The number of MPI processes that can be used is limited by the fact that in the 1-d decomposition, a process should never border a process that is more than one MPI rank away from it. This can happen when UMWM is run for a relatively small domain on a relatively large number of MPI processes. However, instead of warning the user and stopping gracefully, the model proceeds and segfaults, or runs with some other erroneous behavior.

This should be tested and initialization.