[MITgcm-support] Re: viscosity questions

[Martin.Losch at awi.de]

Hi Baylor, Dimitris
thanks for the suggestions. I expect to well below the deformation radius so I will use Smagorinski from now on.
I am using KPP, so that's a good suggestion to try and turn it of
I see noise in T/S and density, not in U/V (so I though I was good with the viscosities).
The noise tends to start near the boundaries/topography, but I have tried JamartWetPoints=.true./.false., without much 
success.

I actually started from Dimitris' suggestion:
http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm_contrib/high_res_cube/input/data_hr
and slowly changed many of the things I found had no impact:
nonlinear free surface back to linear
useareavisclength (I have absolutely square boxes dx=dy)
advection scheme back to 2
multidimadvection is automatically true for 33
vectorinvariantmomentum (does that make a big difference? I couldn't find a positive impact on the noise)
hfacmin=.3 (I tried that, but didn't help)

Martin

Martin Losch
Alfred Wegener Institute 
Postfach 120161, 27515 Bremerhaven, Germany; 
Tel./Fax: ++49(0471)4831-1872/1797



----- Original Message -----
From: Baylor Fox-Kemper <baylor at MIT.EDU>
Date: Thursday, November 16, 2006 8:08 pm
Subject: Re: [MITgcm-support] Re: viscosity questions

> Hi Martin,
>    At 900m or 300m, you probably should be using Smagorinsky, not  
> Leith.  Leith is really only appropriate for the gridscales in the  
> range where one expects the E(k) \propto k^-3 inverse cascade 
> (i.e.,  
> above the deformation radius).  Unless your gulf is quite  
> unstratified, you should be well below the deformation radius, 
> where  
> you will expect a E(k) \propto k^-5/3 scale, and Smagorinsky is the 
> 
> appropriate scaling for that.
>    You could try either c2smag or c4smag or both.
>    As far as diffusivities are concerned, are you seeing noise in  
> T&S but not in U&V?  What about in density?  Also, are you seeing  
> noise near topography (I note you have jamartwetpoints off)?
>    Also, I see you have cadjfreq off, so are you using KPP?  If 
> you  
> are, it will certainly generate gridscale noise during convective  
> events.  You can reduce this by using the "smooth" settings in  
> KPP_OPTIONS.h.
>    I hope that helps.
>       -Baylor
> 
> 
> On Nov 16, 2006, at 12:14 PM, Martin Losch wrote:
> 
> > <data>
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
>