ARRU Phase Picker: Attention Recurrent-Residual U-Net for Picking Seismic P- and S-Phase Arrivals

2021 ◽  
Author(s):  
Wu-Yu Liao ◽  
En-Jui Lee ◽  
Dawei Mu ◽  
Po Chen ◽  
Ruey-Juin Rau
Keyword(s):  
Human Error ◽  
Seismic Waves ◽  
Multiple Scales ◽  
Signal To Noise Ratio ◽  
Source Parameters ◽  
Seismic Source ◽  
Earthquake Location ◽  
P Waves ◽  
Arrival Times ◽  
Unseen Data

Abstract Seismograms are convolution results between seismic sources and the media that seismic waves propagate through, and, therefore, the primary observations for studying seismic source parameters and the Earth interior. The routine earthquake location and travel-time tomography rely on accurate seismic phase picks (e.g., P and S arrivals). As data increase, reliable automated seismic phase-picking methods are needed to analyze data and provide timely earthquake information. However, most traditional autopickers suffer from low signal-to-noise ratio and usually require additional efforts to tune hyperparameters for each case. In this study, we proposed a deep-learning approach that adapted soft attention gates (AGs) and recurrent-residual convolution units (RRCUs) into the backbone U-Net for seismic phase picking. The attention mechanism was implemented to suppress responses from waveforms irrelevant to seismic phases, and the cooperating RRCUs further enhanced temporal connections of seismograms at multiple scales. We used numerous earthquake recordings in Taiwan with diverse focal mechanisms, wide depth, and magnitude distributions, to train and test our model. Setting the picking errors within 0.1 s and predicted probability over 0.5, the AG with recurrent-residual convolution unit (ARRU) phase picker achieved the F1 score of 98.62% for P arrivals and 95.16% for S arrivals, and picking rates were 96.72% for P waves and 90.07% for S waves. The ARRU phase picker also shown a great generalization capability, when handling unseen data. When applied the model trained with Taiwan data to the southern California data, the ARRU phase picker shown no cognitive downgrade. Comparing with manual picks, the arrival times determined by the ARRU phase picker shown a higher consistency, which had been evaluated by a set of repeating earthquakes. The arrival picks with less human error could benefit studies, such as earthquake location and seismic tomography.

Beamform processing for sonic imaging using monopole and dipole sources

Geophysics ◽  
2020 ◽  
Vol 86 (1) ◽  
pp. D1-D14
Author(s):  
Nobuyasu Hirabayashi
Keyword(s):  
Signal To Noise Ratio ◽  
Random Noise ◽  
Dipole Source ◽  
P Waves ◽  
Arrival Times ◽  
Acoustic Reflection ◽  
Back Azimuth ◽  
Sonic Logging ◽  
Logging Tool ◽  
Dipole Sources

New processing techniques are presented that enhance event signals for sonic imaging using monopole and dipole sources. The techniques use the azimuthally spaced receivers of a sonic logging tool. Sonic imaging, which is also known as borehole acoustic reflection surveys, uses a sonic logging tool in a fluid-filled borehole to image geologic structures. Signals from monopole and dipole sources are reflected from geologic interfaces and recorded by arrays of receivers of the same tool. Because the amplitudes of the event signals are very weak compared with the direct waves, borehole modes, and noise, the event signals are often difficult to extract. To enhance the weak event signals, beamforming techniques were developed to stack the waveforms from azimuthally spaced receivers of the tool for given azimuthal directions. For the incident P-waves from the monopole source, phase arrival times for the azimuthal receivers are time shifted for stacking using properties of wave propagation in the borehole. For the incident SH-waves from the dipole source, the signs of waveforms for the receivers are changed for specified azimuths. When the waveforms are stacked for the back azimuth of the event signals, the signal-to-noise ratio of the event signals is significantly improved because the event signals are enhanced whereas the direct waves are relatively smeared, and random noise is canceled. Therefore, the stacked waveforms also provide accurate back azimuths of the incident waves.

