地震预警震级确定方法研究

地震预警技术是减轻地震灾害损失的有效手段之一.地震预警系统中,地震震级计算是最重要也是最困难的部分之一.利用日本KiK-net台网和四川汶川余震共142次地震事件的记录,分别采用τc和Pd方法统计得到了地震预警震级的计算公式,震级计算的方差分别为0.62和0.56个震级单位.为消除震级计算过程中出现的震级饱和现象,作者拓展了Pd方法,提出了一套对位移幅值连续追踪测定的算法.当时间窗长度为10 s时,采用该方法的震级计算方差仅为0.37个震级单位,充分满足地震预警系统的需求.同时,该方法也实现了信息的连续过渡,提高了对现有信息的利用率.最后,还对位移幅值Pd用于地震动峰值PGV的估计以及不同特征参数间的相容性等内容进行了讨论.     

基于2014年云南地区地震的地震预警参数与快速震级估算研究

根据2014年云南地区M6.1盈江地震、M6.5鲁甸地震和M6.6景谷地震的主震、余震P波初期部分的信息,研究了地震震级快速估算中3个预警参数（最大卓越周期τ_p^max、特征周期τ_c和最大位移幅值P_d）与震级的相关性,提出了云南地区的震级估计模型,并对其进行分析,再和其它地区的震级估计模型进行对比和评价。结果表明:3种方法均能在短时间内（2~4 s）有效地进行震级估算,P_d方法估算效果最优,τ_c方法次优,τ_p^max方法较弱。在震级较大的主震震级估计中,3种方法均没有出现明显的震级低估（震级饱和）现象。对于τ_p^max方法,云南地区的估计模型与南加州地区较为接近,但与四川地区区别较大,可能与该方法的计算稳定性有关;而τ_c方法的估计模型则与四川及世界其它地区均较为接近,更具有普适性和稳定性。在地震预警系统的实际应用中,由于云南地区尚未建立密集的地震监测台网系统,因此在短时间内难以得到较为准确的震中距。与震源距相独立的τpmax和τc两种算法则显得较为实用,其中:τ_c方法略优于τ_p^max方法,同时能较好地满足地震预警系统的精度要求,因此推荐使用τc方法应用于云南地区地震预警系统中的快速震级估算。     

地震预警震级计算方法研究综述

地震预警系统是减轻地震灾害的有效手段之一,世界上多个国家和地区都已经建立了地震预警系统,并在实际应用中取得了显著的减灾实效.由于地震预警系统应用中的高度时效性要求,预警震级计算是整个地震预警系统中最重要也是最困难的一部分.本文回顾总结了地震预警研究中采用的一些震级计算方法,并将其归纳为三大类算法:与初始周期相关的算法、与初始幅值相关的算法和与初始强度相关的算法.对每种算法都做了详尽介绍和仔细分析,同时列举出与该算法相关的研究成果.通过对这些算法的分析总结并结合我国地震观测台网的实际情况,作者推荐τc、Pd两种方法作为我国地震预警系统建设中优先采用的两种预警震级算法.     

基于卷积神经网络的地震预警震级持续估算方法研究

为提高地震预警震级快速持续估算结果的准确性,本文构建了基于多种地震动特征参数的卷积神经网络震级估算CNN-M模型.该模型基于日本KiK-net强震动观测记录,利用其P波触发后3～ 10s时间窗内的幅值参数、周期参数、烈度参数、信噪比参数共11种地震波特征参数以及震中距参数作为输入.本文所建立的CNN-M模型随着地震发生...     

Estimation of earthquake parameters in the Alborz seismic zone,Iran using generalized inversion method

Generalized inversion of the S-wave amplitude spectra from the strong-motion network data in the Alborz, Iran has been used to estimate simultaneously source parameters, site response and S-wave attenuation (Qs). To obtain an optimum inverse solution, and also for decreasing the uncertainty level, a frequency-dependent site amplification as a constraint, was imposed to five reference site responses. This constraint removes the undetermined degree of freedom in the inversion. Furthermore, for removing the trade-off between geometrical spreading and anelastic attenuation, a geometrical spreading factor was adopted from the Motazedian [20] study. A point source model has been calibrated against the resulting source terms and consequently source parameters, like corner frequency, moment magnitude and high frequency fall off coefficient, for each event has been determined separately. Also, based on the available data and their connectivity two sub-regions including western and eastern parts of Alborz located east and west of 52.5°E have been considered to see that if there is any possible systematic difference in their seismic source characteristics. The average stress drops obtained are about 182 and 116 bars, respectively for eastern and western Alborz. Another result of the study is the site responses, which have been determined for all of 81 stations individually. Though soil nonlinearity was detected at the Ab-bar station (experienced strong ground shaking, i.e., PGA>0.5 g) near to the epicenter of Manjil M7.4 earthquake of June 20, 1990, but an analysis of residuals showed generally a weak influence of soil nonlinearity (i.e., dependence of amplification on shaking level); probably because of the relatively weak levels of acceleration in our database. Finally, the shear wave quality factor (i.e., Qs) has been determined as a function of frequency represented by a linear equation in logarithmic scale. To evaluate the outcomes of the current study, the results have been compared with similar studies wherever it was available. The results of the current study are of utmost importance for seismic hazard assessment of the metropolitan area of Tehran, where 15 million people live, one-fifth of the population of Iran.     

