Seismic imaging aims at revealing the structural information of the subsurface using the reflected wavefields captured by sensors usually located at the surface. Wave propagation is a complex phenomenon and the measured data contain a set of backscattered events including not only primary reflections, but also surface-related and interbed multiples. Additionally, transmission effects also play an important role in the wave propagation. However, most of the current imaging algorithms, being based on single scattering assumptions, can handle only primary reflections and all other effects are treated as noise that produces false structures (crosstalk) in the resulting image. To avoid this, data used by conventional imaging algorithms is usually preprocessed in a such way that primaries are separated from the rest of the arrivals. However, imaging only the first category of events excludes the available information contained by multiple scattering. Furthermore, as a perfect multiple removal process is a challenge, residual crosstalk is often visible in the final image. The main topic of this thesis is to develop an imaging algorithm that can correctly handle such complex scattering effects. The main motivation is aimed at extracting complete information from the reflection data by using the multiples and, thereby, avoiding their elimination as a preprocessing step. The problem is solved by considering the imaging process as an inverse problem, where the measured data forms the data space and the unknown reflectivities constitute the model space. Solving of the inverse problem requires forward modeling and computing the gradient. The former is based on the modelling approach where amplitudes of the modeled data are driven exclusively by the reflectivity model (to be estimated), whereas travel times are dependent only on the provided migration velocity model. Moreover, because the forward model is based on a recursive scheme (the Bremmer series) it is also possible to efficiently simulate data with any combination of multiple scattering. Therefore, by minimising the misfit between the observed and the modeled data the crosstalk from multiples in the estimated reflectivity model is suppressed, because the process of fitting the data is not based anymore on the single scattering assumption. An important component in the inversion process is extracting a model update for the reflectivities from the data misfit. It is also important to mention that complex wavefields are involved in the ’imaging condition’ step, which clearly shows the contribution of the complex scattering. Therefore, the final inversion-based imaging process is called Full Wavefield Migration (FWM) and it is especially suited for situations where primaries provide a limited illumination of the subsurface, which can be compensated by the multiples. Furthermore, extensions of the method have been proposed as well, like primary/multiple separation, source field estimation, deblending and missing data reconstruction. The virtues of FWM are successfully demonstrated on several numerical and field data examples.
|Award date||6 Dec 2016|
|Publication status||Published - 2016|