Module ocean_vert_mix_mod

OVERVIEW

Time tendency from vertical mixing

OTHER MODULES USED

            constants_mod
         diag_manager_mod
        field_manager_mod
                  fms_mod
          mpp_domains_mod
                  mpp_mod
        ocean_domains_mod
     ocean_parameters_mod
    ocean_tracer_util_mod
          ocean_types_mod
           ocean_util_mod
     ocean_vert_const_mod
      ocean_vert_chen_mod
      ocean_vert_gotm_mod
  ocean_vert_kpp_test_mod
ocean_vert_kpp_mom4p0_mod
ocean_vert_kpp_mom4p1_mod
        ocean_vert_pp_mod
     ocean_vert_tidal_mod
      ocean_vert_util_mod
      ocean_workspace_mod

PUBLIC INTERFACE

PUBLIC ROUTINES

1. ocean_vert_mix_init

   DESCRIPTION

   Initialization for the vertical mixing module.

   
   
   

2. diff_cbt_table_init

   DESCRIPTION

   Read in static diffusivities that have been entered to the diff_cbt_table.

   
   
   

3. bryan_lewis_init

   DESCRIPTION

   Initialize the Bryan-Lewis static background diffusivities.

   
   
   

4. hwf_init

   DESCRIPTION

   Initialize the HWF static background diffusivity. Two forms are available: 1/ Depth dependent form, meant to emulate the Bryan-Lewis approach. This form is not generally used at GFDL. 2/ Depth independent form motivated by use in the CM2G isopycnal ocean climate model at GFDL.

   
   
   

5. diff_cbt_tanh_init

   DESCRIPTION

   Initialize the tanh background diffusivity.

   
   
   

6. vert_friction_init

   DESCRIPTION

   Initialize vertical friction portion of ocean_vert_mix_mod

   
   
   

7. watermass_diag_init

   DESCRIPTION

   Initialization of watermass diagnostic output files. Also determine the logical compute_watermass_diag.

   
   
   

8. vert_mix_coeff

   DESCRIPTION

   This subroutine calls the relevant scheme to compute vertical diffusivity and vertical viscosity. Background values are also incorporated here.

   
   
   

9. vert_diffuse

   DESCRIPTION

   This subroutine computes the thickness weighted and density weighted time tendency for tracer associated with vertical diffusion based on explicit time update of the vertical diffusion equation. MOM only supports aidif==0.0 or aidif==1.0. MOM does not support cases with 0.0 < aidif < 1.0. The watermass diagnostics have not been ported to this subroutine since aidif=0 is rarely used, even in idealized studies.

   
   
   

10. vert_diffuse_implicit

    DESCRIPTION

    Contributions to thickness weighted and density weighted time tendency from time-implicit vertical diffusion.

    
    
    

11. vert_friction_bgrid

    DESCRIPTION

    This subroutine computes the thickness weighted and density weighted acceleration (kg/m^3)*(m^2/s^2) associated with vertical friction. Assumes here that the horizontal grid is B-grid. For aidif=1.0, this module does nothing since all vertical friction is in this case handled implicitly in time, and this is computed elswewhere. MOM only supports aidif==0.0 or aidif==1.0. MOM does not support cases with 0.0 < aidif < 1.0. Note that smf and bmf have units (kg/m^3)*(m^2/s^2) So the vertical diffusive fluxes must be in these units too. For this purpose, we multiply the viscosity by rho0. This is an approximation consistent with the Boussinesq approximation. For non-Boussinesq, we should be using the in situ rho. But to the level of accuracy that we know the vertical viscosity, and to the extent that the ocean density is close to rho0, our use of rho0 for the non-Boussinesq case is of minor consequence for vertical friction calculation. Note: the form drag contribution to vertical viscosity must be handled within aidif=1.0 implicit vertical mixing. It is ignored in this routine, as its contribution would generally cause the model to be unstable. Note: if try to merge this routine with vert_friction_cgrid some machines and compilers will change bits by the mere introduction of extra if-test logic into the calculation. So we define the separate routines to maintain bit-wise agreement with older results, with bit-wise agreement a useful means to check for errors as the model evolves.

