like a central dictionary can access and propagate.

[Steve V.]: The idea is once we get these strings in, even though they are hard coded. They will be very specific all over the places. Hopefully we can simply edit it in the future.

JEDI Model Interface (Francois H.)

  • Liam is working on the optimization for the GPU offload, specifically on the code path inside Atlas, to perform on GPUs sparse matrix multiplications. It enable users to use GPU sparse algebra libraries like cu sparse and hips bars. We are just working inside Atlas to expose that capability.
  • I also participate the variable name code sprint. And I’m also wrapping up to switch FV3-JEDI resolution change interpolations from its old Fortran implementation to the new structured interpolator. That’s the generic interpolation capability. Thanks a lot to Kim and his team for helping test all the edge cases in the regional model. I think this is ready to go. So I just merged it and we can move on from that.

OBS (Hui)

  • We have some PRs related to UFO. The first one is from Ricardo about the hydrometers increment and is merged. The second one is under review about the skin temperature as the control variables. Greg has been testing those code heavily from Skylab as well. The engagement from Ricardo is really important because this helps us to look at some areas that we were not able to test thoroughly in the past. Another PR on the review in the UFO is related to the aircraft bias correction.
  • IODA converters, we have a TPM PR converter completed and it is on review now. And there are two more from COMPO team related GCAS data converter and another one for the surface AQ observation from AirNow stations.
  • Space weather side, we have interface discussion with Atlas team and NRL to having the dual Magnetic Greg’s into the modern interface down the road. Another ongoing work, we started the draft of naming convention for the space weather variables.

[Ricardo]: How does JEDI handle the different versions of Spectral coefficient files in CRTM?

[Ben R.]: It is related to workflow, we want to control it by the YAML file explicitly.

[Ben J.]: Additional context for that. One of the points of confusion was the inconsistencies in the CRTM. Sometimes we have file named V2, but the sensor ID in the file is different. We are trying to slowly get rid of this and just make it where the file name and the sensor ID internally are the same. As we are transitioning toward NetCDF only, we hope some of the issues will go away a little bit. We are trying to make sure everything is consistent with expected behavior. Hopefully this will help smooth over some of the confusion we have had to deal with in the past.

CRTM (Ben J.) 

  • CRTM updates the infrared surface emissivity properties in particular a stretch goal, would be to implement the reference surface emissivity as a fast version CRTM that would be microwave sea surface and inference.
  • The third element is to verify that our polarization schemes are working correctly in CRTM. That would be both surface polarization primarily and how it relates to the orientation of the polarization vector as a sensor travels about it, scan pattern. So across track scanner, the polarization basis rotates for example.
  • There will be some internal structural changes to the CRTM surface properties but that will be completely transparent to the user except hopefully to have more accurate surface emissivity.
  • We will get the aircraft part sorted out soon and then that will be leading into the next phase.

Ricardo and Ben J. have some discussions about the skin temperature increments. Ricardo feels the T-skin increments looks reasonable in GSI but weird in JEDI. Ben J. mention there is a lot of challenges in this topic, if somebody would like to start a discussion topic on it, he would be happy to contribute. Ricardo said there is one already and he will update and add more information on that.

COMPO (Jerome)

  • We are continuing the EDA runs on GEOS-CF and TEMPO to make the demonstration in the publication.
  • Maryam added that run is quarter degree at C360 resolution and 10 C90 ensemble members. And is working on the converter for the Pandora data set.
  • Shih-Wei is continuing to work on the reflectance DA. We are trying to do a sensitivity test over ocean.
  • We are also starting to work on PACE, starting with ingesting observations to do some demonstration. We are waiting for the product to be out there which is not there yet. We will create a converter and test the products.

SOCA (Hamideh)

  • I have been working on adding copied to Skylab experiments, so I issued a couple of PRs in UFO which are on review.
  • At this stage, we have the surface pressure and in this becoming from the atmosphere for the adding the other atmospheric variables, I'm thinking to use composite operator with CRTM.

Ben J. asked for clarification about the meaning of composite operator. Greg added that is to be able to perform two different operators on a set of simulated variables, in which case some variables are using one, operator and some are using the other. And Greg mentioned that Mike Cook may be able to provide solution on this.

[Kriti]

  • I have been working on implementing the HTLM and testing with the quarter degree. We need to have land masks implemented for the norm calculation in OOPS. I know Travis and Francois have been discussing that and working with that.

Francois H. added that a PR is in and it is one step closer for you to put in the mask with more generic way. We can get that work quite soon.

EMC (Dan)

  • Ron is working on the IODA distribution to help with the BUFR reading. He compared all the BUFR reading speed to NetCDF. There was still a couple of issues and some bottlenecks he is working on.
  • Most of our worker are fine tuning the system for GFS v17, I’m working on removing the D-grid winds from FV3-JEDI which is very close to being done. I should have a PR in the next couple of days.
  • We are also continuing to work on the comparison of a regional FV3 system using GSI B and JEDI. Although the regional system is going to be MPAS, it is pretty useful that JEDI and GSI can have the exact same background.

Tom asked about the distribution work is during IO or in memory. Dan responded that is in memory. Francois added that what decouple the file structure and tied to how many nodes are doing IO. Dan added, switching to HDF5 structure in IODA speed up massively and the work is getting there.

GMAO (Ron, Ricardo, Doruk)

[Ricardo]

  • Working on PRs of hydrometer increments and T-skin increments.
  • Can’t pass all the tests with a fresh build, asked for anyone can help with.
  • Working on the sync version of CRTM in GEOS and JEDI. Now there is no difference.

[Doruk]

  • Tested building JEDI with Spack-Stack 1.8 on su-17, it was successfully built and tests are passing.
  • Working on ocean initial conditions and trying to improve our static B. We are running a longer spin-up for ocean runs.

Ron mentioned Virgnia Bouchard’s work on aerosols but she was not in the meeting, so Jerome provided brief update on her work

[Jerome]

  • She is putting a UFO PR on multiwavelength of multiple extinction. Next step is producing increments from that and do a cycling demonstration.

Ricardo added that GMAO is starting to do some comparison to the present system. Jerome added that we should decide on a long-term direction for aerosols. Some discussion about the B

Met Office (Phil U.)

  • Please take a look at OOPS PR 2747 (by Chris and Benjamin M.), which added a state template YAML sections to the configuration of the pseudo model file.
  • Discussion 155 from Reuben Hill, he is considering removing the ability to specify the formulation of a UFO variable transform.

Navy (Sarah)

  • We have been working on ozone assimilation lately. Thanks Jerome for the guidance on hyperspectral IR and then Francois V is helping us by finalizing our gnssro implementation.

ROMS (Hernan)

  • Discovered that you have to specify those parameters in YAML file, otherwise you don’t get the results that you expect. Now I can compare with the diffusion operator and we get the same spread. The remaining issue is the relationship with the Gaussian.

Tom:

  • The AOP was approved. I would like to use the next slot in two weeks to introduce the major pieces of the AOP. If there is any major conflicts of agendas, please let your representative know and we'll try to arrange around.





  • No labels