scholarly journals An automatically generated high-resolution earthquake catalogue for the 2016–2017 Central Italy seismic sequence, including P and S phase arrival times

2020 ◽  
Author(s):  
D Spallarossa ◽  
M Cattaneo ◽  
D Scafidi ◽  
M Michele ◽  
L Chiaraluce ◽  
...  
Keyword(s):  
Real Time ◽  
Central Italy ◽  
Earthquake Catalogue ◽  
Hazard Evaluation ◽  
Depth Range ◽  
Data Set ◽  
Epicentral Area ◽  
Earthquake Parameters ◽  
Waveform Data ◽  
Arrival Times

Summary The 2016–17 central Italy earthquake sequence began with the first mainshock near the town of Amatrice on August 24 (MW 6.0), and was followed by two subsequent large events near Visso on October 26 (MW 5.9) and Norcia on October 30 (MW 6.5), plus a cluster of 4 events with MW > 5.0 within few hours on January 18, 2017. The affected area had been monitored before the sequence started by the permanent Italian National Seismic Network (RSNC), and was enhanced during the sequence by temporary stations deployed by the National Institute of Geophysics and Volcanology and the British Geological Survey. By the middle of September, there was a dense network of 155 stations, with a mean separation in the epicentral area of 6–10 km, comparable to the most likely earthquake depth range in the region. This network configuration was kept stable for an entire year, producing 2.5 TB of continuous waveform recordings. Here we describe how this data was used to develop a large and comprehensive earthquake catalogue using the Complete Automatic Seismic Processor (CASP) procedure. This procedure detected more than 450,000 events in the year following the first mainshock, and determined their phase arrival times through an advanced picker engine (RSNI-Picker2), producing a set of about 7 million P- and 10 million S-wave arrival times. These were then used to locate the events using a non-linear location (NLL) algorithm, a 1D velocity model calibrated for the area, and station corrections and then to compute their local magnitudes (ML). The procedure was validated by comparison of the derived data for phase picks and earthquake parameters with a handpicked reference catalogue (hereinafter referred to as ‘RefCat’). The automated procedure takes less than 12 hours on an Intel Core-i7 workstation to analyse the primary waveform data and to detect and locate 3000 events on the most seismically active day of the sequence. This proves the concept that the CASP algorithm can provide effectively real-time data for input into daily operational earthquake forecasts, The results show that there have been significant improvements compared to RefCat obtained in the same period using manual phase picks. The number of detected and located events is higher (from 84,401 to 450,000), the magnitude of completeness is lower (from ML 1.4 to 0.6), and also the number of phase picks is greater with an average number of 72 picked arrival for a ML = 1.4 compared with 30 phases for RefCat using manual phase picking. These propagate into formal uncertainties of ± 0.9km in epicentral location and ± 1.5km in depth for the enhanced catalogue for the vast majority of the events. Together, these provide a significant improvement in the resolution of fine structures such as local planar structures and clusters, in particular the identification of shallow events occurring in parts of the crust previously thought to be inactive. The lower completeness magnitude provides a rich data set for development and testing of analysis techniques of seismic sequences evolution, including real-time, operational monitoring of b-value, time-dependent hazard evaluation and aftershock forecasting.

Shallow structural setting of an active normal fault zone in the 30 October 2016 Mw 6.5 central Italy earthquake imaged through a multidisciplinary geophysical approach.

