基于深度学习卷积神经网络的地震波形自动分类与识别

发展高效、高精度、普适性强的自动波形拾取算法在地震大数据时代背景下显得越来越重要.波形自动拾取算法的主要挑战来自如何适应不同区域的不同类型地震事件的分类与筛选.本文针对地震事件-噪音分类这一问题,使用13839个汶川地震余震事件建立数据集,应用深度学习卷积神经网络(CNN)方法进行训练,并用8900个新的汶川余震事件作为检测数据集,其训练和检测准确率均达到95%以上.在对连续波形的检测中,CNN方法在精度和召回率上优于STA/LTA和Fbpicker传统方法,并能找出大量人工挑选极易遗漏的微震事件.最后,我们应用训练好的最优模型对选自全国台网的441个台站8天的连续波形数据进行了识别、到时挑取及与参考地震目录关联,CNN检出7016段波形,用自动挑选算法拾取到1380对P,S到时,并与540个地震目录事件成功关联,对1级以上事件总体识别准确率为54%,二级以上为80%,证明了CNN模型具有泛化能力,初步展示了CNN在发展兼具效率、精度、普适性算法,实时地震监测等应用上具有巨大潜力.     

NET-VISA软件研发进展与性能测试

全球台网地震震相有效关联与快速精确定位是地震监测和禁核试核查的核心问题之一。禁核试条约组织(CTBTO)筹备委员会临时技术秘书处(PTS)国际数据中心(IDC)自成立以来,一直致力于台网震相自动关联有效率及自动处理公报质量的提高,以减轻分析员的工作量。NET-VISA作为IDC下一代全球震相自动关联和事件检测定位软件,使用概率模型和贝叶斯方法找到解释检测观测结果的最佳事件集,进而从这些观测结果中生成自动处理地震公报。测试结果表明,NET-VISA产生的公报要比IDC目前使用的全球关联方法生成的自动处理公报SEL3更加完整和精准。本文介绍了NET-VISA方法的基本原理和测试进展,最后对CTBT北京国家数据中心参加NET-VISA测试中朝鲜6次核试验的NETVISA关联结果作了分析对比。     

自动检测地震事件筛选方法研究

重点介绍禁核试北京国家数据中心研究的自动检测地震事件筛选方法.通过分析自动检测事件关联台站分布的合理性、关联台站的类型、关联信号参数的可靠性、多频带平均能量比等方法,判断自动检测事件的真伪.以国际数据中心审核公报事件为参考,研究的筛选方法可有效筛除43％自动检测虚假事件,使得自动检测误检率降低13％,并保持较低的误筛选率.此筛选方法可对国际数据中心以及禁核试北京国家数据中心不同处理平台的自动检测事件进行筛选分析.     

九寨沟M7.0级地震余震目录完备性研究

2017年四川省九寨沟县M 7.0级地震发生后,截止到8月24日共16天内时间内地震台网中心记录到5994个九寨沟地震余震事件,为完备该地震的余震目录,本文利用模板匹配方法对本次地震进行了遗漏地震检测研究,选取震中附近台站记录波形为计算数据,重新检测主震后16天时间内的遗漏地震事件,共得到台网目录外地震目录1053个,与台网给出目录相比增加了0.18倍.为比较本次地震检测前后的最小完整性震级以及活动性b值,利用包络差峰值振幅对检测地震的震级进行了估算,根据检测后的余震目录得到九寨沟地震余震序列的最小完整性震级为M_L 0.9,地震活动性b值为0.80左右,较原目录的M_L 1.2和0.85均有降低.     

地震事件自动判断方法研究

地震事件的自动判断是自动定位系统中至关重要的部分。本文结合数字波形特征的实际模型,提出了一套符合计算机化运行模式的新方法。对波形进行预处理后采用传统STA/LTA方法、波形延续性、台站集中性三者相结合的判断方法,有效的排除了干扰台站,保证了地震事件自动判断的准确性,并在实际运行过程中得到很好的验证。     

模板匹配滤波技术在地震数据处理中的应用

模板匹配滤波技术(Matched Filter Technique,MFT)是一种近几年发展迅速的可用于检测遗漏信号的算法,在图像识别、行星探测等众多领域均有广泛应用。在地震学中,该技术通过扫描连续波形找出模板事件附近发生的、与模板事件相关性较好的遗漏地震事件,可用于完善地震目录,实现对地震活动性以及区域应力状态的更好刻画。本文综述了模板匹配滤波技术在地震数据处理中的应用,首先概述基本原理,其次介绍该方法用于探测各类地震事件的实例,包括前震和余震、远程动态触发地震、诱发地震和非火山震颤等,最后展望了用于提高算法效率及探测精度的一些改进方法。     

