Introduction

Inversion of shot gathers or CMP gathers to a model space that distinguishes multiple and primary energy is a well recognized robust method of filtering multiple energy. Thorson and Claerbout (1985) introduced the method of inverse velocity stacking. Inverse velocity stacking is the representation of data as the superposition of constant amplitude hyperbolas such that the original data can be reconstructed from a model. Hampson (1986) used inverse velocity stacking as a tool for filtering multiple energy, though Hampson improved on the speed of inverse velocity stacking by reformulating the inversion as the modeling of data with constant amplitude parabolas instead of constant amplitude hyperbolas.

In order for a transform to be valuable for multiple suppression the representation in model space of the multiple and primary energy must be such that the multiple and primary energy map into distinct regions of model space. Inverse velocity transforms fit this requirement best when the amplitudes and trajectories of events in the data match the amplitudes and trajectories of the transforming curve. Amplitude variations can be incorporated into inversion schemes to improve the compactness of events in model space. This assumes that the amplitudes of interest are actually known in advance, which is not a realistic assumption in many cases. The PRT method performs well in cases in which events in NMO-corrected gathers approximate constant amplitude parabolas. This situation is not satisfied in real data in all situations. Examples of such situations include cases of: 1) strong AVO on multiples and/or primary events 2) situations in which the primary events are strongly non-hyperbolic, and 3) cases in which the moveout of multiples relative to primaries is so large that the multiples can not be approximated by a parabola in an NMO-corrected gather.

I present a method that utilizes inverse beam stacking for the purpose of filtering multiple energy. The method presented in this paper performs well in cases in which the PRT method produces suboptimal results. The results of the beam stack method are much cleaner and more complete than the results of the PRT method except at the near offsets. The speed of the beam stack method is on the same order as the PRT method.