[MITgcm-support] Re: viscosity questions

[Martin Losch]

Hi there,

Eli, Hezi, and I are trying to set up a high resolution marginal sea  
simulation (900m and 300m horizontal resolution), but for some reason  
we cannot get rid off numerical noise unless we specify very high  
diffusivities. I have carefully read the documentation on eddy  
viscosities (Smagorinski and Leith) and Balor's posting about that:
http://forge.csail.mit.edu/pipermail/mitgcm-support/2006-June/ 
004086.html
and a few other so after a while I am feeling comfortable that I  
don't make any mistakes in specifying parameters for the Leith scheme  
any more.

Now this:
the only way I can get a reasonably smooth simulation is with  
advection scheme 2 with biharmonic diffusion of 1e6 (maybe 1e5) or  
harmonic diffusion of 10. (yes simple central differences, 77, 3, 30,  
33, 4 all give a lot of noise, even with extra diffusivity). An  
example data file for the 900m resolution experiment is attached. I  
have also tried harmonic viscosities/diffusivities and the problem is  
the same. We see the same phenomema in the 300m resolution run (but  
with lower viscosity/diffusivity, naturally), namely that the  
"better" advection schemes 33,77,etc. produce noise. What am I missing?

Martin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: data
Type: application/octet-stream
Size: 6682 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20061116/896b3908/attachment.obj>