[MITgcm-devel] momentum OBCS implemetation

[Jean-Michel Campin]

Hi Dimitris,

OK. 
going to add (in verification/seaice_obcs/input/data):
 diffK4T=1.e11,
 diffK4S=1.e11,
much smaller than the stability limit (arround 2.e15)
but large enough to see changes in monitor stats if I go back 
to old version of OBCS + bi-harmonic diffusivity

Thanks,
Jean-Michel

On Fri, Apr 29, 2011 at 08:18:22AM -0700, Menemenlis, Dimitris (3248) wrote:
> Jean-Michel, suggestion is good.
> Go ahead and change seaice_obcs
> 
> Cheers
> 
> Dimitris Menemenlis
> 
> On Apr 29, 2011, at 8:11 AM, Jean-Michel Campin wrote:
> 
> > Hi,
> > 
> > I did some changes last month (once again) in OBCS implementation for 
> > tracer, using maskInC,W,S in pkg/generic_advdiff/ S/R for all the 
> > wider-stencil scheme (assuming zero gradient beyond the OB), 
> > so that the calls to OBCS_COPY_TRACER (that I added last November !)
> > are no longer needed (just removed yesterday).
> > 
> > The main motivation for this is that it's cleaner (e.g., for biharmonic diffusion), 
> > the OB border-line can be more general (concave interior) with OB closer to 
> > each-other (Olx grid-points separation instead of 2xOlx) and it can be extended 
> > to the momentum whereas the OBCS_COPY_TRACER approach is more difficult for U,V.
> > 
> > Going to apply the same idea to the momentum, starting with biharmonic 
> > viscosity (& leaving the 4th Order coriolis for later).
> > In short, the changes are (like zero gradient beyond the OB):
> > - to multiply divergence (& tension) by maksInC
> > - to multiply horiz. grad of vorticity by maskInW/S
> > Additional changes are needed in mom_calc_visc.F (Leith & Smagorinsky)
> > but even the changes outside mom_calc_visc will already affects the computed 
> > viscosity. So I was going to check-in these changes all at once.
> > 
> > From early tests, it seems that the effects of these changes from the computed 
> > viscosity (Leith) are of the same order that the one from biharmonic viscosity.
> > 
> > The debug parameter I added (OBCS_uvApplyFac) should allow to confirm
> > that the results no longer depend on normal-component being applied 
> > to the 2nd grid-point (=beyond the OB).
> > 
> > But one issue with those changes is that very few test-exp are currently
> > testing momentum wider-stencil scheme with OBCS (in fact, only dome results
> > are affected). Would be good to test biharmonic diffusivity (un-tested),
> > constant biharmonic visc (in flux-form) and zStar could be added to the list.
> > I propose to:
> > - slightly change exp4 (add Non-Hydrostatic and small viscA4)
> > - add 2nd exp4 test with zStar (and time-varying OBC ?), but not yet ready.
> > And was thinking of adding small diffK4T,diffK4S into seaice_obcs to
> > test biharmonic diffusivity. Other exp using OBCS are not as good test,
> > either 2-D or starting with weak T,S horiz gradient or don't have the 
> > 4 types of OB (N,S,E,W).
> > Dimitris, is it OK with you ? (and in this case would add a comment line 
> >  next to these 2 param in seaice_obcs/input/data); 
> > or should I add a secondary test ? (But is it worth to have 2 tests with so 
> > little differences in parameter list ?).
> > 
> > Cheers,
> > Jean-Michel
> > 
> > 
> > _______________________________________________
> > 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