[MITgcm-devel] changing OBC in multiDim advection

[Jean-Michel Campin]

Hi Dimitris,

Unfortunatly, I don't think it's going to fix this problem.
There have been a long discussion about the order of 
the calls to OBCS_APPLY and EXCH, and unless Michael 
 a) put the OB well inside (at least 4 pts inside with AdvScheme=7)
or
 b) disable the exch for the tile edges which have OBC,
it's likely that the problem will still be there.
I was talking to Oliver today to see if we could do something to
fix this one.

And to come back to this future/current time step issue, I should be
able to remove all the OBCS_APPLY in gad_advection, and then, to do 
the same thing in seaice_advection will be very easy (since 
there is none).

Cheers,
Jean-Michel

On Thu, Oct 28, 2010 at 12:21:39PM -0700, Menemenlis, Dimitris (3248) wrote:
> Jean-Michel, does this mean that seaice obcs should
> not be used with multiDim advection?
> 
> Michael is running a hi-res Pine Island Bay configuration.
> He uses seaice_advscheme = 7 and he has some sea ice
> artifacts at the boundary.  I would be curious to know if these
> artifacts go away after you add OBCS_APPLY in seaice_advection.F
> 
> D.
> 
> On Oct 28, 2010, at 7:34 AM, Jean-Michel Campin wrote:
> 
> > Hi,
> > 
> > I am planning to change the OBC in multiDim advection (gad_advection.F):
> > Currently: 
> > 1) calculate "future" OB values (obcs_calc) at the beginning
> >  of the time step.
> > 2) after updating the local tracer field (in gad_advection.F) with
> > X-advective flux, we apply the "future" OB value before computing 
> > fluxes in Y-direction.
> > 
> > There is no clear reason to use, at this stage, the "future" OB value
> > rather than the current OB value (i.e., corresponding to the current time step).
> > And as a matter of fact, I prefer the alternative way (use current OB value)
> > to compute all explicit contribution, and it's also much simpler.
> > 
> > One might think that using value at current-time + deltaT/2 would be 
> > more accurate (would be more complicated for sure), but it's not necessary 
> > the case if other than horizontal advection (e.g., forcing, vertical 
> > processes) dominate the evolution of tracer near OB regions.
> > 
> > I propose to change step-2 to use the current OB value (instead of future 
> > one). This will change the results when multi-dim is used for a tracer 
> > (active or passive) which has time-varying OBC.
> > 
> > I did not see any "OBCS_APPLY" calls within "seaice_advection.F", 
> > and propose, later on, to implement the new method (using "current 
> > OB values") in seaice_advection.F.
> > 
> > Is it OK for everybody ?
> > 
> > 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