Paper Summary

Full waveform inversion (FWI) iteratively minimizes the misfit between the recorded and modelled seismic data to derive an accurate velocity model. The method can produce high-resolution velocity models, provided that the target areas are well-sampled by the seismic waves (e.g. shallow targets with refracted waves and deep targets with reflected waves).Here, we present an application of FWI to a wide-azimuth dataset from the Gulf of Mexico. Due to the large water depth, refracted and diving waves are not well represented in the recorded data. Thus, the inversion must rely on the reflected waves. This in turn places more stringent requirements on the fidelity of the computed amplitudes, so variable density was included in the inversion. Through reflection FWI, we were able to obtain a velocity model with much higher resolution than is shown in the preliminary, ray-tomography model. We observed that the FWI model improve the flatness of common-image gathers and the quality of the stacked, migrated images. Furthermore, the small-scale velocity features revealed by FWI correlated well with the migrated image.