2020 ◽  
Author(s):  
Fabio Villani ◽  
Stefano Maraio ◽  
Pier Paolo Bruno ◽  
Lisa Serri ◽  
Vincenzo Sapia ◽  
...  
Keyword(s):  
Cluster Analysis ◽  
Fault Zone ◽  
Central Italy ◽  
Normal Fault ◽  
Alluvial Fan ◽  
Fault System ◽  
Depth Range ◽  
Inversion Algorithm ◽  
Data Set ◽  
Refraction Tomography

<p>We investigate the shallow structure of an active normal fault-zone that ruptured the surface during the 30 October 2016 Mw 6.5 Norcia earthquake (central Italy) using a multidisciplinary geophysical approach. The survey site is located in the Castelluccio basin, an intramontane Quaternary depression in the hangingwall of the SW-dipping Vettore-Bove fault system. The Norcia earthquake caused widespread surface faulting affecting also the Castelluccio basin, where the rupture trace follows the 2 km-long Valle delle Fonti fault (VF), displaying a ~3 m-high fault scarp due to cumulative surface slip of Holocene paleo-earthquakes. We explored the subsurface of the VF fault along a 2-D transect orthogonal to the coseismic rupture on recent alluvial fan deposits, combining very high-resolution seismic refraction tomography, multichannel analysis of surface waves (MASW), reflection seismology and electrical resistivity tomography (ERT).</p><p>We acquired the ERT profile using an array of 64 steel electrodes, 2 m-spaced. Apparent resistivity data were then modeled via a linearized inversion algorithm with smoothness constraints to recover the subsurface resistivity distribution. The seismic data were recorded by  a190 m-long single array centered on the surface rupture, using 96 vertical geophones 2 m-spaced and a 5 kg hammer source.</p><p>Input data for refraction tomography are ~9000 handpicked first arrival travel-times, inverted through a fully non-linear multi-scale algorithm based on a finite-difference Eikonal solver. The data for MASW were extracted from common receiver configurations with 24 geophones; the dispersion curves were inverted to generate several S-wave 1-D profiles, subsequently interpolated to generate a pseudo-2D Vs section. For reflection data, after a pre-processing flow, the picking of the maximum of semblance on CMP super-gathers was used to define a velocity model (VNMO) for CMP ensemble stack; the final stack velocity macro-model (VNMO) from the CMP processing was smoothed and used for post-stack depth conversion. We further processed Vp, Vs and resistivity models through the K-means algorithm, which performs a cluster analysis for the bivariate data set to individuate relationships between the two sets of variables. The result is an integrated model with a finite number of homogeneous clusters.</p><p>In the depth converted reflection section, the subsurface of the VF fault displays abrupt reflection truncations in the 5-60 m depth range suggesting a cumulative fault throw of ~30 m. Furthermore, another normal fault appears in the in the footwall. The reflection image points out alternating high-amplitude reflections that we interpret as a stack of alluvial sandy-gravels layers that thickens in the hangingwall of the VF fault. Resistivity, Vp and Vs models provide hints on the physical properties of the active fault zone, appearing as a moderately conductive (< 150 Ωm) elongated body with relatively high-Vp (~1500 m/s) and low-Vs (< 500 m/s). The Vp/Vs ratio > 3 and the Poisson’s coefficient > 0.4 in the fault zone suggest this is a granular nearly-saturated medium, probably related to the increase of permeability due to fracturing and shearing. The results from the K-means cluster analysis also identify a homogeneous cluster in correspondence of the saturated fault zone.</p>

A probabilistic, multi-parametric real-time earthquake location method

2020 ◽  
Author(s):  
Alessandro Caruso ◽  
Aldo Zollo ◽  
Simona Colombelli ◽  
Luca Elia ◽  
Grazia De Landro
Keyword(s):  
Real Time ◽  
Central Italy ◽  
Early Warning Systems ◽  
P Wave ◽  
Earthquake Location ◽  
Arrival Times ◽  
The Real ◽  
Location Parameters ◽  
Event Location ◽  
Wave Arrival

<p>For network-based Earthquake Early Warning Systems (EEWS), the real-time earthquake location is crucial for a correct estimation of event location/magnitude and therefore, for a reliable prediction of the potential expected shaking at the target sites in terms of predicted maximum ground shaking. Different approaches have been recently proposed for the real-time location which mainly use absolute (or differential) P-wave travel times at a set of minimum available stations or measurement of the initial P-wave arrival time (Elarms, Presto, Horiuchi), polarization (Eiserman and Bock) or amplitude and time (Yamada). In this work, we propose a new method which is able to exploit the continuous, real-time information available from both time, amplitude and polarization of initial P-wave signals acquired by dense three component arrays deployed in the source zones. The methodology we propose is an evolutionary and Bayesian probabilistic technique that combines three different observed parameters: 1) the differential arrival times of P-waves (which are computed using a 1D velocity model for the estimation of the theoretical arrival times); 2) the differential P-wave amplitudes in terms of P-wave peak velocity) [reference]  (which are computed using an existing P-peak motion prediction equation) and 3) the real-time estimation of back-azimuthal direction, measured shortly after the P-wave arrival. These three parameters are measured in real-time and are used as prior and conditional information to estimate the posterior probability of the event location parameters, e.g. the hypocenter coordinates and the origin time. The method is evolutive, since it updates the location parameters as new data are acquired by more and more distant stations as the P-wavefront propagates across the network. The output is a multi-dimensional Probability Density Function (PDF), which contains the complete information about the maximum likelihood parameter estimation with their uncertainty. The method is computationally efficient and optimized for running in real-time applications, where the earthquake location has to be retrieved in a very short time window (around 1 sec) after data acquisition. We tested the proposed strategy on a sequence of 29 earthquakes of the 2016-2017 central Italy seismic sequence acquired by the RAN (Rete Accelerometrica Nazionale) network with a magnitude range of 4.2-6.5. For the testing phase, we also simulated non-optimal conditions in terms of source-to-receiver geometry. Specifically, we tested the method  by ssimulating the case of “offshore” earthquakes recorded by a coastal network and in the case of a linear “barrier-type” geometry of the network. Our approach turned out to be suitable to work in condition of a sparse network, with a limited number of nodes and poor azimuthal coverage. In most of the cases, reliable location errors, less than 10 km, are achieved within few seconds from the first recorded P wave. As compared to other classical location techniques (i.e RTLOC in PRESTo) our approach shows an improvement of the solutions, especially for the first instants (2 seconds after the first P-wave arrival at network) when a poor number of stations (less than 4) is available.</p>

First-arrival picking with a U-net convolutional network

Geophysics ◽  
2019 ◽  
Vol 84 (6) ◽  
pp. U45-U57 ◽  
Author(s):  
Lianlian Hu ◽  
Xiaodong Zheng ◽  
Yanting Duan ◽  
Xinfei Yan ◽  
Ying Hu ◽  
...  
Keyword(s):  
Seismic Data ◽  
Synthetic Data ◽  
Training Data ◽  
Convolutional Network ◽  
Data Set ◽  
Waveform Data ◽  
Near Surface ◽  
Arrival Times ◽  
Seismic Waveform ◽  
First Arrival

In exploration geophysics, the first arrivals on data acquired under complicated near-surface conditions are often characterized by significant static corrections, weak energy, low signal-to-noise ratio, and dramatic phase change, and they are difficult to pick accurately with traditional automatic procedures. We have approached this problem by using a U-shaped fully convolutional network (U-net) to first-arrival picking, which is formulated as a binary segmentation problem. U-net has the ability to recognize inherent patterns of the first arrivals by combining attributes of arrivals in space and time on data of varying quality. An effective workflow based on U-net is presented for fast and accurate picking. A set of seismic waveform data and their corresponding first-arrival times are used to train the network in a supervised learning approach, then the trained model is used to detect the first arrivals for other seismic data. Our method is applied on one synthetic data set and three field data sets of low quality to identify the first arrivals. Results indicate that U-net only needs a few annotated samples for learning and is able to efficiently detect first-arrival times with high precision on complicated seismic data from a large survey. With the increasing training data of various first arrivals, a trained U-net has the potential to directly identify the first arrivals on new seismic data.

