Direct image inversion of multimodal dispersion spectra and its application to deep seismic reflection data

Nowadays multimodal dispersion spectra can be easily obtained from surface wave data, but sometimes they are actually intractable for the traditional curve-based inversion methods due to the existences of mode-kissing phenomenon and uneven mode-energy distribution. We developed a new spectrum inversion method to circumvent these possible curve-related troubles by directly minimizing the image dissimilarity between the observed and synthetic dispersion spectra. Wherein the synthetic spectrum would be straightforwardly calculated by the Generalized Reflection and Transmission Coefficient method. The new inversion method could rapidly obtain a stable and reliable subsurface velocity structure, even without curve-extracting and mode-identifying in data processing, because it could exploit dispersion energy distribution features to constrain further the velocity structure. As an example of application, we applied this method to deep seismic reflection data in Beijing, and resolved a 2-D S-wave velocity profile above 200 m depth. The strong consistency of structural features between the inversed results of the new and traditional methods shows that the former is effective and practical for realistic data.