Module ocean_nphysics_util_mod
OVERVIEW
Utilities for neutral physics, including the code to compute
space-time dependent diffusivities.
Utilities for neutral physics, including the code to compute
space-time dependent diffusivities and many diagnostics.
OTHER MODULES USED
constants_mod
diag_manager_mod
fms_mod
fms_io_mod
mpp_mod
mpp_domains_mod
time_manager_mod
ocean_domains_mod
ocean_density_mod
ocean_operators_mod
ocean_parameters_mod
ocean_tracer_diag_mod
ocean_tracer_util_mod
ocean_types_mod
ocean_util_mod
ocean_workspace_mod
PUBLIC INTERFACE
PUBLIC ROUTINES
-
ocean_nphysics_util_init
-
DESCRIPTION
-
Initialize the utility module for neutral physics.
-
cabbeling_thermob_init
-
DESCRIPTION
-
Initialize the cabbeling and thermobaricity related diagnostic
fields and register the diagnostics to the diag manager.
-
ocean_nphysics_coeff_init
-
DESCRIPTION
-
Initialize the diffusivities used in neutral physics.
Need to initialize them after the ocean_nphysics_util_init routine,
since need to have the domain parameters known for passing the
array size information into the ocean_nphysics_coeff_init routine.
-
tracer_derivs
-
DESCRIPTION
-
Compute the tracer derivatives.
Horizontal derivatives are taken along surfaces of
constant vertical coordinate (constant k-level)
This approach ensures that when neutral physics defaults to "horizontal" physics
next to boundaries, it will do so as horizontal, defined along surfaces of constant
s-surfaces, and so will not generate spurious extrema.
Additionally, when using generalized vertical coordinates, the neutral diffusion
slope should be computed relative to the s-surfaces. The skew diffusion slope
should ideally be computed with respect to z-surfaces, as z-surfaces define
available potential energy. However, when s and z surfaces are reasonably close,
as they are in the interior for zstar and pstar vertical coordinates, then we
choose to to dissipate thickness as defined relative to the zstar or pstar surfaces.
This should not be such a big deal, and it is certainly easier computationally than
worrying about computing two separate sets of slopes. More on this detail is
discussed in "Elements of MOM".
NOTE: This approach is not appropriate for sigma-models. Indeed, many assumptions
in the neutral physics modules need to be rethought for terrain following vertical
coordinates.
Vertical neutral density derivative for use in fz_terms
and fz_flux, and for use in fx_flux and fy_flux.
Note that the derivative at k=nk vanishes by definition
since these derivatives are at the bottom of tracer cell.
also note the use of -epsln_drhodz ensures the vertical
derivative is always < 0. We also support the same
approach used in the mom4p0d code for legacy purposes.
Comments about smoothing drhodz:
1/ Tests in coupled 1-degree model showed extreme sensitivity
of MOC to smoothing. GFDL users generally do NOT smooth, hence
the default drhodz_smooth_vert=drhodz_smooth_horz=.false.
2/ Smoothing the vertical derivative of drhodzb and drhodzh helps
is greatly needed for producing a regularized (i.e., well behaved)
neutral slope vector.
3/ An attempt was made to smooth dTdz and dSdz rather
than drhodz. The resulting slope was smooth, but not as
smooth as when acting on drhodz itself.
-
neutral_slopes
-
DESCRIPTION
-
Subroutine computes the neutral slopes for the triads associated
with the vertical flux component.
Array tensor_31 initially holds the x-slope used for flux component fz.
Array tensor_32 initially holds the y-slope used for flux component fz.
In subsequent calculations, these arrays will be multipied by the
diffusivities.
No slope tapering is applied in this routine.
slopes are computed over k=1,nk-1, since the slope at k=nk
should be zero.
-
compute_eady_rate
-
DESCRIPTION
-
Finish computing eady growth rate.
-
compute_baroclinicity
-
DESCRIPTION
-
Finish computing baroclinicity, which is defined to be the vertically
averaged magnitude of the horizontal density gradient.
