ALBERT

Description: We present results of a seismic refraction experiment which determines the crustal and upper-mantle structure of an oceanic core complex (OCC) and its conjugate side located south of the 5°S ridge–transform intersection at the Mid-Atlantic Ridge. The core complex with a corrugated surface has been split by a change in location of active seafloor spreading, resulting in two massifs on either side of the current spreading axis. We applied a joint tomographic inversion of wide-angle reflected and refracted phases for five intersecting seismic profiles. The obtained velocity models are used to constrain the magmatic evolution of the core complex from the analysis of seismic layer 3 and crustal thickness. An abrupt increase of crustal velocities at shallow depth coincides with the onset of the seafloor corrugations at the exposed footwall. The observed velocity structure is consistent with the presence of gabbros directly beneath the corrugated fault surface. The thickness of the high-velocity body is constrained by PmP reflections to vary along and across axis between 〈3 and 5 km. The thickest crust is associated with the central phase of detachment faulting at the higher-elevated northern portion of the massif. Beneath the breakaway of the OCC the crust is 2.5 km thick and reveals significantly lower velocities. This implies that the fault initially exhumed low-velocity material overlying the gabbro plutons. In contrast, crust formed at the conjugate side during OCC formation is characterized by an up to 2-km-thick seismic layer 2 overlying a 1.7-km-thick seismic layer 3. Obtained upper-mantle velocities range from 7.3 to 7.9 km s−1 and seem to increase with distance from the median valley. However, velocities of 7.3–7.5 km s−1 beneath the older portions of the OCC may derive from deep fluid circulation and related hydrothermal alteration, which may likely be facilitated by the subsequent rifting. Our velocity models reveal a strongly asymmetric velocity structure across the ridge axis, associated with the accretion of gabbros into the footwall of the detachment fault and upper-crustal portions concentrated at the conjugate side. Our results do not support a substantial increase in the axial ridge's melt supply related to the final phase of detachment faulting. Hence, the footwall rifting at 5°S may be a generic mechanism of detachment termination under very low melt conditions, as predicted by recent numerical models of Tucholke et al.

Description: We present a detailed 3-D P-wave velocity model obtained by first-arrival travel-time tomography with seismic refraction data in the segment boundary of the Sumatra subduction zone across Simeulue Island, and an image of the top of the subducted oceanic crust extracted from depth-migrated multi-channel seismic reflection profiles. We have picked P-wave first arrivals of the air-gun source seismic data recorded by local networks of ocean-bottom seismometers, and inverted the travel-times for a 3-D velocity model of the subduction zone. This velocity model shows an anomalous zone of intermediate velocities between those of oceanic crust and mantle that is associated with raised topography on the top of the oceanic crust. We interpret this feature as a thickened crustal zone in the subducting plate with compositional and topographic variations, providing a primary control on the upper plate structure and on the segmentation of the 2004 and 2005 earthquake ruptures.

Description: We present a new 3-D shear-velocity model for the top 30 km of the crust in the wider Vienna Basin region based on surface waves extracted from ambient-noise cross-correlations. We use continuous seismic records of 63 broad-band stations of the AlpArray project to retrieve interstation Green’s functions from ambient-noise cross-correlations in the period range from 5 to 25 s. From these Green’s functions, we measure Rayleigh group traveltimes, utilizing all four components of the cross-correlation tensor, which are associated with Rayleigh waves (ZZ, RR, RZ and ZR), to exploit multiple measurements per station pair. A set of selection criteria is applied to ensure that we use high-quality recordings of fundamental Rayleigh modes. We regionalize the interstation group velocities in a 5 km × 5 km grid with an average path density of ∼20 paths per cell. From the resulting group-velocity maps, we extract local 1-D dispersion curves for each cell and invert all cells independently to retrieve the crustal shear-velocity structure of the study area. The resulting model provides a previously unachieved lateral resolution of seismic velocities in the region of ∼15 km. As major features, we image the Vienna Basin and Little Hungarian Plain as low-velocity anomalies, and the Bohemian Massif with high velocities. The edges of these features are marked with prominent velocity contrasts correlated with faults, such as the Alpine Front and Vienna Basin transfer fault system. The observed structures correlate well with surface geology, gravitational anomalies and the few known crystalline basement depths from boreholes. For depths larger than those reached by boreholes, the new model allows new insight into the complex structure of the Vienna Basin and surrounding areas, including deep low-velocity zones, which we image with previously unachieved detail. This model may be used in the future to interpret the deeper structures and tectonic evolution of the wider Vienna Basin region, evaluate natural resources, model wave propagation and improve earthquake locations, among others.

