Anisotropy from full waveform inversion of multicomponent seismic data using a hybrid optimization method

Seismic waveforms contain valuable information about the media, but the waveform inversion is a non-linear problem. We present a waveform inversion method that combines a local optimization method with a genetic algorithm to determine the anisotropic parameters of a horizontally stratified medium. Synthetic seismograms for a horizontally stratified anisotropic medium are calculated using the reflectivity technique. In the initial stage of the inversion, the global space-sampling properties of the genetic algorithm are used to direct the search to the region close to the global solution. This solution is then further improved using a conjugate-gradient method. The numerical experiments performed with noisy synthetic data show that our hybrid optimization method satisfactorily reconstructs the anisotropic parameters at a reasonable computing cost while the range of slowness is adequate. We found that (i) for small-angle data neither single- nor multiple-component data are sufficient to determine the anisotropic parameters uniquely; (ii) for medium-angle data the multiple-component data are sufficient to determine the anisotropic parameters exactly whereas the single-component data are not sufficient; and (iii) for wide-angle data, either single- or multiple-component data are sufficient to determine the anisotropic parameters accurately.