[NDSL] Radiation Coupling

[FlorianDeconinck]

Small 3-axis loop fitting the stencil pattern easily to port within the GFDL_1M

Original code: GEOS_GFDL_1M_InterfaceMod.F90:854

Parent: GEOS-ESM/SMT-Nebulae#57

---

• Port radiation coupling 3-axis loop to pyMoist

[FlorianDeconinck]

First port is done.

GEOS Physics runs at 32-bit, which means that the default Float of NDSL is ill-fitted for physics at large.

Port is ok but two fields are seeing ~5% of errors. Analysis is:

• Our system turns every calculation 32-bit, including literal, where we assume in Fortran there's implicit conversion happening
• We might be looking at data that have very low amplitude because the moist has nothing to do (e.g. low precipitation, everything is in liquid and no other tracers, etc.)

We need to explain those 5% a little further and/or generate a more interesting test case

[FlorianDeconinck]

Pushing back into backlog.

To re-open this task requires:

• clear with Fortran dev the 32/64 literal issue (potentially using a py/fn test case)
• clear with help of GMAO that our test case generating data is actually useful for Moist

[mattweil11]

[FlorianDeconinck]

With new multi-modal metric and fixes to the precision of the constants, this is now passing all tests with HPC generated data.