scholarly journals Enhanced timing accuracy for Cluster data

2013 ◽  
Vol 2 (2) ◽  
pp. 323-328
Author(s):  
K. H. Yearby ◽  
S. N. Walker ◽  
M. A. Balikhin
Keyword(s):  
Real Time ◽  
Phase Error ◽  
European Space Agency ◽  
Wide Band ◽  
Time Data ◽  
Timing Error ◽  
Data Set ◽  
Waveform Data ◽  
Timing Accuracy ◽  
Cluster Data

Abstract. The standard timing accuracy for the Cluster mission is ±2 ms. However for inter-spacecraft comparisons of waveform data, a much higher accuracy is needed – for example a timing error of 1 ms results in a phase error of 65° for a signal at 180 Hz. Most Cluster data are recorded on an onboard solid state recorder and time-stamped using an onboard clock which is calibrated to coordinated universal time (UTC). Until recently, the error of this onboard clock was allowed to increase to the ±2 ms limit before a new calibration was applied. However, the timing error for real-time data is estimated to be only ~11 μs, so these data may be used to prepare a time correction data set which allows the standard timing accuracy to be improved considerably. This paper describes the details of the preparation and validation of this data set. Two independent source data sets are used: telemetry to European Space Agency (ESA) ground stations supporting the main operations of the Cluster spacecraft, and the real-time telemetry to the NASA Deep Space Network (DSN) stations supporting the Wide-Band Plasma Wave Investigation.

An accurate relocation of the 2017 Ms 7.0 Sichuan Jiuzhaigou earthquake sequence and the seismicity analysis

2020 ◽  
Author(s):  
Xiangwei Yu ◽  
Qian Song ◽  
Shanquan Deng
Keyword(s):  
P Wave ◽  
Depth Range ◽  
Jiuzhaigou Earthquake ◽  
Waveform Data ◽  
Arrival Times ◽  
Location Method ◽  
Seismic Waveform ◽  
Waveform Cross Correlation ◽  
Seismic Location ◽  
Seismicity Analysis

<p>The 2017 Ms 7.0 Sichuan Jiuzhaigou earthquake occurred at the intersection of the Tazang, Minjiang, and Huya faults on the eastern margin of the Tibetan Plateau. Since it occurred on an unmarked blind fault, it is still a controversial issue whether the fault, which triggered the earthquake, was the extension of the East Kunlun fault or the northern branch of the Huya fault. Therefore, the accurate source location is of great significance for studying the deep distribution of seismogenic faults and seismicity analysis.</p><p>We have not only collected seismic phase arrival data recorded by 24 permanent stations and 6 temporary stations, but also picked up the seismic waveform data recorded by partial permanent stations in this study. Using absolute seismic location method and relative seismic location method, we relocated the earthquake events with magnitude greater than or equal to 1.0 occurred in the Jiuzhaigou area from August to December 2017. In order to ensure reliable data quality, we selected 23422 P-wave absolute arrival times, 24734 S-wave absolute arrival times and 124519 high quality P-waveform cross correlation data of 3449 earthquake events for relocation research.</p><p>The mean value of root mean square residuals of travel time of all earthquakes decrease from 0.21s to 0.08s after relocation. The average location errors in the E-W, N-S, and vertical directions are 0.11km, 0.12km, and 0.16km, respectively. Ninety-nine percent of the earthquake events are distributed in the depth range of 1-25 km, and the dominant distribution range is 5-15 km. The result shows that the earthquakes are distributed along the strike of northwest and southeast, and the Jiuzhaigou mainshock divided these events into two clusters: northwest and southeast. From the parallel strike section, we conclude that the depth of the northwest seismic cluster is shallow with the depth range of 2-15 km, and the depth of the southeast seismic cluster is deeper with the depth range of 6-18 km. Moreover, the number of aftershocks in the northwest cluster is greater than that in the southeast cluster, but after an M 4.9 aftershock occurred in the northwest cluster on the ninety-first day after the Jiuzhaigou mainshock, the number of aftershocks in the northwest cluster began to decrease. The result provides a basis for studying the seismogenic background and seismicity of the Jiuzhaigou earthquake.</p>

scholarly journals Near-surface real-time seismic imaging using parsimonious interferometry

