Merging in Emily Ball's Isca-Mars developments - #302

(cherry picked from commit 46f2ef6)

(cherry picked from commit 0ae2239)

Changed module/routine names etc to lscale_cond_lh

(cherry picked from commit 9d20364)

(cherry picked from commit 958daa1)

Include atmos_param/lscale_cond_lh/lscale_cond_lh.F90

(cherry picked from commit b98deef)

Include atmos_param/lscale_cond_lh/lscale_cond_lh.F90

(cherry picked from commit 1be3277)

Include atmos_param/lscale_cond_lh/lscale_cond_lh.F90

(cherry picked from commit 2ec233a)

(cherry picked from commit 8c61ed5)

Include atmos_param/lscale_cond_lh/lscale_cond_lh.F90

(cherry picked from commit 4ca9008)

(cherry picked from commit b338ff7)

Temperature change to account for CO2 condensation

(cherry picked from commit 11beb1c)

Added dt_tg as input in lscale_cond_lh

(cherry picked from commit bfd934c)

(cherry picked from commit 3bd71ad)

(cherry picked from commit 7f2c985)

(cherry picked from commit 403b943)

(cherry picked from commit eb3ab68)

(cherry picked from commit 22fc5a9)

(cherry picked from commit 32499cc)

(cherry picked from commit ff1ef3a)

(cherry picked from commit 28cb4dc)

(cherry picked from commit ace0b26)

(cherry picked from commit 5792960)

(cherry picked from commit f3dbf06)

(cherry picked from commit 10154d4)

(cherry picked from commit d240f9d)

(cherry picked from commit 84cfa81)

(cherry picked from commit 6f7bfb8)

(cherry picked from commit 1363cf5)

The dust profile computation in run_socrates was gated on
some_dust_condition, a flag that defaulted to .TRUE. and was never
added to the namelist, so it could not be turned off and silently fed
a nonzero dust aerosol mixing ratio (via dust_mix_ratio's default of
1.0e-06) into every Socrates radiation call, regardless of whether
Mars dust forcing was actually wanted. Renamed to do_dust_forcing,
defaulted to .FALSE., and added to socrates_rad_nml, matching the
gating pattern used for the other new physics options on this branch.

…test case

Moves the ported socrates_mars_dust_lh.py into
exp/test_cases/socrates_mars_dust/socrates_mars_dust_test_case.py, matching the
current test-case conventions (module-level Experiment/diag_table/namelist,
compile in __main__, no per-loop Experiment construction since all the swept
lists were single-element). Adds do_dust_forcing:True, now required after the
gating fix. Drops the superseded intermediate socrates_mars_lh.py (LH-only, no
dust) in favour of this combined dust+LH version. Documents the required
externally-sourced spectral/dust-climatology files in input/README.md,
following the same pattern as the existing socrates_mars test case.

Mirrors the existing socrates_mars entry: an import block in
get_experiment_details_from_test_case_name(), and a commented-out entry in
the default test-case list pointing at input/README.md, since it also needs
externally-sourced files before it can run.

Follow-up to the previous commit, which removed the old files but missed
staging the new ones.

Missed during conflict resolution when merging in the do_lscale_cond_lh
dry-convection consistency check - gfortran rejects comparing two
logicals with .eq., it must be .eqv. Caught by a full compile.

Matches the existing explicit lscale_cond_nml block for consistency, since
this test case actually uses do_lscale_cond_lh rather than do_lscale_cond.

Replaces the standalone lscale_cond_lh_mod with a new subroutine,
lscale_cond_co2_mars, living alongside lscale_cond in the same module -
one file, one namelist, rather than a near-duplicate copy of the whole
module for a single alternative subroutine. Renamed for clarity: this is
Mars/CO2-specific physics (temperature-threshold condensation against a
pressure-dependent frost point), not a general Earth-applicable option
like the water-vapor lscale_cond it sits beside.

The four CO2 thermodynamic constants (L_c0, c_pg, c_pc, c_p) are scoped
locally to the new subroutine rather than at module level, with comments
explaining what each represents - including a note that c_pg/c_p (~736-770
J/kg/K) don't match CP_AIR (~845 J/kg/K for the Mars test cases' configured
rdgas/kappa, used everywhere else in the model's energy budget:
two_stream_gray_rad, socrates_interface, spectral_dynamics, mixed_layer,
hs_forcing), and should be checked for consistency before the implied
condensed mass is trusted quantitatively.

Also drops the do_evap/precip_evap call this subroutine inherited from
lscale_cond_lh: it depends on qsat/dqsat/hlcp, none of which this
temperature-threshold formulation ever computes, so it was reading
uninitialized memory regardless of the do_evap namelist setting - as was
the qdel/precip output more generally, now explicitly zeroed rather than
left as whatever was previously on the stack. tdel/lh_rel (the only
physics currently live in this scheme) are unchanged.

lscale_cond_init()/lscale_cond_end() (already called unconditionally by
idealized_moist_phys) now serve both subroutines - no separate init/end
or namelist needed for the Mars variant.

