Viscoelastic Modeling and Inversion of a Marine Data Set

Deterministic waveform inversion in the p ? domain permits estimation of the long-wavelength p-wave background velocity, the short-wavelength elastic parameter uctuations, and the energy source for a marine data set from the Gulf of Mexico. The forward map is a convolutional model for p-wave primaries-only seismograms of viscoelastic, layered media. Two types of inversion (output least squares and diierential semblance optimization) allow eecient estimation of various parameter combinations. Background velocities computed by diierential semblance optimization were as consistent kinematically as those produced by interactive migration velocity analysis and gave somewhat more accurate depths. Source parameters obtained by output least squares inversion permitted considerably better t to data than did those produced by an air gun modeling package. Output least squares inversion gave very short-scale detail of major reeectors closely resembling the detail derived from well logs. Despite evidence of considerable lateral heterogeneity in the shallow subsurface, estimates of deeper reeectors were remarkably consistent from midpoint to midpoint.