[NDSL] Hystpdf/Evap/Subl loop

[FlorianDeconinck]

Self-contained 3-axis loop to port with the GFDL_1M parametrization

Original code: GEOS_GFDL_1M_InterfaceMod.F90:L578

Parent: GEOS-ESM/SMT-Nebulae#57

---

• Port Hystpdf/Eval/Subl 3-axis loop to ndsl

[CharlesKrop]

Flowchart to track progress on issue here

[CharlesKrop]

evap3, subl3, and initial calculations within the loop have been ported and verified. hystpdf has been ported and compiles, but is not yet verified.

A QSat module has been developed to handle the numerous calls to various QSat functions throughout GEOS, currently on branch dsl/QSat.

Functions available in this module:

• get_table: called on init, creates the three qsat tables to be referenced from subsequent functions
• QSat_FloatField: takes 3D data as input and computes QSat at each according to Fortran QSAT0/DQSAT0. Interpolates QSat values between ice and liquid water at/around 0C based on TMIX parameter
• QSat_Float: identical to QSat_FloatField but for single point calculations (can be called from within a stencil)
• QSat_Float_Liquid/QSat_Float_Ice: Fortran QSATLQU0/QSATICE0, these functions are similar to QSat_Float, but abandones the interpolation and always calculates for ice/liquid water

KNOWN ISSUES:

• QSat_Float/FloatField fails verification at 86% of points, metric diff O[10^-4]. Errors traces back to differences in saturation tables, which fail at 97% of points. These errors are also small and dependent on temperature. Differences in temperature used within the calculation may be cause of some of these errors. Potential rounding error.
• issue casting to int with dace:cpu backend. have to use intermediary buffer to avoid error

[FlorianDeconinck]

Port is done - might be re-adjusted after validation is ran