PhaseNet and EQTransformer are two state-of-the-art earthquake detection methods that have been increasingly applied worldwide.To evaluate the generaliz-ation ability of the two models and provide insights for the dev...PhaseNet and EQTransformer are two state-of-the-art earthquake detection methods that have been increasingly applied worldwide.To evaluate the generaliz-ation ability of the two models and provide insights for the development of new models,this study took the sequences of the Yunnan Yangbi M6.4 earthquake and Qinghai Maduo M7.4 earthquake as examples to compare the earthquake detection effects of the two abovementioned models as well as their abilities to process dense seismic sequences.It has been demonstrated from the corresponding research that due to the differences in seismic waveforms found in different geographical regions,the picking performance is reduced when the two models are applied directly to the detection of the Yangbi and Maduo earthquakes.PhaseNet has a higher recall than EQTransformer,but the recall of both models is reduced by 13%-56%when compared with the results rep-orted in the original papers.The analysis results indicate that neural networks with deeper layers and complex structures may not necessarily enhance earthquake detection perfor-mance.In designing earthquake detection models,attention should be paid to not only the balance of depth,width,and architecture but also to the quality and quantity of the training datasets.In addition,noise datasets should be incorporated during training.According to the continuous waveforms detected 21 days before the Yangbi and Maduo earthquakes,the Yangbi earthquake exhibited foreshock,while the Maduo earthquake showed no foreshock activity,indicating that the two earthquakes’nucleation processes were different.展开更多
Earthquake detection and location are essential in earthquake studies,which generally consists of two main classes:waveform-based and pick-based methods.To evaluate the ability of two different methods,a graphicsproce...Earthquake detection and location are essential in earthquake studies,which generally consists of two main classes:waveform-based and pick-based methods.To evaluate the ability of two different methods,a graphicsprocessing-unit-based Match&Locate(GPU-M&L)method and a rapid earthquake association and location(REAL)method are applied to continuous seismic data recorded by 24 digital seismic stations from Jiangsu Seismic Network during 2013 for comparison.GPU-M&L is one of waveform-based methods by waveform cross-correlations while REAL is one of pick-based method to associate arrivals of different seismic phases and locate events through counting the number of P and S picks and travel time residuals.Twenty-six templates are selected from the Jiangsu Seismic Network local catalog by using the GPU-M&L.The number of newly detected and located events is about 2.8 times more than those listed in the local catalog.We both utilize a deep-neural-network-based arrival-time picking method called PhaseNet and a shortterm/long-term average(STA/LTA)trigger algorithm for seismic phase detection and picking by applying the REAL.We then refine seismic locations using a least-squares location method(VELEST)and a high-precision relative location method(hypoDD).By applying STA/LTA and PhaseNet,1006 and 1893 events are associated and located,respectively.The newly detected events are mainly clustered and show steeply dipping fault planes.By analyzing the performance of these methods based on long-term continuous seismic data,the detected catalogs by the GPU-M&L and REAL show that the magnitudes of completeness are 1.4 and 0.8,respectively,which are smaller than 2.6 given by the local catalog.Although REAL provides improvement compared with GPU-M&L,REAL is highly dependent on phase detection and picking which is strongly affected by signal-noise ratio(SNR).Stations at southeast of the study region with low SNR may lead to few detections in the same area.展开更多
This article extends a signal-based approach formerly proposed by the authors, which utilizes the fractal dimension of time frequency feature (FDTFF) of displacements, for earthquake damage detection of moment resis...This article extends a signal-based approach formerly proposed by the authors, which utilizes the fractal dimension of time frequency feature (FDTFF) of displacements, for earthquake damage detection of moment resist frame (MRF), and validates the approach with shaking table tests. The time frequency feature (TFF) of the relative displacement at measured story is defined as the real part of the coefficients of the analytical wavelet transform. The fractal dimension (FD) is to quantify the TFF within the fundamental frequency band using box counting method. It is verified that the FDTFFs at all stories of the linear MRF are identical with the help of static condensation method and modal superposition principle, while the FDTFFs at the stories with localized nonlinearities due to damage will be different from those at the stories without nonlinearities using the reverse-path methodology. By comparing the FDTFFs of displacements at measured stories in a structure, the damage-induced nonlinearity of the structure under strong ground motion can be detected and localized. Finally shaking table experiments on a 1:8 scale sixteen-story three-bay steel MRF with added frictional dampers, which generate local nonlinearities, are conducted to validate the approach.展开更多
