[CIG-SEISMO] Topography simulation : numerical instability

[Alan Schiemenz]

Dear SPECFEM3D users,
I am having problems running simulations with a mountain topography. 
The simulation is unstable despite taking time steps much smaller than
the CFL limit would require.  Furthermore, the numerical instabilities
are originating at the corners of the model, whereas the source is
located in the center.   The link below contains velocity snapshots of
two different topographies at the point where instability is obvious. 
On the left is a rough topography whereas on the right is a smooth
topography :
https://www.geophysik.uni-muenchen.de/~schiemenz/topo_examples.jpg

In case we are doing something incorrect CUBIT, I also have the
following .cub file for the smooth topography case.
https://www.geophysik.uni-muenchen.de/~schiemenz/gauss_500_166_red3.cub

Thank you for your help,
Alan Schiemenz