Use of autoconvolution to suppress first‐order, long‐period multiples

Abstract

A common problem in interpreting marine seismic data is the interference of water‐bottom multiples with primary reflections containing the structural or stratigraphic information. In deep ‐water areas, where considerable primary energy arrives before the first simple water‐bottom multiple, weak and deep crustal reflections are often obscured by the first‐order water‐bottom multiples. In order to obtain a more interpretable section, a technique involving a two‐step process was developed to suppress the first‐order water‐bottom multiples. First, the relation between the zero‐order, water‐bottom primary and its first‐order, simple water‐bottom multiple is used to derive statistically an inverse of the seismic wavelet in order to remove its effect, i.e., to wavelet‐shape the data. This wavelet processing provides a band‐limited estimate of the subsurface impulse response. The second step consists of using the autoconvolution of the wavelet‐shaped primary energy to estimate deterministically and subtract the actual first‐order, water‐bottom multiples, The method was applied to field data from the deep Gulf of Mexico. Different incidence angles for the input primaries and multiples, as well as dipping reflecting interfaces, introduce uncompensated traveltime errors. These errors reduce the ability to suppress multiples, thus restricting the validity of the method to low frequencies where common‐depth‐point stacking is less effective. On the other hand, curved interfaces may also cause amplitude prediction problems. In spite of this, the first‐order, water‐bottom multiple energy is significantly reduced (by up to 18 dB) on dip‐filtered, single‐channel data.