[MITgcm-devel] spurious vertical diffusion

[Jean-Michel Campin]

Hi Dimitris,

As Martin suggests, you can diagnose the numerical diffusion with
those 2 diagnostics: ADVr_TH & WTHMASS
There are some area involved, and would also need to divide by 
the time-average vertical gradient of T (d.Theta/dz) to get 
something that looks like an effective diffusivity:
 ( ADVr_TH/rAc - WTHMASS ) / ( d.Theta/dz ) : in m^2/s

Jean-Michel

On Tue, May 08, 2007 at 09:30:37AM +0200, Martin Losch wrote:
> I agree with Chris, KPP is most likely the culprit. Try using simple  
> convective adjustment instead of KPP and see what happens.
> 
> I have used advScheme = 7 (DST7 with flux limiter, require OLx=OLy=4)  
> successfully.
> 
> The diagnostics are called
> DFrI_TH and DFrE_TH for theta and diffusion (including KPP), but you  
> want
> ADVr_TH (the total vertical advective flux including numerical  
> diffusion) minus WTHMASS (vertical flux of theta = w*theta), right?
> 
> 
> Martin
> 
> On 8 May 2007, at 02:06, chris hill wrote:
> 
> >D.,
> >
> > Have you tried putting in a tracer that is horizontally uniform  
> >but that has a vertical profile? KPP will probably be doing  
> >something gross, so the background diffusivity may not be so  
> >relevant just below the ice.
> >
> >Chris
> >Dimitris Menemenlis wrote:
> >>Alistair and Jean Michel, are any of your super-duper new tracer- 
> >>advection schemes ready for prime time on the cube?
> >>In the Arctic domain of the 18-km cube configuration, there seems  
> >>to be too much vertical diffusion (or perhaps horizontal diffusion  
> >>across sloping isopycnals).  We have started an integration where  
> >>the explicit background vertical diffusion is much reduced (from  
> >>1.4e-5 to 1e-6 m^2/s) but that does not seem to impact the  
> >>solutions very much.
> >>Right now we use
> >> tempAdvScheme=33,
> >> saltAdvScheme=33,
> >>and we have zero explicit horizontal diffusivity.
> >>Two specific questions:
> >>1) Is there a way to diagnose what is the effective vertical  
> >>diffusivity in the integration?
> >>2) Is there an advection scheme that will lead to less spurious  
> >>vertical diffusion than flux-limited DST3?
> >>Dimitris
> >>_______________________________________________
> >>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