GNGTS 2013 - Atti del 32° Convegno Nazionale

IMAGING OF MULTIPLE REFLECTIONS C. Fortini, V. Lipari Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano, Italy Introduction. Conventional seismic imaging algorithms are based on single-scattering hypothesis. The reverberations that generates during the wavefield propagation in the subsurface of a seismic experiment ( multiple reflections ) are usually considered as unwanted noise. In marine seismic acquisition, both primary reflections and sea surface-related multiples are recorded. The most common approach used to handle with the presence of multiple reflections in the acquired seismic data is to try to eliminate them. First, a model of the multiples is built by means of either model-based or data-driven methods and then the noise model is adaptively subtracted from the original data (Verschuur, 2006; Verschuur et al. , 1992; Lipari et al. , 2012). Althoughmultiple reflections affect the recorded primary reflections, they also had interacted with the subsurface discontinuities and they thus carry information about the Earth’s structure. Recently, researches on multiples (especially surface-related multiples) had shifted their focus on the exploitation on what has been often considered only as noise. Berkhout and Verschuur (2006) showed that multiple reflections can be used to re-create the associated primaries and then exploited for imaging purposes. They also proved that multiples carry useful information and can thus enhance the overall understanding of the subsurface structure. Guitton (2002) and Muijs (2007) showed that WE (Wavefield Extrapolation) migration algorithms can be effectively used for the imaging of the multiple reflections. Alkhalifah (2011) showed that it is possible to use the whole recorded seismic field to correctly image the subsurface discontinuities, avoiding the costly steps of multiples prediction and subtraction. In Shaoping et al. (2011) the new dual-sensor acquisition devices are exploited to separate up-going and down-going fields. The separated wavefields are then used for the migration of both primaries and surface-related multiples. Moreover, also in the context of seismic full waveform inversion multiples are now included in the inversion process as a source of valuable information. In this work we review how it is possible to adapt WE migration techniques to effectively exploit the information contained in the surface-related multiples recorded in a seismic experiment. We developed a new WE migration algorithm that provides the ability to migrate both primary and multiple reflections. The migrated section coming from the migration of multiples is shown to provide a wider illumination of the subsurface in most of the cases or, at least, to illuminate the Earth with different angles with respect of those of the primary reflections. In order to prove this statement, we analyzed the migration results also in the A-CIG (Angle Common Image Gather) domain. Moreover, some preliminary results about the discrimination of the artifacts that generates in the multiples imaging are shown. This document is organized as follows: in first section we explain the underlying interpretation of the seismic experiment used throughout the whole work. In sections two and three we show, respectively, the modifications we made to adapt the WE migration algorithm for dealing with the surface-related multiples and the main issues involved in the proposed imaging technique. Section four contains the results obtained on two different synthetic dataset. One was created ad-hoc for the validation of the new migration while the other is the more realistic Sigsbee2b synthetic dataset, often used as benchmark for the multiples prediction and attenuation. Section five concludes the document with some final remarks and the outlook about the possible future developments. Earth’s impulse response. In Yilmaz (2001) the concept of the convolutional model for the characterization of the seismic experiment was introduced. According with the proposed interpretation, the recorded seismogram can be modeled as the convolution of the Earth’s impulse response with the injected source wavelet. The injected source is the elastic (or acoustic for marine seismic surveys) impulse sent into the ground, while the recorded primaries (the reflections) constitute the Earth’s response to the injected source. Also the surface-related 35 GNGTS 2013 S essione 3.1