scholarly journals Magnitude Estimation for Earthquake Early Warning Using a Deep Convolutional Neural Network

2021 ◽  
Vol 9 ◽  
Author(s):  
Jingbao Zhu ◽  
Shanyou Li ◽  
Jindong Song ◽  
Yuan Wang
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Magnitude Estimation ◽  
Early Warning ◽  
Epicentral Distance ◽  
Estimation Error ◽  
Deep Convolutional Neural Network ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Wave Arrival

Magnitude estimation is a vital task within earthquake early warning (EEW) systems (EEWSs). To improve the magnitude determination accuracy after P-wave arrival, we introduce an advanced magnitude prediction model that uses a deep convolutional neural network for earthquake magnitude estimation (DCNN-M). In this paper, we use the inland strong-motion data obtained from the Japan Kyoshin Network (K-NET) to calculate the input parameters of the DCNN-M model. The DCNN-M model uses 12 parameters extracted from 3 s of seismic data recorded after P-wave arrival as the input, four convolutional layers, four pooling layers, four batch normalization layers, three fully connected layers, the Adam optimizer, and an output. Our results show that the standard deviation of the magnitude estimation error of the DCNN-M model is 0.31, which is significantly less than the values of 1.56 and 0.42 for the τc method and Pd method, respectively. In addition, the magnitude prediction error of the DCNN-M model is not affected by variations in the epicentral distance. The DCNN-M model has considerable potential application in EEWSs in Japan.

Magnitude Estimation for Earthquake Early Warning with Multiple Parameter Inputs and a Support Vector Machine

2021 ◽  
Author(s):  
Jingbao Zhu ◽  
Shanyou Li ◽  
Jindong Song
Keyword(s):  
Support Vector Machine ◽  
Magnitude Estimation ◽  
Early Warning ◽  
Epicentral Distance ◽  
Estimation Error ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Support Vector ◽  
Multiple Parameter ◽  
Wave Arrival

Abstract Accurately estimating the magnitude within the initial seconds after the P-wave arrival is of great significance in earthquake early warning (EEW). Over the past few decades, single-parameter approaches such as the τc and Pd methods have been applied to EEW magnitude estimation studies considering the first 3 s after the P-wave onset. However, these methods present considerable scatter and are affected by the signal-to-noise ratio (SNR) and epicentral distance. In this study, using Japanese K-NET strong-motion data, we propose a machine-learning method comprising multiple parameter inputs, namely, the support vector machine magnitude estimation (SVM-M) model, to determine earthquake magnitudes and resolve the aforementioned problems. Our results using a single seismological station record show that the standard deviation of the magnitude prediction errors of the SVM-M model is 0.297, which is less than those of the τc (1.637) and Pd (0.425) methods. The magnitudes estimated by the SVM-M model within 3 s after the P-wave arrival are not obviously affected by the SNR or epicentral distance, and not overestimated for MJMA≤5. In addition, in an offline EEW application, the magnitude estimation error of the SVM-M model gradually decreases with increasing time after the first station is triggered, and the underestimation of event magnitudes for 6.5≤MJMA gradually improves. These results demonstrate that the proposed SVM-M model can robustly estimate earthquake magnitudes and has potential for EEW.

A Comparison of Two Magnitude Estimation Methods Using Predominant Period in Earthquake Early Warning

2012 ◽  
Vol 256-259 ◽  
pp. 2193-2199
Author(s):  
Jin Dong Song ◽  
Shan You Li
Keyword(s):  
Magnitude Estimation ◽  
Early Warning ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Estimation Methods ◽  
Pass Filter ◽  
Low Pass Filter ◽  
Predominant Period ◽  
Low Pass ◽  
Wave Arrival

Currently, there are two magnitude estimation approaches using predominant period for earthquake early warning, Tpmax method and Tc method. We compared Tpmax method with Tc method from the NSMP strong motion records of 22 earthquakes in United States with moment magnitude ranging from 4.1 to 7.9, to explore which method was with higher precision and could be suitable for the earthquake early warning system. Our results show that scaling relations between the two predominant period parameters, Tpmax and Tc, calculated from P-wave arrivals and earthquake magnitude are consistent with previous research. It was found that Tc method had higher precision than Tpmax method with the same filter band, and had best result in magnitude estimation resulting in a correlation coefficient of 0.78 and a standard deviation of 0.16 using 3 seconds signal after the P-wave arrival. We also found that Tc method without low-pass filter had higher accuracy than Tpmax method without low-pass filter. We recommended Tc method using 3 seconds signal after the P-wave arrival as the priority magnitude estimation method for earthquake early warning.