2020年自动地震速报系统结果分析

选取2020年3月1日至12月31日速报地震,通过对比自动地震速报和人工地震速报结果,对自动地震速报系统产出结果进行偏差分析。该时段内自动速报系统共产出国内地震事件216次,漏报49个地震,未发生误报地震。自动速报地震震中位置平均偏差为6.7 km,震级平均偏差为0.15,平均用时119 s。震中位置偏差较大主要由台站分布不均匀、缺少近台、台站空隙角过大及系统自动拾取的震相误差较大所致。震级偏差较大主要是震级转换算法不合理等造成的。自动速报漏报地震与系统在地震定位时初至到时信噪比较低、台站分布不均匀和地震的孤立性等有关。此外,还对全国自动地震速报系统的改进提出了建议。     

基于多输入卷积神经网络的天然地震和爆破事件识别

天然地震和爆破事件识别是地震监测预警的重要内容.近年来,快速发展的深度学习算法以其强大的数据特征挖掘和图像识别能力,能够较快并准确地约束地震事件属性.利用多输入卷积神经网络算法构建天然地震和爆破事件自动分类网络模型,其中输入信息包括多台站地震波形和单台站的地震时频数据,使得卷积神经网络同时获取事件的波形、频谱和极性特征...     

2017年九寨沟MS7.0地震序列活动特征分析

中强地震序列的主震发生后,短时间内受台站距震中较远、尾波干扰和波形重叠等因素的影响,往往会遗漏大量的地震,而地震目录的完整性会直接影响到震后趋势判定和余震序列特征分析的科学性和可靠性。本文利用基于GPU加速的模板匹配方法对2017年8月1～12日的连续波形进行扫描计算,检测九寨沟M_S7.0地震前后遗漏的地震事件,选取台网目录中信噪比较高的1033个地震事件作为模板,在主震前7天至震后5天期间识别出4854个检测地震事件,为台网可定位目录的3.3倍,除去对台网单台地震事件的修正外,还检测到1797个遗漏地震事件,将完备震级从1.6级降低到1.4级。基于补充了遗漏地震的完整地震目录,对2017年8月8日九寨沟M_S7.0地震序列活动特征进行分析。结果表明,前震序列在主震前短时间内出现了地震活动的密集增强,b值也显示为低值状态,可能是深部断层发生破裂之前的加速蠕动的结果。随着时间的推移,余震序列的完备震级逐渐下降并趋于稳定,b值存在缓慢升高的趋势,未来较长时期内余震序列仍将处于持续衰减的状态。     

台网系统监视平台

广东数字遥测地震台网的台网系统监视平台,是为监视台网中心的运行和台站仪器状况而设计的,通过网络对实时地震数据收集、处理系统运行和台站状态进行图形化自动监视,当发生地震事件、断记或系统故障时该平台能自动报警。实践表明,它既减轻了监测值班人员的工作负担,又为仪器维修人员跟踪台站仪器故障提供了帮助。     

Perspectives of Cross-Correlation in Seismic Monitoring at the International Data Centre

We demonstrate that several techniques based on waveform cross-correlation are able to significantly reduce the detection threshold of seismic sources worldwide and to improve the reliability of arrivals by a more accurate estimation of their defining parameters. A master event and the events it can find using waveform cross-correlation at array stations of the International Monitoring System (IMS) have to be close. For the purposes of the International Data Centre (IDC), one can use the spatial closeness of the master and slave events in order to construct a new automatic processing pipeline: all qualified arrivals detected using cross-correlation are associated with events matching the current IDC event definition criteria (EDC) in a local association procedure. Considering the repeating character of global seismicity, more than 90 % of events in the reviewed event bulletin (REB) can be built in this automatic processing. Due to the reduced detection threshold, waveform cross-correlation may increase the number of valid REB events by a factor of 1.5–2.0. Therefore, the new pipeline may produce a more comprehensive bulletin than the current pipeline—the goal of seismic monitoring. The analysts’ experience with the cross correlation event list (XSEL) shows that the workload of interactive processing might be reduced by a factor of two or even more. Since cross-correlation produces a comprehensive list of detections for a given master event, no additional arrivals from primary stations are expected to be associated with the XSEL events. The number of false alarms, relative to the number of events rejected from the standard event list 3 (SEL3) in the current interactive processing—can also be reduced by the use of several powerful filters. The principal filter is the difference between the arrival times of the master and newly built events at three or more primary stations, which should lie in a narrow range of a few seconds. In this study, one event at a distance of about 2,000 km from the main shock was formed by three stations, with the stations and both events on the same great circle. Such spurious events are rejected by checking consistency between detections at stations at different back azimuths from the source region. Two additional effective pre-filters are f–k analysis and F _prob based on correlation traces instead of original waveforms. Overall, waveform cross-correlation is able to improve the REB completeness, to reduce the workload related to IDC interactive analysis, and to provide a precise tool for quality check for both arrivals and events. Some major improvements in automatic and interactive processing achieved by cross-correlation are illustrated using an aftershock sequence from a large continental earthquake. Exploring this sequence, we describe schematically the next steps for the development of a processing pipeline parallel to the existing IDC one in order to improve the quality of the REB together with the reduction of the magnitude threshold.     