State-of-the art and future of earthquake early warning in the European region

European researchers and seismic networks are active in developing new approaches to earthquake early warning (EEW), implementing and operating test EEW systems, and in some cases, offering operational EEW to end users. We present the key recent developments in EEW research in Europe, describe the networks and regions where EEW is currently in testing or development, and highlight the two systems in Turkey and Romania that currently provide operational systems to a limited set of end users.     

Towards earthquake early warning for the Rion-Antirion bridge,Greece

Patras is the third largest city in Greece and an ideal candidate for earthquake early warning (EEW) applications due to its high seismic hazard, its existing research infrastructure and the presence of critical structures such as the Rion-Antirion bridge. Patras is located a few hundred kilometres from the Hellenic Arc, where very strong and potentially damaging earthquakes occur. This distance is large enough to allow a few tens of seconds of warning time prior to significant shaking, provided earthquakes are timely detected by a dense seismic network. Within the framework of the EC-funded project REAKT, the Virtual Seismologist (VS) EEW software was installed at Patras Seismological Laboratory. Its initial performance evaluation is presented here. In general VS provides magnitudes similar to the official, manually revised ones. Given the current station density and network telemetry, the average time that VS needs to deliver the first magnitude estimate is rather large, of the order of tens of seconds and not yet satisfactory for routine operational use of EEW. Even so, the system is able to provide up to 10 s of warning time prior to S-wave arrivals for events occurring on the Hellenic Arc. Our results indicate that the seismic networks in Greece need enhancements for regional EEW, either by adding stations or by upgrading the hardware to reduce delays. The application of an EEW system in the area is promising and, once operational, capable of mitigating earthquake risk.     

地震预警中两种利用卓越周期估算震级方法的比较

利用地震P波初始信息快速估算震级是地震预警系统中的重要部分之一,文中详细介绍了地震预警研究中常用的两种卓越周期估算震级方法:τpmax方法和τc方法。为了评估这两种方法哪一种震级估算精度更高,文中通过设置不同的估算震级时间窗以及不同的滤波频带,利用震级范围4.1～7.9的22次地震的强震观测记录对这两种方法进行了比较分析。比较的结果表明:本文计算得到的τpmax值和τc值与震级之间均存在线性比例关系,与前人的研究结果一致;在相同的0.075～3 Hz带通滤波条件下,利用P波触发后3 s地震记录计算的τc值可获得最优的震级估计结果,τc与震级的相关系数为0.78,标准差为0.16。仅经过0.075 Hz高通滤波得到的τpmax值与震级的线性相关度比0.075～3Hz带通滤波得到的τpmax值与震级的线性相关度低,而是否经过低通滤波对τc值与震级的线性比例关系没有影响。建议在地震预警系统中优先采用3 sτc方法作为震级估算方法。     

Continuous estimates on the earthquake early warning magnitude by use of the near-field acceleration records

In this article, the seismic records of Japan’s Kik-net are selected to measure the acceleration, displacement, and effective peak acceleration of each seismic record within a certain time after P wave, then a continuous estimation is given on earthquake early warning magnitude through statistical analysis method, and Wenchuan earthquake record is utilized to check the method. The results show that the reliability of earthquake early warning magnitude continuously increases with the increase of the seismic information, the biggest residual happens if the acceleration is adopted to fit earthquake magnitude, which may be caused by rich high-frequency components and large dispersion of peak value in acceleration record, the influence caused by the high-frequency components can be effectively reduced if the effective peak acceleration and peak displacement is adopted, it is estimated that the dispersion of earthquake magnitude obviously reduces, but it is easy for peak displacement to be affected by long-period drifting. In various components, the residual enlargement phenomenon at vertical direction is almost unobvious, thus it is recommended in this article that the effective peak acceleration at vertical direction is preferred to estimate earthquake early warning magnitude. Through adopting Wenchuan strong earthquake record to check the method mentioned in this article, it is found that this method can be used to quickly, stably, and accurately estimate the early warning magnitude of this earthquake, which shows that this method is completely applicable for earthquake early warning.     

