[MITgcm-support] SSH drift and obcs again

[Alistair Adcroft]

Thomas Haine wrote:
> OK, I've made a little more progress. The obc corners are the problem:
> 
> i) I can get zero cg2d_b and eta drift if I have obcs on opposite
> boundaries (N&S or E&W). This requires me to put a solid wall on the
> boundaries where I don't specify obcs. (Incidentally, I have zero eta
> drift with N&S obcs and periodic E/W bcs but *not* vice versa. This I
> don't understand but presumably it depends on my specific
> configuration). So, obcs with no obc corners works OK - meaning I can
> correct for net total convergence across the obcs accurately.
> 
> ii) I can get zero cg2d_b and eta drift with an obc corner (I've tried 1
> and 3 corners) only if I put zero depth at each corner. I.e. adding a
> single-cell island to block the obc corner works.
> 
> I've checked that: my obc forcing files have consistent speeds at the
> corners (as Martin's message of 24 October) and that my bathy file has
> no gradient at the corners.
> 
> So, how should I calculate the total convergence of fluid across my obcs
> when the obcs have a corner?  If I can calculate this exactly then I can
> correct for it and guarantee zero eta drift.

It's just occured to me that the calculation we modified in obcs_balance 
is including the corners while the solver probably isn't; that is, 
obcs_balance calculates the zonal flux across all of the eastern OB but 
should only consider the part of the eatern OB between the northerna and 
southern OBs.

I think your (ii) above ought to solve this so why not stick with it?

A.
-- 
Dr Alistair Adcroft            http://www.mit.edu/~adcroft
MIT Climate Modeling Initiative        tel: (617) 253-5938
EAPS 54-1624,  77 Massachusetts Ave,  Cambridge,  MA,  USA