[MITgcm-devel] ggl90_calc

[David Ferreira]

Martin Losch wrote:
> David,
>
> I think that the ALLOW_GGL90_SMOOTH and ALLOW_GGL90_HORIZDIFF should 
> really be exclusive options. The first (your addition) is a 9-point 
> running average smoother, whereas the HORIZDIFF is in effect a 5-point 
> smoother (but maybe a bit more physical).
>
> what does mskCor do (as far as I can see, it's 1 everywhere)?
> what's the rationale behind dividing by TKEPrantlNumber?
Hi Martin,
Jean-Michel and I had a look a ggl90_cal.F. A couple of his remarks were 
due to some
confusion about the Dirichlet BC, but there were also a few weird stuffs 
in the subroutine.
In the end, we changed a couple of masks, treated the upper boundary 
condition as the
bottom one, put a deltaT in the TKE diffusion.
Also, you are right: ALLOW_GGL90_SMOOTH and ALLOW_GGL90_HORIZDIFF should
be exclusive (there is a check in ggl90_check.F now).
About mskCor: it is not 1 everywhere on the cube sphere (it is zero in 
the weird overlaps near
the corners).
About TKEPrantlNumber in the smoothing: it is because the viscosity is 
averaged to obtain
the diffusivity. This is exactly the OPA implementation (initially I 
tried my own flavor by averaging
directly the diffusivities, and it doesn't work, apparently the Prandlt 
number itself should not be
averaged).
And finally, about having some convective adjustment with ggl90: it is 
often necessary to deal
with unstable profile in the ocean interior. OPA has convective 
adjusment on top of ggl90. And,
(admittedly a weak reason), KPP has one hidden somewhere.

To be checked in soon.
Cheers,
david













> Martin
>
> On Nov 2, 2009, at 2:36 PM, Martin Losch wrote:
>
>> Hi Jean-Michel and David,
>>
>> here are my  comments (after a relatively quick glance);
>> On Oct 31, 2009, at 4:31 AM, Jean-Michel Campin wrote:
>>
>>> Hi Martin and David,
>>>
>>> I look at S/R ggl90_calc.F (more precisely, to the implicit diffusion
>>> of TKE) and noticed few thing that I don't like too much
>>> (at least misleading, but some might also be more like a bugg).
>>>
>>> It's all checked-in as comments in the code ("C- jmc:"):
>>>> danton{ggl90}% egrep '^C- jmc:' ggl90_calc.F
>>>> C- jmc: concerned about missing a deltaT since gTKE is already the 
>>>> future TKE.
>> I agree, there is a deltaT missing. By default, the horizontal mixing 
>> is turned off, and nobody ever noticed that it does not work 
>> properly, maybe we don't even _want_ horizontal diffusion?
>>>> C- jmc: concerned that a(k=2) should always be zero
>>>> C- jmc: concerned that c(k) from k=kLow to k=Nr should always be zero
>>>> C- jmc: this is dangerous since klowC could be zero if land column
>>>> C- jmc: concerned about conservation when a or c are changed after 
>>>> computing b
>> I agree, the implicit solver does not look clean (I remember that I 
>> changed things around a lot before I got them working properly, but 
>> that does not mean that they are actually correct). Maybe it's a good 
>> idea to replace this code with a call of S/R solve_tridiagonal, as 
>> you have already tried? I do not think that we can use impldiff 
>> (which would even be better), because GGL90TKE is defined at the cell 
>> interfaces (and not the cell centers as theta/salt/uVel/vVel). What 
>> do you think.
>>>> C- jmc: would be much better to update the provisional TKE (i.e. 
>>>> gTKE) at k=2
>> I don't understand this part, We do I want to impose the surface 
>> boundary conditions below the surface?
>>
>> M.
>>
>> PS. More comment for David:
>> I don't understand this:
>>> C     GGL90 needs convection turned on, but keep warning message
>>>      IF (cAdjFreq.NE.0.  .OR. ivdc_kappa.NE.0. ) THEN
>>>         WRITE(msgBuf,'(A)')
>>>     &  'GGL90_CHECK: WARNING: Some form of convection has been enabled'
>>>         CALL PRINT_MESSAGE( msgBuf, errorMessageUnit,
>>>     &                        SQUEEZE_RIGHT , myThid)
>>>      ENDIF
>>
>> Do you want to use GGL90 along with convective adjustment or the 
>> "implicit diffusion" convection scheme? I think that the stop 
>> statement should go back in and that the comment should read
>>> C     GGL90 needs other types of parameterized convection (i.e. 
>>> convective adjustment and "implicit diffusion") to be turned off
>>
>> Also it would be nice if you could include some comments about the 
>> different choices for the mixing length. Would be nice if I know what 
>> choices I have and what they mean (instead of just 0-3).
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel