[MITgcm-support] SSH drift and obcs again

[Thomas Haine]

Hi Martin, That's a great idea but, confusingly, it doesn't fix the
issue. If I correct over a fraction of the southern wall excluding the
corners it works to machine precision with land corner points, but fails
with ocean corner points. This has me baffled.

Alistair's good suggestion about computing transport excluding corners
also doesn't work even with land at the corners.

Any final ideas?  I'm ready to stick land in at the corners and run it
(and/or explore other domains without obcs).

Thanks for your help, Tom.


On Thu, 2005-10-27 at 09:23 +0200, Martin Losch wrote:
> Hi Tom,
> my guess is, that you need to satisfy BOTH of the following:
> 1. you need you corner points to have consistent u and v velocities, so 
> that they do not introduce vertical velocitites due to horizontal 
> divergences.
> 2. At the same time you need your total flow to balance
> Have you tried this:
> After specifying consistent velocities at the corners, compute the 
> total balance without changing the corner velocities, that is, you take 
> these points into account, when you compute the imbalance, but when you 
> compute average boundary velocity corrections, you don't take into 
> account the area occupied by the corner points and you don't correct 
> the corner points.
> 
> I wonder if this can be included in the code in general way ...
> 
> Martin