CCPP Framework Meeting Minutes 2019 08 22

Updates:

• Steve wrote a new tool that analyzes a Fortran file and creates a prototype of the metadata file

WRF-MPAS-GFS physics interoperability:

• both WRF-MPAS and GFS have YSU
• Dave and Chunxi looked at the code diffs and identified places that need to be combined (fields needed by WRF-MPAS not in GFS version and vice versa)
• fields with different units, some fields (tracers) in one big array versus split out
• temperature versus potential temperature
• demonstration deadlines coming up soon - when?
  • portability aspect of YSU scheme to GFS is not in a critical path
  • but portability to CESM early/mid September (when YSU needs to run through CCPP in MPAS)
• will start with the version that in in WRF-MPAS, not with what is in ccpp-physics today
  • this version has one horizontal dimension and one vertical dimension
  • needs starting and ending for horizontal dimension
• in ccpp_prebuild.py, optional arguments are used to pass the available data (e.g. temperature or potential_temperature) to the scheme
• another option would be to augment the SDF and add information about optional arguments in there (instead of ccpp_prebuild.py)
• what about a host model that doesn't have combined tracer arrays
  • in the past, we had discussions that these combined arrays would be constructed based on information in the metadata
• todo: open issues on ccpp-framework
  • schemes that take constituent arrays with host models that don't have those arrays
    • is this a real problem? none of the models we are working with now/in the near future has this problem
  • CCPP constants / constant values for a model run:
    • prebuild system could create a parameter file accessible by both the host model and the physics
    • especially useful for handling constituent arrays
  • in addition, there are some scientific questions (related to cloud ice/water and longwave/shortwave radiation fields)
• do we really want to have two different versions in ccpp-physics? can we agree on a common core and move the rest to model-specific interstitials? this is our first study problem
• Thompson would be easier but that is not in the targeted MPAS suite to port to CCPP first

Standard names discussion (cont'd):

• chemistry folks ok with "interface" and "layer", "level" was termed to be too ambiguous
• "layer" is the default for a 3-d variable, if not need to decorate with "interface", "at_lowest_model_level", ...
• dimensions should follow this terminology: "vertical_layer_dimension", "vertical_interface_dimension"
• cell averages would take another "average" decorator