GLORIA

GEOMAR Library Ocean Research Information Access

feed icon rss

Your email was sent successfully. Check your inbox.

An error occurred while sending the email. Please try again.

Proceed reservation?

Export
  • 1
    Publication Date: 2021-11-24
    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.
    Type: Article , NonPeerReviewed
    Format: text
    Location Call Number Limitation Availability
    BibTip Others were also interested in ...
  • 2
    Publication Date: 2021-11-23
    Description: Machine Learning (ML) methods have seen widespread adoption in seismology in recent years. The abilityof these techniques to efficiently infer the statistical properties of large datasets often provides significantimprovements over traditional techniques when the number of data are large(»millions of examples). Withthe entire spectrum of seismological tasks, e.g., seismic picking and detection, magnitude and source propertyestimation, ground motion prediction, hypocentre determination; among others, now incorporating ML ap-proaches, numerous models are emerging as these techniques are further adopted within seismology. To evaluatethese algorithms, quality controlled benchmark datasets that contain representative class distributions are vital.In addition to this, models require implementation through a common framework to facilitate comparison.Accessing these various benchmark datasets for training and implementing the standardization of models iscurrently a time-consuming process, hindering further advancement of ML techniques within seismology. Thesedevelopment bottlenecks also affect ’practitioners’ seeking to deploy the latest models on seismic data, withouthaving to necessarily learn entirely new ML frameworks to perform this task. We present SeisBench as a soft-ware package to tackle these issues. SeisBench is an open-source framework for deploying ML in seismology.SeisBench standardises access to both models and datasets, whilst also providing a range of common processingand data augmentation operations through the API. Through SeisBench, users can access several seismologicalML models and benchmark datasets available in the literature via a single interface. SeisBench is built to beextensible, with community involvement encouraged to expand the package. Having such frameworks availablefor accessing leading ML models forms an essential tool for seismologists seeking to iterate and apply the nextgeneration of ML techniques to seismic data.
    Type: Article , NonPeerReviewed
    Format: text
    Location Call Number Limitation Availability
    BibTip Others were also interested in ...
  • 3
    Publication Date: 2021-07-03
    Description: The Klyuchevskoy Volcanic Group is a cluster of the world's most active subduction volcanoes, situated on the Kamchatka Peninsula, Russia. The volcanoes lie in an unusual off‐arc position within the Central Kamchatka Depression (CKD), a large sedimentary basin whose origin is not fully understood. Many gaps also remain in the knowledge of the crustal magmatic plumbing system of these volcanoes. We conducted an ambient noise surface wave tomography, to image the 3‐D shear wave velocity structure of the Klyuchevskoy Volcanic Group and CKD within the surrounding region. Vertical component cross correlations of the continuous seismic noise are used to measure interstation Rayleigh wave group and phase traveltimes. We perform a two‐step surface wave tomography to model the 3‐D Vsv velocity structure. For each inversion stage we use a transdimensional Bayesian Monte Carlo approach, with coupled uncertainty propagation. This ensures that our model provides a reliable 3‐D velocity image of the upper 15 km of the crust, as well as a robust assessment of the uncertainty in the observed structure. Beneath the active volcanoes, we image small slow velocity anomalies at depths of 2–5 km but find no evidence for magma storage regions deeper than 5 km—noting the 15 km depth limit of the model. We also map two clearly defined sedimentary layers within the CKD, revealing an extensive 8 km deep sedimentary accumulation. This volume of sediments is consistent with the possibility that the CKD was formed as an Eocene‐Pliocene fore‐arc regime, rather than by recent (〈2 Ma) back‐arc extension.
    Description: Plain Language Summary: The Klyuchevskoy Volcanic Group is a cluster of 13 volcanoes on the Kamchatkan corner of the Pacific ring of fire. The volcanoes regularly produce large eruptions, but good knowledge of the magma plumbing system beneath the surface is still lacking. Why the Klyuchevskoy Volcanic Group volcanoes lie in the location they do, in a large low‐lying depression, is also unexplained. We undertook a seismic experiment and used the data to produce a 3‐D velocity image of the subsurface beneath the volcanoes and the depression. We found that small regions of slow seismic velocity are located beneath the active volcanoes, at 2–5 km depth below sea level. This slower velocity is probably caused by magma lying within the porous fracture spaces in this rock. The seismic velocities are much faster beneath the dormant volcanoes, suggesting they have no magma beneath them. With our velocity image, we also find that the Central Kamchatka Depression is very deep, filled with over 8 km of sediments. This supports an idea that the sediments accumulated as a fore‐arc basin over many millions of years, since 40 Ma, when the active line of volcanoes was found 100 km to the west.
