Langmuir_EFactor doesnt seems to have any value

[patrickscholz]

Hi everyone i have some trouble with the Langmuir parameterisation in cvmix_kpp.F90...

CVMix-src/src/shared/cvmix_kpp.F90

Lines 644 to 657 in a4a6fb0

	call cvmix_coeffs_kpp(new_Mdiff, new_Tdiff, new_Sdiff, &
	CVmix_vars%zw_iface, CVmix_vars%zt_cntr, &
	CVmix_vars%Mdiff_iface, CVmix_vars%Tdiff_iface, &
	CVMix_vars%Sdiff_iface, &
	CVmix_vars%BoundaryLayerDepth, &
	CVmix_vars%kOBL_depth, &
	CVmix_vars%kpp_Tnonlocal_iface, &
	CVmix_vars%kpp_Snonlocal_iface, &
	CVmix_vars%SurfaceFriction, &
	CVmix_vars%SurfaceBuoyancyForcing, &
	nlev, max_nlev, &
	CVmix_vars%LangmuirEnhancementFactor, &
	CVmix_vars%StokesMostXi, &
	CVmix_kpp_params_user)

... there the Langmuir Enhancment Factor is setted as in input parameter

CVMix-src/src/shared/cvmix_kpp.F90

Line 655 in a4a6fb0

CVmix_vars%LangmuirEnhancementFactor, &

... its given as an optional parameter to

CVMix-src/src/shared/cvmix_kpp.F90

Lines 676 to 680 in a4a6fb0

	subroutine cvmix_coeffs_kpp_low(Mdiff_out, Tdiff_out, Sdiff_out, zw, zt, &
	old_Mdiff, old_Tdiff, old_Sdiff, OBL_depth, &
	kOBL_depth, Tnonlocal, Snonlocal, surf_fric,&
	surf_buoy, nlev, max_nlev, Langmuir_EFactor,&
	StokesXI,CVmix_kpp_params_user)

CVMix-src/src/shared/cvmix_kpp.F90

Line 703 in a4a6fb0

real(cvmix_r8), intent(in), optional :: Langmuir_EFactor

It looks to me that the variable CVmix_vars%LangmuirEnhancementFactor, never get filled up with a sensible value but is later used in the computation at multiple locations. There seems to be no line with cvmix_put('LangmuirEnhancementFactor', val,....)

CVMix-src/src/shared/cvmix_kpp.F90

Lines 994 to 999 in a4a6fb0

	if (CVMix_KPP_Params_in%Langmuir_Mixing_Opt &
	.eq. LANGMUIR_MIXING_LWF16) then
	! enhance the turbulent velocity scale
	wm_OBL = wm_OBL * Langmuir_EFactor
	ws_OBL = ws_OBL * Langmuir_EFactor
	end if

CVMix-src/src/shared/cvmix_kpp.F90

Lines 1202 to 1210 in a4a6fb0

	select case (CVMix_KPP_Params_in%Langmuir_Mixing_Opt)
	case (LANGMUIR_MIXING_LWF16)
	MixingCoefEnhancement = Langmuir_EFactor
	case (LANGMUIR_MIXING_RWHGK16)
	MixingCoefEnhancement = cvmix_one + ShapeNoMatchAtS/NMshapeMax * &
	(Langmuir_EFactor - cvmix_one)
	case default
	MixingCoefEnhancement = cvmix_one
	end select

thus i have my doubts that the options Langmuir_mixing_str = 'LWF16' and 'RWHGK16' can work, only Langmuir_mixing_str='NONE' is possible. Can you verify this?!

Thanks for you help !

kind regards

[mnlevy1981]

Users can specify the Langmuir mixing scheme in cvmix_init_kpp(); you should use the optional character string argument Langmuir_mixing_str:

  subroutine cvmix_init_kpp(ri_crit, minOBLdepth, maxOBLdepth, minVtsqr,         &
                            vonkarman, Cstar, zeta_m, zeta_s, surf_layer_ext,    &
                            CVt2, Cv, interp_type, interp_type2, MatchTechnique, &
                            old_vals, lEkman, lStokesMOST, lMonOb, lnoDGat1,     &
                            lenhanced_diff, lnonzero_surf_nonlocal,              &
                            Langmuir_mixing_str, Langmuir_entrainment_str,       &
                            l_LMD_ws, CVmix_kpp_params_user)

