Deep Water Case Histories - SRME / HR Radon / Diffracted Multiple Attenuation

Abstract

Abstract Through several case histories, we present what has became a standard demultiple sequence for deep water data. These data are generally contaminated with strong regular multiples, which are aliased on the far offset traces, and often by diffracted multiples. Given the variety of problems, no single tool can attenuate all the multiples, but rather a sequence of tools: Surface Related Multiple Elimination (SRME), High Resolution Radon (HR Radon), and Diffracted Multiple Attenuation (DIMAT). This sequence can be easily extended to data with medium water depth. Introduction In the recent years, several new techniques have come up in the subject of multiple attenuation. The most famous one is the SRME method, which was a real technological breakthrough. Particularly, its fully data-driven principle makes it very attractive. HR Radon brought fundamental improvements to the classical parabolic Radon filtering, with the ability to go beyond aliasing and with improved discrimination. SRME and HR Radon appear to be remarkably complementary techniques, the former working better on the near offset range, the latter working better on the far offset range: thatâ??s why they are often associated into a "hybrid method". However, still many multiples are not properly attenuated by SRME or HR Radon, including the diffracted multiples. For them, we propose a statistical approach, based on hypothesis of amplitude and frequency content. This approach (DIMAT) is easy to apply even to huge 3D surveys and exhibits an excellent efficiency-cost ratio. These three methods, SRME, HR Radon, and DIMAT, are particularly suited to deep water data, and their association can lead to a near-optimum multiple attenuation. For data with medium water depth, a simple variation of SRME, the Partial SRME, can help extending the sequence by improving the modeling of the amplitudes on the high order multiples. For medium water depths, however, the data may not fulfill the DIMAT assumptions as nicely as in deep water. SRME - Surface Related Multiple Elimination The SRME method predicts and attenuates the whole surface multiple wavefield (the multiples that experience a reflection on the free surface), without any knowledge of the subsurface (velocities, horizons,...): it is fully data-driven. The usual implementation consists in an iterative 2-step [modeling / adaptive subtraction] loop (Verschuur and Berkhout, 1997). Although there is no theoretical obstacle to a 3D SRME, it requires dense data (recorded or simulated) over the whole free surface. Therefore, virtually all the current operational implementations are 2D, and hence cannot handle the multiples that follow out-of-the-plane travel paths (3D effects). SRME can work very nicely on the near-offset range; it does not rely on any assumption, such as periodicity or curvature discrimination. It is, however, usually less efficient on the far-offset range, not because of any failure in the theory (the theory can predict and subtract the multiples equally well on the near and on the far offset traces), but rather because of practical reasons:Spatial aliasing: in deep water, the tails of the hyperbolas are often aliased on the far offset range, especially in modern acquisition geometries (e.g. flip-flop shooting) which lead to undersampled CMP gathers.