GNGTS 2013 - Atti del 32° Convegno Nazionale

upgoing wavefield reconstruction we use an estimation of the surface-related multiples M ( t , x ) in place of the recorded primaries and we propagate this field backward in time. This way the downgoing wavefield would contain the primary reflections, regarded as a virtual source for the surface related multiples M ( t , x ). The multiples field can be retrieved using one of the many multiples prediction methods. In our work we used both the well-known SRME algorithm (Verschuur et al. , 1992) and an original model-based prediction algorithm (Lipari et al. , 2012). Eq. (1) for the case of multiples imaging becomes: (3) Surface-related multiples had followed different propagation paths with respect to the primary reflections and then they carry a different information about the Earth’s structure. Moreover, multiples can produce better illumination. For the case of conventional migration, the source illumination is basically determined by the source-receivers geometry and covers approximately half of the receivers coverage. On the other hand, in imaging of multiples, every receiver is treated as a source and almost the whole region covered by the receivers is illuminated (see Fig. 1). In particular, multiples tend to reach the subsurface discontinuities with lower angle on incidence with respect to the primaries (Fig. 3). These considerations makes the use of multiples reflections attractive for improving the overall picture of the subsurface. Cross-talk. Each one of the two terms involved in Eq. (3) can be re-written expanded in its components: (4) (5) where the subscript letters ( A,B,…,n ) refer to different interfaces in the subsurface, while the superscript numbers indicate the order of the multiple events. We indicated the primary events with the letter P and the multiples with M . IM indicates all the others events not accounted in the present work (diffractions, internal multiples, etc.). Useful contributions to the migrated section only come from the combination of primaries and multiples corresponding to the same reflector. All the others terms in the summation of Eq. (6) have to be regarded as noise and they are commonly called cross-talk terms . We can consider the seismic image obtained from the multiples migration to be composed by two terms: the desired image Ȋ ( x ) and the image of the artifacts I N ( x ): (6) (7) In Eq. (7) it is possible to notice that the cross talk terms are of two types: those events that come from the interaction of primaries and multiples of order higher that the first and those events generated by the cross-correlation of primary reflections of one interfaces and multiples of a different reflector. The former produce the same effect of the cross-correlation of a source function S ( t , x ) with the surface-related multiples recorded in the original data. The latter show up as unwanted virtual interfaces in the final migrated image. Both type of artifacts seem to have a different moveout with respect to the one of the desired events in the A-CIG domain (see Fig. 3) even if more studies are needed for their identification and removal. As a matter of fact, cross-talk attenuation is still an open topic in the area of seismic imaging. Tests. For the testing phase we developed a WE migration code based on the use of Kirchhoff extrapolation for the simulation of the downgoing and upgoing wavefields. Being based on Kirchhoff extrapolation it lacks of the accuracy provided by the more effective finite difference extrapolation engines commonly used in WEM and RTM. Nevertheless, 37 GNGTS 2013 S essione 3.1