Multiple attenuation by explicit wave extrapolation to an interpreted horizon

Abstract

One of the oldest approaches to multiple attenuation is to predict the multiples using their periodic nature as their identifying trait and then simply subtract them from the original data. It is done routinely as a trace‐by‐trace, time‐axis deconvolution on nearly every exploration seismic data set to attenuate the short‐period multiples. This simple prediction method loses its effectiveness, however, when the multiple period gets to be significantly longer than the source wavelet. The problem is that the longer‐period multiples do not occur in an evenly spaced train. The delay between cycles depends on the wave propagation angle and the dip of the multiple‐generating reflectors. Thus the period varies both with source‐receiver offset and the subsurface geology.