[MITgcm-support] hFac's and obcs

[Thomas Haine]

Dear MITgcm Support, 

I'm confused by the following issue and suspect a bug, or at least the
potential for unwary users to be seriously mislead:

I'm running checkpoint57p_post with pkg/obcs and realistic bathymetry
(Denmark Strait/Irminger Sea). My bathy looks roughly like:

                    N
   _______________________________________
  |                  _______________/     |
  |  LAND       ____/                     |
  |        ____/                          |
  |       /                              /|
  |      /                              /l|
W |    _/                               |a| E
  |   /               OCEAN             |n|
  |  /                                  |d|
  | /                                    \|
  |/                                      |
  |                                       |
  |_______________________________________|

                      S

There are open bcs on all four walls, but notice that they're all
different. The problem is that hFacS(:,j=1,:) is set wrong (i.e. the
southernmost wall should be entirely open). It's set to the hFacS of the
*northern* wall, which is almost closed. The other bounding hFacS/hFacW
fields are OK (but presumably the problem would occur in different
places with different bathy).

The symptom, in my case, is that SSH rises by about 1m/day until the r*
code crashes after a few days. This makes sense because the western
Irminger Sea is draining the domain, but with the wrong hFacS only a
small part of my obc is being applied here (at the eastern end). My
calculations of the volume convergence with the wrong hFac's gives a few
Sv, which gives O(1)m/day increase in SSH, as I see. The obcs I'm
applying with the correct boundaries have very small convergence,
however, as I expect.

The problem seems to be ini_masks_etc.F where the hFac's are set.
There's no logic to deal with the case of obcs in a complex bathy such
as mine. With obcs off wrapping the hFac's is correct (periodic bcs).
With obcs on and simple bathy (same at opposite boundaries) it also
works correctly. But wrapping for cases like mine looks either wrong or
very misleading. 

Can I fiddle with the bathy and obc definitions (data.obcs) to fool the
current code?  Or maybe ini_masks_etc.F needs to be more sophisticated?
Other users should beware in any case. Previous versions of this config.
included a hack that relaxed mean ssh to zero to deal with drifting ssh.
Clearly, we should have worked harder to understand the cause of this
problem, however.

Cheers, Tom.
P.S. Incidentally, with Leith, r*, and vec inv. mom eqns, my solutions,
as far as they go, look exquisite!
-- 
--------------------------------------------
 Thomas W. N. Haine
                                                                                                                                                              
 Associate Professor of Physical Oceanography,
 Department of Earth & Planetary Sciences,
 329 Olin Hall, The Johns Hopkins University,
 Baltimore, MD 21218, USA.
 Tel : 410 516 7048,   Fax : 410 516 7933
 Thomas.Haine at jhu.edu
 http://www.jhu.edu/~eps/faculty/haine
--------------------------------------------