Description: The dense AlpArray network allows studying seismic wave propagation with high spatial resolution. Here we introduce an array approach to measure arrival angles of teleseismic Rayleigh waves. The approach combines the advantages of phase correlation as in the two-station method with array beamforming to obtain the phase-velocity vector. 20 earthquakes from the first two years of the AlpArray project are selected, and spatial patterns of arrival-angle deviations across the AlpArray are shown in maps, depending on period and earthquake location. The cause of these intriguing spatial patterns is discussed. A simple wave-propagation modelling example using an isolated anomaly and a Gaussian beam solution suggests that much of the complexity can be explained as a result of wave interference after passing a structural anomaly along the wave paths. This indicates that arrival-angle information constitutes useful additional information on the Earth structure, beyond what is currently used in inversions.

Description: What controls subduction megathrust seismogenesis downdip of the mantle wedge corner (MWC)? We propose that, in the region of the 2010 Mw=8.8 Maule, Chile, earthquake, serpentine minerals derived from the base of the hydrated mantle wedge exert a dominant control. Based on modeling, we predict that the megathrust fault zone near the MWC contains abundant lizardite/chrysotile‐rich serpentinite that transforms to antigorite‐rich serpentinite at greater depths. From the MWC at 32–40 km depth to at least 55 km, the predominantly velocity‐strengthening megathrust accommodated dynamic propagation of the 2010 rupture but with small slip and negative stress drop. The downdip distribution of interplate aftershocks exhibits a gap around the MWC that can be explained by the velocity‐strengthening behavior of lizardite/chrysotile. Interspersed velocity‐weakening and dynamic weakening antigorite‐rich patches farther downdip may be responsible for increased abundance of aftershocks and possibly for some of the high‐frequency energy radiation during the 2010 rupture.

Description: Seismic event detection and phase picking are the base of many seismological workflows. In recent years, several publications demonstrated that deep learning approaches significantly outperform classical approaches and even achieve human-like performance under certain circumstances. However, as most studies differ in the datasets and exact evaluation tasks studied, it is yet unclear how the different approaches compare to each other. Furthermore, there are no systematic studies how the models perform in a cross-domain scenario, i.e., when applied to data with different characteristics. Here, we address these questions by conducting a large-scale benchmark study. We compare six previously published deep learning models on eight datasets covering local to teleseismic distances and on three tasks: event detection, phase identification and onset time picking. Furthermore, we compare the results to a classical Baer-Kradolfer picker. Overall, we observe the best performance for EQTransformer, GPD and PhaseNet, with EQTransformer having a small advantage for teleseismic data. Furthermore, we conduct a cross-domain study, in which we analyze model performance on datasets they were not trained on. We show that trained models can be transferred between regions with only mild performance degradation, but not from regional to teleseismic data or vice versa. As deep learning for detection and picking is a rapidly evolving field, we ensured extensibility of our benchmark by building our code on standardized frameworks and making it openly accessible. This allows model developers to easily compare new models or evaluate performance on new datasets, beyond those presented here. Furthermore, we make all trained models available through the SeisBench framework, giving end-users an easy way to apply these models in seismological analysis.

