[CIG-LONG] Extension - Compressible Layer in Gale

[Walter Landry]

"atyl0789 at uni.sydney.edu.au" <atyl0789 at uni.sydney.edu.au> wrote:
> As a first step I have been attempting to create a simple
> extensional model consisting of 8km of brittle crust overlying 2km
> of weak viscous material, extending at a constant rate of
> 1cm/year. I have been so far using the Gale_1_4_1 binary and have
> encountered a number of problems in getting this model to run. Using
> either iterative or direct solver, Gale generally crashes after
> around 50-60 timesteps coinciding with the development of high
> strain and the first glimpses of faulting. I initially increased
> resolution to 100m (x and y) in order to determine whether this was
> the issue, however the same problem was encountered.

I have found that at the surface, I often get very low viscosity
regions that looked like landslides.  I have at least a partial fix
for it in the upcoming release.

But as a general rule, increasing resolution only makes stability
problems worse.

> I read that John Naliboff encountered a similar sounding problem in
> the cig-long archive (March 2010), and following Guillaume and
> Walter's advice I attempted to implement a compressible air layer
> above the crust and added a basal velocity. The compressible 'air'
> layer was 2km thick, had a density of zero, a viscosity of 1e19, and
> compressibility factor of 10. The model seems to run smoothly,
> however I observe some extreme deformation of the air layer as
> extension proceeds, and I am unsure whether this is expected and/or
> whether it affects the deformation of the layers beneath it. I am
> also unsure whether this would begin to cause problems if I were to
> run the model longer.

The only real problem you may have is from having such a deformed
grid.  For runs with an air layer, I would recommend making the top of
the simulation fixed.  Air will flow in and out of the simulation, but
the mesh will stay the same.  And, as Louis Moresi said, it does not
really matter what the air does, since it really only acts as a
stabilization term.

> Finally, I am unsure how to use the hydrostatic function with the
> addition of this compressible layer. Is it possible to make it so
> that it affects only the lithosphere? I have attached an image of
> what my model output looks like after 152 timesteps (arbitrary, its
> just when i stopped the model).

In HydrostaticTerm, there is a variable 'height'.  Set that to the top
of your material layer, not the top of the simulation.

For an example of how to do this, see the input file I posted at

  http://www.geodynamics.org/pipermail/cig-long/2010-May/000376.html

> I have corresponded with John and he has told me that he was unable
> to get the compressible layer working with the hydrostatic term, and
> ultimately ended up using an a weak incompressible layer with a
> density of 1000, simulating a body of water above the crust. Is this
> a more effective approach for me to follow?

That really should not be necessary.  An air layer should work for you
right now.

Cheers,
Walter Landry
walter at geodynamics.org