[MITgcm-support] CheapAML diffusion at only one wall

[Chris Horvat]

Hey all,

I'm running ocean channel experiments using the CheapAML package, with
walls on the N/S boundary of the domain. I've been running into a
convective instability at the southern channel wall and think I've
discovered an issue I can't quite sort out, hoping for some help.

There is a wall with HFacW=HFacW=HfacS=0 and Depth=0 at points with j=1 and
j=300. CheapAML computes fluxes there differently than over the channel
where Depth=75.

When I allow for diffusion in the CheapAML code (cheapAML_kdiff~=0), there
is a diffusive flux computed over the j=1th meridional level into the j=2th
meridional level. This isn't ideal, but if this was the only problem I
would survive. The issue is that there is not a flux from the j=300th point
into the j=299th.

The attached image shows the air temperature after 12 hours (diffusion is
high in this case ~ k = 200 m^2/s with dY = 100 m). Clearly there's
something different about the southern boundary and the northern. The
diffusive flux is just upwind: F_Y ~ T(j+1) - T(j) and is handled in
pkg/cheapaml/gad_diff_2d_y .

I previously thought CheapAML knew enough about boundaries to prevent this
but am unsure now... I am running with CheapAMLXPeriodic=.TRUE.,
CheapAMLYPeriodic=.FALSE.,

Is there something about the way I'm defining the initial topography?
Hoping someone can help me sort this out.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20180307/81cabfe5/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: examp.pdf
Type: application/pdf
Size: 6403 bytes
Desc: not available
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20180307/81cabfe5/attachment.pdf>