The growth process of spark channels recently becomes accessible
through complementary methods: experiments with nanosecond
resolution, and computations with adaptive grid refinement
resolving three well separated scales of the process.
This scale separation motivates a hierarchy of models valid
on different length scales. The focus of the talk will
lie on a moving boundary approximation for the ionization
fronts that generate the conducting channel.
This moving boundary problem shows strong similarities, e.g.,
with classical viscous fingering.