Full waveform inversion of multimode surface wave data: numerical insights

Raul Cova, Kristopher A. Innanen

Full waveform inversion has been demonstrated to be a powerful tool for high resolution velocity model building. However, using surface wave data in FWI presents many challenges. In particular, the dispersive nature of surface waves results in the amplification of cycle skipping problems. Here, we propose decomposing surface waves into their fundamental and higher order modes and inverting them in a sequential approach to mitigate this problem. Even though the fundamental mode amplitudes are typically larger than the higher order modes, the latter ones can travel in the deeper parts of the near-surface at higher frequencies. Therefore, we use the fundamental mode to produce an initial approximation to the near-surface S-wave velocities and then perform another step of FWI using the higher order modes to produce a more detailed velocity profile, particularly at larger depths. We also argue that although each individual higher mode is less energetic than the fundamental mode, as a group, the combination of all higher modes surpasses the energy and reach of the fundamental mode. We performed the modal separation by designing a mask in the F-K domain that roughly contains the energy of each mode. Results obtained from synthetic data demonstrate the potential of this approach to avoid cycle skipping and to improve the resolution of inverted S-wave velocity models.