End-to-end PGA estimation for earthquake early warning using transformer networks

2020 ◽  
Author(s):  
Jannes Münchmeyer ◽  
Dino Bindi ◽  
Ulf Leser ◽  
Frederik Tilmann
Keyword(s):  
Neural Network ◽  
Real Time ◽  
Ground Motion ◽  
Early Warning ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Lead Times ◽  
Ground Shaking ◽  
Target Sites ◽  
End To End

<p>The key task of earthquake early warning is to provide timely and accurate estimates of the ground shaking at target sites. Current approaches use either source or propagation based methods. Source based methods calculate fast estimates of the earthquake source parameters and apply ground motion prediction equations to estimate shaking. They suffer from saturation effects for large events, simplified assumptions and the need for a well known hypocentral location, which usually requires arrivals at multiple stations. Propagation based methods estimate levels of shaking from the shaking at neighboring stations and therefore have short warning times and possibly large blind zones. Both methods only use specific features from the waveform. In contrast, we present a multi-station neural network method to estimate horizontal peak ground acceleration (PGA) anywhere in the target region directly from raw accelerometer waveforms in real time.</p><p>The three main components of our model are a convolutional neural network (CNN) for extracting features from the single-station three-component accelerograms, a transformer network for combining features from multiple stations and for transferring them to the target site features and a mixture density network to generate probabilistic PGA estimates. By using a transformer network, our model is able to handle a varying set and number of stations as well as target sites. We train our model end-to-end using recorded waveforms and PGAs. We use data augmentation to enable the model to provide estimations at targets without waveform recordings. Starting with the arrival of a P wave at any station of the network, our model issues real-time predictions at each new sample. The predictions are Gaussian mixtures, giving estimates of both expected value and uncertainties. The model can be used to predict PGA at specific target sites, as well as to generate ground motion maps.</p><p>We analyze the model on two strong motion data sets from Japan and Italy in terms of standard deviation and lead times. Through the probabilistic predictions we are able to give lead times for different levels of uncertainty and ground shaking. This allows to control the ratio of missed detections to false alerts. Preliminary analysis suggest that for levels between 1%g and 10%g our model achieves multi-second lead times even for the closest stations at a false-positive rate below 25%. For an example event at 50 km depth, lead times at the closest stations with epicentral distances below 20 km are 6 s and 7.5 s. This suggests that our model is able to effectively use the difference between P and S travel time and accurately assess the future level of ground shaking from the first parts of the P wave. It additionally makes effective use of the information contained in the absence of signal at other stations.</p>

Practical Site-Specific Earthquake Early Warning Application

2009 ◽  
Vol 4 (4) ◽  
pp. 579-587 ◽  
Author(s):  
Katsuhisa Kanda ◽  
◽  
Tadashi Nasu ◽  
Masamitsu Miyamura
Keyword(s):  
Early Warning ◽  
Estimation Error ◽  
Seismic Intensity ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Japan Meteorological Agency ◽  
Estimation Accuracy ◽  
Intensity Estimation ◽  
Site Specific ◽  
Wide Range

Real-time hazard mitigation we have developed using earthquake early warning (EEW) (1) enhances seismic intensity estimation accuracy and (2) extends the interval between when an EEW is issued and when strong tremors arrive. We accomplished the first point (enhancing seismic intensity estimation) by reducing estimation error to less than that commonly used based on an attenuation relationship and soil amplification factor by considering source-location and wave propagation path differences based on site-specific empiricism. We accomplished the second point (shortening the time between warnings and when tremors arrive) using a high-speed, reliable communication network for receiving EEW information from the Japan Meteorological Agency (JMA) and quickly transmitting warning signals to users. In areas close to quake epicenters, however, warnings may not arrive before the arrival of strong ground motions. The on-site warning system we developed uses P-wave pickup sensors that detect P-wave arrival at a site and predict seismic intensity of subsequent S-waves. We confirmed the on-site warning prototype’s feasibility based on numerical simulation and observation. We also developed an integration server for combining on-site warnings with JMA information to be applied to earthquakes over a wide range of distances. We installed a practical prototype at a construction site near the 2008 Iwate-Miyagi Inland Earthquake epicenter to measure its aftershocks because JMA EEW information was too late to use against the main shock. We obtained good aftershock results, confirming the prototype’s applicability and accuracy. Integration server combination logic was developed for manufacturing sites requiring highly robust, reliable control.

scholarly journals P-wave picking for earthquake early warning: Refinement of a T pd method

2021 ◽  
Author(s):  
Masumi Yamada ◽  
Jim Mori
Keyword(s):  
Real Time ◽  
Early Warning ◽  
Arrival Time ◽  
Early Warning Systems ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Warning Systems ◽  
Wave Detection ◽  
Earthquake Early Warning Systems ◽  
Wave Arrival

Summary Detecting P-wave onsets for on-line processing is an important component for real-time seismology. As earthquake early warning systems around the world come into operation, the importance of reliable P-wave detection has increased, since the accuracy of the earthquake information depends primarily on the quality of the detection. In addition to the accuracy of arrival time determination, the robustness in the presence of noise and the speed of detection are important factors in the methods used for the earthquake early warning. In this paper, we tried to improve the P-wave detection method designed for real-time processing of continuous waveforms. We used the new Tpd method, and proposed a refinement algorithm to determine the P-wave arrival time. Applying the refinement process substantially decreases the errors of the P-wave arrival time. Using 606 strong motion records of the 2011 Tohoku earthquake sequence to test the refinement methods, the median of the error was decreased from 0.15 s to 0.04 s. Only three P-wave arrivals were missed by the best threshold. Our results show that the Tpd method provides better accuracy for estimating the P-wave arrival time compared to the STA/LTA method. The Tpd method also shows better performance in detecting the P-wave arrivals of the target earthquakes in the presence of noise and coda of previous earthquakes. The Tpd method can be computed quickly so it would be suitable for the implementation in earthquake early warning systems.

