The Prediction And Attenuation Of Surface-related Multiple

In marine seismic exploration, since both the sea surface and the seabed are usually strong acoustic impedance interfaces, most multiples in corresponding seismic records belong to surface-related multiple. These strong multiples may seriously impair the authenticity and reliability of the seismic imaging, and even mislead subsequent seismic-geologic interpretation.Surface-related multiple elimination (SRME) method based on feedback formulation has made remarkable progress in recent years. This method does not depend on velocity information and can well attenuate the multiples in near offset. Therefore it has become one of the preferred multiple suppression methods. In practical applications, however, there still exists a few of problems of the SRME approaches, which may mainly be summarized as follows:①Since the method usually adopts the iterative processing of multiple removal, it may increase the probability of damaging primary events and decrease the operation efficiency;②The missed seismic traces, especially in the near-offset, can damage the result of multiple suppression severely;③The existing multiple matching attenuation approaches can hardly maintain the characteristics of primary waves while suppress multiples effectively, especially in the far-offset.Traditional feedback theory has just studied the surface-related multiple prediction using the seismic records of primary events. Since a seismic record which consists of primary waves only can hardly be obtained in reality, the source or raw seismic records are used as initial iterative wavefield instead. For the real case, however, the traditional feedback theory has not given the result of corresponding forward model. Based upon the assumption of the source wavefield consists of not only the primaries but also all the orders of surface multiples, the thesis derives and proves that a predictive wave field which consists of the information of all orders of "surface multiple" can be obtained by only one iterative operation. The energy of "surface multiple" of order n+1(n≥1) in the predictive wave field has increased to n times comparing with that in the source seismic records. The conclusion provides an important foundation for a suitable multiple matching and attenuation approach using only one iterative operation.In the SRME method, both seismic trace extrapolation (or interpolation) and multiple matching filtering are important steps for surface multiple suppression. Thus to develop a multiple event tracing technique has important practical significance for the accurate seismic trace extrapolation (or interpolation) and multiple matching filtering. Based upon the relationship between event in seismic data and the spectrum energy of velocity records, an auto-tracing technique for main multiple events on common-midpoint (CMP) gathers is developed. This technique provides an important fundamental tool for accurate seismic trace extrapolation (or interpolation) and multiple matching filtering.The result of surface multiple prediction and attenuation will be severely affected by the trace missing effect, especially in the near-offset. The existing seismic trace extrapolation or interpolation methods may still suffer the problems of accuracy and efficiency in the wavefield rebuilding. In this thesis, a seismic traces extrapolation or interpolation method through velocity weighted stacking is presented. The method not only takes the hyperbolic events into account but also the relationship of amplitude versus offset (AVO). Therefore the events extrapolated or interpolated appear with not only the true geometry but also the amplitude rule in the rebuilt seismic traces. Compared to the least-square parabolic Radon transform which is widely used at the present time, the new method may have following advantages:①It may be more closer to the real event distribution because of the hyperbolic (not parabolic) event assumption;②Since the AVO effect of seismic wavefield has been taken into account, the extrapolated or interpolated signal may, to some extent, have a relative true amplitude relationship close to a real seismic record;③The normal moveout (NMO) stretching effect may be avoided since no NMO correction processing is needed;④It can improve the efficiency significantly. The computation time of the new method is much smaller than that of the least-square parabolic Radon transform.General speaking, there always exist some differences between the modeled multiples and those in the source seismic records. This may cause that the conventional multiple matching attenuation methods can hardly maintain the primary waves while suppress the multiples effectively. In addition, conventional approaches may fail to suppress the multiples in far-offset. In order to solve these problems, a combined matching attenuation method based on event tracing is presented in this thesis. The basic idea of the new approach is that tracing the multiple events in CMP gathers on the modeled records, and then applying an expanded Wiener type filtering in the near-offset and the apparent velocity filtering in the far-offset respectively to suppress multiples. In practice, comparing to traditional single matching filtering methods, it can improve the multiple suppression result significantly in both near and far offset by using only one iterative operation.The processing experiments of Pluto 1.5 synthetic dataset and real seismic data demonstrate the effectiveness of the presented methods.