Arrival-time picking uncertainty: Theoretical estimations and their application to microseismic data

Geophysics ◽  
2020 ◽  
Vol 85 (4) ◽  
pp. U65-U76
Author(s):  
Ivan Abakumov ◽  
Aurelian Roeser ◽  
Serge A. Shapiro
Keyword(s):  
Seismic Waves ◽  
Arrival Time ◽  
Signal To Noise Ratio ◽  
Estimation Error ◽  
Additive White Gaussian Noise ◽  
Accurate Determination ◽  
Microseismic Monitoring ◽  
High Signal ◽  
Arrival Times ◽  
Velocity Models

Traveltime-based methods depend on the accurate determination of the arrival times of seismic waves. They further benefit from information on the uncertainty with which the arrival times are determined. Among other applications, arrival-time uncertainties are used to weight data in inversion algorithms and to define the resolution of reconstructed velocity models. The most physically meaningful approaches for the estimation of arrival-time uncertainties are based on probabilistic formulations. The two approaches for the assessment of the lower bound of arrival-time uncertainties, the Cramér–Rao Bound (CRB) and the Ziv–Zakai Bound (ZZB), have been reviewed. The CRB determines the minimum-achievable estimation error under the assumption of a high signal-to-noise ratio (S/N) but underestimates said error for small S/N. The ZZB provides a better result for noisy data because it utilizes a priori information. The CRB and ZZB require knowledge of the spectral variance of the signal, which often is hard to determine in seismic experiments. Furthermore, both bounds assume additive white Gaussian noise (AWGN), which does not hold for seismic data. To overcome these problems, alternative expressions have been proposed, which yield comparable estimates as CRB and ZZB but are solely based on the S/N and the dominant period in the data. Moreover, a recipe to correct the S/N and account for the difference between the seismic noise and AWGN has been provided. For a case study of downhole microseismic monitoring, it is determined that the new expressions provide station-dependent arrival-time uncertainties, which are used as weights to improve source location uncertainties.

Seismic waves from the underground atomic explosion in Nevada

1959 ◽  
Vol 49 (1) ◽  
pp. 11-32
Author(s):  
B. F. Grossling
Keyword(s):  
Time Scale ◽  
Seismic Waves ◽  
Vertical Component ◽  
Short Wave ◽  
Apparent Velocity ◽  
Frequency Range ◽  
P Waves ◽  
Frequency Spectra ◽  
Arrival Times ◽  
Absolute Amplitude

>Abstract Seismic waves from the underground atomic explosion of September 19, 1957, were recorded for 45 minutes on multichannel magnetic tape at a point 25 miles north of Holbrook, Arizona, about 370 miles from their source. Twelve vertical-component seismometers were laid out in the form of an L 5,280 ft. by 1,600 ft. to permit determinations of apparent velocity and of direction of arrival. The frequency range recorded, about 6 to 40 cps, was higher than usual in earthquake seismology. During tape playback, supplementary filtering, gain adjustments, and changes in time scale served to improve the quality and legibility of the records. The playback seismograms reveal strikingly well not only to Pn and P* waves transmitted and refracted by the crust, but also many others as well. Some of these, as clearly indicated by their directions of arrival, did not originate from the explosion. We have attempted only an elementary interpretation, our main purpose being to make the data available to anyone who might be interested in them. The relatively short wave lengths of the recorded events may make them of unusual significance. In addition to arrival times, we made a few measurements of absolute amplitude and of frequency spectra.

scholarly journals Autocorrelation of the ground vibration recorded by the SEIS-InSight seismometer on Mars for imaging and monitoring applications

2021 ◽  
Author(s):  
Nicolas Compaire ◽  
Ludovic Margerin ◽  
Raphaël F. Garcia ◽  
Marie Calvet ◽  
Baptiste Pinot ◽  
...  
Keyword(s):  
Seismic Waves ◽  
Signal To Noise Ratio ◽  
Ground Vibration ◽  
Velocity Model ◽  
Temporal Variations ◽  
Seismic Events ◽  
Ambient Vibrations ◽  
Ground Response ◽  
Arrival Times ◽  
Single Station

