[MITgcm-devel] cost function terms for obcs

[Matthew Mazloff]

Hi Martin,

I already do this.  And yes, I think this is how it should be done and  
would be useful to others -- so I vote to check it in.  Here is my  
code which I believe I acquired from I Hoteit -- it likely needs a  
little cleaning

-Matt


-------------- next part --------------
A non-text attachment was scrubbed...
Name: ObcsCostCode.tar
Type: application/x-tar
Size: 30720 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20100401/35d2683f/attachment-0001.tar>
-------------- next part --------------







On Apr 1, 2010, at 6:41 AM, Martin Losch wrote:

> Hi Patrick, Matt and who-ever is affected,
>
> I'd like the reformulate the obcs cost terms. Currently the some  
> mean velocity, e.g., ubar (which is a monthly mean, as far as I can  
> see) is compared to the prescribed u-component field at the open  
> boundary, say OBNu. While this is perfectly valid, it makes more  
> sense to me (and is more general) to constrain the deviation from  
> the first guess (OBNu), that is, to use the control variable stored  
> in xx_obcsn.*.data directly, ie., sum up all xx_obcsn^2. This way,  
> the contribution from the open boundaries is zero at iteration zero.
>
> Patrick, you have already pointed me to how to do this, but I did  
> not get to it immediately. Could you please remind me, where I can  
> see examples of how to read these files? Was it cost_forcing_gen.F?
>
> The other question is whether this is a formulation useful to others  
> (I think so) and if so how should I include into the repository? As  
> additional files, or CPP-flags in cost_obcs?.F? Or even replace the  
> old formulation?
>
> Martin
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel