[MITgcm-devel] Re: arctic with thsice

[Martin Losch]

Hi Dimitris,

I don't think it's time to compare these distributions yet, we just  
wanted to see wether the thsice+seaice is stable (and it appears so).

For a direct comparision you'd need to
a. start from the same (or similar) initial conditions
b. use the same advection scheme for both models (which would mean  
thSIceAdvScheme=SeaiceAdvScheme=1 (+ diff1=0. in data.seaice)  as  
this the only one that both models support. "1" means 1st order  
upwind so you can bet that it is very diffusive.)

For a. (initial conditions) you probably need to generate fields for
actual thickness (just use the HEFF of seaice_initialheff),  
fractional cover (=area=1 where there is ice), snow height (20cm  
where there is ice) and two enthalpies (uniform latent heat of  
freezing where is ice should be sufficient).

Martin

PS. Have you tried some runs with EVP? Maybe reducing the evp step  
may help, too.
Otherwise I am thinking of trying a slightly different discretisation.

On 26 Apr 2007, at 21:52, Dimitris Menemenlis wrote:

> Jean Michel and Martin, comparison of thsice and seaice  
> thermodynamics after 5 month integration attached.  tshice looks a  
> lot more diffuse at the edges?  Is this the advection scheme or  
> something else?  For seaice I use the default scheme.  Is this the  
> relevant parameter for thsice: thSIceAdvScheme=1,
> Is there any added diffusion to that also?  Maybe AdvScheme=1 is  
> too diffuse?
>
> D.
>
> <thsice_vs_seaice11304.ps.gz>