There was an evident increase in the number of earthquakes in the Xinfengjiang Reservoir from June to July 2014 after the landing of Typhoon Hagibis.To understand the spatial and temporal evolution of this microseismi...There was an evident increase in the number of earthquakes in the Xinfengjiang Reservoir from June to July 2014 after the landing of Typhoon Hagibis.To understand the spatial and temporal evolution of this microseismicity,we built a high-precision earthquake catalog for 2014 and relocated 2275 events using recently developed methods for event picking and catalog construction.Seismicity occurred in the southeastern part of the reservoir,with the preferred fault plane orientation aligned along the Heyuan Fault.The total seismic energy peaked when the typhoon passed through the reservoir,and seismicity correlated with typhoon energy.In contrast,a limited seismic response was observed during the later Typhoon Rammasun.Combining data regarding the water level in the Xinfengjiang Reservoir and seismicity frequency changes in the Taiwan region during these two typhoon events,we suggest that typhoon activity may increase microseism energy by impacting fault stability around the Xinfengjiang Reservoir.Whether a fault can be activated also depends on how close the stress accumulation is to its failure point.展开更多
We present a high-resolution seismic catalog for the 2021 M_(S)6.4/M_(W)6.1 Yangbi sequence.The catalog has a time range of 2021-05-01 to 2021-05-28,and contains~8,000 well located events.It captures the features of t...We present a high-resolution seismic catalog for the 2021 M_(S)6.4/M_(W)6.1 Yangbi sequence.The catalog has a time range of 2021-05-01 to 2021-05-28,and contains~8,000 well located events.It captures the features of the whole foreshock sequence and the early aftershocks.We designed a detection strategy incorporating both an artificial intelligent(AI)picker and a matched filter algorithm.Here,we adopt a hybrid AI method incorporating convolutional and recurrent neural network(CNN&RNN)for event detection and phase picking respectively(i.e.CERP),a light-weight AI picker that can be trained with small volume of data.CERP is first trained with detections from a STA/LTA and Kurtosis-based method called PAL,and then construct a rather complete template set of~4,000 events.Finally,the matched filter algorithm MESS augments the initial detections and measures differential travel times with cross-correlation,which finally results in precise relocation.This process gives 9,026 detections,among which 7,943 events can be well relocated.The catalog shows as expected power-law distribution of frequency magnitude and reveals detailed pattern of seismicity evolution.The main features are:(1)the foreshock sequence images simple fault geometry with consistent strike,but also show a variable event depth along strike;(2)the mainshock ruptures the same fault of the foreshock sequence and activate conjugate faults further to the southeast;(3)complex seismicity are developed in the post-seismic period,indicating complex triggering mechanisms.Thus,our catalog provides a reliable basis for further investigations,such as b-value studies,rupture process,and triggering relations.展开更多
The April 25, 2015 Mw7.8 Nepal earthquake was successfully recorded by Crustal Movement Observation Network of China (CMONOC) and Nepal Geodetic Array (NGA). We processed the high-rate GPS data (1 Hz and 5 Hz) b...The April 25, 2015 Mw7.8 Nepal earthquake was successfully recorded by Crustal Movement Observation Network of China (CMONOC) and Nepal Geodetic Array (NGA). We processed the high-rate GPS data (1 Hz and 5 Hz) by using relative kinematic positioning and derived dynamic ground motions caused by this large earthquake. The dynamic displacements time series clearly indicated the displacement amplitude of each station was related to the rupture directivity. The stations which located in the di- rection of rupture propagation had larger displacement amplitudes than others. Also dynamic ground displacement exceeding 5 cm was detected by the GPS station that was 2000 km away from the epicenter. Permanent coseismic displacements were resolved from the near-field high-rate GPS stations with wavelet decomposition-reconstruction method and P-wave arrivals were also detected with S transform method. The results of this study can be used for earthquake rupture process and Earthquake Early Warning studies.展开更多
