We present a seismic model of the African Plate, made with the technique of full-waveform inversion. The purpose of our model is to become a foundation for future use and research, such as quantitative geodynamic interpretations, earthquake-induced ground motion predictions, and earthquake source inversion. Starting from the first-generation Collaborative Seismic Earth Model (CSEM), we invert seismograms filtered to a minimum period of 35 s and compute gradients of the misfit function with respect to the model parameters using the adjoint state method. In contrast to the conventional FWI approach, we use dynamically changing data subsets (mini-batches) of the complete dataset to compute approximate gradients at each iteration. This approach has three significant advantages: (1) it reduces computational costs for model updates and the inversion, (2) it enables the use of larger datasets without increasing iteration costs, and (3) it makes it trivial to assimilate new data since we can add it to the complete dataset without changing the misfit function, thereby enabling “evolutionary FWI". We perform 130 mini-batch iterations and invert data from 397 unique earthquakes and 184,356 unique source-receiver pairs at the cost of approximately 10 full-data iterations. We clearly image tectonic features such as the Afar triple junction. Particularly interesting are the low-velocity zones below the Hoggar, Aïr, and Tibesti Mountains, pronounced more than in earlier works. Finally, we introduce a new strategy to assess model uncertainty. We deliberately perturb the final model, perform additional mini-batch iterations, and compare the result with the original final model. This test uses actual seismic data instead of artificially generated synthetic data and requires no assumptions about the linearity of the inverse problem.

Avalanches and other hazardous mass movements pose a danger to the population and critical infrastructure in alpine areas. Hence, understanding and continuously monitoring mass movements is crucial to mitigate their risk. We propose to use Distributed Acoustic Sensing (DAS) to measure strain rate along a fiber-optic cable to characterize ground deformation induced by avalanches. We recorded 12 snow avalanches of various dimensions at the Vallée de la Sionne test site in Switzerland, utilizing existing fiber-optic infrastructure and a DAS interrogation unit during the winter 2020/2021. By training a Bayesian Gaussian Mixture Model, we automatically characterize and classify avalanche-induced ground deformations using physical properties extracted from the frequency-wavenumber and frequency-velocity domain of the DAS recordings. The resulting model can estimate the probability of avalanches in the DAS data and is able to differentiate between the avalanche-generated seismic near-field, the seismo-acoustic far-field and the mass movement propagating on top of the fiber. By analyzing the mass-movement propagation signals, we are able to identify group velocity packages within an avalanche that propagate faster than the phase velocity of the avalanche front, indicating complex internal structures. Importantly, we show that the seismo-acoustic far-field can be detected before the avalanche reaches the fiber-optic array, highlighting DAS as a potential research and early warning tool for hazardous mass movements.

We demonstrate the logistic feasibility and scientific potential of Distributed Acoustic Sensing (DAS) in alpine volcano-glacial environments that are subject to a broad range of natural hazards. Our work considers the Mount Meager massif, an active volcanic complex in British Columbia, estimated to have the largest geothermal potential in Canada, and home of Canada’s largest recorded landslide in 2010. From September to October 2019, we acquired continuous strain data, using a 3 km long fiber-optic cable, deployed on a ridge of Mount Meager and on the uppermost part of a glacier above 2000 m altitude. The data analysis detected a broad range of unexpectedly intense, low-magnitude, local seismicity. The most prominent events include long-lasting, intermediate-frequency (0.01 - 1 Hz) tremor, and high-frequency (5 - 45 Hz) earthquakes that form distinct spatial clusters and often repeat with nearly identical waveforms. We conservatively estimate that the number of detectable high-frequency events varied between several tens and nearly 400 per day. We also develop a beamforming algorithm that uses the signal-to-noise ratio (SNR) of individual channels, and implicitly takes the direction-dependent sensitivity of DAS into account. Both the tremor and the high-frequency earthquakes are most likely related to fluid movement within Mount Meager’s geothermal reservoir. Our work illustrates that DAS carries the potential to reveal previously undiscovered seismicity in challenging environments, where comparably dense arrays of conventional seismometers are difficult to install. We hope that the logistics and deployment details provided here may serve as a starting point for future DAS experiments.

We develop a Hamiltonian Monte Carlo (HMC) sampler which solves a multi-parameter elastic full-waveform inversion (FWI) in a probabilistic setting for the first time. This gives novel access to the full posterior distribution for this type of highly non-linear inverse problem. Typically, FWI has focused on using gradient descent methods with proper regularization to iteratively update models to a minimum misfit value. Non-uniqueness and uncertainties are mostly in this approach. Bayesian inversions offer an alternative by assigning a probability to each model in model space given some data and prior constraints. The drawback is the need to evaluate a very large number of models. Random walks from Markov chains counter this effect by only exploring regions of model space where probability is significant. HMC method additionally incorporates gradient information, i.e. local structure, typically available for numerical waveform tomography experiments. So far, HMC has only been implented for acoustic FWI. We implement HMC for multiple 2D elastic FWI set-ups. Using parallelized wave propagation code, wavefields and kernels are computed on an regular numerical grid and projected onto basis functions. These gradients are subsequently used to explore the posterior space of different target models using HMC. The free parameters in these experiments are P and S velocity, and density. Although simulating Hamiltonian dynamics in the resulting phase space is approximated numerically, the results of the Markov chain are nevertheless very insightful. No prior tuning of kernels, data or model space is required, under the constraint that the sampler is properly tuned. After a burn in phase during which the mass matrix is iteratively optimized, the Markov chain is run on multiple nodes. After approximately 100,000 samples (combined from all nodes) the Markov chain mixes well. The resulting samples give access to the full posterior distribution, including the mean and maximum-likelihood models, conditional probabilities, inter-parameter correlations and marginal distributions.

