P3 CRM Integration

Steps for integration into samxx and PAM:

• Identify why ice_fall is performed before micro_proc in sam (Kyle will talk to Mikhial O.)

  • Walter is reaching out to Marat and Peter Bl. about this
  • For now, we can just assume it doesn't matter, and ifdef for SAM microphysics only
  • Kyle theorizes it's because of AB + MPDATA
  • Walter theorizes it's because of micro effects on liquid-ice static energy. (This was likely debunked though)
  • Order:
    • Buoyancy
    • Forcing
    • radiation
    • damping
    • ice fall
    • surface fluxes
    • SGS
    • dynamics
    • compute fields at (t+1/2)
    • advection (MPDATA)
    • micro

• Find the ideal way to get P3 source code into E3SM

  • Keep Peter Caldwell and Jim Foucar, Jeff Johnson (EKAT expert among other things), and Rob Jacob in the loop for this
    • The main problem with this is multiple copies of YAKL and Kokkos, etc.
    • We would probably have to use our own CMake rather than SCREAM's. Not sure how hard that would be.
  • Include the entire SCREAM as a submodule in externals or something
    • This automagically brings in EKAT and other stuff
    • Already has SHOC, and Peter C will want to do this eventually anyway
      • SHOC is BL and SGS
      • Need P3 and SHOC to interact similarly to SCREAM
      • BUT! Let's ignore this for now.

• SAM will require both CRM-facing and E3SM-facing interfaces at the same time

  • Walter will own E3SM-facing samxx interface for P3 stuff for MMF forcing
  • Xingqiu and Kyle will own P3-facing interface for SAM++
  • Example in Marat's Fortran SAM already with Fortran P3
    • We could potentially just GPU-port to YAKL the Fortran interface code...
    • Really helps us with thermodynamics being done correctly

• PAM can integrate P3 in true standalone with no E3SM-facing interface.

  • Isaac will own P3-facing interface for PAM

• Create a CRM-facing interface layer on top of P3 source code to transform to / from SCREAM "pack" data structures

  • Matt will own with Kyle's help, the creation of unit tests to interface with SPACKs
  • This will differ between SAM and PAM with potential common files as well.

• Might have inconsistency in current interface where number concentration can have non-zero tendency while mass is zero

  • {Mass , number} --> particle size for sedimentation according to some distribution

• Testing the interface for P3

  • We only need to test the interface, not P3, and Matt will create unit tests for P3 --> YAKL data structures
  • Might need a few more unit tests, but it does not, in general, require realistic input / output data for P3