[MITgcm-support] viscAh & viscAz!

[Baylor Fox-Kemper]

Van Thinh and Dimitris,
I appreciate the compliment, but sometimes it is useful to have a 
constant viscosity for simplicity.

Van Thinh, I think you need to be careful about your grid-cell Reynolds 
number.  That is, you should be aware that you will get noise from the 
advection of momentum unless:

U*deltaX/Ah~1
and
W*deltaZ/Az~1

where U and W are typical velocities and deltaX and deltaZ are the grid 
spacing where you're getting the noise.

The schemes that Dimitris mention automatically do a similar scaling 
for which adjusts "on the fly" for the magnitude of U and W, and so one 
just sets a nondimensional coefficient (or two in the case of viscLeith 
and viscLeithD).

For an internal wave problem, I'd recommend trying the Smagorinsky 
scheme with

  viscC2Smag=4.,
  viscAhGridMax=0.5,

You'll want some vertical viscosity, use the above scaling for your 
grid size and the W values you got from your run that didn't blow up.

You may want some background horizontal viscosity.  This can be pretty 
tiny, you could probably use 2e-6 as you suggest.

Then, if your run is too viscous, just reduce viscC2Smag until you're 
happy...

Cheers,
    -Baylor


On Jun 29, 2005, at 12:29 PM, Dimitris Menemenlis wrote:

>> If anyone would know how can I improve this noise, please give an 
>> advice.
>
> Van Thinh, you need to turn on the Leith or Smagorinsky schemes. 
> Alistair's Leith together with Baylor's Leithd are absolutely, 
> miraculously, fantasmagorically great!!!
>
> D.
>
> -- 
> Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
> Jet Propulsion Lab, California Institute of Technology
> MS 300-323, 4800 Oak Grove Dr, Pasadena CA 91109-8099
> tel: 818-354-1656;  fax: 818-393-6720
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support