[MITgcm-support] RE: smag

[Alistair Adcroft]

The scheme is due to Leith, not Smagorinsky.

I wasn't referring to cube tile edge but physical boundaries - the no-slip
condition does not use the new variable viscosity (yet). It shouldn't matter
in your calculations with real coasts...

You turn it on by setting
  vectorInvariantMomentum=.TRUE.
  viscAh=0
  viscAhMax=unset (should be 1.D21 by default)
  viscAhGrid=0
  viscC2Leith=0
  viscA4=0 (or small background, prefer not)
  viscA4Max=something sensible (i.e. whatever you were able to run with
before as an upper limit)
  viscA4Grid=0 (or small background, prefer not)
  viscC2Leith=1.

Setting viscA4Max is only a safety feature - you should be able to leave it
unset (but not zero).

The last parameter is non-dimensional and I was able to get it as large as
1.7 in a 2D decaying turbulence test last night. I suspect this will be
application dependent.
The viscosity goes like

   Ah = C2 |grad zeta| dx^3
   A4 = C4 |grad zeta| dx^5 

A4 scales like   C4 U dx^5/L^2   or   C4 Ro f/L dx^5

Procedure: find largest stable value for C4, find smallest stable value. Use
a value nearer the lower limit.

A.
--
Dr Alistair Adcroft            http://www.mit.edu/~adcroft
MIT Climate Modeling Initiative        tel: (617) 253-5938
EAPS 54-1624,  77 Massachusetts Ave,  Cambridge,  MA,  USA

-----Original Message-----
From: Dimitris Menemenlis [mailto:menemenlis at jpl.nasa.gov] 
Sent: Thursday, May 27, 2004 10:16 AM
To: Alistair Adcroft
Subject: smag


Alistair, just tried calling.  How do you turn on Smag?  I'd like to give it
a try in a 32x32x6 config.

The tile-edge issues that you mention in your phone message, are they
related to the missing exchange routines that Chris is working on, or is
there more?

We've had very little access to the altix for past few months (and none for
the last week because of the new password system that they are implementing
;-).  Also Benny and Ichiro are using up the little of the altix that's
available to ECCO.  So Smag needs to be working well at low res first before
we can justify trying it out on the hi-res cube.

Right now, Benny is inching along with 300-s time steps.
So a key motivation for using Smag is to speed things up
by getting back up to 1200-s time steps or more.  It's a
bonus if the physics are also improved in the process.

Any developments with implicit vertical advection?  Typical
CFL numbers for the hi-res cube are 0.6 in the vertical, compared to 0.2 in
the horizontal.

D.


-- 
Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
Jet Propulsion Lab, California Insitute of Technology
MS 300-323, 4800 Oak Grove Dr, Pasadena CA 91109-8099
tel: 818-354-1656;  fax: 818-393-6720