… mismatch

Same bug as socrates_mars_test_case.py (which this test case's namelist was
originally copied from, on the socrates_mars_tcmax_fix branch off master):
sat_vapor_pres.F90 declares tcmin/tcmax as INTEGER, but tcmax was set with a
trailing '.' (350.), making it a Python float that gets serialized as
350.0 - a namelist type mismatch gfortran correctly rejects, surfaced by
FMS's check_nml_error as a FATAL naming sat_vapor_pres_nml.

Confirmed by bisection (removing/varying individual namelist assignments)
and by directly running this test case to completion after the fix.

Once socrates_mars_tcmax_fix lands on master and this branch is synced
with it, this commit's diff should already match - nothing further to do.

Found via trip_test: plain socrates_aquaplanet (no dust, standard ga7
spectral file, n_aerosol=0) crashed with "lookup_es_1d: saturation vapor
pressure table overflow" on this branch but not on master, in an isolated,
cache-cleared, reproducible A/B comparison - a real behavioural difference
between the upstream set_aer_mod (which only allocates the aerosol
structure and never touches mr_source when there's nothing to set) and
this branch's replacement socrates_set_aer, which unconditionally set
aer%mr_source even for spectral files with zero aerosol species.

Wrapping the mr_source assignment and mix_ratio loop in
`n_aerosol > 0` restores exact equivalence to the original set_aer_mod
whenever a spectral file (like the standard, non-Mars ga7 files) defines
no aerosols, while leaving the dust-forcing behaviour for Mars-dust
spectral files (which do define aerosol species) unchanged.

The real fix for the socrates_aquaplanet crash the previous commit
(gating socrates_set_aer's mr_source assignment) didn't fully resolve -
read_control.F90 unconditionally flipped control%l_aerosol from .FALSE.
to .TRUE. for every Socrates run, Mars/dust or not. This is the top-level
switch that tells Socrates' core radiative transfer to apply aerosol
effects at all, so forcing it on unconditionally activated aerosol
processing even for spectral files that define zero aerosol species
(the standard, non-Mars ga7 files) - destabilizing an ordinary aquaplanet
run enough to blow the saturation-vapor-pressure lookup table within the
first few days of a cold start.

Same gating condition as the previous commit: only switch it on when
spectrum%aerosol%n_aerosol > 0.

…osol

The standard ga7 spectral files used by every non-Mars Socrates run also
carry aerosol species blocks (inherited from the Met Office UM) even
though Isca never wires up their parametrizations, so
spectrum%aerosol%n_aerosol > 0 is true for those too. Gating l_aerosol on
that check switched aerosol radiative effects on for ordinary runs,
crashing Socrates' opt_prop_aerosol with "i_aerosol_parametrization ...
illegal value". do_dust_forcing (the namelist flag for the Mars dust
scheme) is the correct switch: it reflects what the user actually asked
for, independent of what a given spectral file happens to contain.

main_nml's days/seconds units, and DiagTable's 'days' time_units, always
mean Earth's fixed 86400s/day regardless of planet. The Mars test cases
(grey_mars, radiative_eq_mars, socrates_mars, socrates_mars_dust) instead
use a calendar Mars day of 88440s (dt_atmos=110s, 804 steps/day) so that
an integer number of days fits into a Mars year - and 86400 is not a
multiple of 110, so the generic 3-Earth-day short-run override crashes
with "run length must be multiple of atmosphere time step" for these
cases. Override to 3*88440s instead, with a matching Mars-specific diag
table using the same day length for its output cadence.

Same fix as radiative_eq_mars_nml_fix (for master): do_newtonian_cooling_as_rad
and newt_relax_surface aren't declared Fortran namelist variables anywhere in
the source, so idealized_moist_phys_nml fails FMS's mistyped-namelist check.
Applied here directly too since trip_test's driver script always reads test
case definitions from the live GFDL_BASE checkout regardless of which commit
is being compiled, so this branch needs its own copy of the fix to run
radiative_eq_mars at all.

Same fix as radiative_eq_mars_nml_fix: update_sst_from_fluxes is not a
declared Fortran namelist variable in mixed_layer_nml. Found by actually
running the test case, which failed on this key next after the previous
commit's fix.

Neither exists anywhere as a declared Fortran namelist variable (confirmed
via full-source grep) - same class of bug as the previous two commits.
With these gone, radiative_eq_mars now runs cleanly and matches master
bit-for-bit, confirming no Mars-dust-physics regression leaked into this
unrelated test case.

These keys (do_newtonian_cooling_as_rad, newt_relax_surface,
update_sst_from_fluxes, rh_target, delta_t_relax) aren't declared on
current master's Fortran namelists, but they likely originate from - and
work correctly on - the much older top_down_with_moisture branch, which
predates the code they depend on being removed/renamed. Rather than
deleting them here, the plan is to reconcile a modernized version of
top_down_with_moisture instead, so this test case is expected to fail
until that lands.

Matches do_dust_forcing's default and the "everything new defaults off"
pattern used throughout this branch: previously, setting do_dust_forcing
alone (without touching account_for_effect_of_dust) was enough to
silently get dust radiative heating, since the latter defaulted .TRUE.
like its water/ozone counterparts. Now both flags must be explicitly set
to get the normal dust-on case, as socrates_mars_dust_test_case.py already
does. No effect on any currently-passing test case - it's the only place
do_dust_forcing is used, and it already sets account_for_effect_of_dust
explicitly.