Seismologists working with fiber-optic sensing, commonly referred to as Distributed Acoustic Sensing (DAS), have yet to find an established way of automatically detecting signals of interest within its recordings. We propose a new research perspective within the field by examining the output of a DAS array as an image and processing the image to find signals of interest. In this manuscript, we show an example of such a method, where we automatically detect seismic events of interest within two different DAS datasets, finding, respectively 99 % and 96 % of the local earthquakes previously identified within the data by manual analysis. The method is based on simple image processing and computer vision techniques, which clean the image, and, ideally, leave nothing but the signal of interest. These simple image processing steps yield promising results, indicating that computer vision and image processing might have an immediate impact in geophysical applications of fiber-optic sensing.

A new seismic velocity model for the south Central Andes is derived from full waveform inversion, covering the Pampean flat and adjacent Payenia steep subduction segments. Strong focused crustal low-velocity anomalies indicate partial melts in the Payenia segment along the volcanic arc, whereas weaker low-velocity anomalies covering a wide zone in Pampean possibly indicates remnant melts in the past. Thinning and tearing of the flat Nazca slab below the Pampean is inferred by gaps in the high-velocity slab along the inland projection of the Juan-Fernandez-Ridge. A high-velocity anomaly in the upper mantle below the flat slab is interpreted as a relic Nazca slab segment, which indicates an earlier slab break-off during the flattening process, triggered by the buoyancy of the Juan-Fernandez-Ridge. In Payenia, large-scale low-velocity anomalies atop and below the re-steepened Nazca slab are associated with the re-opening of the mantle wedge and sub-slab asthenospheric flow, respectively.

We present a new seismic tomography model for the crust and upper-mantle beneath the Central Andes based on multi-scale full seismic waveform inversion, proceeding from long periods (40–80~s) over several steps down to 12–60~s. The spatial resolution and trade-offs among inversion parameters are estimated through the multi-parameter point-spread functions. P and S wave velocity structures with a spatial resolution of 30–40 km for the upper mantle and 20 km for the crust could be resolved in the central study region. In our study, the subducting Nazca slab is clearly imaged in the upper mantle, with dip-angle variations from the north to the south. Bands of low velocities in the crust and mantle wedge indicate intense crustal partial melting and hydration of the mantle wedge beneath the frontal volcanic arc, respectively and they are linked to the vigorous dehydration from the subducting Nazca plate and intermediate depth seismicity within the slab. These low velocity bands are interrupted at 19.8º–21°S, both in the crust and uppermost mantle, hinting at the lower extent of crustal partial melting and hydration of the mantle wedge. The variation of lithospheic high velocity anomalies below the backarc from North to South allows insight into the evolutionary foundering stages of the Central Andean margin. A high velocity layer beneath the southern Altiplano suggests underthrusting of the leading edge of the Brazilian Shield. In contrast, a steeply westward dipping high velocity block and low velocity lithospheric uppermost mantle beneath the southern Puna plateau hints at the ongoing lithospheric delamination.

Seismic ambient noise sources have received increased attention recently, creating new possibilities to study the Earth’s subsurface and the atmosphere-ocean-solid Earth coupling. In efforts to locate such noise sources using nonlinear finite-frequency inversions, methodological developments such as pre-computed wavefields and spatially variable grids were necessary. These make inversions feasible for the secondary microseismic sources in a frequency range up to 0.2 Hz on a daily basis. By obtaining a starting model for the inversion using Matched Field Processing (MFP) we are able to further steer the inversion towards acceptable global noise source models and improve the final result. Analysis of one year of daily inversions shows the seasonal variations of the secondary microseisms and their dependence on the atmosphere-ocean-solid Earth coupling due to storm-induced ocean waves. We present a web framework, SANS (Seismic Ambient Noise Sources, sans.ethz.ch), where daily regional- to global-scale seismic ambient noise source maps are made available to the public. This eases the implementation of time-variable noise source distributions into full-waveform ambient noise tomography and time-dependent subsurface monitoring methods. Additionally, it encourages other studies to verify if changes in the seismic data are due to changes in the subsurface velocity or spatio-temporal variations of noise sources.

We present probabilistic centroid-moment tensor solutions inferred from the combination of Hamiltonian Monte Carlo sampling and a 3-D full-waveform inversion Earth model of the Japanese archipelago. While the former provides complete posterior probability densities, the latter allows us to exploit waveform data with periods as low as 15 s. For the computation of Green’s functions, we employ spectral-element simulations through the radially anisotropic and visco-elastic model, leading to substantial improvements of data fit compared to layered models. Focusing on Mw 4.8 - Mw 5.3 offshore earthquakes with a significant non-double-couple component, we simultaneously infer the centroid location, time and moment tensor without any a priori constraints on the faulting mechanism. Furthermore, we perform the inversions across several period bands, varying the minimum period between 15 s and 50 s. Accounting for 3-D Earth structure at shorter periods can increase the double-couple component of an event, compared to the GCMT solution, by tens of percent. This suggests that at least some of the non-double-couple events in the GCMT catalog might result from unmodeled Earth structure. We also observe that significant changes in source parameters, and the double-couple component in particular, may be related to only small waveform changes, thereby accentuating the importance of a reliable Earth model. Posterior probability density distributions become increasingly multimodal for shorter-period data that provide tighter constraints on source parameters. This implies, in our specific case, that stochastic approaches to the source inversion problem are required for periods below ~ 20 s to avoid trapping in local minima.