Skip to content

CCPP Framework Meeting Minutes 2019 06 27

Jump to bottom
Dom Heinzeller edited this page Jun 27, 2019 · 5 revisions

CCPP usage at NCAR

  • NCAR chemistry folks are furthest with using CCPP and implementing a model based on it
    • MUSICA is now using the new metadata standard
  • Dave Gill tasked to manage transition of WRF physics to CCPP and make it work in MPAS
    • MPAS will be the first target model
    • RRTMG, YSU, NTIEDTKE, WSM6: standard_names identified, about 1/3 are already in GFS physics, about 1/3 names can be inferred from GFS physics, 1/3 are unclear
    • starting with YSU PBL scheme, use it in MPAS via cap_gen
    • option: put caps in place that call 2d schemes as opposed to 3d; arrays are more traditional and less WRF/MPAS specific
    • how much code will be rolled into scheme, interstitial code, host model cap
    • new metadata: attribute "state variable", use intent information to infer whether it needs updating later or not
    • radical suggestion: force schemes to return tendencies and never update the state, gives full control to the suite designer - but requires buy-in from all physics developers
    • index ordering MPAS is swapped compared to FV3: MPAS (k,i) versus FV3 (i,k) or WRF (i,k,j)
      • cap_gen will handle this automatically
    • vertical direction and vertical coordinate: we need to decide on the metadata standard to describe these

Strategy for hosting physics in different repositories: ccpp-physics, sima-physics, plus additional separate repositories for schemes under heavy development (e.g. MG microphysics)

  • need to minimize duplication
  • hierarchy of curation (schemes used in operation, development, ...)
  • NOAA-NCAR discussion about using a single repository
  • what are the governance rules for code review, which changes get accepted
    • NCAR has a governance process for SIMA
    • extending this to NOAA is more complicated
    • NOAA is still working on this internally

How to progress with ccpp-framework development

  • what else needs to be done by GMTB for CCPP at EMC
    • coupled model: build system, correct interstitial code, different metadat, ...
    • new parameterizations
  • metadata, cap_gen, GFS data structures (how urgent?)
    • removing the GFS data structures requires functionality from cap_gen
    • it may be possible to change the memory layout on the host model side

Recap of features in new cap_gen:

  • variables created by cap_gen are allocated in the right place and made persistent
  • time-split and process-split: if just listed in SDF one after the other, time-split is assumed

CCPP Framework wiki

Clone this wiki locally