[MITgcm-devel] Check for OLx/OLy for AdvScheme

[Jean-Michel Campin]

Hi Dimitris,

I have something almost ready to check-in (still few error 
msg to test). It's little bit more complicate than before 
but also more accurate. I've check that I am getting the minimum 
overlap size in a) tutorial_global_oce_latlon + GM advecForm 
+ Visbeck + advScheme=20 (olx=2); b) in global_ocean.cs32x15
with advScheme=20 (olx=2); c) global_ocean.cs32x15 + GM advecForm 
+ Visbeck + advScheme=20 (olx=3).

We next thing will be to extend this to check also for 
seaice advection-scheme (similar to what is done in ptracers).

Cheers,
Jean-Michel

On Mon, Nov 15, 2010 at 04:29:41PM -0500, Jean-Michel Campin wrote:
> Hi Dimitris,
> 
> I agree with you, we should put this check in the code.
> 
> A long time ago, I thought about having a clean (and reliable)
> check of the overlap size (and I guess it's what is commented
> out in gad_check.F). Because right now, it's not
> strictly right (few cases in GMRedi + some advection scheme
> limitations and some due to multi-dim advection).
> And with advection scheme 7 (9 points stencil), it's more
> important to have it right (e.g., it's more likely that
> a user will forget to set Olx=Oly=5 with GM-visbeck
> and advective form).
> 
> I will try to remember what I had in mind, and will
> get back to you soon.
> 
> Cheers,
> Jean-Michel
> 
> On Mon, Nov 15, 2010 at 10:33:55AM -0800, Menemenlis, Dimitris (3248) wrote:
> > Shall I add a check that OLx/OLy>3 for AdvScheme=7 ?
> > 
> > If yes where should that check be added,
> > in gad_check.F or in config_check.F ?
> > 
> > Recently I configured a channel with AdvScheme=7
> > but forgot to increase OLx/OLy to 4.
> > A homogeneous, unforced channel
> > will start moving spontaneously at the edges and
> > eventually blow up if AdvScheme=7 and OLx/OLy=3
> > 
> > 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