Log In | Users | Register
COSMO Web
Edit | Attach | New | Raw | Diff | Print | | Tools
You are here: ICON » ICONmeet20191107

ICON meeting November 7, 2019

Participants:

Zane Dedekind (ZD); Remo Dietlicher (RD); Saskia Drossaart van Dusseldorp (SD); Xavier Lapillonne (XL); Katie Osterried (KO); Colombe Siegenthaler (CS); Michael Steiner (MS); Colin Tully (CT); André Walser (AW).

Reports:

* André Walser (AW):
  • AW is leading the project of transition from COSMO to ICON at Meteoswiss . Guy and another colleague will begin in January to work on this project.
  • DWD will use ICON for their operational forecast at small scale in about 1 year.
  • AW did not begin yet. AW needs to woks substantially on another project which will end next Spring.
  • AW will run ICON-LAM (on CPUs) at ECMWF till end of next year, afterwards on GPUS on cscs machines.
  • Current Spring/Summer, AW will set-up an ICON-LAM on Swiss domain, with 1km resolution.

* Colombe Siegenthaler (CS):

  • CS worked on setting up the ICON-C2SM organization. It contains the dwd-tools, ICON and ICON-ART for C2SM members.
  • CS worked on documenting on the C2SM wiki :
  • The summary of which compiler version is used on daint for different compilers was up-to-date last week. This morning (after the big upgrade on Monday) ICON could compile but not run because of a problem in unloading modules from the compute nodes.

* Colin Tully (CT):

  • CT is running ICON-HAM on daint.
  • CT is validating ICON-HAM against ECHAM-HAM. CT is looking for the same variables in both models (lots of variables change names and location between ICON and ECHAM). RD suggests to use the fact that for CMIP6 simulations, the variable names have been standardized, and both ECHAM and ICON will participate to CMIP6. CS says it could be a good idea, but the post-processing of ECHAM variables to stick to the standardized names is done by MPI-Met and still on-going.

* Saskia Drossaart van Dusseldorp (SD):

  • SD would like to work on hail storm simulation with ICON (NWP physics) over Switzerland at 1km resolution.
  • SD needs latent heat nudging which is not available yet.
    • SD would need to run her last simulation next summer/autumn.
    • XL says there is hope that the latent heat nudging SD needs will be available.
    • SD needs the product of the radar simulator which is now used in COSMO for use in ICON. AW will let SD know about the Meteoswiss plans.
  • SD works with COSMO on another simulation in the mean time.
  • It becomes difficult to get computing time on daint with COSMO on CPU.

* Xavier Lapillonne (XL):

  • XL is the only ENIAC person at Meteoswiss left, since Philippe Marti and Valentin Clément left this summer.
  • Before leaving, Philippe ported all the physics parameterizations which will be used in the QUBICC experiment, except for the radiation. A version of the radiation scheme in itself already exists on GPU, but a wrapper is needed.
    • The parameterizations which are ported are : cloud microphysics (comes from the NWP part, so called " Graupel scheme"), JSBACH (surface model), vdiff (turbulence).
    • There will be no convection scheme in QUBICC, so no convection scheme is ported.
  • The QUBICC experiment is a PRACE project. The proposal has been submitted already. Since the radiation is not ported yet, the proposal will have to be updated. The run should be made next year.

* Michael Steiner (MS):

  • MS began an internship by Dominik Brunner for 3 months on 1st of November. The internship will be followed by a PhD project beginning in February if a new project in collaboration with China is funded.
  • The PhD project is about studying the formation of secondary inorganic aerosol over china.
  • MS will work with setting-up ICON-ART during the next 3 months.
  • MS downloaded and compiled ICON-ART from the C2SM-ICON organization on daint on Monday (before the upgrade)

* Remo Dietlicher (RD):

  • RD worked on the set-up of all the infrastructure needed for the porting of NWP physics on GPU.
  • For example, RD wrote a testsuite which is working on Buildbot. The testsuite tests if the performance and physics are not changed when someone change the code.
  • RD is working on porting the infrastructure code on GPU. The NWP routines are mostly the same as the one in COSMO, which were already ported on GPU during the COSMO porting.

* Zane Dedekind (ZD):

  • ZD will eventually work with ICON-LAM over Switzerland and compare the results with COSMO.
  • In the case Zane works with ICON, he will have to implement the same changes he did in COSMO into ICON.

* Katie Osterried (KO):

  • KO began to port the radar simulator on GPU. The radar simulator is used in COSMO, and will be used in ICON as well. The radar simulator will go into the DAS (data assimilation) library and be linked by ICON. KO will first test the ported radar simulator in the context of COSMO.
  • KO will have an intern who will work on Extpar, cleaning up the code. The intern will begin next January.

Topic of common interest:

* ICON license:
  • The ICON license has been signed by ETHZ. C2SM is now allowed to distribute the official ICON tags to ETHZ employees and C2SM members. C2SM members who are not at an ETHZ institution need their own license. This signed ICON license applies for the moment only to official ICON tags.

* Workflow in C2SM-ICON organization in Github: Branch or Fork ?

  • CS raises the following question: "Based on your experience with COSMO, would it be better to let people have their own branch in the ICON repository, or should they have their own Fork with merge requests?"
    • XL : at Meteoswiss, the following rule is used: If person is working on a project, then this person should work in a fork; as soon as 2 persons are working on the same project, they are allowed to have a branch in the repository. In an idea of continuity, the more permanent people should also get a branch.

* Tagging in ICON repository in the C2SM-ICON

  • There should be a C2SM specific tag number for ICON. For example, in COSMO, for the official COSMO version 1.1, the C2SM COSMO tag is COSMO 1.1.1.

* COSMO releases

  • There will be a final major release of COSMO (6.0) by the end of next year.

* ICON-COSMO branch

  • Ulrich Schaettler will be the source code administrator of the ICON-COSMO branch in the German repository.

Next meeting:

January-February 2020, before the next ICCARUS meeting which will take place (March 2 to 6, 2020). spacer
This site is managed by the Center for Climate Systems Modeling (C2SM).
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors. Ideas, requests, problems? Send feedback!