Wave field autofocusing and applications to multidimensional deconvolution and imaging with internal multiples

Abstract

One of the most challenging tasks of exploration geophysics is to build a quantitatively accurate image of the structures inside the Earth from reflection data measured at the Earth's surface. When the subsurface is structurally complicated, accurate images are needed to locate energy sources, such as hydrocarbon reservoirs. Conventional migration algorithms rely on the single-scattering assumption. This restrictive assumption requires that the recorded data do not include waves that have bounced multiple times between layers before reaching the receivers. Standard imaging algorithms incorrectly image the multiple reflections as ghost reflectors and these artifacts can mislead the interpreters in locating potential energy sources. The objective of this thesis is to investigate a new method in reflection seismology for building ghost-free images of the subsurface. The goal is to take advantage of multiply-scattered waves in order to produce more correct images in complicated geological subsurface environments. The method I propose, defined as wave field autofocusing, is based on inverse methods originally used in quantum scattering. The first part of this thesis presents the connection between such inverse methods and Green's function retrieval for a one-dimensional medium. I emphasize that the importance of these inversion methods is linked to the retrieval of the wave field propagating in an unknown medium and not to the retrieval of the impedance profile of the same medium. Based on this connection, I extend the method to two-dimensional media and apply it to multidimensional deconvolution to obtain a ghost-free image, hence presenting ad advantage over standard imaging techniques. Then, using the unitarity of the scattering matrix, I show the consistency between the new approach (requiring data on only one side of the medium) and existing methods that require measurement on a closed boundary, such as seismic interferometry. In the last part of this thesis, I test the robustness of the proposed method with respect to errors in the background model used to estimate the first-arriving waves (a required input for the autofocusing process).