利用高频GPS资料研究2016年新西兰凯库拉地震的地表形变及预警震级

2016年11月13日,新西兰凯库拉地区发生了M_W7.8级地震.本文利用1 Hz高频GPS观测数据,基于GAMIT track解算模块,采用主分量分析（PCA）空间滤波方法获取了地震地表形变.同时从trackRTr模拟实时解算的动态位移中,提取P波5秒峰值位移（P_d）和地面峰值位移（PGD）,并根据震级统计回归模型计算预警震级.结果表明：测站的动态形变时间长达2 min,距震源最近的HANM和KAIK站出现二次剧烈形变,震源北部测站的形变幅度大于南部,而高频GPS静态同震形变场表现出先逆冲后走滑的震源机制特征.不同GPS台站的P_d预警震级相差较大,最大震级差为M_W2.5.综合考虑预警震级发布的时效性和可靠性,采用顾及空间分布的四台站PGD联合预警方法,其预警震级在震后23 s达到初始稳定（M_W7.56）,在震后110 s达到最终稳定（M_W7.78）,该震级与USGS矩张量反演震级（M_W7.8）基本一致.

     

用于地震预警的P波震相到时自动拾取

P波震相的自动拾取可用于地震预警中地震事件判别和地震定位,是实现基于地震台网地震预警的首要条件.针对地震预警中P波震相拾取的特点,本文发展了一套基于长短时平均(STA/LTA)和池赤准则(AIC)算法的多步骤P波自动拾取技术,应用Delaunay三角剖分提出了一种非几何相关的干扰信号剔除方法,并应用福建省数字地震台网记录对方法进行了验证,目前方法已经用到了福建省地震预警试验系统中.     

Evaluation of earthquake signal characteristics for early warning

This paper evaluates different characteristics for earthquake early warning.The scaling relationships between magnitude,epicenter distance and calculated parameters are derived from earthquake event data from USGS.The standard STA/LTA method is modified by adding two new parameters to eliminate the effects of the spike-type noise and small pulsetype noise ahead of the onset of the P-wave.After the detection of the P-wave,the algorithm extracts 12 kinds of parameters from the first 3 seconds of the P-wave.Then stepwise regression analysis of these parameters is performed to estimate the epicentral distance and magnitude.Six different parameters are selected to estimate the epicentral distance,and the median error for all 419 estimates is 16.5 km.Four parameters are optimally combined to estimate the magnitude,and the mean error for all events is 0.0 magnitude units,with a standard deviation of 0.5.Finally,based on the estimation results,additional work is proposed to improve the accuracy of the results.     

Progress of the earthquake early warning system in Fujian, China

In this article, we systematically introduce the latest progress of the earthquake early warning (EEW) system in Fujian, China. We focus on the following key technologies and methods: continuous earthquake location and its error evaluation; magnitude estimation; reliability judgment of EEW system information; use of double-parameter principle in EEW system information release threshold; real-time estimation of seismic intensity and available time for target areas; seismic-monitoring network and data sharing platform; EEW system information release and receiving platform; software test platform; and test results statistical analysis. Based on strong ground motion data received in the mainshock of the Wenchuan earthquake, the EEW system developed by the above algorithm is simulated online, and the results show that the system can reduce earthquake hazards effectively. In addition, we analyzed four earthquake cases with magnitude greater than 5.5 processed by our EEW system since the online-testing that was started one year ago, and results indicate that our system can effectively reduce earthquake hazards and have high practical significance.     

An alternative approach for the Istanbul earthquake early warning system

Two recent catastrophic earthquakes that struck the Marmara Region on 17 August 1999 (M_w=7.4) and 12 November 1999 (M_w=7.2) caused major concern about future earthquake occurrences in Istanbul and the Marmara Region. As a result of the preparations for an expected earthquake may occur around Istanbul region, an earthquake early warning system has been established in 2002 with a simple and robust algorithm, based on the exceedance of specified thresholds of time domain amplitudes and the cumulative absolute velocity (CAV) levels (Erdik et al., 2003 [1]). In order to improve the capability of Istanbul earthquake early warning system (IEEWS) for giving early warning of a damaging earthquake in the Marmara Region, we explored an alternative approach with the use of a period parameter (τ_c) and a high-pass filtered vertical displacement amplitude parameter (Pd) from the initial 3 s of the P waveforms as proposed by Kanamori (2005) [2] and Wu and Kanamori (2005) 3 and 4. The empirical relationships both between τ_c and moment magnitude (M_w), and between Pd and peak ground velocity (PGV) for the Marmara Region are presented. These relationships can be used to detect a damaging earthquake within seconds after the arrival of P waves, and can provide on-site warning in the Marmara Region.     

