[CIG-LONG] [gale] model equilibration (thermal and isostatic) + boxShape initial geometry

[Walter Landry]

<Guillaume.Duclaux at csiro.au> wrote:
> 
> Hello,
> 
> I have a couple of questions regarding (1) model equilibration in Gale,
> and (2) model box initial geometry.
> 
> (1) I'm working on a model involving oceanic and continental
> lithospheres. 
> 
> I first ran non-dynamic models in order to let them equilibrate by
> isostasy and let geotherms equilibrate for the different materials too. 
> Then, I ran dynamic models in compression, starting with a non
> equilibrated initial geometry identical to the one of the previous
> models. 
> 
> I'd like now to start a model with a non equilibrated initial geometry,
> lets it equilibrate for some 100 Myr, and turn the velocity ON. This
> could be done using a velocity function pretty similar to the
> StepFunction (a ramp function like: V=0 for t<t{-},
> V=V((t-t{-})/(t{+}-t{-})) for t{-}<t<t{+}, and V=V for t>t{+}). However,
> the StepFunction depends of (x,y,z) and not time (or number of steps).
> 
> Is there a smart way to do such thing with Gale?

The best way to do this is to modify the source code to put in that
step function.  Specifically, in

  src/StgFEM/plugins/StandardConditionFunctions/StandardConditionFunctions.c

add the lines (untested!)

        if(dim==3)
          {
            dim=0;
            coord=&(context->currentTime);
          }

at line 861.  Then, you can specify "StepFunctionDim" as 3, and it
will use time instead of space.

> Alternatively, is it possible to use some output files as input
> parameters for a new model?
> For example, I guess it might be possible to define Initial Conditions
> from a file rather than a double value. Thus, I could use equilibrated
> values for particles and nodes data (i.e. isostatic adjustments and
> equilibrated geotherms) and directly impose a velocity to a model
> boundary.

Checkpointing is something that does not work right now.

> (2) "By default Gale starts the simulation in a rectangular box" - from
> Gale's user manual. Is it possible to setup an alternative initial
> geometry such as a polygon shape rather than a box?
> 
> i.e.:
> <struct name="bottomBoundaryShape">
>       <param name="Type">PolygonShape</param>     instead of Box
> 
> <struct name="boxShape">
>       <param name="Type">PolygonShape</param>	 instead of Box

At present, you can only modify the surface using a SurfaceAdaptor.
The file input/examples/tibet.xml has an example.  In the near future,
I plan to make it possible to have an arbitrarily deformed bottom
boundary.  Completely arbitrary shapes are not coming soon.  It
becomes difficult to put down a rectangular mesh.

However, you can fix the degrees of freedom of an arbitrary shape in
the box, leaving only the polygon that you want.  That is detailed in
the cookbook and the input file

  input/cookbook/viscous_fixed.xml

Another example, where a cylinder is cut out of a box, is in

  input/benchmarks/falling_sphere/sphere_in_cylinder.xml

Cheers,
Walter Landry
walter at geodynamics.org