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Sherif M. Hanafy ◽  
Hussein Hoteit ◽  
Jing Li ◽  
Gerard T. Schuster
Keyword(s):  
Fluid Flow ◽  
Real Time ◽  
Temporal Resolution ◽  
Seismic Imaging ◽  
Time Lapse ◽  
Rock Properties ◽  
Recording Time ◽  
Near Surface ◽  
Arrival Times ◽  
Order Of Magnitude

AbstractResults are presented for real-time seismic imaging of subsurface fluid flow by parsimonious refraction and surface-wave interferometry. Each subsurface velocity image inverted from time-lapse seismic data only requires several minutes of recording time, which is less than the time-scale of the fluid-induced changes in the rock properties. In this sense this is real-time imaging. The images are P-velocity tomograms inverted from the first-arrival times and the S-velocity tomograms inverted from dispersion curves. Compared to conventional seismic imaging, parsimonious interferometry reduces the recording time and increases the temporal resolution of time-lapse seismic images by more than an order-of-magnitude. In our seismic experiment, we recorded 90 sparse data sets over 4.5 h while injecting 12-tons of water into a sand dune. Results show that the percolation of water is mostly along layered boundaries down to a depth of a few meters, which is consistent with our 3D computational fluid flow simulations and laboratory experiments. The significance of parsimonious interferometry is that it provides more than an order-of-magnitude increase of temporal resolution in time-lapse seismic imaging. We believe that real-time seismic imaging will have important applications for non-destructive characterization in environmental, biomedical, and subsurface imaging.

MobileNet-Yolo Based wildlife detection model: A case study in yunnan tongbiguan nature reserve, China

2021 ◽  
pp. 1-11
Author(s):  
Tingting Zhao ◽  
Xiaoli Yi ◽  
Zhiyong Zeng ◽  
Tao Feng
Keyword(s):  
Real Time ◽  
Nature Reserve ◽  
Complex Model ◽  
Training Data ◽  
Mean Average Precision ◽  
Detection Accuracy ◽  
High Definition ◽  
Average Precision ◽  
Data Set ◽  
Real Time Detection

YTNR (Yunnan Tongbiguan Nature Reserve) is located in the westernmost part of China’s tropical regions and is the only area in China with the tropical biota of the Irrawaddy River system. The reserve has abundant tropical flora and fauna resources. In order to realize the real-time detection of wild animals in this area, this paper proposes an improved YOLO (You only look once) network. The original YOLO model can achieve higher detection accuracy, but due to the complex model structure, it cannot achieve a faster detection speed on the CPU detection platform. Therefore, the lightweight network MobileNet is introduced to replace the backbone feature extraction network in YOLO, which realizes real-time detection on the CPU platform. In response to the difficulty in collecting wild animal image data, the research team deployed 50 high-definition cameras in the study area and conducted continuous observations for more than 1,000 hours. In the end, this research uses 1410 images of wildlife collected in the field and 1577 wildlife images from the internet to construct a research data set combined with the manual annotation of domain experts. At the same time, transfer learning is introduced to solve the problem of insufficient training data and the network is difficult to fit. The experimental results show that our model trained on a training set containing 2419 animal images has a mean average precision of 93.6% and an FPS (Frame Per Second) of 3.8 under the CPU. Compared with YOLO, the mean average precision is increased by 7.7%, and the FPS value is increased by 3.