    Description: Key Points: Three‐dimensional shear velocity structure of the Klyuchevskoy area was determined using coupled transdimensional Monte Carlo inversions. Slow velocity anomalies suggest magma storage beneath active volcanoes at 2–5 km depth (below sea level) but not in the midcrust. Sediments filling the Central Kamchatka Depression are 8 km deep, consistent with an origin of the depression as a fore‐arc basin.
    Description: European Union Horizon 2020 Research and Innovation Programme http://dx.doi.org/10.13039/501100007601
    Description: Russian Ministry of Education and Science http://dx.doi.org/10.13039/501100003443
    Description: Alexander von Humboldt Foundation http://dx.doi.org/10.13039/100005156
    Keywords: 551.1 ; tomography ; Central Kamchatka Depression ; transdimensional ; Bayesian ; ambient noise ; Klyuchevskoy Volcanic Group
    Type: article
    Location Call Number Limitation Availability
    BibTip Others were also interested in ...
  • 4
    Publication Date: 2021-07-03
    Description: We study the crustal structure of Sri Lanka by analyzing data from a temporary seismic network deployed in 2016–2017 to shed light on the amalgamation process from a geophysical perspective. Rayleigh wave phase dispersion curves from ambient noise cross correlation and receiver functions were jointly inverted using a transdimensional Bayesian approach. The Moho depths in Sri Lanka range between 30 and 40 km, with the thickest crust (38–40 km) beneath the central Highland Complex (HC). The thinnest crust (30–35 km) is found along the west coast, which experienced crustal thinning associated with the formation of the Mannar Basin. VP/VS ratios lie within a range of 1.60–1.82 and predominantly favor a felsic to intermediate bulk crustal composition with a significant silica content of the rocks. A major intracrustal (18–27 km), slightly westward dipping (∼4.3°) interface with high VS (∼4 km/s) underneath is prominent in the central HC, continuing into the western Vijayan Complex (VC). The discontinuity might have been part of the respective units prior to the collision and could be an indicator for the proposed tilting of the Wanni Complex/HC crustal sections. It might also be related to the deep crustal HC/VC thrust contact with the VC as an indenting promontory of high VS. A low‐velocity zone in the central HC could have been caused by fluid influx generated by the thrusting process.
    Description: Key Points: Sri Lanka has mostly isostatically compensated 30–40 km thick crust. VP/VS ratios are between 1.60 and 1.82 and predominantly favor a felsic to intermediate bulk crustal composition. A midcrustal westward dipping interface could be related to the thrust contact between the Highland Complex and the Vijayan Complex.
    Keywords: 551.1 ; 551.8 ; Sri Lanka ; crustal structure
    Type: article
    Location Call Number Limitation Availability
    BibTip Others were also interested in ...
  • 5
    Publication Date: 2021-07-03
    Description: Subduction zone processes and the resulting geometries at depth are widely studied by large‐scale geophysical imaging techniques. The subsequent interpretations are dependent on information from surface exposures of fossil subduction and collision zones, which help to discern probable lithologies and their structural relationships at depth. For this purpose, we collected samples from Holsnøy in the Bergen Arcs of western Norway, which constitutes a well‐preserved slice of continental crust, deeply buried and partially eclogitized during Caledonian collision. We derived seismic properties of both the lower crustal granulite‐facies protolith and the eclogite‐facies shear zones by performing laboratory measurements on cube‐shaped samples. P and S wave velocities were measured in three perpendicular directions, along the principal fabric directions of the rock. Resulting velocities agree with seismic velocities calculated using thermodynamic modeling and confirm that eclogitization causes a significant increase of the seismic velocity. Further, eclogitization results in decreased VP/VS ratios and, when associated with deformation, an increase of the seismic anisotropy due to the crystallographic preferred orientation of omphacite that were obtained from neutron diffraction measurements. The structural framework of this exposed complex combined with the characteristic variations of seismic properties from the lower crustal protolith to the high‐pressure assemblage provides the possibility to detect comparable structures at depth in currently active settings using seismological methods such as the receiver function method.
    Description: Key Points: Eclogitization of continental crust increases seismic velocities (isotropic averages up to 8.21 km/s) and decreases VP/VS ratios by ~0.04. Eclogitization coeval with deformation causes a high P wave anisotropy of up to 9%. Shear zone formation coeval with eclogitization causes changes of the seismic response of the structure.
    Description: Deutsche Forschungsgemeinschaft (DFG) http://dx.doi.org/10.13039/501100001659
    Keywords: 551.1 ; subducted continental crust ; seismic properties
    Type: article
    Location Call Number Limitation Availability
    BibTip Others were also interested in ...
Close ⊗
This website uses cookies and the analysis tool Matomo. More information can be found here...