Many models (for example, aerosol and chemistry models, but potentially others as well) use very specific variables that will realistically only be used in a single context. For example, an aerosol scheme may use custom bins for various constituents that are based on the specific logic used in that model, and not general physical laws. Not only will a variable like "mass fraction for dust bin 3" not be useful outside of the specific context of that model, but it may be ambiguous and confusing in the context of other models, schemes, etc. that use different binning strategies.
Potential solutions
We had some initial discussion during a CCPP Framework meeting, which included a few initial suggestions:
Separate host- / model-specific variables into a dedicated section
Add a new attribute for variables indicating which models make use of that specific variable
Include model in variable names
Further discussion is needed on this, hopefully we can use this issue as a jumping-off point.
Through discussion at last week's CCPP framework meeting, we decided that, at least in the short term, the first option above is preferable. The option to allow subsections is being discussed in Issue #48.
The issue
Many models (for example, aerosol and chemistry models, but potentially others as well) use very specific variables that will realistically only be used in a single context. For example, an aerosol scheme may use custom bins for various constituents that are based on the specific logic used in that model, and not general physical laws. Not only will a variable like "mass fraction for dust bin 3" not be useful outside of the specific context of that model, but it may be ambiguous and confusing in the context of other models, schemes, etc. that use different binning strategies.
Potential solutions
We had some initial discussion during a CCPP Framework meeting, which included a few initial suggestions:
Further discussion is needed on this, hopefully we can use this issue as a jumping-off point.