In recent years,artificial intelligence technology has exhibited great potential in seismic signal recognition,setting off a new wave of research.Vast amounts of high-quality labeled data are required to develop and a...In recent years,artificial intelligence technology has exhibited great potential in seismic signal recognition,setting off a new wave of research.Vast amounts of high-quality labeled data are required to develop and apply artificial intelligence in seismology research.In this study,based on the 2013–2020 seismic cataloging reports of the China Earthquake Networks Center,we constructed an artificial intelligence seismological training dataset(“DiTing”)with the largest known total time length.Data were recorded using broadband and short-period seismometers.The obtained dataset included 2,734,748 threecomponent waveform traces from 787,010 regional seismic events,the corresponding P-and S-phase arrival time labels,and 641,025 P-wave first-motion polarity labels.All waveforms were sampled at 50 Hz and cut to a time length of 180 s starting from a random number of seconds before the occurrence of an earthquake.Each three-component waveform contained a considerable amount of descriptive information,such as the epicentral distance,back azimuth,and signal-to-noise ratios.The magnitudes of seismic events,epicentral distance,signal-to-noise ratio of P-wave data,and signal-to-noise ratio of S-wave data ranged from 0 to 7.7,0 to 330 km,–0.05 to 5.31 dB,and–0.05 to 4.73 dB,respectively.The dataset compiled in this study can serve as a high-quality benchmark for machine learning model development and data-driven seismological research on earthquake detection,seismic phase picking,first-motion polarity determination,earthquake magnitude prediction,early warning systems,and strong ground-motion prediction.Such research will further promote the development and application of artificial intelligence in seismology.展开更多
Seismic phase pickers based on deep neural networks have been extensively used recently,demonstrating their advantages on both performance and efficiency.However,these pickers are trained with and applied to different...Seismic phase pickers based on deep neural networks have been extensively used recently,demonstrating their advantages on both performance and efficiency.However,these pickers are trained with and applied to different data.A comprehensive benchmark based on a single dataset is therefore lacking.Here,using the recently released DiTing dataset,we analyzed performances of seven phase pickers with different network structures,the efficiencies are also evaluated using both CPU and GPU devices.Evaluations based on F1-scores reveal that the recurrent neural network(RNN)and EQTransformer exhibit the best performance,likely owing to their large receptive fields.Similar performances are observed among PhaseNet(UNet),UNet++,and the lightweight phase picking network(LPPN).However,the LPPN models are the most efficient.The RNN and EQTransformer have similar speeds,which are slower than those of the LPPN and PhaseNet.UNet++requires the most computational effort among the pickers.As all of the pickers perform well after being trained with a large-scale dataset,users may choose the one suitable for their applications.For beginners,we provide a tutorial on training and validating the pickers using the DiTing dataset.We also provide two sets of models trained using datasets with both 50 Hz and 100 Hz sampling rates for direct application by end-users.All of our models are open-source and publicly accessible.展开更多
The pre-earthquake ionospheric anomalies in Wenchuan,China(21°-41°N,93°-113°E)are studied and analyzed using the summer nighttime data from 2005 to 2008 measured by DEMETER(Detection of Electro-Mag...The pre-earthquake ionospheric anomalies in Wenchuan,China(21°-41°N,93°-113°E)are studied and analyzed using the summer nighttime data from 2005 to 2008 measured by DEMETER(Detection of Electro-Magnetic Emission Transmitted from Earthquake Regions)satellite detectors ICE(Internet Communications Engine),IAP(In Application Programming),and ISL(Interior Switching Link).In this paper,we take the 12 May 2008 Wenchuan earthquake as an example,use the spatial gridding method to construct the background field over the epicenter,analyze the background characteristics of very low frequency(VLF)electric field components,low-energy particle parameters,and plasma parameters,and define the perturbation intensity index of each parameter before the earthquake to extract each parameter anomaly in both space and time dimensions.The results show that the background values of some ionospheric parameters in the Wenchuan area are related to spatial distribution.Moreover,anomalous enhancement of low-frequency electric field power spectral density,H+concentration,He+concentration and ion concentration with different intensities and anomalous weakening of ion temperature were extracted in the fifteen days before the Wenchuan earthquake.After filtering the data to exclude external interference,such as solar activity,this paper concludes that there is some connection between these anomalies and the Wenchuan earthquake.展开更多
基金funded by the National Key R&D Program of China(No.2021YFC3000702)the National Natural Science Foundation of China(No.41774067)the Fundamental Research Funds for the Institute of Geophysics,China Earthquake Administration(Nos.DQ JB21Z05,DQJB20X07).
