Data-driven focusing and two-way wave modeling with applications to seismic processing and imaging
- Doctoral Thesis
Rights / licenseIn Copyright - Non-Commercial Use Permitted
Our knowledge of the Earth’s internal structure and composition is to a large extent based on recordings of seismic waves. In exploration seismology, such waves are exploited to extract in- formation about the subsurface. Seismic waves, artificially generated by sources placed at the Earth’s surface, travel through the subsurface and are recorded as seismograms by large sensor arrays also placed at the surface. A profound understanding of how waves propagate and interact with geological structures is necessary in order to turn seismograms into structural subsurface images that can be geologically interpreted. For this purpose, elaborate algorithms have been developed, both to numerically simulate wave propagation as well as to process and image the vast amounts of data recorded during seismic exploration campaigns. This thesis contributes to developments in modeling and processing of acoustic seismograms, in- troducing new and more sophisticated algorithms. A key objective is to preserve the full waveform information that is contained in seismic data and to consider both high-order wave interactions as well as complex interference patterns. This is achieved by introducing new approaches for seismic modeling and an augmented formulation of a data-driven focusing technique, commonly known as the Marchenko method. A hybrid modeling approach is developed that combines the benefits of state-of-the-art modeling and data-driven focusing. This work demonstrates how modeling schemes that solve the full (two-way) wave equation can be used to isolate individual wavefield constituents, which is of key importance for seismic data analysis and processing. Although such methods are often considered as only being capable to compute the full wavefield, here pseudo one-way wave propagation is implemented for layered media, resulting in the retrieval of separated first- and higher-order wavefield constituents. More- over, one (or a series of) target interfaces can be cloaked in a wave simulation, enabling the retrieval (or removal) of all reflections from that target. Similar cloaking can be achieved in a data-driven way, permitting the identification of target-related reflections from surface data. Another important aspect of this work is the Marchenko method and its fundamental solutions, the so-called focusing functions. Their ability to focus recorded reflection data to an arbitrary point at depth is often interpreted as creating virtual sources. Among other applications, this can be exploited for seismic imaging to remove effects due to a scattering overburden. Here, a demon- stration is given for how physical arguments can be used to incorporate energy conservation and a minimum phase property into the scheme, resulting in an augmented Marchenko method. A detailed analysis of the effects of band-limitation is presented, which is important for any appli- cation that uses field data (given the band-limited nature of all recorded data). Furthermore, it is shown how the additional arguments can be used to account for the effects of fine-layering, removing a limitation that is usually introduced by time-windowing individual events. Results for a synthetic data example that represents realistic geological challenges, such as those faced in the Middle East, show significant improvements, making this the first demonstration of a data-driven method that successfully accounts for the effects of fine layering. Combining data-driven focusing with state-of-the-art modeling gives rise to a novel concept for immersive modeling. This work demonstrates how the results of the Marchenko method can be used as so-called immersive boundaries. In previous applications such boundaries required detailed knowledge of the subsurface. Since the Marchenko method requires relatively little a priori knowledge, the approach presented here might be preferable. It permits wave simulations of a target domain (e.g., a reservoir) to be embedded in an ‘unknown’ environment. Moreover, the target-oriented approach makes it possible to limit the wave simulations to a relatively small domain, which is beneficial for applications where many simulations of a limited target region are required, such as time-lapse modeling and target-oriented inversion. Finally, a way to compute focusing functions in complex media is introduced here, based on a mod- eling algorithm that propagates a wavefield through subsequent depth levels instead of through subsequent time steps. These may be used to study focusing mechanisms in realistic geologic environments or to reveal perspectives and limitations of the Marchenko method. A brief analysis of the linear properties of the focusing functions suggests a new strategy for ‘direct’ imaging. The concept is demonstrated with a horizontally layered medium, of which the velocity and density distributions are accurately recovered. Show more
External linksSearch via SFX
ContributorsExaminer: Robertsson, Johan O.A.
Examiner: van Manen, Dirk-Jan
Examiner: Dukalski, Marcin S.
Examiner: Wapenaar, C.P.A.
SubjectSEISMIC WAVES/PROPAGATION (GEOPHYSICS); SEISMIC IMAGING, SEISMIC TOMOGRAPHY (APPLIED GEOLOGY AND GEOPHYSICS); Focusing; numerical modeling; Acoustic waves; Data processing
Organisational unit03953 - Robertsson, Johan / Robertsson, Johan
633172-1 - EuroMix (Horizon 2020) (SBFI)
MoreShow all metadata