Introduction

Multiples are often the most significant impediment to the successful construction and interpretation of marine seismic and ocean bottom seismic data. There are several techniques for multiple removal. The selection of the most appropriate method depends on, among other factors, the geology, the acquisition methods and the processing costs. Examples of possible methods include: 1) multiple separation with a high-resolution hyperbolic Radon Transform Kostov and Nichols (1995); Lumley et al. (1995), which depends on a observable difference between the moveout of the primaries and multiples; and 2) surface-related multiple elimination (SRME) Verschuur et al. (1992), which works better on areas with a difficult-to-distinguish difference between the moveout of the primaries and multiples.

Where to perform the multiple elimination, in the data space (before imaging) or in the image space (after imaging), is also a variable to consider when attenuating the multiples. Sava and Guitton (2003) conclude that multiples can be eliminated after migration, in the angle-domain, using Radon Transforms. Guitton (2004) suggests that the image space should be used as much as possible for the multiple-suppression process, since one of the final products of the seismic processing workflow is a migrated image. This paper compares the performance of surface-related multiple elimination in the data and image spaces.

For this purpose, we use a 2D/4C real data set, acquired with an ocean-bottom cable in the Mahogany field, located in the Gulf of Mexico. The dataset provides an interesting test, because conventional multiple-removal methods fail. There is not enough difference between the moveout of primaries and multiples, and the water depth is relatively shallow, thus producing strong multiple reflections at deep targets.

We first present a review of the theory behind multiple suppression by adaptive subtraction. Then, we present the results of multiple suppression both in the data space and in the image space. We conclude that multiple elimination in the image space yields a better final pre-stack image. However, the image space approach has a cost disadvantage, since a full migration of the multiple model is needed.