We apply three moveout-based methods for multiple suppression to simple synthetic common midpoint (CMP) gathers and compare their performance in terms of level of multiple rejection and, more importantly, in terms of (1) suitability of processed output for use in analysis of amplitude variations with offset (AVO), (2) primary-to-multiple amplitude (p/m) ratio of the data after CMP stacking, and (3) preservation of primary wavelet shape and amplitude on the stack. The three approaches are filtering in the frequency–wavenumber f–k domain, the Hampson method of filtering in the parabolic Radon transform domain, and a hybrid method that improves upon Hampson's approach by using a variation of Harlan's statistical pattern recognition approach to separate primaries from multiples in the parabolic Radon transform domain.
The f–k approach is unsuitable prior to AVO analysis; moreover, it is little better in suppressing the multiples than the CMP stack itself. Hampson's method performs considerably better and is suitable for both purposes, although it can yield distorted AVO response where multiples are relatively strong on input. At about 50% additional computation effort over that of Hampson's method, the hybrid approach has superior treatment of amplitude behavior with offset. For many situations, the hybrid method yields up to twice the primary-to-multiples amplitude ratio on the CMP stack.