Description: We deployed a dense geodetic and seismological network in the Atacama seismic gap in Chile. We derive a microseismicity catalog of 〉30,000 events, time series from 70 GNSS stations, and apply a transdimensional Bayesian inversion to estimate interplate locking degree. We identify two highly locked regions of different sizes whose geometries appear to control seismicity patterns. Interface seismicity concentrates beneath the coastline just downdip of the highest locking. A region of lower interplate locking around 27.5ºS coincides with higher seismicity levels, a high number of repeating earthquakes and events extending further towards the trench. Having shown numerous signs of aseismic deformation (slow-slip events and earthquake swarms), this area is situated where the Copiapó Ridge is subducted. While these findings suggest that the structure of the downgoing oceanic plate prescribes patterns of interplate locking and seismicity, we note that the Taltal Ridge further north lacks a similar signature.

Description: We present the results of P-to-S receiver function analysis to improve the 3D image of the sedimentary layer, the upper crust, and lower crust in the Pannonian Basin area. The Pannonian Basin hosts deep sedimentary depocentres superimposed on a complex basement structure and it is surrounded by mountain belts. We processed waveforms from 221 three-component broadband seismological stations. As a result of the dense station coverage, we were able to achieve so far unprecedented spatial resolution in determining the velocity structure of the crust. We applied a three-fold quality control process; the first two being applied to the observed waveforms and the third to the calculated radial receiver functions. This work is the first comprehensive receiver function study of the entire region. To prepare the inversions, we performed station-wise H-Vp/Vs grid search, as well as Common Conversion Point migration. Our main focus was then the S-wave velocity structure of the area, which we determined by the Neighborhood Algorithm inversion method at each station, where data were sub-divided into back-azimuthal bundles based on similar Ps delay times. The 1D, nonlinear inversions provided the depth of the discontinuities, shear-wave velocities and Vp/Vs ratios of each layer per bundle, and we calculated uncertainty values for each of these parameters. We then developed a 3D interpolation method based on natural neighbor interpolation to obtain the 3D crustal structure from the local inversion results. We present the sedimentary thickness map, the first Conrad depth map and an improved, detailed Moho map, as well as the first upper and lower crustal thickness maps obtained from receiver function analysis. The velocity jump across the Conrad discontinuity is estimated at less than 0.2 km/s over most of the investigated area. We also compare the new Moho map from our approach to simple grid search results and prior knowledge from other techniques. Our Moho depth map presents local variations in the investigated area: the crust-mantle boundary is at 20–26 km beneath the sedimentary basins, while it is situated deeper below the Apuseni Mountains, Transdanubian and North Hungarian Ranges (28–33 km), and it is the deepest beneath the Eastern Alps and the Southern Carpathians (40–45 km). These values reflect well the Neogene evolution of the region, such as crustal thinning of the Pannonian Basin and orogenic thickening in the neighboring mountain belts.

Description: Detecting phase arrivals and pinpointing the arrival times of seismic phases in seismograms is crucial for many seismological analysis workflows. For land station data, machine learning methods have already found widespread adoption. However, deep learning approaches are not yet commonly applied to ocean bottom data due to a lack of appropriate training data and models. Here, we compiled an extensive and labeled ocean bottom seismometer (OBS) data set from 15 deployments in different tectonic settings, comprising ∼90,000 P and ∼63,000 S manual picks from 13,190 events and 355 stations. We propose PickBlue, an adaptation of the two popular deep learning networks EQTransformer and PhaseNet. PickBlue joint processes three seismometer recordings in conjunction with a hydrophone component and is trained with the waveforms in the new database. The performance is enhanced by employing transfer learning, where initial weights are derived from models trained with land earthquake data. PickBlue significantly outperforms neural networks trained with land stations and models trained without hydrophone data. The model achieves a mean absolute deviation of 0.05 s for P-waves and 0.12 s for S-waves, and we apply the picker on the Hikurangi Ocean Bottom Tremor and Slow Slip OBS deployment offshore New Zealand. We integrate our data set and trained models into SeisBench to enable an easy and direct application in future deployments. Key Points We assembled a database of ocean Bottom Seismometer (OBS) waveforms and manual P and S picks, on which we train PickBlue, a deep learning picker Our picker significantly outperforms pickers trained with land-based data with confidence values reflecting the likelihood of outlier picks The picker and database are available in the SeisBench platform, allowing easy and direct application to OBS traces and hydrophone records