Scoring IDC SEL1 events for identifying misformed events

The International Data Center (IDC) produces several automatic seismic event lists followed by a reviewed event list as part of the verification regime for the Comprehensive Nuclear-Test-Ban Treaty. For the time period from 2005 to 2011, on average, approximately 160 events per day were included in the first IDC automatic Standard Event List (SEL1). For the same time period, “only” 86 events per day were included in the IDC Reviewed Event Bulletin (REB). Several possible reasons may explain why an event included in SEL1 is excluded from the REB; for example: misassociation, namely, the association process generated an artificial event by associating information from stations that recorded several different events. Reducing the number of SEL1 events that are ultimately excluded from the REB would improve the automatic process relying on the SEL1 Bulletin and would reduce the workload of the analysts at the IDC. In this work, we present a scoring method that indicates the probability that an SEL1 event will “survive” the analysis process and be included in the REB. The method tries to imitate “good” analyst practice. Namely, based on the experience gained in analyzing information from a specific region with a seismic network, the analyst can decide if an event is a “real” event or not. The score was created for the International Monitoring System primary seismic stations based on the REB for the time period of 2005 to the beginning of 2011. The score is designed to indicate if the set of stations contributing/not contributing to the event is consistent with what is expected based on the analysis of past events that occurred in the same region. A byproduct of the analysis needed for the score is a statistical measure that assesses the typical contribution of a specific station to a specific region. This information can be useful in its own right, for example, to design subnetworks that will increase automatic processing efficiency.     

2013年前郭MS5.8震群矩张量研究

2013年吉林前郭M_S5.8震群为爆发性震群,目前余震活动仍然在持续.基于吉林、辽宁、黑龙江、内蒙古四省地震台网记录的前郭震群波形资料,利用波形信噪比、震源类型、台站及速度模型组合的指标选择最佳的反演方案,应用矩张量的三种反演模式,对序列中5个M_S≥5.0地震进行矩张量反演研究,获得了全矩张量、偏矩张量和纯双力偶的矩张量.使用F-test对地震的三种模式的矩张量反演结果进行显著性检验来确定最佳反演模式.结果显示,5个地震的最优矩张量解均为全矩张量模式反演获得的结果,其双力偶分量仅有20%~65%,矩心深度位于地下3~4 km处,地震在Hudson震源类型图上的投影远离双力偶震源类型区域.这些结果表明,震源类型并非典型的构造地震,推断前郭地震可能是与人类活动有关的诱发地震.     

CDSN日常分析工作辅助软件设计

通过修改Wgsnplot的源代码,改进了Wgsnplot软件地震参数的计算方法和菜单功能,使之更简单实用。同时,设计了一个集成CDSN台站日常工作基本软件为一体的统一界面辅助软件,它直接采用Wgsnplot分析软件产生的震相文件计算地震参数、编辑地震报告和进行大震速报等。本软件是CDSN台站人员测震分析工作的有力工具,减轻了工作人员的负担。     

The East Thuringian Seismic Network

The East Thuringian Seismic Network (OTSN) was installed in 1997. It started its operation with five and now consists of six seismic stations, the GRSN (German Regional Seismic Network) station MOX and a control and analysis centre. All stations are equipped with 3-component GÜRALP and short-period seismometers, RefTek 24-bit data acquisition systems (dynamic range 23.5 bit), hard disks, GPS-receivers, modems and communication computers for dial-up purposes. The seismic signals are sampled at 100 Hz and stored on the hard disk. Simultaneously, the signals are processed by a STA/LTA detector which generates an extended event list. The central station calls these event lists once per day, analyses them, produces a list of real seismic events and calls the waveform data for these events only from the single stations. All stations operate completely autonomously and the whole system works automatically, but all operations can also be carried out interactively. The event analysis is performed manually using common seismic analysis programs. The main purpose of installing the seismic network is to investigate the local seismicity, its relation to recent tectonics, the stress field and structure of the upper crust in order to render more precisely the seismic hazard of East Thuringia. A further aim of the network is to improve the seismic monitoring situation for the neighbouring regions, especially the Vogtland/Northern Bohemia and the Western Saxony area.     

