[MITgcm-support] icebergs

[Jean-Michel Campin]

Hello Martin,

> Hi Jean-Michel,
> 
> we have an idea here let an ice-berg float around in a high-resolution
> (dx<100m) idealized domain, melt and interact with the ocean. One major
> concern is that such an ice-berg has a draught of up to 500m. In a
> sigma-coordinate model this would be simple to implement, so that the
> to layer is at the surface away from the ice-berg and abuts the
> ice-berg underneath the ice-berg. In a z-coordinate model, there it's
> not so easy. How would you see the chances that we could use the z-star
> coordinate in the MITgcm successfully? I have not looked into that part
> of the code at all so I don't know whether there are any limitations on
> the size of the possible depression of the free surface. What's your
> take on this?
> 
> Martin

With a flat bottom, sigma is identical to z*. 
With bathymetry, the coordinate are much less squeeze with z* 
than with sigma, so z* might end-up being even better than sigma.
The real strong limitation is that the ocean must always be 
deeper than the iceberg draught.

Now, practically speaking, we don't have in MITgcm any of the various 
tricks that are generally added in a sigma model to deal with the 
pressure gradient error. This is good to keep in mind, when the 
depression is much larger that the thickness of the first level.

I have been using the thermodynamic sea-ice package with
real fresh water and z*, with eta being the height anomaly
of the sea / sea-ice interface. The mass of sea-ice is added
(like the atmospheric loading) as a sea-ice loading.
I need (very soon) to add this forcing file that is presently
missing.
However, sea-ice is never as thick as an iceberg, and where it
is thicker, it's generally surrounded by grid points that are also
ice-covered, so that the gradient of eta (sigma type pressure error
gradient) is not a problem.
I don't have any experience with z*+realFreshWaterflux+ DYNAMIC sea-ice
so I go to your 2nd e-mail.

> Hi,
> 
> we are trying something interesting: Have an iceberg float in a domain.
> As a first step we want to have a stationary iceberg that melts at a
> constant rate: specify fresh water flux and pressure load. Later there
> may be a "real" iceberg model that drifts and melts according to a
> realistic equation of state.
> In a sigma model this would be simple, with the toplayer always being
> beneath the iceberg. What about the z*-coordinate? From the
> Adcroft+Campin2004 paper I don't see any real limitations on the size
> of the depression of the free surface. Am I right? Could I place an
> iceberg with a draught of, say, 200m (with delZ<<200m) into a basin by
> specifying the pressure load p=2000000Pa? What parameter setting would
> I need?
> nonlinFreeSurf=4,
> exactConserv=.true.,
> select_rstar=2,
> ?
> Finding the initial conditions may be a problem.
> I have not quite understood how the free surface Eta should be
> interpreted in such an experiment.
> 
> Any opinions?
> Martin

I recommend to use also
useRealFreshWaterFlux=.TRUE.

so that the ice-loading time-variations
are consistent with the BC at the surface:
when the iceberg is melting,
ice-loading decrease <-> fresh water is added

And since I didn't put yet the ice-loading file (but the array
exist:n FFIELDS.h: 

C  sIceLoad :: sea-ice loading, expressed in Mass of ice+snow / area unit
C                Units are           kg/m^2
C              Note: only used with Sea-Ice & RealFreshWater formulation

you can use the atmospheric loading instead, since then have a
similar effect (units are different).

Hope this help,

See you,

Jean-Michel