[MITgcm-support] EVP stability

[Jinlun Zhang]

Matt,

I don't have much experience with EVP, but here is a paper 
http://www.mrcc.uqam.ca/V_f/Publications/articles/Saucier2004_ClimDyn.pdf, 
that shows a time step of about 2 s for a resolution of 10km. So perhaps 
you could try a smaller dt for your 3.8km grid. Another way to test 
whether resolution is an issue is to go to the model to artificially 
enlarge the grid size, while not changing anything else.

Jinlun

Matthew Mazloff wrote:

> Hi Dimitris,
>
> As you said, for her EVP model Hunke used 30s time-step with grid  
> spacing 16km....at 1/6 deg. resolution my min spacing is  
> 3.8km...about 1/4 this..so I tried 7.5 sec timesteps and the model  
> blew up on the third timestep.  Very unstable.  I have a 4x2 test  
> setup and I ran EVP fine on this so I don't think its a bug in the  
> configuration.  Could it be possible that EVP is just not stable at  
> these resolutions?  Any advice?  Should I make a small high- 
> resolution test setup?  Or does anyone else have one?
>
> Thanks
> -Matt
>
>
>
>
> On Jan 27, 2007, at 9:36 AM, Dimitris Menemenlis wrote:
>
>> Matt, stability criterion for EVP solver depends on internal ice  
>> pressure, strain rates, a damping parameter for the elastic waves,  
>> and other such things.  So it's not easy to compute explicitly.
>>
>> In "Journal of Computational Physics 170, 18–38 (2001)", Elizabeth  
>> Hunke used a time step of 30 s for a domain with horizontal grid  
>> spacing of 16 km.  So 45 s for your 1/6-deg domain seems too  large.  
>> What is your smallest horizontal grid spacing, 4 km?  I  suggest that 
>> you try a 5-s to 10-s time step.
>>
>> Dimitris
>