MOM_ice_shelf_dynamics.F90  modified
initialize_ice_shelf_boundary_channel in MOM_ice_shelf_initialze.F90

Merge branch 'revert_drivers' of https://github.com/MJHarrison-GFDL/MOM6 into matts-revert-drivers

…e ice-shelf_fields consistent

with diag_table
Modifications to MOM_ice_shelf.F90 to apply melting to the case of a
dynamic ice shelf.

v_bdry_val through halo

due to sub-ice-shelf melting

- Tag v0.93-beta of CVMix is the last tag on their "master" branch before
  an API change within CVMix.
- Answers reproduce with the three-year prior commit being used, as tested
  in MOM6-examples single_column cases. These do not cover all CVMix code.

- This tag involves an API change that requires changes to MOM_tidal_mixing.F90.
- Changes are dropped arguments so presumably that data is not needed.
- Answers reproduce in so far as they are covered by MOM6-examples.
  Not all of CVMix is covered in these tests and this code in particular
  is not.

- v0.98-beta is the latest tag of CVMix that is available a.t.t.
- There were no API changes since the v0.94b-beta that affected MOM6.
- No answer changes for MOM6-examples, but these do not exercise much
  of CVMix.

…90; MOM_ice_shelf_initialize.F90

This patch fixes an issue with the vertical array bounds of the Ferrari
et al. streamfunction.  The array is bounded across interfaces, from 1
to nz+1, but only the interior values need to be determined due to the
arbitrary boundary value (set here to zero).

In the current source, the streamfunction is rescaled before calling
streamfn_solver, but need not be applied to the boundary values.  This
is unlikely to cause errors in production, since the values are later
reset to zero, but the rescaling can raise errors in more aggressive
debugging builds, such as when the arrays are initialized with NaN
values.

This commit fixes the issue that dirs%restart_input_dir is not initiated before
"ocean_solo.res" and "ice_solo.res" are read.

Bugfix: empty restart_input_dir for *_solo.res

Merge from main

  Added code to explicitly set a (1x1) io_domain when no other io_layout is
specified, complying with changing requirements for 2020 and later versions of
FMS, and following the default behavior of previous versions.  Also corrected
three module use statements in infra/FMS1 to eliminate any dependencies of code
in config_src/infra/FMS1 on code in src/framework.  These use statements
eventually point to the same place as before, but with less indirection.  This
change should facilitate later steps to compile everything in and under the
config_src/infra directories as libraries.  All answers are bitwise identical.

* if true, allow the topog overrides to change the land mask

Explicitly set (1x1) io_domain as a default

- Units of isoneutral or interface slope were recorded as "nondim". While
  true in SI units, not so for MOM6 units. MOM6 distinguishes between units
  of length in the vertical (Z) and horizontal (L) the slopes should have
  units of "Z L-1 ~> nondim".
- This has consequences for other variables in calc_isoneutral_slopes().
  - An internal constant, G_Rho0, was defined differently from elsewhere in
    the code. "g" has units of "L2 Z-1 T-2 ~ m s-2" because it is the
    vertical component of the gradient of geopotential in "L2 T-2 ~ m2 s-2".
    Everywhere else `G_Rho0 = g_Earth/Rho0` but in this routine it was
    different in order render N2 (the Brunt-Vaisala frequency) in units of
    "T-2" (s-2).
  - N2 is a quantity associated with dispersion relations and defined
    N2 = - g/Rho0 d/dz rho and either way acquires units of "L2 Z-2 T-2"
    and not just "T-2". In SI units L2 Z-2 = 1. So I have also changed
    the units of N2 in this, and connected, modules.
- The changes also propagate to MOM_lateral_mixing_coeffs.F90 and
  MOM_thickness_diffuse.F90.
- Changing the definition of G_Rho0 in calc_isoneutral_slopes(), and
  its units to "L2 Z-1 T-2", the slope and N2 calculations then require
  many less inline conversions. Many of the one-line changes in this commit
  remove factors like US%Z_to_L.There is one exception:
  - In the calculation of slope, we use in the denominator a mostly
    non-vanishing replacement for d/dz rho, the magnitude of grad rho from
    mag_grad2 = ( d/dx rho )^2 + ( d/dz rho )^2. In code this had
    `mag_grad2 = drdy**2 + (L_to_Z*drdz)**2` since this is mixing
    gradients in the horizontal and vertical. The result should be
    in "R2 Z-2" so now `mag_grad2 = (Z_to_L*drdy)**2 + drdz**2`
- A few diagnostics needed new, or changed, conversion factors.
- One run-time parameter needed a conversion parameter.
- For the most part this commit moves inline conversions of units to
  the I/O stage, which is an indicator that it is the right thing to do.

Update CVMix submodule

Change units of slope returned from calc_isoneutral_slopes() to "Z L-1"

Bugfix: FGNV streamfunction vertical bounds

dynamic ice shelf

Two minor changes to the .testing build:

- We now apply `--recurse-submodules` to the target build in the
  regression test.  This is required after an update to the submodules,
  when the target submodule is out of sync with the main branch (e.g.
  dev/gfdl at NOAA-GFDL).

- A LDFLAGS_USER hook was added to the `.testing/Makefile`
  configuration, similar to the FCFLAGS_* hooks.  This is required if
  the library dependencies do not reside in the default directores.

  For example, this may be needed for a custom netCDF or MPI library.

Testing: Recurse target submodules, LDFLAGS hook

EMC update 20210322

…ate-2021-03-26

dev/gfdl main candidate 2021-03-26