-
compute_rossby_radius
-
DESCRIPTION
-
Subroutine computes the first baroclinic Rossby radius of deformation.
Employ WKB approach described by Chelton et al. In particular,
use formulae (2.2), (2.3a) and (2.3b) from their paper.
Place a max and min value on the Rossby radius.
Compute buoyancy frequency in terms of vertical gradient of
locally referenced potential density. Place the reference point
at the interface between the tracer cells, which is also where
the vertical derivative of neutral density is located. This amounts
to a centered difference computation similar to that used by
Chelton et al. equation (B.4).
-
compute_bczone_radius
-
DESCRIPTION
-
Subroutine computes the radius of the baroclinic zone in a manner
suggested by the Hadley Centre approach (Malcolm Roberts, personal
communication).
Algorithm is used in MOM3 and documented in the MOM3 Manual.
-
compute_diffusivity
-
DESCRIPTION
-
Subroutine computes flow dependent diffusivity.
Allow for an added dimensionless tuning factor as well as a
minimum and maximum diffusivity.
-
transport_on_nrho_gm
-
DESCRIPTION
-
Classify horizontal GM mass transport according to neutral density classes.
NOTE: This diagnostic works with transport integrated from bottom to
a particular cell depth. To get transport_on_nrho_gm, a remapping is
performed, rather than the binning done for trans_rho.
Code history
2008: algorithm based (incorrectly) on transport_on_rho
2009: algorithm corrected to be consistent with remapping
used in tracer_on_rho algorithm
-
transport_on_rho_gm
-
DESCRIPTION
-
Classify horizontal GM mass transport according to potential density classes.
Algorithm based on linear interpolation of function on s-surfaces to
function on rho-surfaces.
Diagnostic makes sense when potrho is monotonically increasing with
depth, although the algorithm does not explicitly make this assumption.
NOTE: This diagnostic works with transport integrated from bottom to
a particular cell depth. To get transport_on_rho_gm, a remapping is
performed, rather than the binning done for trans_rho.
Code history
2008: algorithm based (incorrectly) on transport_on_rho
2009: algorithm corrected to be consistent with remapping
used in tracer_on_rho algorithm
-
transport_on_theta_gm
-
DESCRIPTION
-
Classify horizontal GM mass transport according to potential temp classes.
Algorithm based on linear interpolation of function on s-surfaces to
function on rho-surfaces.
Diagnostic makes sense when potential temp is monotonically increasing
with depth, although the algorithm does not explicitly make this assumption.
NOTE: This diagnostic works with transport integrated from bottom to
a particular cell depth. To get transport_on_theta_gm, a remapping is
performed, rather than the binning done for trans_rho.
Code history
2009: algorithm based on transport_on_rho_gm
-
compute_eta_tend_gm90
-
DESCRIPTION
-
Diagnose contribution to global mean sea level evolution arising
from analytic form of Gent-McWilliams scheme.
This routine computes the diagnostic based on an analytic form of the
GM90 contribution. The raw numerical form is computed inside the respective
nphysics modules. The raw numerical form is more accurate and thus
recomended for purposes of sea level budgets.
Compute an averaged slope using tensor_31 and tensor_32.
Then compute the neutral divergence of this slope vector, just
as if each component of the slope was a scalar.
To avoid stencil issues with bottom cells, mask to zero
contributions from cells next to the bottom in either of the
three directions.
Send output to diagnostic manager.
Subroutine history:
Feb2010 version 1.0: Stephen.Griffies
-
cabbeling_thermob_tendency
-
DESCRIPTION
-
Compute tendencies from cabbeling and thermobaricity.
To avoid stencil issues with bottom cells, mask to zero
contributions from cells next to the bottom in either of the
three directions.
Set cabbeling and thermobaricity to zero in regions where
vertical stratification is gravitationally unstable. The idea is
that in such regions, convective mixing will wash away the impact of
along-neutral diffusion, in which case cabbeling and thermobaricity
are not relevant anyhow.
Set the vertical stratification drho_dz to a negative number
with lower bound on its magnitude as set by epsln_drhodz_diagnostics.
Send output to diagnostic manager.
Subroutine history:
Jan2010 version 1.0: initial coding by Stephen.Griffies
Mar2010 version 2.0: tweaks on the division by drho_dz and strat_mask
-
pressure_derivs
-
DESCRIPTION
-
Compute the pressure derivatives for use in thermobaricity diagnostic.
-
calc_gradx_gamma_scalar
-
DESCRIPTION
-
Subroutine computes the i-gradient along neutral surfaces
for a scalar field. For use in the cabbeling and thermobaricity
diagnostic. Thus, when slope is steep, the full derivative is
tapered, rather than just the off-diagonal term.
We only compute neutral i-gradient for interior regions,
since will be setting thermobaricity and cabbeling to zero at
top and bottom levels; do not trust the calculation at top
and bottom boundaries due to stencil truncations.
Algorithm slightly modified from ocean_nphysC calculation of
neutral diffusion x-flux.
-
calc_grady_gamma_scalar
-
DESCRIPTION
-
Subroutine computes the y-gradient along neutral surfaces
for a scalar field. For use in the cabbeling and thermobaricity
diagnostic. Thus, when slope is steep, the full derivative is
tapered, rather than just the off-diagonal term.
We only compute neutral y-gradient for interior regions,
since will be setting thermobaricity and cabbeling to zero at
top and bottom levels; do not trust the calculation at top
and bottom boundaries due to stencil truncations.
Algorithm slightly modified from ocean_nphysC calculation of
neutral diffusion y-flux.
-
watermass_diag_init
-
DESCRIPTION
-
Initialization of watermass diagnostic output files.
-
watermass_diag
-
DESCRIPTION
-
Diagnose effects from nphysics on the watermass transformation.
For use with nphysicsA and nphysicsB.
-
watermass_diag_ndiffuse
-
DESCRIPTION
-
Diagnose watermass transformation from neutral diffusion.
For use with the nphysicsC scheme.
-
watermass_diag_sdiffuse
-
DESCRIPTION
-
Diagnose watermass transformation from skew diffusion.
For use with the neutral physics C scheme.
-
ocean_nphysics_util_restart
-
DESCRIPTION
-
Write out restart files registered through register_restart_file
-
ocean_nphysics_coeff_end
-
DESCRIPTION
-
Write to restart.
NAMELIST
&ocean_nphysics_util_nml
-
debug_this_module
For printing starting and ending checksums for restarts
[logical]
-
nphysics_util_zero_init
For Time%init=.true. and wishing to ensure starting with a clean
suite of nphysics_util fields, even if ocean_neutral.res.nc exists.
[logical]
-
epsln_drhodz
For computing drhodz used in slope calculation.
We must keep drhodz < 0 in order to maintain integrity of the
quasi-Stokes streamfunction as well as computation of buoyancy frequency.
Default epsln_drhodz=1e-30.
[real, units: kg/m^3]
-
drhodz_mom4p1
For computing the vertical deriviative of locally referenced
potrho as in the preferred MOM algorithm rather than the
earlier mom4p0 approach. Default drhodz_mom4p1=.true.
[logical]
-
drhodz_smooth_horz
For horizontal laplacian smoothing the vertical derivative
of density prior to its use in computing the neutral slopes.
This smoothing helps to produce regularized slopes.
Note that this option breaks the integrity of the triads
and is thus NOT generally recommended.
Default drhodz_smooth_horz=.false.
[logical]
-
drhodz_smooth_vert
For vertical 1-2-1 smoothing the vertical derivative of
density prior to its use in computing the neutral slopes.
This smoothing helps to produce regularized slopes.
Note that this option breaks the integrity of the triads
and is thus NOT generally recommended.
Default drhodz_smooth_vert=.false.
[logical]
-
vel_micom_smooth
Velocity scale that is used for computing the MICOM Laplacian mixing
coefficient used in smoothing of drhodzb.
Default vel_micom_smooth=0.2.
[real, units: m/sec]
-
num_121_passes
For number of 1-2-1 passes through to smooth drhodz or
eady_rate in vertical. Default num_121_passes=1.
[integer]
-
aredi
Neutral diffusivity used for experiments using a constant diffusivity.
[real]
-
agm
GM-skew diffusivity used for experiments using a constant diffusivity.
[real]
-
aredi_equal_agm
Will set aredi_array=agm_array, over-riding any other specification
of aredi_array. Default aredi_equal_agm=.true.
[logical]
-
smax
Value of the maximum neutral direction slope above which the neutral fluxes are
either tapered to zero or saturated. Typical value is smax=0.01 or smaller.
[real]
-
swidth
Width in slope over which use tanh with dm_taper scheme to taper fluxes in
steep sloped regions. Typical value swidth=0.1*smax
[real]
-
smax_grad_gamma_scalar
For calculation of gradients of scalars along a neutral direction, then
when abs(slope) > smax_grad_gamma_scalar, will compute the gradient using
only the vertical scalar gradient, since the slopes are so large they are
effectively infinite.
Default smax_grad_gamma_scalar=.01
[real]
-
neutral_horz_mix_bdy
If .true., then use a horizontal diffusivity in the neutral boundary layer.
[logical]
-
vel_micom_bdy
Velocity scale that is used for computing the MICOM horizontal diffusivity
within the neutral boundary layer.
[real, units: m/sec]
-
ah_bdy
Constant horizontal diffusivity for the boundary layer.
Default ah_bdy=0.0.
[real, units: m^2/sec]
-
tracer_mix_micom
If .true., then the GM-skew diffusivity is set according to a velocity scale
times the grid spacing.
[logical]
-
vel_micom
Velocity scale that is used for computing the MICOM diffusivity.
[real, units: m/sec]
-
agm_lat_zones
If true, will set agm_array as constant within two latitudinal zones.
The idea is that one may wish to use a larger agm in the ACC than
elsewhere.
[logical]
-
agm_lat_zones_boundary
Boundary between agm in the south and north zones.
[real]
-
agm_lat_zones_ratio
Ratio between the large agm used in the southern latitudinal zone
to that used in the north.
agm_array(north) = agm
agm_array(south) = agm*agm_lat_zones_ratio
[real]
-
bryan_lewis_aredi
Set bryan_lewis_aredi=.true. when want to have aredi a function of depth
according to the Bryan and Lewis (1979) profile. Maintained for legacy
purposes, and not recommended for new models.
[logical]
-
ahs
ahs = adjustable parameter at the surface for bryan_lewis_aredi
[real]
-
ahb
ahb = adjustable parameter at the bottom for bryan_lewis_aredi
[real]
-
agm_read_restart
For those cases with agm_closure=.false. where we wish to read in
the agm_array from restart files and keep the value from the restart.
This approach allows us to read in a spatially dependent agm_array
that may have been computed from another integration, but to leave
the coefficient static in time.
Default agm_read_restart=.false.
[logical]
-
agm_closure
If .true. then will compute the GM-skew diffusivity as a function of the flow.
The length scale is determined by the Rossby radius and the time scale is
determined by the Eady growth rate. Diffusivities are depth independent.
[logical]
-
agm_closure_max
Maximum GM diffusivity allowed when using agm_closure=.true.
[real, units: m^2/sec]
-
agm_closure_min
Minimum GM diffusivity allowed when using agm_closure=.true.
[real, units: m^2/sec]
-
agm_closure_scaling
Dimensionless tuning parameter for computing flow dependent diffusivities.
[logical, units: dimensionless]
-
agm_closure_upper_depth
Upper depth where start the depth integration to compute the Eady
growth rate and/or baroclinicity.
[real, units: m]
-
agm_closure_lower_depth
Deeper depth where finish the depth integration to compute the Eady
growth rate and/or baroclinicity.
