[MITgcm-devel] stevens boundary conditions

[Martin Losch]

Hi Jean-Michel,

I thinks I know what's going on: in timestep_tracer the gtracer is masked by maskInC. That of course spoils my idea. Do you think we can get rid of this mask? Here's your check-in message: "leave tracer unchanged outside OB interior region: This has no effect
 on the solution but just to prevent unrealistic tracer value ouside OB."

Martin

On Dec 12, 2011, at 5:52 PM, Martin Losch wrote:

> Hi Jean-Michel,
> 
> before I check in my modified obcs_calc_stevens, I need to understand, why I do not see any effect  of those changes  in exp4. 
> I expect for the western boundary of exp4 (i=1) the advective fluxes "af" (even in the diagnostic DFxE_TH) are zero, but this does not seem to be so, why? Instead, they even vary with time, what am I missing? I tried tempAdvScheme 2, 4, 33
> 
> Martin
> 
> On Dec 12, 2011, at 3:43 PM, Jean-Michel Campin wrote:
> 
>> Martin,
>> 
>> It's not a bad idea to put those masks in gmredi routines, I agree. 
>>> so I am ready to check this in, if you do not object.
>> Yes, go for it.
>> 
>> Jean-Michel
>> 
>> On Mon, Dec 12, 2011 at 03:34:38PM +0100, Martin Losch wrote:
>>> Hi Jean-Michel,
>>> 
>>> rather than introducing another df-field, I decided to apply the maskInC directly in the gmredi routines (gmredi_rtransport and gmredi_calc_diff). All the changes that I did to implement option 2 did not change any results, so I am ready to check this in, if you do not object.
>>> 
>>> Martin
>>> 
>>> Index: generic_advdiff/gad_advection.F
>>> ===================================================================
>>> RCS file: /u/gcmpack/MITgcm/pkg/generic_advdiff/gad_advection.F,v
>>> retrieving revision 1.69
>>> diff -r1.69 gad_advection.F
>>> 810c810
>>> <      &        )*rkSign
>>> ---
>>>>    &        )*rkSign*maskInC(i,j,bi,bj)
>>> Index: generic_advdiff/gad_calc_rhs.F
>>> ===================================================================
>>> RCS file: /u/gcmpack/MITgcm/pkg/generic_advdiff/gad_calc_rhs.F,v
>>> retrieving revision 1.58
>>> diff -r1.58 gad_calc_rhs.F
>>> 522c522
>>> <           fVerT(i,j,kUp) = fVerT(i,j,kUp) + af(i,j)
>>> ---
>>>>         fVerT(i,j,kUp) = fVerT(i,j,kUp) + af(i,j)*maskInC(i,j,bi,bj)
>>> 718a719
>>>> C     for Stevens OBC: keep only vertical diffusive contribution on boundaries
>>> 724,725c725,726
>>> <      &   *( (fZon(i+1,j)-fZon(i,j))
>>> <      &     +(fMer(i,j+1)-fMer(i,j))
>>> ---
>>>>    &   *( (fZon(i+1,j)-fZon(i,j))*maskInC(i,j,bi,bj)
>>>>    &     +(fMer(i,j+1)-fMer(i,j))*maskInC(i,j,bi,bj)
>>> 730c731
>>> <      &                  )
>>> ---
>>>>    &                  )*maskInC(i,j,bi,bj)
>>> Index: gmredi/gmredi_calc_diff.F
>>> ===================================================================
>>> RCS file: /u/gcmpack/MITgcm/pkg/gmredi/gmredi_calc_diff.F,v
>>> retrieving revision 1.10
>>> diff -r1.10 gmredi_calc_diff.F
>>> 58a59
>>>>    &           *maskInC(i,j,bi,bj)
>>> 64a66
>>>>    &           *maskInC(i,j,bi,bj)
>>> 75a78
>>>>    &           *maskInC(i,j,bi,bj)
>>> 81a85
>>>>    &           *maskInC(i,j,bi,bj)
>>> Index: gmredi/gmredi_rtransport.F
>>> ===================================================================
>>> RCS file: /u/gcmpack/MITgcm/pkg/gmredi/gmredi_rtransport.F,v
>>> retrieving revision 1.16
>>> diff -r1.16 gmredi_rtransport.F
>>> 133c133
>>> <      &      - _rA(i,j,bi,bj)
>>> ---
>>>>    &      - _rA(i,j,bi,bj)*maskInC(i,j,bi,bj)
>>> 137c137
>>> <      &      - _rA(i,j,bi,bj)
>>> ---
>>>>    &      - _rA(i,j,bi,bj)*maskInC(i,j,bi,bj)
>>> 167a168
>>>>    &        *maskInC(i,j,bi,bj)
>>> 184a186
>>>> c    &    *maskInC(i,j,bi,bj)
>>> 
>>> 
>>> On Dec 12, 2011, at 3:03 PM, Jean-Michel Campin wrote:
>>> 
>>>> Hi Martin,
>>>> 
>>>> I did some tests to check that seaice obcs were clean
>>>> (but don't remember how I did them !),
>>>> and did not pass unless I put useKPP=F (would need to
>>>> investigate within KPP code how to make diffusivity & viscosity
>>>> independant of value from the other side). I think GGL90
>>>> might also have similar issues but should be easier to fix.
>>>> 
>>>> Cheers,
>>>> Jean-Michel
>>>> 
>>>> On Mon, Dec 12, 2011 at 01:34:39PM +0100, Martin Losch wrote:
>>>>> On Dec 12, 2011, at 12:41 PM, Martin Losch wrote:
>>>>> 
>>>>>>> (in fact, in seaice_obcs test.exp. 
>>>>>>> which is using KPP, some of the western OB values already 
>>>>>>> influence the Eastern side).
>>>>> 
>>>>> How do you see that? Do I need to modify the exp to see that? Or is the flag OBCS_uvApplyFac enough?
>>>>> 
>>>>> M.
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> 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
>>> 
>>> 
>>> _______________________________________________
>>> 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
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

Martin Losch
Martin.Losch at awi.de