    
    
    

12. vert_friction_cgrid

    DESCRIPTION

    This subroutine computes the thickness weighted and density weighted acceleration (kg/m^3)*(m^2/s^2) associated with vertical friction. Assumes horizontal grid is C-grid. For aidif=1.0, this module does nothing since all vertical friction is instead handled implicitly in time, and this is computed in the velocity module. MOM only supports aidif==0.0 or aidif==1.0. MOM does not support cases with 0.0 < aidif < 1.0. Note that smf and bmf have units (kg/m^3)*(m^2/s^2) So the vertical diffusive fluxes must be in these units too. For this purpose, we multiply the viscosity by rho0. This is an approximation consistent with the Boussinesq approximation. For non-Boussinesq, we should be using the in situ rho. But to the level of accuracy that we know the vertical viscosity, and to the extent that the ocean density is close to rho0, our use of rho0 for the non-Boussinesq case is of minor consequence for vertical friction calculation. Note: use visc_cbt for both C-grid velocity components, even though the velocity components sit at different sides of the tracer cell. This choice is for simplicity. It also acknowledges that the alternative of introducing distinct visc_cbt_u and visc_cbt_v would presume knowledge of subgrid scale features that we really do not have. So again, visc_cbt is used for both u,v C-grid velocity components. Note: the form drag contribution to vertical viscosity must be handled within aidif=1.0 implicit vertical mixing. It is ignored in this routine, as its contribution would generally cause the model to be unstable. Note: if try to merge this routine with vert_friction_cgrid some machines and compilers will change bits by the mere introduction of extra if-test logic into the calculation. So we define the separate routines to maintain bit-wise agreement with older results, with bit-wise agreement a useful means to check for errors as the model evolves.

    
    
    

13. vert_friction_implicit_bgrid

    DESCRIPTION

    Contributions to thickness weighted and density weighted acceleration from implicit vertical friction. Assume that the horizontal grid is B-grid. Note that smf and bmf have units N/m^2. These are the natural units for surface stress. To include these stresses as boundary terms in the call to invtri, it is necessary to use vertical viscosities with units (kg/m^3)*(m2^/s) = N/m^2. This is achieved by multiplying visc_cbu by rho0 when sent to invtri. For depth-like vertical coordinates, this is cancelled exactly by the rho0 in rho_dzu. For pressure-like vertical coordinates, the rho0*visc_cbu introduces a negligible change in the vertical viscosity that is well within uncertainty in this coefficient. Include visc_cbu_form_drag to each of the velocity components vertical friction. Note: if try to merge this routine with vert_friction_cgrid some machines and compilers will change bits by the mere introduction of extra if-test logic into the calculation. So we define the separate routines to maintain bit-wise agreement with older results, with bit-wise agreement a useful means to check for errors as the model evolves.

    
    
    

14. vert_friction_implicit_cgrid

    DESCRIPTION

    Contributions to thickness weighted and density weighted acceleration from implicit vertical friction. Assume that the horizontal grid is C-grid. Note that smf and bmf have units N/m^2. These are the natural units for surface stress. To include these stresses as boundary terms in the call to invtri, it is necessary to use vertical viscosities with units (kg/m^3)*(m2^/s) = N/m^2. This is achieved by multiplying visc_cbu by rho0 when sent to invtri. For depth-like vertical coordinates, this is cancelled exactly by the rho0 in rho_dzu. For pressure-like vertical coordinates, the rho0*visc_cbu introduces a negligible change in the vertical viscosity that is well within uncertainty in this coefficient. Include visc_cbu_form_drag to each of the velocity components vertical friction. Note: if try to merge this routine with vert_friction_cgrid some machines and compilers will change bits by the mere introduction of extra if-test logic into the calculation. So we define the separate routines to maintain bit-wise agreement with older results, with bit-wise agreement a useful means to check for errors as the model evolves. Note: use visc_cbt for both C-grid velocity components, even though the velocity components sit at different sides of the tracer cell. This choice is for simplicity. It also acknowledges that the alternative of introducing distinct visc_cbt_u and visc_cbt_v would presume knowledge of subgrid scale features that we really do not have. So again, visc_cbt is used for both u,v C-grid velocity components.

    
    
    

15. on_comp_domain

    DESCRIPTION

    Determine if the point is in comp-domain for the processor.

    
    
    

16. invcosh

    DESCRIPTION

    Inverse cosh function. Argument must be >=1.

    
    
    

17. vmix_min_dissipation

    DESCRIPTION

    Impose a floor to the dissipation arising from vertical tracer diffusion.

    
    
    

18. watermass_diag

    DESCRIPTION

    Diagnose watermass transformation diagnostics for --sbc --bbc --vmix = diff_cbt + K33-implicit neutral diffusion Estimations of the contributions from diff_cbt alone, and from K33-implicit alone, are computed in vert_diffuse_watermass_diag. The diagnostic for vdiffuse computes all processes in one invtri call, which is actually how the model prognostically updates tracers from vertical diffusion. The sum of the sbc + bbc + vmix should equal to the full vdiffuse diagnostic, so that, for example, neut_rho_vdiffuse = neut_rho_sbc + neut_rho_bbc + neut_rho_vmix Additionally, we should have neut_rho_vmix = neut_rho_diff_cbt + neut_rho_k33 with the terms neut_rho_diff_cbt and neut_rho_k33 computed in routine vert_diffuse_watermass_diag. watermass_diag is called prior to implicit update of the tracer fields, so that the initial taup1 value contains only explicit in-time tendencies. The incremental tendencies are diagnosed in this routine by various calls to invtri using same methods as for the prognostic calculation. This routine requires the logical compute_watermass_diag=.true., which is determined inside watermass_diag_init.

    
    
    

19. vert_diffuse_implicit_diag

    DESCRIPTION

    Diagnose contributions from time-implicit vertical diffusion. We perform the diagnostics using the taup1 value of the tracer concentration, obtained after performing the invtri step. The fluxes are diagnosed as if we are performing an explicit-time update, but using the taup1 values of the tracer concentrations. This diagnostic suffers from time truncation errors relative to the prognostic calculation, since the prognostic time update is performed using an invtri calculation. However, the errors are small. The alternative method, which is to separate the combined invtri step into individual physical mixing processes is not an option, since this step is not equivalent algorithmically to the single invtri mixing step. It is for this reason that we perform the diagnostic step using this "explicit flux computed with time-implicit computed concentration" approach. In contrast to the interior mixing, the boundary fluxes can be diagnostically split from the invtri step, as these fluxes are determined prior to the vert_diffuse_implicit routine. Hence, the boundary flux contributions can be diagnosed either by calling an invtri step passing just the stf and btf terms, or by using the even simpler methods in this subroutine.

    
    
    

20. vert_diffuse_watermass_diag

    DESCRIPTION

    Diagnose contributions from time-implicit vertical diffusion acting on watermasses from diff_cbt and k33. We perform the diagnostics using the taup1 value of the tracer concentration, obtained after performing the invtri step. The fluxes are diagnosed as if we are performing an explicit-time update, but using the taup1 values of the tracer concentrations. This diagnostic suffers from time truncation errors relative to the prognostic calculation, since the prognostic time update is performed using an invtri calculation. However, the errors are small. The alternative method, which is to separate the combined invtri step into individual physical mixing processes is not an option, since this step is not equivalent algorithmically to the single invtri mixing step. It is for this reason that we perform the diagnostic step using this "explicit flux computed with time-implicit computed concentration" approach.

    
    
    

21. ocean_vert_mix_restart

    DESCRIPTION

    Write out restart files registered through register_restart_file

    
    
    

22. ocean_vert_mix_end

    DESCRIPTION

    Chen Scheme requires output of Krauss mixed layer for reproducible results. GOTM requires fields for advection tendency.

    
    
    

NAMELIST

REFERENCES

NOTES