[MITgcm-devel] Re: [MITgcm-support] obcs_apply_ptracer

[Martin Losch]

OK, let's talk on the phone next week.

My personal solution to this is, that I always use a bathymetry that  
has no cross-boundary gradients. I don't want to impose this solution  
to anyone else, but if we had been more strict about it, Taka would  
not have run into his problem which he is trying to fix with a hack.

Martin

On 15 Apr 2008, at 14:09, chris hill wrote:
> I agree with Jean-Michel here. It seems risky to fix ob bathy too  
> automatically. If it is fixed automatically you could, for example,  
> take the bc's from the Bering Strait and attach them to the  
> Atlantic side of a limited area Arctic setup, and it would still  
> run happily without making it obvious that something was wrong. Not  
> that we would ever be that dumb, but it could happen!
>
> Can we talk on the phone when Jean-Michel is back from EGU.
>
> Chris
> Jean-Michel Campin wrote:
>> Hi Martin and Dimitris,
>> Sounds little risky/arbitrary and, at the end bringing some  
>> unexpected
>> situation to change the bathy with obcs ; I would have favour a  
>> check & stop,
>> that you can disable (if, e.g., lab-sea is happy with a non
>> strait coasline).
>> I would propose (if you continue in this way) to tart a list of  
>> thing to check (tracer initial condition, divergence of the  
>> initial flow before a
>> nd after the bathy-changes...) and then we can figure out where  
>> this piece of obcs should go.
>> Jean-Michel
>> On Fri, Apr 11, 2008 at 08:07:56PM -0700, Dimitris Menemenlis wrote:
>>> Martin, still thinking about the obcs problem.  Am I correct in  
>>> understanding that all that is needed is to change the bathymetry  
>>> "outside" the obcs?  Another option, if it is easier and would  
>>> save time, would be to provide a matlab tool in utils/matlab that  
>>> cleans up the obcs bathymetry in the correct way, while still  
>>> allowing for arbitrary calvings of larger domains, and then  
>>> pointing the user to this utility in your warning message.  Again  
>>> if you instruct me, I can take care of this.  D.
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel