Waves in complex media

We accommodate wave propagation in highly complex media using spectral-element methods. The crux to honor such realistic settings relies on the meshing process which maps an assumed heterogeneous velocity model into a hexahedral grid. We apply this to scales of interest to exploration seismology (salt bodies, overthrust faults), seismic hazard (Los Angeles, Christchurch), and continental tomography (ambient noise in Europe) using the spectral-element solver SPECFEM3D.

Seismic hazard in Los Angeles & Christchurch

(A. Tesoniero, T. Nissen-Meyer, E. Casarotti, B. Fry)

As a consequence of the Mw 6.7 1994 Northridge earthquake, the area of Santa Monica in the Los Angeles basin experienced concentrated and unexpectedly high values of ground shaking. Elastic focusing wave effects have been proposed by several authors as a cause for the enhanced ground shaking. Acting as an elastic lense, the Santa Monica overthrust has been attributed as a possible contributor to this enhanced, localized shaking. We model 3D wave propagation with kinematic source models for this setting including external and internal interface topography and vary local, smoothed properties of this overthrust to constrain its effect on wave propagation. Peak-ground velocities in Santa Monica vary by as much as 40% depending on the local properties of the overthrust, matching also the actual amplitudes recorded at the Santa Monica City Hall seismic station. We conclude that in regions of complex 3D subsurface settings, where complex wave phenomena such as elastic focusing might happen, honoring these structures both numerically (by the mesh model) and physically (by the velocity model) is important for a proper representation of the seismic ground shaking.
Together with colleagues from GNS New Zealand, we are also in the process of conducting similar studies around the M6.3 Christchurch earthquake in light of the 2g ground accelerations measured in the city. This is an extraordinary set- ting, including complex local structures (volcanic layers as possible waveguides) and high-quality recordings of near-field ground motion of the entire aftershock sequence. As such, this sequence represents an invaluable site to assess dynamic rupture, near-field wave propagation, and 3D structure from a full-wave propagation standpoint with unprecedented data quality and quantity.

Exploration seismology

(T. Nissen-Meyer, B. Artman, J. Tromp, A. Plesch, J. Shaw)

We apply our wave propagation and adjoint imaging techniques to shallow structures of high complexity (surface topography, overthrust faults, salt domes, wedges, lenses, thin layers, offshore settings), and draw connections between full-wave finite-frequency sensitivity and classical imaging principles of the oil industry. Our results highlight the importance of choosing sensible misfit functions (e.g., band-passed time windows for normalized adjoint sources) as well as the most appropriate set of model parameterization. Impedance images show great promise in illuminating structures much like reverse-time migration, but within the more general framework of subsequently inverting volumetric and reflecting contributions. Time-lapse sensitivity kernels highlight the potential to use appropriate physics in the forward problem and sophisticated inversion techniques. Further work relates to constraining the benefit of low frequencies and multi-component broadband seismometers for full-waveform inversions in exploration settings. Of particular interest to the exploration industry here is the potential to predict optimal design for acquisition surveys.

Adjoint-based nonlinear tomography using ambient noise

(P. Basini, T. Nissen-Meyer, L. Boschi, L. Gaudio, O. Schenk)

The European crust and upper mantle are key to understanding the tectonic environment of the Alpine orogene, continental collision, and plate boundaries. Since earthquakes happen rarely in western Europe, we use ambient noise generated in the oceans as a basis for iterative, non-linear, adjoint-based tomography beneath Europe. We compiled a dense regional database of European station-station surface-wave dispersion between seismic periods of 8-35 seconds using noise interferometry. It has recently been shown how adjoint techniques can be applied to ambient-noise data, overcoming the often severe nonuniformity in the geographic distribution of noise sources, and the subsequent discrepancies between the recorded noise cross-correlation and the theoretical Green’s function.
Our initial model is composed of two contributions: EPcrust, a new 3D crustal model for the European plate, derived from collection of numerous independent previous studies of multiple scale lengths, and a new adaptive-grid surface-wave tomography of the uppermost mantle down to periods of 35 seconds. This model is discretized with irregular meshes (using Cubit) that honor all relevant discontinuities and are adaptive within the inversion procedure. The misfit function between modeled and data-based cross-correlations that defines the adjoint source is based on a multitaper traveltime difference, allowing us to iteratively march from coarse to fine scale. We address the peculiar issue of non-uniform noise sources by including the frequency-dependent noise distribution in the inversion process. The inversion is a computationally intensive, heavily parallel procedure for which we have allocations on national supercomputers. We also embark on studying the behavior of objective functions within such realistic 3D models regarding convergence, preconditioning, appropriate misfit choices, and the limitations of assuming ambient-noise cross correlations to represent Green’s functions.