Applying Full Waveform Inversion to Wide-Angle Seismic Surveys

The frequency domain waveform inversion method of Pratt [1999] was applied to data from crustal refraction and wide-angle reflection seismic surveys across the plate-bounding San Andreas Fault (SAF), California, and the Chesapeake Bay Impact (CBI) structure, Virginia. This presentation discusses technical aspects and lessons learned for this relatively new application. Both long-offset (46 km and 30 km) surveys were acquired using similar parameters: sparse (0.2 1.5 km spaced) explosion seismic sources were recorded on a stationary receiver line at 50 m spacing. The CBI data were acquired in flat terrain with relatively homogenous sedimentary surface strata and show only minor static time shifts and amplitude variations. In contrast, the SAF data contain large time shifts due to rough topography and geology and huge variations in amplitude due to attenuation and surface scattering. Observed amplitudes of the CBI data were easily scaled to match synthetic 2D spreading. For the SAF data, only the phases could be inverted. First-arrival traveltime tomography was used to create starting velocity models for waveform inversion. To account for noise in the SAF data, weighting factors were derived from the local coherence of the phases at each frequency in the dense receiver domain. For CBI, noisy data at longer offsets was simply not used, as they did not contribute greater depth penetration. The resulting velocity models show a marked improvement in resolution as compared to traveltime tomography models, justifying the application of the computationally expensive waveform inversion method.