Effects of surface scattering in full-waveform inversion

In full-waveform inversion of seismic body waves, often the free surface is ignored on grounds of computational efficiency. Asynthetic study was performed to investigate the effects of this simplification. In terms of size and frequency, the test model and data conform to a real long-offset surv...

Full description

Bibliographic Details
Main Authors: Bleibinhaus, Florian, Rondenay, Stephane
Other Authors: Massachusetts Institute of Technology. Earth Resources Laboratory
Published: Massachusetts Institute of Technology. Earth Resources Laboratory 2012
Subjects:
Online Access:http://hdl.handle.net/1721.1/68593
Description
Summary:In full-waveform inversion of seismic body waves, often the free surface is ignored on grounds of computational efficiency. Asynthetic study was performed to investigate the effects of this simplification. In terms of size and frequency, the test model and data conform to a real long-offset survey of the upper crust across the San Andreas fault. Random fractal variations are superimposed on a background model with strong lateral and vertical velocity variations ranging from 1200 to 6800 m/s. Synthetic data were computed and inverted for this model and different topographies. A fully viscoelastic time-domain code was used to synthesize the seismograms, and a viscoacoustic frequency-domain code was utilized to invert them. The inversion was focused on early arrivals, which are dominated by P-waves but also contain strong P-Rayleigh wave conversions from the near-field of the receiver. Resulting waveform models show artifacts and a loss of resolution from neglecting the free surface in the inversion, but the inversions are stable, and they still improve the resolution of kinematic models. The extent of deterioration depends more on the subsurface than on the surface structure. Inversion results were improved at no additional expense by introducing a weak contrast along a staircase function above shots and receivers.