Added RRTMG and RATS toggles and settings

[msl3v]

MSL

[mathomp4]

Looks like no setup script is providing @CH4_PROVIDER (and I'm guessing @CO2_PROVIDER), so you probably need edits in those.

[msl3v]

Ah. Yes. Thanks. Will do.

[msl3v]

@mathomp4 OK. Done. Do I need to create a new PR? The changes are pushed to the mslong1/rrg+rad branch

[mathomp4]

No. You did right, though I found a small bug. I just pushed a fix and we'll see how CI goes.

[sdrabenh]

MSL

@msl3v can you please provide a description on what is intended behind this PR and what configurations these changes will support? Will there be corresponding changes to the setup scripts?

[msl3v]

There are two primary changes in this set of PRs.

Radiatively Relevant Gases (RRG) Gridcomp: This gridcomp was developed to supercede GOCART for GHG modeling in GEOS. This PR is the 1st formal introduction of RRG into the GEOS-ESM base. There are no gridcomps that currently depend on RRG by default. It lives in @GEOSchem_GridComp

RRTMG:
a) The RRTMG longwave radiation module was modified to receive 3D CO2 fields for LW forcing. CO2 can be provided by either GOCART or RRG. The default settings were not modified
b) A "double-call" was added to the RRTMG LW module that enables online computation of GHG-specific LW radiative forcing diagnostics for any of the RAT gases (CO2, CH4, CFCs, N2O, etc...)

The setup will be (should have been already :-/) modified to populate relevant AGCM.rc settings to default values that preserve a zero-diff result.

[mathomp4]

@msl3v Well, first up the GNU build is not happy:

/root/project/GEOSgcm/src/Components/@GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/@GEOSchem_GridComp/@RRG/RRGMod.F90:350:45:

  350 |     type (MAPL_MetaComp), pointer     :: mapl
      |                                             1
Error: Symbol 'gridcomprun1' at (1) cannot have a type

here is a permalink for this: https://github.com/GEOS-ESM/RRG/blob/2324ec4cdb54cc961f0e868d762ad319e6448da7/RRGMod.F90#L350

I honestly do not know why it's unhappy. You are just declaring a pointer. @tclune any ideas?

But beyond that, Intel builds but the run is dying with:

exception: unknown provider none                                                                                                                                                                                                                                           
pe=00003 FAIL at line=01520    GEOS_ChemGridComp.F90                    <unknown error>

and the issue seems to be that none is not an allowed provider:

https://github.com/GEOS-ESM/GEOSchem_GridComp/blob/d7cf3052142bdd0156a3f1cd30bad6b80bb5cf93/GEOS_ChemGridComp.F90#L1483

[tclune]

I don't know the precise problem, but do have some general recommendations. (And some might have an impact.)

1. All modules should have a default private to avoid re-exporting things. Makes the compilation more efficient at the very least.
2. Consider adding some namespace prefix to your module names. Some of them are rather generic (e.g., global_mod)and could cause a collision in other contexts.

[mathomp4]

I'm going to try building the RRG model with GNU on discover. See if I can reproduce this...