Parametrization study of the land multiparameter VTI elastic waveform inversion. Issue 3 (13th March 2018)

• Record Type: Journal Article
• Authors: He, W; Plessix, R -É; Singh, S
• Is Part Of: Geophysical journal international. Volume 213:Issue 3(2018:Jun.)
• Page Start: 1660
• Page End: 1672
• Publication Date: 2018-03-13
• Subjects: Numerical approximations and analysis -- Waveform inversion -- Computational seismology -- Seismic anisotropy -- Inverse theory -- Wave propagation; Geophysics -- Periodicals; 550
• DOI: 10.1093/gji/ggy099
• Languages: English
• ISSNs: 0956-540X

Abstract

SUMMARY: Multiparameter inversion of seismic data remains challenging due to the trade-off between the different elastic parameters and the non-uniqueness of the solution. The sensitivity of the seismic data to a given subsurface elastic parameter depends on the source and receiver ray/wave path orientations at the subsurface point. In a high-frequency approximation, this is commonly analysed through the study of the radiation patterns that indicate the sensitivity of each parameter versus the incoming (from the source) and outgoing (to the receiver) angles. In practice, this means that the inversion result becomes sensitive to the choice of parametrization, notably because the null-space of the inversion depends on this choice. We can use a least-overlapping parametrization that minimizes the overlaps between the radiation patterns, in this case each parameter is only sensitive in a restricted angle domain, or an overlapping parametrization that contains a parameter sensitive to all angles, in this case overlaps between the radiation parameters occur. Considering a multiparameter inversion in an elastic vertically transverse isotropic medium and a complex land geological setting, we show that the inversion with the least-overlapping parametrization gives less satisfactory results than with the overlapping parametrization. The difficulties come from the complex wave paths that make difficult to predict the areas of sensitivity of each parameter. This shows that the parametrization choice should not only be based on the radiation pattern analysis but also on the angular coverage at each subsurface point that depends on geology and the acquisition layout.