<p>Since early February 2019, the SEIS seismometer deployed at the surface of Mars in the framework of the NASA-InSight mission has been continuously recording the ground motion at Elysium Planitia. In this work, we take advantage of this exceptional dataset to put constraints on the crustal properties of Mars using seismic interferometry (SI). This method use the seismic waves, either from background vibrations of the planet or from quakes, that are scattered in the medium in order to recover the ground response between two seismic sensors. Applying the principles of SI to the single-station configuration of SEIS, we compute, for each Sol (martian day) and each local hour, all the components of the time-domain autocorrelation tensor of random ambient vibrations in various frequency bands. A similar computation is performed on the diffuse waveforms generated by more than a hundred Marsquakes. For imaging application a careful signal-to-noise ratio analysis and an inter-comparison between the two datasets are applied. These analyses suggest that the reconstructed ground responses are most reliable in a relatively narrow frequency band around 2.4Hz, where an amplification of both ambient vibrations and seismic events is observed. The average Auto-Correlation Functions (ACFs) from both ambient vibrations and seismic events contain well identifiable seismic arrivals, that are very consistent between the two datasets. We interpret the vertical and horizontal ACFs as the ground reflection response below InSight for the compressional waves and the shear waves respectively. We propose a simple stratified velocity model of the crust, which is most compatible with the arrival times of the detected phases, as well as with previous seismological studies of the SEIS record. The hourly computation of the ACFs over one martian year also allows us to study the diurnal and seasonal variations of the reconstructed ground response with a technique call Passive Image Interferometry (PII). In this study we present measurements of the relative stretching coefficient between consecutive ACF waveforms and discuss the potential origins of the observed temporal variations.</p>

scholarly journals Fully probabilistic seismic source inversion – Part 1: Efficient parameterisation

2013 ◽  
Vol 5 (2) ◽  
pp. 1125-1162 ◽  
Author(s):  
S. C. Stähler ◽  
K. Sigloch
Keyword(s):  
Moment Tensor ◽  
Source Parameters ◽  
Practical Importance ◽  
Principal Component ◽  
Seismic Source ◽  
Empirical Orthogonal Functions ◽  
Time Function ◽  
Source Inversion ◽  
Source Time Function ◽  
Earthquake Parameters

Abstract. Seismic source inversion is a non-linear problem in seismology where not just the earthquake parameters themselves, but also estimates of their uncertainties are of great practical importance. Probabilistic source inversion (Bayesian inference) is very adapted to this challenge, provided that the parameter space can be chosen small enough to make Bayesian sampling computationally feasible. We propose a framework for PRobabilistic Inference of Source Mechanisms (PRISM) that parameterises and samples earthquake depth, moment tensor, and source time function efficiently by using information from previous non-Bayesian inversions. The source time function is expressed as a weighted sum of a small number of empirical orthogonal functions, which were derived from a catalogue of >1000 STFs by a principal component analysis. We use a likelihood model based on the cross-correlation misfit between observed and predicted waveforms. The resulting ensemble of solutions provides full uncertainty and covariance information for the source parameters, and permits to propagate these source uncertainties into travel time estimates used for seismic tomography. The computational effort is such that routine, global estimation of earthquake mechanisms and source time functions from teleseismic broadband waveforms is feasible.

A periodic seismic isolation foundation with an extremely broad low-frequency attenuation zone: Theoretical analysis and experimental verification

2021 ◽  
pp. 136943322110646
Author(s):  
Peng Zhou ◽  
Shui Wan ◽  
Xiao Wang ◽  
Yingbo Zhu ◽  
Muyun Huang
Keyword(s):  
Seismic Isolation ◽  
Seismic Waves ◽  
Periodic Structures ◽  
Finite Size ◽  
Low Frequency ◽  
Bridge Structure ◽  
Engineering Structures ◽  
P Waves ◽  
New Type ◽  
Dominant Frequencies

The attenuation zones (AZs) of periodic structures can be used for seismic isolation design. To cover the dominant frequencies of more seismic waves, this paper proposes a new type of periodic isolation foundation (PIF) with an extremely wide low-frequency AZ of 3.31 Hz–17.01 Hz composed of optimized unit A with a wide AZ and optimized unit B with a low-frequency AZ. The two kinds of optimized units are obtained by topology optimization on the smallest periodic unit with the coupled finite element-genetic algorithm (GA) methodology. The transmission spectra of shear waves and P-waves through the proposed PIF of finite size are calculated, and the results show that the AZ of the PIF is approximately the superposition of the AZs of the two kinds of optimized units. Additionally, shake tests on a scale PIF specimen are performed to verify the attenuation performance for elastic waves within the designed AZs. Furthermore, numerical simulations show that the acceleration responses of the bridge structure with the proposed PIF are attenuated significantly compared to those with a concrete foundation under the action of different seismic waves. Therefore, the newly proposed PIF is a promising option for the reduction of seismic effects in engineering structures.

