Removal of scattered surface waves using multicomponent seismic data

摘要

We have discussed previously a method for removing scattered surface waves from single-component seismic data based on the estimation of a near-surface scattering model, and we have applied it to both simulated and field data. In the present paper, we extend this method to the multicomponent case. From tests on simulated data, we have found that the multi-component method is more selective in differentiating between body waves and scattered surface waves since it exploits the fact that surface waves and body waves have different states of polarization. The crossline component turns out to be particularly effective for removing the crossline-scattered noise from the vertical component. We have applied the method to a multicomponent field data set. For this case, we have estimated the source waveform and Green's function directly from the data itself. The method results in a good image of the scatterer but can probably be improved by using better estimates of the horizontal and vertical waveforms. There are many ways in which the method, both for single-component and multicomponent data, can be extended. One can think of extending it to laterally varying background models. The method can, in principle, be used to eliminate other near-surface scattering effects like scattered head waves and guided waves, provided that a high-frequency asymptotic expression for these waves is used, or that their propagation characteristics can be estimated from the data. Furthermore, it can be useful to extend the method such that it takes conversions into account, especially head waves converted into Rayleigh waves and vice versa. The method can possibly be used also for obtaining information about shallow heterogeneities for construction purposes and might also be applied to the nondestructive evaluation of materials.