文摘PhaseNet and EQTransformer are two state-of-the-art earthquake detection methods that have been increasingly applied worldwide.To evaluate the generaliz-ation ability of the two models and provide insights for the development of new models,this study took the sequences of the Yunnan Yangbi M6.4 earthquake and Qinghai Maduo M7.4 earthquake as examples to compare the earthquake detection effects of the two abovementioned models as well as their abilities to process dense seismic sequences.It has been demonstrated from the corresponding research that due to the differences in seismic waveforms found in different geographical regions,the picking performance is reduced when the two models are applied directly to the detection of the Yangbi and Maduo earthquakes.PhaseNet has a higher recall than EQTransformer,but the recall of both models is reduced by 13%-56%when compared with the results rep-orted in the original papers.The analysis results indicate that neural networks with deeper layers and complex structures may not necessarily enhance earthquake detection perfor-mance.In designing earthquake detection models,attention should be paid to not only the balance of depth,width,and architecture but also to the quality and quantity of the training datasets.In addition,noise datasets should be incorporated during training.According to the continuous waveforms detected 21 days before the Yangbi and Maduo earthquakes,the Yangbi earthquake exhibited foreshock,while the Maduo earthquake showed no foreshock activity,indicating that the two earthquakes’nucleation processes were different.
基金This research is co-supported by National Key R&D Program of China(No.2017YFC1500402)National Natural Science Foundation of China(Nos.41874063 and U1939203)Shanghai Sheshan National Geophysical Observatory(No.2020K02)。
文摘Earthquake detection and location are essential in earthquake studies,which generally consists of two main classes:waveform-based and pick-based methods.To evaluate the ability of two different methods,a graphicsprocessing-unit-based Match&Locate(GPU-M&L)method and a rapid earthquake association and location(REAL)method are applied to continuous seismic data recorded by 24 digital seismic stations from Jiangsu Seismic Network during 2013 for comparison.GPU-M&L is one of waveform-based methods by waveform cross-correlations while REAL is one of pick-based method to associate arrivals of different seismic phases and locate events through counting the number of P and S picks and travel time residuals.Twenty-six templates are selected from the Jiangsu Seismic Network local catalog by using the GPU-M&L.The number of newly detected and located events is about 2.8 times more than those listed in the local catalog.We both utilize a deep-neural-network-based arrival-time picking method called PhaseNet and a shortterm/long-term average(STA/LTA)trigger algorithm for seismic phase detection and picking by applying the REAL.We then refine seismic locations using a least-squares location method(VELEST)and a high-precision relative location method(hypoDD).By applying STA/LTA and PhaseNet,1006 and 1893 events are associated and located,respectively.The newly detected events are mainly clustered and show steeply dipping fault planes.By analyzing the performance of these methods based on long-term continuous seismic data,the detected catalogs by the GPU-M&L and REAL show that the magnitudes of completeness are 1.4 and 0.8,respectively,which are smaller than 2.6 given by the local catalog.Although REAL provides improvement compared with GPU-M&L,REAL is highly dependent on phase detection and picking which is strongly affected by signal-noise ratio(SNR).Stations at southeast of the study region with low SNR may lead to few detections in the same area.
基金National Natural Science Foundation under Grant No.51161120359Ministry of Education under Grant No.20112302110050Special Fund for Earthquake Scientific Research in the Public Interest under Grant No.201308003
文摘This article extends a signal-based approach formerly proposed by the authors, which utilizes the fractal dimension of time frequency feature (FDTFF) of displacements, for earthquake damage detection of moment resist frame (MRF), and validates the approach with shaking table tests. The time frequency feature (TFF) of the relative displacement at measured story is defined as the real part of the coefficients of the analytical wavelet transform. The fractal dimension (FD) is to quantify the TFF within the fundamental frequency band using box counting method. It is verified that the FDTFFs at all stories of the linear MRF are identical with the help of static condensation method and modal superposition principle, while the FDTFFs at the stories with localized nonlinearities due to damage will be different from those at the stories without nonlinearities using the reverse-path methodology. By comparing the FDTFFs of displacements at measured stories in a structure, the damage-induced nonlinearity of the structure under strong ground motion can be detected and localized. Finally shaking table experiments on a 1:8 scale sixteen-story three-bay steel MRF with added frictional dampers, which generate local nonlinearities, are conducted to validate the approach.
