CCPP Framework Meeting Minutes 2021 10 07

Agenda

capgen not yet not implemented inside of MPAS
CCPP physics schemes in MPAS that are b4b with existing non-CCPP schemes
Schemes have been ported but get called directly/explicitly
Tinkering with schemes to get them beyond compliant and more portable (naming conventions etc.)
- Tracking down a b4b issue in a cumulus scheme
Once schemes are ready Dave will hand them over to Steve who will try to get them into CESM
- CESM will use capgen to call this schemes, no direct calls
At the same time, capgen will be introduced in MPAS
Requires some follow-up work in MPAS (namelist options vs pre-compiled code)
- Process identifiers - easy for most primary schemes, difficult for schemes that do multiple things / multiple schemes doing parts of the same process
- We need to follow up with Steve next week on this proposal
Single version of MPAS that calls capgen and/or existing physics or not? Not yet decided
MPAS has reversed dimensions (k,i) but ccpp-physics has (i,k) and WRF has (i,k,j)
- Automatic array transformations (part of DTC's statement of work for this performance period) will address this problem
Suite definition file will be specific to MPAS (which groups exist, how they get called) - same for UFS, SCM, ..., not a problem

GitHub project "capgen unification" https://github.com/NCAR/ccpp-framework/projects/3
Steve on leave until mid October
Created a new issue to capture the work on implementing the var_action class
Steve and Dom need to come up with estimated on how long the different tasks take
Navy would like to use capgen in their first operational implementation, but the implementation window is closing fast (when?)
- Dom sent an email to Alex (Oct 5) to ask about this
Navy also wants to add RAMS microphysics to CCPP physics repository
- Dom also asked Alex about this (who ports it, who tests and maintains it)

Being tested and maintained by someone using one or more models
Regression testing: what does it mean/encompass?
Follow CCPP requirements
Runtime performance and other computational criteria?
Scientific value? This is very difficult to decide, may depend on host model, ...
Relates to ongoing CCPP physics governance discussion

New standard names from MPAS porting efforts
- Every name in every scheme is now uniquely identified across all schemes for MPAS
Dave will create a PR for their variables to CCPPStandardNames
- How do we handle duplicates? By definition there can't be any duplicates
- Differences w.r.t. rules for composing standard names (prefix vs suffix)
- Can we start with schemes that are already in both NCAR ccpp-physics main and MPAS
  - For conflicting names, decide which is better and update the dictionary, existing modeling systems and rules accordingly

Quasi agreement on how to interact between CCPP and the yet-to-be-designed-and-coded-up solution for the particular problems that chemistry has (names of chemical species, assumptions of what they are outside a mechanism, overlap between schemes, schemes handling part of a functionality)
Follow-up meetings with the NCAR chemistry group every 2-3 months
Standard names for explicit chemistry species that the host models need to know about (ozone, ...)
- Matt and Francis have been working on a draft for an abstract interface for aerosols so that various aerosol packages can be communicated through the CCPP

From discussion last week: CCPP metadata is becoming more complex, we should dedicate a section to it in the CCPP technical documentation
With Julie retiring, we will give this task to the new CCPP doc lead (yet to be determined)

CCPP Framework wiki