Internal multiple prediction and removal is a critical component of seismic data processing prior to imaging, inversion, and quantitative interpretation. Inverse scattering series methods predict multiples without identification of generators, and without requiring a velocity model. Land environments present several challenges to the inverse scattering series prediction process. This is particularly true for algorithm versions that explicitly account for elastic conversions and incorporate multicomponent data. The theory for elastic reference medium inverse scattering series internal multiple prediction was introduced several decades ago, but no numerical analysis or practical discussion of how to prepare data for it currently exists. We have focused our efforts on addressing this gap. We extend the theory from 2D to 3D, analyze the properties of the input data required by the existing algorithm, and, motivated by earlier research results, reformulate the algorithm in the plane-wave domain. The success of the prediction process relies on the ordering of events in either pseudodepth or vertical traveltime being the same as the ordering of reflecting interfaces in true depth. In elastic-multicomponent cases, it is difficult to ensure that this holds true because the events to be combined may have undergone multiple conversions as they were created. Several variants of the elastic-multicomponent prediction algorithm are introduced and examined for their tendency to violate ordering requirements (and create artifacts). A plane-wave domain prediction, based on elastic data that have been prepared (1) using variable, “best-fit” velocities as reference velocities, and (2) with an analytically determined vertical traveltime stretching formula, is identified as being optimal in the sense of generating artifact-free predictions with relatively small values of the search parameter [Formula: see text], while remaining fully data driven. These analyses are confirmed with simulated data from a layered model; these are the first numerical examples of elastic-multicomponent inverse scattering series internal multiple prediction.
Data-driven internal multiple elimination and its consequences for imaging: A comparison of strategies