Abstract
A seismic depth-imaging project starts from an initial depth model of interval velocities. From time processing of reflection seismic data, a set of stacking parameters or kinematic data attributes usually is available for an initial model building at little effort. Two methods for initial model building from time-processing attributes are compared in this case study, using 3D seismic land data from the coast of the Gulf of Mexico. Conventional normal moveout (NMO)/dip moveout (DMO) time processing performs one-parametric stacking using stacking velocity as the parameter. The stacking velocity field can be converted into a depth model by the well-known vertical Dix inversion, which is very fast and robust but degrades with increasing dip. Common-reflection surface (CRS) time processing, on the contrary, isbased on the multiparametric CRS stacking approach, providing several volumes of CRS-stacking attributes that include the wavefield dip, or horizontal slowness. Inversion of CRS attributes by CRS tomography incorporates this dip information in depth model building. In this case study, CRS or normal-incidence point (NIP) wave tomography is presented as a model-building link between high-resolution CRS time processing and subsequent depth processing. The CRS tomography model shows a better adaptation to the dipping subsurface structures than the Dix model and a good fit to well data. The smooth tomography model is well suited for further use in poststack and prestack depth migrations. It provides a good starting point for iterative model enhancement and salt-body definition in prestack depth migration.