[real, units: m]
-
agm_closure_n2_scale
For computing the agm coefficient using a 3-dimensional
scaling by (N/Nref)^2, with N=buoyancy frequency and
Nref the buoyancy frequency at the base of the neutral
blayer. Default agm_closure_n2_scale=.false.
[logical]
-
agm_closure_n2_scale_coeff
Coefficient setting the scale for the diffusivity computed from
agm_closure_n2_scale.
Default agm_closure_n2_scale_coeff=1e3.
[real, units: m^2/s]
-
agm_closure_n2_scale_nref_cst
For taking the reference buoyancy frequency as agm_closure_buoy_freq
for the (N/Nref)^2 scaling.
Default agm_closure_n2_scale_nref_cst=.false.
[logical]
-
agm_closure_baroclinic
For computing the agm coefficient using only the vertically
averaged magnitude of the horizontal density gradient
(i.e., the "baroclinicity").
[logical]
-
agm_closure_buoy_freq
For computing the agm coefficient using only the vertically
averaged horizontal density gradient, we need to specify a
buoyancy frequency, which is taken to be fixed over all space-time.
[real, units: sec^-1]
-
agm_closure_length_cap
For setting a maximum length scale for the agm_closure calculation.
Default agm_closure_length_cap=.false.
[logical]
-
agm_closure_length_max
Maximum length scale used for computing agm_closure.
Default agm_closure_length_max=50e3.
[real, units: metre]
-
agm_closure_length_fixed
Use fixed length scale for computing agm_closure diffusivity
[logical]
-
agm_closure_length
Fixed length scale for use with agm_closure_fixed_length
[real, units: meter]
-
agm_closure_length_rossby
For computing the agm_closure length scale according to Rossby radius.
[logical]
-
rossby_radius_max
Maximum Rossby radius used for agm_closure_length_rossby and
the neutral_sine_taper. Default = 100e3 m.
[real, units: meter]
-
rossby_radius_min
Minimum Rossby Radius used for agm_closure_length_rossby and
the neutral_sine_taper. Default = 15e3 m.
[real, units: meter]
-
agm_closure_length_bczone
For computing the agm_closure length scale according to radius of baroclinic zone.
[logical]
-
bczone_max_pts
Max number of horizontal grid points for use in computing the baroclinic zone radius.
[integer]
-
agm_closure_bczone_crit_rate
Critical growth rate for determining width of the baroclinic zone.
[real, units: sec^-1]
-
agm_closure_growth_scale
Dimensionless scaling used to set a maximum for agm_growth.
[real, units: dimensionless]
-
agm_closure_eden_greatbatch
For computing the agm_closure length scale according to minimum
of the Rhines scale and the Rossby radius, and using 3d Eady
growth rate.
[logical]
-
agm_closure_eden_gamma
For use in regularizing the growth rate used in the eden/greatbatch approach.
Default agm_closure_eden_gamma=200. Setting to zero removes the regularization.
[real, units: dimensionless]
-
agm_closure_eden_length_const
To set the length scale for agm_closure_eden_greatbatch to constant.
Default agm_closure_eden_length_const=.false.
[logical]
-
agm_closure_eden_length
Length scale for use with agm_closure_eden_length_const=.true.
Default agm_closure_eden_length=10e3.
[real, units: metre]
-
agm_closure_eady_smooth_vert
For vertical 1-2-1 smoothing the eady_rate
Default agm_closure_eady_smooth=.false.
[logical]
-
agm_closure_eady_smooth_horz
For horizontal Laplacian smoothing of eady growth rate.
Default agm_closure_eady_smooth_horz=.false.
[logical]
-
agm_closure_eady_cap
For capping the eady growth rate to avoid huge values.
Default agm_closure_eady_cap=.false.
[logical]
-
agm_closure_eady_ave_mixed
To set the Eady growth rate to its average within mixed layer region.
This is used to avoid spuriously large values which often appear just
in the upper regions of the ocean mixed layer.
Default agm_closure_eady_ave_mixed=.false.
[logical]
-
agm_smooth_space
For smoothing the agm diffusivity in space when nonconstant diffusivity used.