Station Correction of P-Alert Network to Improve Magnitude Estimation for Earthquake Early Warning

2021 ◽  
Author(s):  
Yu-Ting Wu ◽  
Yih-Min Wu
Keyword(s):  
Regression Model ◽  
Magnitude Estimation ◽  
Early Warning ◽  
Latent Variables ◽  
Regression Method ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Station Correction ◽  
Hypocentral Distance ◽  
Station Corrections

<p>Magnitude estimation for earthquake early warning has been shown that it can be achieved by utilizing the relationship among the first three seconds P-wave amplitude, hypocentral distance and magnitude. However, the regression models in previous studies about P-Alert didn't include station correction factors, which may cause non-negligible effects. Thus, to improve the precision of magnitude estimation, we take station corrections into consideration when building the regression model. For the reason that station corrections are the unobserved latent variables of the model, we adopt the iteration regression method, which is based on the expectation-maximization algorithm, to determine them. By using this method, we are able to approach the values of both the station corrections and the coefficients of the regression model after several iterations. Our preliminary results show that after utilizing the iteration regression method, the standard deviation reduces from 0.30 to 0.26, and the station corrections we get range from -0.70 to 0.66.</p>

An Improved Magnitude Estimation Method Using Combination of Predominant Period and Peak Amplitude for Earthquake Early Warning

2012 ◽  
Vol 256-259 ◽  
pp. 2775-2780
Author(s):  
Jin Dong Song ◽  
Shan You Li
Keyword(s):  
Magnitude Estimation ◽  
Early Warning ◽  
Estimation Method ◽  
Strong Motion ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Peak Amplitude ◽  
Combined Method ◽  
Predominant Period ◽  
Great Magnitude

The critical technology of Earthquake Early Warning (EEW) is determining the size of an earthquake and the predicted ground motion at given site, from the first few seconds of the P wave arrivals. Currently, there were two different approaches to the EEW magnitude estimation, the predominant period method and the peak amplitude method. However, both methods mentioned above had some disadvantages, such as significant uncertainty and saturation at great magnitude. To improve the results of magnitude estimation, a combined method using predominant period τc and peak amplitude of acceleration Pmax was introduced. Compared with the predominant period method and the peak amplitude method, the estimation standard deviation level of the combined method is 0.42 using NSMP strong motion data. The magnitude estimation results of the first three seconds P wave indicate that, the estimation precision of combined method is higher than those of the two methods, the predominant period method and the peak amplitude method, and the saturation at great magnitude is improved.

A LSTM Neural Network for On-site Earthquake Early Warning

2020 ◽  
Author(s):  
Chia Yu Wang ◽  
Ting Chung Huang ◽  
Yih Min Wu
Keyword(s):  
Neural Network ◽  
Early Warning ◽  
Short Term Memory ◽  
Earthquake Early Warning ◽  
P Wave ◽  
False Alarms ◽  
Time Step ◽  
P Waves ◽  
Ground Shaking ◽  
S Waves

<p>On-site Earthquake Early Warning (EEW) systems estimate possible destructive S-waves based on initial P-waves and issue warnings before large shaking arrives. On-site EEW plays a crucial role to fill up the “blind zone” of regional EEW systems near the epicenter, which often suffers from the most disastrous ground shaking. Previous studies show that peak P-wave displacement amplitude (Pd) may provide a possible indicator of destructive earthquakes. However, the attempt to use a single indicator with fixed thresholds suffers from inevitable misfits, since the diversity in travel paths and site effects for different stations introduce complex nonlinearities. To overcome the above problem, we present a deep learning approach using Long-Short Term Memory (LSTM) neural networks. By utilizing the properties of multi-layered LSTM, we are able to train a highly non-linear neural network that takes initial waveform as input and gives an alert probability as the output on every time step. It is then tested with several major earthquake events, giving the results of a missed alarm rate less than 0.03 percent and false alarm rate less than 15 percent. Our model shows promising outcomes in reducing both missed alarms and false alarms while also providing an improving warning time for hazard mitigation procedures.</p>

Deep learning for P-wave arrival picking in earthquake early warning

2021 ◽  
Vol 20 (2) ◽  
pp. 391-402
Author(s):  
Wang Yanwei ◽  
Li Xiaojun ◽  
Wang Zifa ◽  
Shi Jianping ◽  
Bao Enhe
Keyword(s):  
Deep Learning ◽  
Early Warning ◽  
Earthquake Early Warning ◽  
P Wave ◽  
Wave Arrival
Sign in / Sign up

Export Citation Format

Share Document