基金Strategic Priority Research Program(B)of the Chinese Academy of Sciences(No.XDB42020304)National Natural Science Foundation of China(No.42074059).
文摘There was an evident increase in the number of earthquakes in the Xinfengjiang Reservoir from June to July 2014 after the landing of Typhoon Hagibis.To understand the spatial and temporal evolution of this microseismicity,we built a high-precision earthquake catalog for 2014 and relocated 2275 events using recently developed methods for event picking and catalog construction.Seismicity occurred in the southeastern part of the reservoir,with the preferred fault plane orientation aligned along the Heyuan Fault.The total seismic energy peaked when the typhoon passed through the reservoir,and seismicity correlated with typhoon energy.In contrast,a limited seismic response was observed during the later Typhoon Rammasun.Combining data regarding the water level in the Xinfengjiang Reservoir and seismicity frequency changes in the Taiwan region during these two typhoon events,we suggest that typhoon activity may increase microseism energy by impacting fault stability around the Xinfengjiang Reservoir.Whether a fault can be activated also depends on how close the stress accumulation is to its failure point.
基金supported jointly by National Key R&D Program of China(No.2018YFC1503400)National Natural Science Foundation of China projects(Nos.41774067,U2039204,and 42074046)+2 种基金Science for Earthquake Resilience(No.XH20082Y)US National Science Foundation(No.1941719)University of California at Riverside.
文摘We present a high-resolution seismic catalog for the 2021 M_(S)6.4/M_(W)6.1 Yangbi sequence.The catalog has a time range of 2021-05-01 to 2021-05-28,and contains~8,000 well located events.It captures the features of the whole foreshock sequence and the early aftershocks.We designed a detection strategy incorporating both an artificial intelligent(AI)picker and a matched filter algorithm.Here,we adopt a hybrid AI method incorporating convolutional and recurrent neural network(CNN&RNN)for event detection and phase picking respectively(i.e.CERP),a light-weight AI picker that can be trained with small volume of data.CERP is first trained with detections from a STA/LTA and Kurtosis-based method called PAL,and then construct a rather complete template set of~4,000 events.Finally,the matched filter algorithm MESS augments the initial detections and measures differential travel times with cross-correlation,which finally results in precise relocation.This process gives 9,026 detections,among which 7,943 events can be well relocated.The catalog shows as expected power-law distribution of frequency magnitude and reveals detailed pattern of seismicity evolution.The main features are:(1)the foreshock sequence images simple fault geometry with consistent strike,but also show a variable event depth along strike;(2)the mainshock ruptures the same fault of the foreshock sequence and activate conjugate faults further to the southeast;(3)complex seismicity are developed in the post-seismic period,indicating complex triggering mechanisms.Thus,our catalog provides a reliable basis for further investigations,such as b-value studies,rupture process,and triggering relations.
基金supported by Director Foundation of Institute of Seismology,China Earthquake Administration(IS201426142)National Natural Science Foundation of China(41541029,41574017, 41274027)+1 种基金Natural Science Foundation of HuBei Province (2015CFB642)provided by Crustal Movement Observation Network of China(CMONOC) and UNAVCO
文摘The April 25, 2015 Mw7.8 Nepal earthquake was successfully recorded by Crustal Movement Observation Network of China (CMONOC) and Nepal Geodetic Array (NGA). We processed the high-rate GPS data (1 Hz and 5 Hz) by using relative kinematic positioning and derived dynamic ground motions caused by this large earthquake. The dynamic displacements time series clearly indicated the displacement amplitude of each station was related to the rupture directivity. The stations which located in the di- rection of rupture propagation had larger displacement amplitudes than others. Also dynamic ground displacement exceeding 5 cm was detected by the GPS station that was 2000 km away from the epicenter. Permanent coseismic displacements were resolved from the near-field high-rate GPS stations with wavelet decomposition-reconstruction method and P-wave arrivals were also detected with S transform method. The results of this study can be used for earthquake rupture process and Earthquake Early Warning studies.
