[MITgcm-devel] free drift ice model

[Matthew Mazloff]

Hi Dimitris,

thanks for the advice.   Regarding 1 and 2:  communication is 75% of  
my wallclock time.  I need to minimize exchanges, and though I have  
not tried it, I am afraid that the ice dynamics solver has too many  
exchanges.  Switching to EVP would not remedy this but if I could get  
away from using the solver every timestep that would be great.  My  
model dt = 900;  Is  SEAICE_deltaTdyn = 3600 feasible.  How do you  
find the seaice performance burden in the high-res cube results?   
Regarding 3:  I would love to investigate how the ice moves wrt to  
ocean in the high-res cube results.  Can you point me towards some  
files I could analyze?

Thanks,
Matt







On Nov 30, 2006, at 11:10 AM, Dimitris Menemenlis wrote:

> Matt, a few more options for you to consider:
>
> 1) A wonderful recent addition to pkg/seaice, courtesy of Martin,  
> is EVP sea-ice dynamic solver, which I suspect would be more  
> efficient in your configuration than LSR running at every time step.
> 2) Another way to speed code in your configuration is to set
>    SEAICE_deltaTdyn = 3600
> in data.seaice.  This is because LSR is implicit in everything  
> except Coriolis.  That way cost of LSR will become negligible  
> relative to other stuff.
>
> 3) For diagnosing how ice moves wrt to ocean, the high-res cube  
> results may be more appropriate than the coarser-resolution ECCO- 
> GODAE integrations.
>
> 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-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel