Coherence-based imaging and inversion on large-scale seismic arrays
Final Report Abstract
In contrast to earthquake seismology, seismic exploration deals with spatially and temporally localized and highly repeatable sources. In addition, receiver distributions are dense and in the vast majority of cases, full wavefields are recorded and can be utilized for the imaging and inversion of subsurface structure. Owing to the limited capability of achieving dense coverage over a large region at the Earth's surface, maximum source-receiver distances are typically small and reflected and back-scattered wavefields are the primary carriers of information. In earthquake seismology on the other hand, natural sources excited e.g. by tectonic processes are used, which are not controllable and rarely reveal repeated radiation characteristics. As an additional complication, inter-station distances are large and the wavefleld can rarely be fully reconstructed in the frequency range typically investigated. Accordingly, tomographic inversion of subsurface properties, rather than wavefleld-based imaging has evolved to the method of choice on the earthquake scale. In recent years, however, rapid improvements in instrumentation through the deployment of vast densely spaced arrays spanning whole continents could be observed, which promise to make highly-resolved wavefleld-based imaging of large-scale Earth structure feasible for the first time. As a consequence, collective properties still typically ignored in earthquake studies can and should be taken into account and, aside from the commonly utilized transmitted contributions and surface waves, earthquake reflection can be systematically investigated to enable complementary imaging of the Earth's major discontinuities. This project had the ultimate objective of strengthening the still quite feeble conceptual interface between earthquake and exploration seismology. Originating in optics, coherence can be viewed as an inherent property of wavefields and its usefulness and applicability was thoroughly investigated over the past two years. Drawing from the vast experience with wavefields on the exploration scale, controlled-source coherence analysis represents a considerably richer framework than beamforming and, for the first time, makes wavefront curvature, which for continent-sized arrays cannot be neglected any more, a freshly exploitable observable leading to interesting and novel applications in wavefleld reconstruction, separation and singlearray source localisation. Somewhat unexpectedly, I had to discover within the first months after my arrival in Oxford, that the way wavefrom data is handled and stored is fundamentally different from what I was used to in crustal exploration. As a consequence, my pre-developed software, which has the logistics of wavefields engrained in its often non-modular structure, had to be fully rewritten. This additional effort, however, led to the development of a software package (termed SeismicBeam) that I believe will be widely usable by both communities, providing joint support for exploration's Seismic Unix or SEG-Y formats, as well as being able to handle individual seismological SAC files and the according meta information. With the capability of translating between the different standards, joint, more seamless efforts in method development and geoscientiflc research now become possible. Being written in the modern and flexible highperformance programming language Julia, intuitive future code development and straight-forward extensions to the accomplished research are expected. It was observed that even earthquake wavefields that were densely sampled by North America's USArray reveal in parts strong spatial aliasing that needs to be circumvented for wavefleld imaging to be successful. In reaction to this observation, I introduced the concept of differential traveltime corrections, which allow coherence-based wavefleld analysis and reconstruction even in cases of severe under-sampling. As an additional major challenge, I observed that the faint but highly resolving reflected and back-scattered wavefields are typically masked by the more dominant and commonly utilized transmitted contributions and surface waves. In light of these findings, a striking resemblance of the process of diffraction and passive-source excitation became obvious, leading to a joint development of techniques for diffracted events on the exploration scale and a systematic treatment of large-scale earthquake back-scattering and reflection. Forming an integral part of the julia toolbox, interfering coherent wavefields - let it be weak diffraction in the crust or faint reflection in the mantle transition zone - can now be efficiently separated by making use of the novel extracted wavefront curvature information. Owing to the integrated treatment of diffracted and passive-source wavefields, joint improvements could be achieved in simultaneous source / scatterer localisation and velocity model building, data enhancement and regularization, coherence-assisted finite-frequency measurements for large-scale tomography, and high resolution diffraction imaging of faults and erosional unconformities, supporting a complementary, multi-scale view of the Earth's interior.
Publications
- (2017). ‘A generalized view on normal moveout'. In: Geophysics 82.5, pp. V335-V349
B. Schwarz and D. Gajewski
(See online at https://doi.org/10.1190/GE02017-0159.1) - (2017). ‘Accessing the diffracted wavefield by coherent subtraction'. In: Geophysical Journal International 211.1, pp. 45-49
B. Schwarz and D. Gajewski
(See online at https://doi.org/10.1093/gji/ggx291) - (2019). ‘Coherent wavefield subtraction for diffraction separation'. In: Geophysics 84.3, pp. 112
B. Schwarz
(See online at https://doi.org/10.1190/GE02018-0368.1) - (2019). ‘Unsupervised event identification and tagging for diffraction focusing'. In: Geophysical Journal International
Bauer, A., B. Schwarz, T. Werner and D. Gajewski
(See online at https://doi.org/10.1093/gj i/ggzl06)