Strong ground motion of mine tremors: Some implications for near-source ground motion parameters

1981 ◽  
Vol 71 (1) ◽  
pp. 295-319
Author(s):  
A. McGarr ◽  
R. W. E. Green ◽  
S. M. Spottiswoode
Keyword(s):  
Ground Motion ◽  
Source Parameters ◽  
Seismic Source ◽  
Shear Velocity ◽  
Recording System ◽  
Ground Acceleration ◽  
Hypocentral Distance ◽  
Earthquake Size ◽  
Mine Tremors ◽  
Stress Drops

abstract Ground acceleration was recorded at a depth of about 3 km in the East Rand Proprietary Mines, South Africa, for tremors with −1 ≦ ML ≦ 2.6 in the hypocentral distance range 50 m < R ≦ 1.6 km. The accelerograms typically had predominant frequencies of several hundred Hertz and peak accelerations, a, as high as 12 g. The peak accelerations show a dependence on magnitude, especially when expressed as dynamic shear-stress differences, defined as σ˜ = ρRa, where ρ is density. For the mine tremors, σ˜ varies from 2 to 500 bars and depends on magnitude according to log σ˜ = 1.40 + 0.38 · ML. Accelerograms for 12 events were digitized and then processed to determine velocity and, for seven events with especially good S/N, displacement and seismic source parameters. Peak ground velocities v ranged up to 6 cm/sec and show a well-defined dependence one earthquake size as measured by ML or by seismic moment, Mo. On the basis of regression fits to the mine data, with −0.76 ≦ ML ≦ 1.45, log Rv = 3.95 + 0.57 ML, where Rv is in cm2/sec, and log Rv = −4.68 + 0.49 log Mo. These regression lines agree excellently with the corresponding data for earthquakes of ML up to 6.4 or Mo to 1.4 × 1026 dyne-cm. At a given value of ML or Mo, a, at fixed R, shows considerably greater variation than v and appears to depend on the bandwidth of the recording system. The peak acceleration at small hypocentral distances is broadly consistent with ρRa = 1.14 Δτrofs/β, where Δτ is stress drop, ro is the source radius, β is shear velocity, and fs is the bandwidth of the recording system. The peak velocity data agree well with Rv = 0.57 βΔτro/μ, where μ is the modulus of rigidity; both expressions follow from Brune's model of the seismic source and were compared with data for events in the size range 5 × 1016 ≦ Mo ≦ 1.4 × 1026 dyne-cm. Measurements of the source parameters indicated that, as for earthquakes, the stress drops for the tremors range from 1 to 100 bars and show no consistent dependence on Mo down to Mo = 5 × 1016 dyne-cm.

Novel Method for Detecting Weak AE Signals Based on the Similarity of Time-Frequency Spectra

Geophysics ◽  
2021 ◽  
pp. 1-62
Author(s):  
Wencheng Yang ◽  
Xiao Li ◽  
Yibo Wang ◽  
Yue Zheng ◽  
Peng Guo
Keyword(s):  
Mathematical Morphology ◽  
Signal To Noise Ratio ◽  
Critical Role ◽  
Real Data ◽  
Monitoring Method ◽  
Frequency Spectra ◽  
Arrival Times ◽  
Time Frequency ◽  
Fracturing Process ◽  
Ae Signals

As a key monitoring method, the acoustic emission (AE) technique has played a critical role in characterizing the fracturing process of laboratory rock mechanics experiments. However, this method is limited by low signal-to-noise ratio (SNR) because of a large amount of noise in the measurement and environment and inaccurate AE location. Furthermore, it is difficult to distinguish two or more hits because their arrival times are very close when AE signals are mixed with the strong background noise. Thus, we propose a new method for detecting weak AE signals using the mathematical morphology character correlation of the time-frequency spectrum. The character in all hits of an AE event can be extracted from time-frequency spectra based on the theory of mathematical morphology. Through synthetic and real data experiments, we determined that this method accurately identifies weak AE signals. Compared with conventional methods, the proposed approach can detect AE signals with a lower SNR.

Sign in / Sign up

Export Citation Format

Share Document