Extrapolated supervirtual refraction interferometry

Abstract

SUMMARY Accurate picking of head-wave arrival times is an important component of first-arrival traveltime tomography. Far-offset traces in particular have low signal-to-noise ratio (SNR), but picking on these traces is necessary in order to obtain velocity information at depth. Furthermore, there is often an insufficient number of far-offset traces for obtaining reliable models at depth. We present here an extrapolation method for increasing the number of first arrivals beyond the maximum recorded offset, thereby extending the supervirtual refraction interferometry (SVI) method. We refer to the method as extrapolated SVI (ESVI). It is a novel attempt to extrapolate first arrivals using a fully data-driven method. We first test the methodology on synthetic data sets, and we then apply ESVI to two published real data sets over the Pärvie fault system in northern Sweden. These data sets were acquired along the same profile at different times with different acquisition parameters and noise levels. The results show that ESVI enhances the SNR of head waves when the noise level is high. That is the same as the conventional SVI. ESVI also increases the number of pickable first arrivals by extrapolating head waves past the original maximum offset of each shot. We also show that the significant increase in first-arrival traveltime picks is beneficial for improving resolution and penetration depth in the tomographic imaging and, consequently, better revealing the subsurface velocity distribution. The tomographic images show higher velocities in the hanging walls of the main Pärvie fault and another subsidiary fault, as interpreted relative to migrated images from previous seismic reflection processing.