Valid values are "LWF16", "RWHGK16", or "NONE", with NONE as default, and it is converted to the integer Langmuir_Mixing_Opt with calls to cvmix_put_kpp():

    if (present(Langmuir_mixing_str)) then
       select case (trim(Langmuir_mixing_str))
       case ("LWF16")
          call cvmix_put_kpp('Langmuir_Mixing_Opt', LANGMUIR_MIXING_LWF16 ,   &
               CVmix_kpp_params_user)
       case ("RWHGK16")
          call cvmix_put_kpp('Langmuir_Mixing_Opt',                           &
               LANGMUIR_MIXING_RWHGK16, CVmix_kpp_params_user)
       case ("NONE")
          call cvmix_put_kpp('Langmuir_Mixing_Opt',                           &
               NO_LANGMUIR_MIXING, CVmix_kpp_params_user)
       case DEFAULT
          print*, "ERROR: ", trim(Langmuir_mixing_str), " is not a valid ",   &
                  "option for Langmuir_mixing_str!"
          stop 1
       end select
    else
       call cvmix_put_kpp('Langmuir_Mixing_Opt',                              &
            NO_LANGMUIR_MIXING, CVmix_kpp_params_user)
    end if

[apologies for the poor formatting of the first version; I was hustling to a meeting and didn't realize blocks of code wouldn't parse correctly from the links provided]

[patrickscholz]

Yes I understand what you want to say but, the options Langmuir_mixing_str='LWF16' and 'RWHGK16' require that CVmix_vars%LangmuirEnhancementFactor has a value!!! But in

CVMix-src/src/shared/cvmix_kpp.F90

Lines 226 to 232 in a4a6fb0

	subroutine cvmix_init_kpp(ri_crit, minOBLdepth, maxOBLdepth, minVtsqr, &
	vonkarman, Cstar, zeta_m, zeta_s, surf_layer_ext, &
	CVt2, Cv, interp_type, interp_type2, MatchTechnique, &
	old_vals, lEkman, lStokesMOST, lMonOb, lnoDGat1, &
	lenhanced_diff, lnonzero_surf_nonlocal, &
	Langmuir_mixing_str, Langmuir_entrainment_str, &
	l_LMD_ws, CVmix_kpp_params_user)

There is no option to set the Variable LangmuirEnhancementFactor, This variable seems to remain only allocated but never initialised with a value

[mnlevy1981]

Oh, I see what you're saying -- CVmix_kpp_params_out%Langmuir_Mixing_opt is used to determine which Langmuir turbulence parameterization to use, but some of those parameterizations use CVmix_vars%LangmuirEnhancementFactor and that value is not set inside CVMix.

There might be a spatial structure to LangmuirEnhancementFactor and CVMix is a single column model, therefore the OGCM needs to set this value on a column by column basis. Looking back at old POP2 code, we used the following:

CVmix_vars(ic)%LangmuirEnhancementFactor = LAMULT(i,j,bid)

CVMix does provide a helper routine for setting this parameter, and POP2 could have also used:

call cvmix_put(CVmix_vars(ic), "LangmuirEnhancementFactor", LAMULT(i,j,bid))

but again, the ocean model should explicitly set this value before calling cvmix_coeffs_kpp() if your choice of Langmuir_Mixing_opt requires it.

[mnlevy1981]

I see that we do not provide an example of this in the stand-alone driver / CVMix test suite; I'll try to find time in the next month or two to update the KPP example to demonstrate this feature.

[mnlevy1981]

@patrickscholz did you delete a comment asking for details on how the LAMULT variable is calculated in POP? It comes from the wave model:

integer :: index_x2o_Sw_lamult       ! wave model langmuir multiplier

This is just setting up the index into the buffer passed from the coupler to the ocean model, but if you follow the breadcrumbs eventually you get to the actual copy into POP memory:

! langmuir mixing variables
! apply enhancement factor to ice-free region
if (IFRAC(i,j,iblock) <= 0.05_r8) then
   ! import enhancement factor (unitless)
   WORKB(i,j       ) = x2o(index_x2o_Sw_lamult,n)
   LAMULT(i,j,iblock) = WORKB(i,j)*RCALCT(i,j,iblock)
else
   LAMULT(i,j,iblock) = c1
endif

I believe the paper that describes all of this is Li, et al., 2017; if that's not the right reference maybe @qingli411 could join the conversation and point us in the right direction :) I just looked through his publication list looking for papers from the 2015 - 2018 time frame that deal with Langmuir and KPP...