Capturing, Preserving, and Digitizing Legacy Seismic Data from the Montserrat Volcano Observatory Analog Seismic Network, July 1995–December 2004

2020 ◽  
Vol 91 (4) ◽  
pp. 2127-2140 ◽  
Author(s):  
Glenn Thompson ◽  
John A. Power ◽  
Jochen Braunmiller ◽  
Andrew B. Lockhart ◽  
Lloyd Lynch ◽  
...  
Keyword(s):  
Real Time ◽  
Seismic Data ◽  
Original Data ◽  
South Florida ◽  
Seismic Network ◽  
Quality Data ◽  
Spectral Amplitude ◽  
Assistance Program ◽  
Waveform Data ◽  
Amplitude Measurement

Abstract An eruption of the Soufrière Hills Volcano (SHV) on the eastern Caribbean island of Montserrat began on 18 July 1995 and continued until February 2010. Within nine days of the eruption onset, an existing four-station analog seismic network (ASN) was expanded to 10 sites. Telemetered data from this network were recorded, processed, and archived locally using a system developed by scientists from the U.S. Geological Survey (USGS) Volcano Disaster Assistance Program (VDAP). In October 1996, a digital seismic network (DSN) was deployed with the ability to capture larger amplitude signals across a broader frequency range. These two networks operated in parallel until December 2004, with separate telemetry and acquisition systems (analysis systems were merged in March 2001). Although the DSN provided better quality data for research, the ASN featured superior real-time monitoring tools and captured valuable data including the only seismic data from the first 15 months of the eruption. These successes of the ASN have been rather overlooked. This article documents the evolution of the ASN, the VDAP system, the original data captured, and the recovery and conversion of more than 230,000 seismic events from legacy SUDS, Hypo71, and Seislog formats into Seisan database with waveform data in miniSEED format. No digital catalog existed for these events, but students at the University of South Florida have classified two-thirds of the 40,000 events that were captured between July 1995 and October 1996. Locations and magnitudes were recovered for ∼10,000 of these events. Real-time seismic amplitude measurement, seismic spectral amplitude measurement, and tiltmeter data were also captured. The result is that the ASN seismic dataset is now more discoverable, accessible, and reusable, in accordance with FAIR data principles. These efforts could catalyze new research on the 1995–2010 SHV eruption. Furthermore, many observatories have data in these same legacy data formats and might benefit from procedures and codes documented here.

scholarly journals Research on Real-Time Multiple Single Garbage Classification Based on Convolutional Neural Network

2020 ◽  
Vol 2020 ◽  
pp. 1-6
Author(s):  
Jian-ye Yuan ◽  
Xin-yuan Nan ◽  
Cheng-rong Li ◽  
Le-le Sun
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Real Time ◽  
Environmental Science ◽  
Fine Tuning ◽  
Classification Model ◽  
Classification Models ◽  
Data Set ◽  
Classification Tasks ◽  
Accuracy Rates

Considering that the garbage classification is urgent, a 23-layer convolutional neural network (CNN) model is designed in this paper, with the emphasis on the real-time garbage classification, to solve the low accuracy of garbage classification and recycling and difficulty in manual recycling. Firstly, the depthwise separable convolution was used to reduce the Params of the model. Then, the attention mechanism was used to improve the accuracy of the garbage classification model. Finally, the model fine-tuning method was used to further improve the performance of the garbage classification model. Besides, we compared the model with classic image classification models including AlexNet, VGG16, and ResNet18 and lightweight classification models including MobileNetV2 and SuffleNetV2 and found that the model GAF_dense has a higher accuracy rate, fewer Params, and FLOPs. To further check the performance of the model, we tested the CIFAR-10 data set and found the accuracy rates of the model (GAF_dense) are 0.018 and 0.03 higher than ResNet18 and SufflenetV2, respectively. In the ImageNet data set, the accuracy rates of the model (GAF_dense) are 0.225 and 0.146 higher than Resnet18 and SufflenetV2, respectively. Therefore, the garbage classification model proposed in this paper is suitable for garbage classification and other classification tasks to protect the ecological environment, which can be applied to classification tasks such as environmental science, children’s education, and environmental protection.

Sign in / Sign up

Export Citation Format

Share Document