CCPP Framework Meeting Minutes 2019 03 07

Metadata PR

• three main issues
  • km+1 problem:
  • do we want to support this? no
  • or limit to every dimension being a defined variable with standard name? yes
  • cleanup work to be done by GMTB
  • missing intents - cleanup work to be done by GMTB
  • module use dimensions - cleanup work to be done by GMTB
• beef up preprocessor if handling - to be done by Steve
• accept PR after preprocessor handling is improved, tag for conversion process with maintenance branch

Bit-for-bit versus other comparison methods

• MPAS also doing testing non bit-for-bit
  • various regions that are impacted differently by different physics
  • this system identifies stuff that should be "the same" as "the same" and flags others as different
• EMC aware that there method is not the right way forward, but don't have other tools
  • if GMTB gave them something that works, it is likely that they will adopt it
  • EMC's butterfly test was used for the GSM, not for FV3 - and it was only used for computer purchases

Automatic unit conversion etc.

• state converter prototype from CGD, for example temperature to potential temperature
  • goal is to have a number of those converters
  • unit change will require generating temporary variables before the scheme
    • udunits backend? drawbacks like many variations for conversions?
    • Raffaele has something as well - constant library/database that is robust and can be shared, does this have conversion units?
    • start with something simpler, can we do that by June?

standard name library for future auto-generation of code:

• total tendency of STANDARD NAME
• tendency of STANDARD NAME
• STANDARD NAME ...
• from previous timestep

Subcycling versus iteration:

• standard name for timestep remains the same, just gets the subcycling time step dtp/nsteps
• add ITERATION construct in ADDITION to SUBCYCLING which doesn't change the time step

Other work to be done by GMTB for CCPP transition:

• ports
• coupled system
• infrastructure changes, separation of repositories

Other questions by Dave:

• physics routines for WRF
  • inside either k or k+horizontal dimension
  • convention/best practice for dimension?
  • (i,k) versus (k,i) versus k only?