[MITgcm-devel] Re: [MITgcm-support] noise in high resolution run

[Martin Losch]

Hi Dimitris, Patrick,
I am almost gone (o:

To be honest I do not know too much about viscA4grid. You are  
probably right, but nevertheless IF one uses viscA4/viscAh, for  
whatever reasons, she/he will encounter broken code (as I did). I  
think I will check in this code anyway, but will add a comment in  
CPP_OPTIONS.h that these options only apply to constant viscosity  
parameters viscAh and viscA4.

You are also right about the diffusion parameters. They do need  
isotropic scaling as well, but there everything is fine (as far as I  
can see), and they have some defaults set in GAD_OPTIONS.h.

Anyway these changes do not affect the verification experiments and  
without them running on an "high" resolution isotropic grid is not  
really possible, so I'll go ahead and check this in.

Martin

On Apr 5, 2006, at 10:50 PM, Dimitris Menemenlis wrote:

> Hi Patrick,
>
>> Martin wants to use the adjoint so can't use
>> Leith/Smag., at least not for the time being.
>
> but he should be able to use viscAhGrid and viscA4Grid,
> which depend only on grid dimension and do not require
> any of the cosine CPP option flags.
>
> Maybe the problem is horizontal diffusivity?  Are the cosine
> option flags also applied to diffusivity?  For the 1/8-1/16
> integrations we use
>  diffKhT=0.,
>  diffK4T=0.,
>  diffKhS=0.,
>  diffK4S=0.,
>  tempAdvScheme=33,
>  saltAdvScheme=33,
> but I think that you are not able to use 33 for adjoint.
> Is that the problem?
>
> D.
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel