We are developing a system, called SISMONAUTE, that supports a geophysicist in studying the propagation of waves in a postulated model of the subsoil. The system tracks the wave fronts in a sequence of 2D-snapshots produced by a numerical simulator. Furthermore, it recovers the history of a wave, i.e. the sequence of reflections, transmissions, and diffractions which represents qualitative changes of the wave. This history is essential for comparing simulated and observed seismograms. We use a knowledge-driven approach to interpret the simulated data. Using causal reasoning, the system predicts new wave fronts in each step. These predicted fronts enable a guided detection of wave fronts in the snapshots.