Summary
Seismic exploration is increasingly challenged to image complex subsuface targets, such as subalt and overburden velocity inversion based on reverse-time migration. A yet unsolved problem is direct exploration of multipathing that exists within complex structures. When using the cross-correlation image condition in 2D prestack migration for each source, the contributions are summed over all the time steps to give one migrated image. The stacked image includes contributions from all primary and prismatic waves, multiples, converted waves and artifacts. Thus the final image summation contains a variety of wavefield distortions. However, if the image time slices are saved for each image time, and are sorted (across sources) into a three-parameter (incident angle, depth, image time) volume for which separate images can be constructed using any desired subset of the migrated data in the three data dimensions, much valuable information from multipath arrivals can be obtained. The separate partial images can by displayed for the primary contributions, or any combination of different types of waves (with or without artifacts). A numerical example for a simple model with two sets of source-to-reflector paths show how primary and prismatic contributions merge into a single incident angle vs image time trajectory. A second example using synthetic data from the Sigsbee2 model shows that the relative contributions of multipath arrivals to subsalt images are different.
Introduction
Prestack reverse-time migration (RTM) is effective for generation of prestack depth migrated images from common-source data. Combining the migrated depths with incident (or reflection) angle information produces angle-domain commonimage gathers (ADCIGs) from any prestack migration. Raybased prestack migration velocity analysis (Stork, 1992), is efficient, and the propagation angle information is already inherently available as part of the ray paths, but it based on high frequency assumption and often fails when we image complex structures. Wave-based prestack ADCIG algorithms have one of three different forms to extract angle information; directionvector- based methods (e.g., Yoon and Marfurt (2006); Zhang and McMechan (2011); Dickens and Winbow (2011); Vyas et al. (2011)), local plane-wave decomposition methods (e.g., Xu et al. (2011); Rui and Xie (2012)), and local-shift imaging condition methods (Sava and Fomel (2003)). Calculating the propagation angles from the wavefronts is a substantial extra computational effort. Xu et al. (2001) show that ADCIGS are ’cleaner’ than offset domain CIGs as the offset domain contains triplications, whereas ordering the image contributions by angle does not. However, the amplitudes in ADCIGS can still have data gaps, or rapid variations, discontinuities discontinuities and biases if both the source and receiver wavefields are not adequately corrected for all propagations effects (Deng and McMechan (2008)). Wave-based ADCIGs usually use the crosscorrelation (or source-normalized crosscorrelation) image condition, so all possible wave types and paths automatically contribute to the image. A prismatic path also satisfies an image time, but for it’s unique path; Cavalca and Lailly (2005) show that RTM images with multipaths provide more complete target information in complex geology, as multipaths usually have different incident angles, and amplitudes, from the primary reflections.