基金the National Natural Science Foundation of China(Nos.41804047 and 42111540260)Fundamental Research Funds of the Institute of Geophysics,China Earthquake Administration(NO.DQJB19A0114)the Key Research Program of the Institute of Geology and Geophysics,Chinese Academy of Sciences(No.IGGCAS-201904).
文摘In recent years,artificial intelligence technology has exhibited great potential in seismic signal recognition,setting off a new wave of research.Vast amounts of high-quality labeled data are required to develop and apply artificial intelligence in seismology research.In this study,based on the 2013–2020 seismic cataloging reports of the China Earthquake Networks Center,we constructed an artificial intelligence seismological training dataset(“DiTing”)with the largest known total time length.Data were recorded using broadband and short-period seismometers.The obtained dataset included 2,734,748 threecomponent waveform traces from 787,010 regional seismic events,the corresponding P-and S-phase arrival time labels,and 641,025 P-wave first-motion polarity labels.All waveforms were sampled at 50 Hz and cut to a time length of 180 s starting from a random number of seconds before the occurrence of an earthquake.Each three-component waveform contained a considerable amount of descriptive information,such as the epicentral distance,back azimuth,and signal-to-noise ratios.The magnitudes of seismic events,epicentral distance,signal-to-noise ratio of P-wave data,and signal-to-noise ratio of S-wave data ranged from 0 to 7.7,0 to 330 km,–0.05 to 5.31 dB,and–0.05 to 4.73 dB,respectively.The dataset compiled in this study can serve as a high-quality benchmark for machine learning model development and data-driven seismological research on earthquake detection,seismic phase picking,first-motion polarity determination,earthquake magnitude prediction,early warning systems,and strong ground-motion prediction.Such research will further promote the development and application of artificial intelligence in seismology.
基金jointly supported by the National Natural Science Foundation of China (No. 42074060)the Special Fund, Institute of Geophysics, China Earthquake Administration (CEA-IGP) (Nos. DQJB19B29, DQJB20B15, and DQJB22Z01)supported by XingHuo Project, CEA (No. XH211103)
文摘Seismic phase pickers based on deep neural networks have been extensively used recently,demonstrating their advantages on both performance and efficiency.However,these pickers are trained with and applied to different data.A comprehensive benchmark based on a single dataset is therefore lacking.Here,using the recently released DiTing dataset,we analyzed performances of seven phase pickers with different network structures,the efficiencies are also evaluated using both CPU and GPU devices.Evaluations based on F1-scores reveal that the recurrent neural network(RNN)and EQTransformer exhibit the best performance,likely owing to their large receptive fields.Similar performances are observed among PhaseNet(UNet),UNet++,and the lightweight phase picking network(LPPN).However,the LPPN models are the most efficient.The RNN and EQTransformer have similar speeds,which are slower than those of the LPPN and PhaseNet.UNet++requires the most computational effort among the pickers.As all of the pickers perform well after being trained with a large-scale dataset,users may choose the one suitable for their applications.For beginners,we provide a tutorial on training and validating the pickers using the DiTing dataset.We also provide two sets of models trained using datasets with both 50 Hz and 100 Hz sampling rates for direct application by end-users.All of our models are open-source and publicly accessible.
文摘The pre-earthquake ionospheric anomalies in Wenchuan,China(21°-41°N,93°-113°E)are studied and analyzed using the summer nighttime data from 2005 to 2008 measured by DEMETER(Detection of Electro-Magnetic Emission Transmitted from Earthquake Regions)satellite detectors ICE(Internet Communications Engine),IAP(In Application Programming),and ISL(Interior Switching Link).In this paper,we take the 12 May 2008 Wenchuan earthquake as an example,use the spatial gridding method to construct the background field over the epicenter,analyze the background characteristics of very low frequency(VLF)electric field components,low-energy particle parameters,and plasma parameters,and define the perturbation intensity index of each parameter before the earthquake to extract each parameter anomaly in both space and time dimensions.The results show that the background values of some ionospheric parameters in the Wenchuan area are related to spatial distribution.Moreover,anomalous enhancement of low-frequency electric field power spectral density,H+concentration,He+concentration and ion concentration with different intensities and anomalous weakening of ion temperature were extracted in the fifteen days before the Wenchuan earthquake.After filtering the data to exclude external interference,such as solar activity,this paper concludes that there is some connection between these anomalies and the Wenchuan earthquake.