Default is agm_smooth_space=.false.
[logical]
-
agm_smooth_time
For smoothing the agm diffusivity in time when nonconstant diffusivity used.
Default is agm_smooth_time=.false.
[logical]
-
agm_damping_time
The damping time used for time smoothing agm_array.
Default agm_damping_time=10days.
[real, units: days]
-
agm_closure_grid_scaling
For an overall scaling of the agm coefficient, according to
the relative resolution of the grid and deformation radius.
Default is agm_closure_grid_scaling=.false.
[logical]
-
agm_closure_grid_scaling_power
Power used to scale the agm_closure diffusivity.
Default is agm_closure_grid_scaling_power=2.0
[real]
-
aredi_diffusivity_grid_scaling
For an overall scaling of the aredi coefficient, according to
the relative resolution of the grid and deformation radius.
This option is used only when aredi_equal_agm=.false.
Default is aredi_diffusivity_grid_scaling=.false.
[logical]
-
epsln_drhodz_diagnostics
For drhodz used in calculation of dianeutral velocity component
from cabbeling and thermobaricity.
Default epsln_drhodz_diagnostics=1e-7.
[real, units: kg/m4]
-
wdianeutral_smooth
For smoothing the diagnosed dianeutral velocity component using a
horizontal 1-2-1 smoother. Default is wdianeutral_smooth=.true.
[logical]
-
smooth_eta_tend_gm90
For smoothing the diagnosed contribution to steric sea level
time tendency associated with the GM90 scheme.
Default is smooth_eta_tend_gm90=.false.
[logical]
REFERENCES
-
D.B. Chelton, R.A. deSzoeke, M.G. Schlax, K.E. Naggar, N. Siwertz
Geographical Variability of the First Baroclinic Rossby Radius of Deformation
Journal of Physical Oceanography (1998) vol 28 pages 433-460
-
G. Danabasoglu and J. C. McWilliams
Sensitivity of the global ocean circulation to
parameterizations of mesoscale tracer transports
Journal of Climate (1995) vol 8 pages 2967--2987
-
Held and Larichev
A scaling theory for horizontally homogeneous baroclinically
unstable flow on a beta plane
Journal of Atmospheric Sciences (1996) vol 53 pages 946-952
-
M. Visbeck, J.C. Marshall, T. Haine and M. Spall
Specification of eddy transfer coefficients in coarse resolution ocean
circulation models
Journal of Physical Oceanography (1997) vol 27 pages 381--402
-
D. Ferreira, J. Marshall, and P. Heimbach,
Estimating eddy stresses by fitting dynamics to observations
using a residual-mean ocean circulation omdel and its adjoint.
Journal of Physical Oceanography (2005) vol 35 pages 1891-1910.
-
K. Eden, Eddy length scales in the North Atlantic, 2007,
Preprint.
-
K. Eden and R. Greatbatch, 2008: Towards a mesoscale eddy closure,
Ocean Modelling, vol. 20, pages 223-239
NOTES
Diffusivities can be determined in a number of manners
TIME INDEPENDENT
Various methods are available for specifying a time
independent diffusivity, either globally uniform or
with selections of spatial dependence.
TIME DEPENDENT (as a function of the flow)
Various methods are available for determining the
diffusivity that changes in time according to the
properties of the fluid. There are various means
for specifying the length and time scales needed
to set the diffusivity.
LENGTH SCALES
1. First baroclinic Rossby radius (estimated as per Chelton etal).
Equatorial Rossby radius is used within 5deg of the equator.
2. Width of the baroclinic zone, as done in the Hadley Centre
model and documented in the MOM3 Manual.
3. Specified length scale set independent of the flow.
TIME SCALE
When using either of the above for the length scale,
the time scale is determined by the Eady growth rate.
COMBINED LENGTH/TIME SCALE
Another option, used in the GFDL CM2.X coupled climate models,
is to set the diffusivity proportional to the depth averaged
absolute value of the horizontal density gradient.