True 3D Data-driven Multiple Removal: Acquisition & Processing Solutions

Abstract

Abstract This paper describes the issues and possible solutions involved in the application of 3D data-driven multiple removal in a marine production environment. The optimization of marine data acquisition for 3D SRME processing solutions is discussed. Secondly, the impact of marine data acquisition on the ability to predict 3D multiples is analyzed. Results from a controlled modeling experiment indicate that a dense grid of saillines and streamers are beneficial to predicting the full 3D characteristics of the multiples. Thirdly, a three-step processing sequence is proposed to predict and remove 3D multiples for each sail line using only the recorded data around the output streamer. The proposed strategy does not rely on any a priori information or model of the subsurface to remove 3D multiples. Finally, results for a field data set from the Norwegian Sea are shown. The application of 3D SRME leads to results that could not be obtained using its 2D equivalent, or any other method. Introduction The removal of free-surface multiples from seismic reflection data remains an essential processing step before the application of prestack migration. Slowly decaying water layer multiples arising from a strong impedance contrast at the sea floor severely degrade the quality of the seismogram. In addition, peg legs are generated from structurally complex 3D sedimentary bodies to create a complex set of reverberations that can easily obscure weak primary reflections. In other areas, diffracted multiples from shallow point diffractors generate a 'cloud' of multiple energy masking underlying primaries. Over the years, many different methods have been developed to attack these problems; these may be subdivided into four major types:Statistical methods, where assumptions on the statistical properties of the Earth's impulse response are used to predict multiples.Differential move-out methods, where primaries and multiples are assumed to have different move-out properties.Model driven methods, where a priori information about the multiple-generating interface is used to predict multiples.Surface-related multiple elimination methods (SRME), where the measured data itself are used to predict multiples. For many years, statistical methods (like predictive deconvolution), differential move-out methods (Radon, f-k methods), and model-driven methods (often based on prestack wave-field extrapolations using a model of the subsurface) have been able to attenuate multiple-energy successfully, and they will continue to do so. However, the reliability of the inherent assumptions and/or the accuracy of the user-provided a priori information, as well as the level of user-interaction required make these methods less suitable for true 3D multiple removal. In recent years, surface-related multiple elimination (SRME) has gained the interest of the geophysical community, primarily for three reasons:SRME provides a theoretically correct solution to the prediction of multiples, and can therefore take into account the full complexity of the earth;SRME does not rely on any information about the subsurface, and only uses the data itself to predict and subtract multiples. As such, results are never biased towards any assumptions about the subsurface or to any a priori information provided by the user;SRME does not rely extensively on user interaction. As it relies much more on compute times, SRME can fully benefit from increased computational efficiency obtained from advances in computer technology.