CD数据流至miniSEED数据流近实时转换的方法及实现

禁核试北京国家数据中心（NDC）是我国与CTBTO/IDC进行数据交换的唯一节点,完成汇集、保存和处理我境内各IMS台站的监测数据,同时向国内用户提供境内IMS台站监测数据。NDC目前主要接入、处理、转发CD数据,但国内履约技术支持单位和协作单位多采用miniSEED数据,并使用JOPENS系统的流服务来交换数据,为了与其进行数据交换或向其分发监测数据,NDC需要具备分发miniSEED数据流的能力。文中设计的软件,是NDC自主开发的应用工具软件之一,可作为JOPENS系统流服务器的本地或远程仪器适配器,将CD数据流准实时转换为miniSEED流或将历史CD数据转换为miniSEED数据流,并通过JOPENS流服务器分发境内IMS台站数据,扩展了NDC的实时数据服务能力。     

用sPn震相计算震源深度的初步分析与应用

用近震震相sPn与Pn波的到时差测定壳内近震深度的方法简捷准确。本文对2004年3月24日内蒙古东乌珠穆沁旗5.9级地震震中1 000 km范围内不同台站记录的sPn震相进行了初步分析,计算了该次地震的震源深度。还讨论了用sPn测定震源深度的可能性及该方法在日常工作中的应用问题。     

天津数字测震台网系统建设

“十五”期间,为完善与强化首都圈地震监测及速报能力,天津数字测震台网在原有数字遥测台基础上,新建及改造升级部分测震台站,并建立环渤海虚拟监测台网,以加强对环渤海地区的地震监测。建成后丰富了记录信息,扩大了观测动态范围,地震监测能力明显提高,速报能力得到进一步加强。     

智能化网络地震监控和实时数据处理系统

中国数字地震台目前所采用的数据处理软件存在下列不足:1)不能对含lg震相的地震定位;2)不能计算mB、mb震级;3)不能自动生成月报;4)不能实现数据共享和独立完成地震速报;5)近震参数计算误差较大;6)地震定位效果不好。为此,我们基于VisualBasic软件平台,开发出智能化网络地震监控和实时数据处理系统。该系统除克服了以上不足外,还实现了数字地震仪工作状况的实时监控、地震参数计算、地震速报、各种地震单台定位、地震事件检索查询、地震编目、震相反推分析、数据传输等功能。该软件囊括了数字地震台全部日常工作,减轻了台站人员的工作量,有助于提高地震监测质量,充分发挥了数字地震仪的优越性。该系统在洛阳台近1年的试用中取得了较好的效果     

Relocation of Wyoming Mine Production Blasts Using Calibration Explosions

—?An important requirement for a comprehensive seismic monitoring system is the capability to accurately locate small seismic events worldwide. Accurate event location can improve the probability of determining whether or not a small event, recorded predominantly by local and regional stations, is a nuclear explosion. For those portions of the earth where crustal velocities are not well established, reference event calibration techniques offer a method of increased locational accuracy and reduced locational bias.¶In this study, data from a set of mining events with good ground-truth data in the Powder River Basin region of eastern Wyoming are used to investigate the potential of event calibration techniques in the area. Results of this study are compared with locations published in the prototype International Data Center's Reviewed Event Bulletin (REB). A Joint Hypocenter Determination (JHD) method was applied to a s et of 23 events. Four of those events with superior ground-truth control (mining company report or Global Positioning System data) were used as JHD reference events. Nineteen (83%) of the solutions converged and the resulting set of station-phase travel-time corrections from the JHD results was then tested. When those travel-time corrections were applied individually to the four events with good ground-truth control, the average locational error reduced the original REB location error from 16.1?km to 5.7?km (65% improvement). The JHD locations indicated reduced locational bias and all of the individual error ellipses enclosed the actual known event locations.¶Given a set of well-recorded calibration events, it appears that the JHD methodology is a viable technique for improving locational accuracy of future small events where the location depends on arrival times from predominantly local and/or regional stations. In this specific case, the International Associ ation of Seismology and the Physics of the Earth's Interior (IASPEI) travel-time tables, coupled with JHD-derived travel-time corrections, may obviate the need for an accurately known regional velocity structure in the Powder River Basin region.