Body wave attenuation characteristics in the crust of Alborz region and North Central Iran

Attenuation of P and S waves has been investigated in Alborz and north central part of Iran using the data recorded by two permanent and one temporary networks during October 20, 2009, to December 22, 2010. The dataset consists of 14,000 waveforms from 380 local earthquakes (2 < M _L < 5.6). The extended coda normalization method (CNM) was used to estimate quality factor of P (Q _P) and S waves (Q _S) at seven frequency bands (0.375, 0.75, 1.5, 3, 6, 12, 24 Hz). The Q _P and Q _S values have been estimated at lapse times from 40 to 100 s. It has been observed that the estimated values of Q _P and Q _S are time independent; therefore, the mean values of Q _P and Q _S at different lapse times have been considered. The frequency dependence of quality factor was determined by using a power-law relationship. The frequency-dependent relationship for Q _P was estimated in the form of (62 ± 7)f ^{(1.03 ± 0.07)} and (48 ± 5)f ^{(0.95 ± 0.07)} in Alborz region and North Central Iran, respectively. These relations for Q _S for Alborz region and North Central Iran have estimated as (83 ± 8)f ^{(0.99 ± 0.07)} and (68 ± 5)f ^{(0.96 ± 0.05)}, respectively. The observed low Q values could be the results of thermoelastic effects and/or existing fracture. The estimated frequency-dependent relationships are comparable with tectonically active regions.     

Earthquake relocation in the Central Alborz region of Iran using a non-linear probabilistic method

In this study, we calculate accurate absolute locations for nearly 3,000 shallow earthquakes (≤20 km depth) that occurred from 1996 to 2010 in the Central Alborz region of northern Iran using a non-linear probabilistic relocation algorithm on a local scale. We aim to produce a consistent dataset with a realistic assessment of location errors using probabilistic hypocenter probability density functions. Our results indicate significant improvement in hypocenter locations and far less scattering than in the routine earthquake catalog. According to our results, 816 earthquakes have horizontal uncertainties in the 0.5–3.0 km range, and 981 earthquakes are relocated with focal-depth errors less than 3.0 km, even with a suboptimal network geometry. Earthquake relocated are tightly clustered in the eastern Tehran region and are mainly associated with active faults in the study area (the Mosha and Garmsar faults). Strong historical earthquakes have occurred along the Mosha and Garmsar faults, and the relocated earthquakes along these faults show clear north-dipping structures and align along east–west lineations, consistent with the predominant trend of faults within the study region. After event relocation, all seismicity lies in the upper 20 km of the crust, and no deep seismicity (>20 km depth) has been observed. In many circumstances, the seismicity at depth does not correlate with surface faulting, suggesting that the faulting at depth does not directly offset overlying sediments.     

地震预警实质的探讨

随着国家地震烈度速报与预警工程的开展,“地震预警”越来越进入公众的视线,国内外关于“什么是地震预警”的问题越来越多,对地震预警技术的解释也多种多样。本文论述了我国地震预警的进展,地震预警的实质和局限性。地震预警实质是地震观测进入密集观测新阶段,地震速报从分钟级发展到秒级超快地震速报,也就是地震警报。由于在地震预警实际应用中受预警盲区和地震强度估算不准确的局限,地震科学家对地震预警技术应用效能的认识也在不断加深和变化,逐步认识了发挥地震预警的警报作用的重要性。同时,地震预警是复杂的社会工程,引导公众认识地震预警的局限性,才能有效发挥减灾效能。      

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

Fast and accurate P-wave arrival picking significantly affects the performance of earthquake early warning(EEW)systems.Automated P-wave picking algorithms used in EEW have encountered problems of falsely picking up noise,missing P-waves and inaccurate P-wave arrival estimation.To address these issues,an automatic algorithm based on the convolution neural network(DPick)was developed,and trained with a moderate number of data sets of 17,717 accelerograms.Compared to the widely used approach of the short-term average/long-term average of signal characteristic function(STA/LTA),DPick is 1.6 times less likely to detect noise as a P-wave,and 76 times less likely to miss P-waves.In terms of estimating P-wave arrival time,when the detection task is completed within 1 s,DPick′s detection occurrence is 7.4 times that of STA/LTA in the 0.05 s error band,and 1.6 times when the error band is 0.10 s.This verified that the proposed method has the potential for wide applications in EEW.