[qingli411]

Yes, that is the correct reference (it was published in 2016 but the link is correct).

As @mnlevy1981 mentioned, LangmuirEnhancementFactor is an input parameter for CVMix that is used by Langmuir turbulence parameterization. It is generally a function of time and space that needs to be updated regularly in the ocean model that calls CVMix (similar to the friction velocity as an input parameter). In addition to assigning a value to CVmix_vars%LangmuirEnhancementFactor and passing CVmix_vars to cvmix_coeffs_kpp(), I believe you can also use the more explicit interface (by assigning a value to Langmuir_EFactor):

CVMix-src/src/shared/cvmix_kpp.F90

Lines 676 to 680 in a4a6fb0

	subroutine cvmix_coeffs_kpp_low(Mdiff_out, Tdiff_out, Sdiff_out, zw, zt, &
	old_Mdiff, old_Tdiff, old_Sdiff, OBL_depth, &
	kOBL_depth, Tnonlocal, Snonlocal, surf_fric,&
	surf_buoy, nlev, max_nlev, Langmuir_EFactor,&
	StokesXI,CVmix_kpp_params_user)

[patrickscholz]

Thanks for all the infos, i took now the path to input the Langmuir parameter via the explicite interface of cvmix_coeffs_kpp_low. Since we dont have a fully coupled wave model yet in FESOM, i used a rough approximation for the stokes drift only depending on surface wind speed to compute stokes drift over depth, stokes drift @ the SL_depth and averaged over the SL_depth. Than with the stokes drift averaged over SL_depth i can compute the Surface Layer Langmuir Number LaSL and with that the Langmuir Entrainment Enhancement Factor (EFactor). However i found two ways to compute the Langmuir Entrainment Enhancement Factor (EFactor)

if (kpp_langmuir_entrainment=='LF16') then 
    ! According to Li et al. (2016) --> Mc Williams and Sullivan (2000) proposed 
     kpp_EFactor(node) =  sqrt(1.0_wp + 0.08 / kpp_LaSL(node)**4)
            
else ! (kpp_langmuir_entrainment=='LF17') then 
     ! According to Li et al. (2017) --> Harcourt and D’Asaro (2008) and
     ! Van Roekel et al. (2012) proposed
     kpp_EFactor(node) =  sqrt(1.0_wp + (1.0_wp / 1.5_wp**2) / kpp_LaSL(node)**2 + &
                                                   (1.0_wp / 5.4_wp**4) / kpp_LaSL(node)**4)
            

Based on your experience is there one to favor over the other?

[qingli411]

The second option to compute the Langmuir enhancement factor is much better than the first based on our experience -- the first tend to overestimate the effects (see the comparison between MS2K and VR12-MA in Figs. 3 and 4 in Li et al., 2016).

One way to estimate the Langmuir enhancement factor (and the Stokes drift) from the wind speed without coupling with a wave model is described in Li et al., 2017, which is implemented as a function in CVMix that can be called in the ocean model:

CVMix-src/src/shared/cvmix_kpp.F90

Lines 3151 to 3157 in a4a6fb0

	! !IROUTINE: cvmix_kpp_EFactor_model
	! !INTERFACE:
	function cvmix_kpp_EFactor_model(u10, ustar, hbl, CVmix_params_in)
	! !DESCRIPTION:
	! This function returns the enhancement factor, given the 10-meter
	! wind (m/s), friction velocity (m/s) and the boundary layer depth (m).

[patrickscholz]

Yes i saw the functions

function cvmix_kpp_EFactor_model(u10, ustar, hbl, CVmix_params_in)
function cvmix_kpp_ustokes_SL_model(u10, hbl, CVmix_params_in)

The problem here that i encountered is, that function cvmix_kpp_ustokes_SL_model() only gives back the stokes drift averaged over the SLdepth. However the routine subroutine cvmix_kpp_compute_StokesXi(...) to compute the Stokes similarity parameter StokesXI that is needed by subroutine cvmix_coeffs_kpp(...), requires the stokes drift over depth, stokes drift @SL_depth and stokes drift averaged over SL_depth as input.

I wrote my self a routine based on function cvmix_kpp_ustokes_SL_model(u10, hbl, CVmix_params_in) that gives me out all the necessary stokes drift parameters parameters .