Once primarily the domain of geophysical exploration, FWI has evolved into a versatile tool, finding groundbreaking applications across a wide spectrum of disciplines—from deep Earth imaging and reservoir characterization to environmental monitoring and even medical diagnostics.
This webinar program brings together leading researchers from ETH Zurich sharing recent results achieved for real-world applications of FWI. Through a series of presentations and discussions, we would like to foster a cross-disciplinary dialogue and inspire new collaborations that could push the boundaries of this powerful technique further.
Speakers: Andreas Fichtner, Josef Patrick Marty and Hansruedi Maurer (ETH Zurich)
Agenda
REVEAL: Data-adaptive global full-waveform inversion
by Andreas Fichtner
Andreas Fichtner will present REVEAL, a transversely isotropic full-waveform inversion (FWI) model of the Earth's crust and mantle that assimilates more than 6 million three-component seismograms, including all body- and surface wave phases. REVEAL resolves previously unknown large-scale features that challenge the standard interpretation of global tomographic Earth models in terms of thermally dominated mantle convection.
The construction of REVEAL rests on the combination of stochastic mini-batch optimisation and wavefield-adapted spectral-element meshes. While the former exploits redundancies in the dataset, the latter reduces the cost of wavefield simulations by lowering the effective dimension of the numerical grid. As a consequence, the average cost of an iterative model update is only around 0.62 % of a standard update that uses the complete dataset in combination with a cubed-sphere-type mesh. We calculated 3-D synthetic seismograms using a GPU-accelerated spectral-element wave propagation solver that accounts for anelasticity, topography, bathymetry, ocean loading, and ellipticity. The ensemble of methodological improvements allows us to incorporate 6,005,727 three-component waveforms from 2366 earthquakes and to perform 305 quasi-Newton iterations; an order of magnitude more than all previous global-scale FWI models. For a diverse range of wave paths, REVEAL explains complete seismograms at 30 s period that have not been included in the inversion.
Tomographic models are paramount to unravel the Earth's interior dynamics. Most previous studies found positive wave speed anomalies that spatially correlate with the expected locations of subducted slabs. This correlation has been widely applied in plate reconstructions and geodynamic modelling. Thanks to the unprecedented amount of data included in REVEAL, the model resolves numerous previously undetected positive wave speed anomalies in the lower mantle. Many of these anomalies are situated below major oceans and continental interiors, with no geologic record of subduction, such as beneath the western Pacific Ocean. Moreover, we find no statistically significant correlation of positive anomalies in REVEAL and past subduction. This suggests more diverse origins for large-scale anomalies in Earth’s lower mantle, unlocking FWI as an indispensable tool for mantle exploration.
Challenges of FWI applied to near-surface problems
by Hansruedi Maurer
FWI applications have the potential to be a game changer in near surface applications, but there are a number of problems that still need to be resolved. Most problems are related with the complexity of the shallow subsurface. This will require 3D elastic or even viscoelastic problems to be solved. Furthermore, the variable coupling of the seismic sources and receivers needs to be considered. Finally, near surface projects often operate on limited budgets. This requires not only the data acquisition, but also the inversion procedures to be optimized. In this presentation, we will show a few options, how these problems can be addressed.
Full-Waveform Inversion in Medicine: Imaging the Human Brain with Ultrasound
by Josef Patrick Marty
Transcranial ultrasound imaging is an emerging modality for non-invasively recovering the internal structure of the brain through the intact skull. Imaging the human brain with ultrasound has been notoriously challenging due to the complexities introduced at the soft tissue-bone interfaces. The high impedance contrast between soft tissue and bone tends to result in the wavefield becoming strongly distorted as it propagates across this interface; these challenges are analogous to those encountered in subsalt imaging within exploration geophysics. Furthermore, the fluid-solid interactions between these adjacent materials introduce the presence of acoustic-elastic coupling, in a comparable way to those observed between the Earth’s fluid outer core and the solid mantle within global-scale seismology. In order to enable the use of transcranial ultrasound imaging, a number of techniques such as the spectral-element method, full-waveform inversion, optimal transport, and reverse time migration are adapted from seismic geophysics for use within medical ultrasound.
