Seismic imaging by nonlinear inversion

Oct 30, 2023 | CWP Blog

Posted by Werter Silva

Seismic imaging is a valuable technique for understanding subsurface geological structures. It has been shown that this process can be formulated as an inverse problem, where the recorded seismic data are linearly related to model perturbations through the Born approximation. Consequently, seismic events such as ghosts and multiples are attenuated prior to inversion to meet the single scattering assumption implied by this approximation. In contrast, nonlinear imaging seeks solutions to the inverse problem without the need for multiple attenuation. This approach uses seismic waves that scatter multiple times before reaching receivers. Unlike traditional methods, nonlinear imaging doesn’t require removing multiples beforehand. As a result, it provides cost-effective and higher-quality images, enhancing the efficiency of the imaging process.
The figure below compares the seismic image obtained by nonlinear inversion with LSRTM (linear inversion) for the Marmousi model. The input data for both methods comprises shots containing primaries, ghosts, surface, and internal multiples. The LSRTM image (b) shows the distorted reflector phase due to the ghost “derivative” effect, best seen at the water bottom. A first-order surface multiple parallel to the water bottom intersecting with the reflectors is also visible in the image. This image is noisy in general because unrelated events are cross-correlated. On the other hand, the nonlinear image (c) can handle multiples and ghosts in addition to primaries and generates a clearer image comparable to the true reflectivity (a).