Earthquake detection in the Jiangsu region,China using graphics-processing-unit-based Match & Locate and rapid earthquake association and location*

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 graphics-processing-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 short-term/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.     

Comparison of the earthquake detection abilities of PhaseNet and EQTransformer with the Yangbi and Maduo earthquakes

PhaseNet and EQTransformer are two state-of-the-art earthquake detection methods that have been increasingly applied worldwide. To evaluate the generalization 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 reported in the original papers. The analysis results indicate that neural networks with deeper layers and complex structures may not necessarily enhance earthquake detection performance. 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.     

基于深度学习的地震检测模型在区域台网的泛化性研究

将识别地震的深度学习算法PhaseNet应用于四川台网和首都圈台网, 对该模型的泛化能力进行了测试和评估。 首先利用2010年1月至2018年10月首都圈台网199个地震台站记录的29328个事件(M_L0~M_L4)所对应的126761段事件波形, 以及 2019年4—9月四川及邻省部分台网227个地震台站记录的16595个事件(M_L0~M_L6.0)所对应的120233段事件波形分别建立了SC和CA测试数据集, 并用预训练好的PhaseNet模型进行P、 S震相自动识别和到时拾取, 并将拾取结果与人工拾取结果在不同误差阈值下进行对比。 测试结果表明, PhaseNet在两个数据集上具有良好的震相检测能力(误差阈值为0.5 s), 其P、 S震相检测的F₁值都超过0.75, 具有比较稳定的准确拾取P波到时能力(误差阈值0.1 s), 其检测F₁值均超过0.6, 而S波到时拾取的F₁值分别为0.33(SC)和0.53(CA)。 进一步分析了测试结果与震中距、 震级、 信噪比、 台站所处地域之间的关系, 为下一步继续训练更优化的模型指明了方向。 研究结果表明, PhaseNet算法在区域台网地震自动检测和到时拾取方面有很大的应用潜力和提升空间, 可以为区域台网的自动编目工作提供辅助。     

基于深度学习到时拾取自动构建长宁地震前震目录

将深度学习到时拾取、震相关联技术与传统定位方法联系起来,构建一套连续波形自动化处理与地震目录自动构建流程,对于高效充分利用地震资料,提升微震检测能力具有十分重要的意义.我们应用最新发展的迁移学习震相识别技术、震相自动关联技术,对长宁M S6.0地震震中附近21个台站震前半个月(6月1日—6月17日)的连续记录波形进行P、S震相识别、震相自动关联和初步定位,并应用传统绝对定位和相对定位技术得到了长宁地震震前微震活动的绝对和相对定位目录.其中绝对定位目录能在较小的误差范围匹配85%的人工处理目录,其发震时刻平均误差为0.36±0.07 s,震级平均误差为0.15±0.024级,水平定位平均误差为1.45±0.028 km,其识别的1.0级以下微震数目是人工的8倍以上,将长宁地震震前微震目录的检测下限提升至M L-1左右,证明了基于深度学习到时识取和REAL(Rapid Earthquake Association and Location,快速震相关联和定位技术)震相自动关联来构建微震目录具有较好的实用性.我们的自动地震目录揭示了长宁M S6.0主震所发生的区域震前异常频繁的微震活动,以及与区域内盐矿注水井的关联性,更好地描绘了这些微震活动的时空演化特征,其空间活动性分布特征与长宁M S6.0余震序列的分布一致.     

2012年9月7日彝良地震及余震序列双差定位研究

本文提出了时域多通道相关检测函数并用其计算波形互相关走时差数据,采用双差定位法对2012年9月7日云南彝良地震和余震序列共944个地震进行重定位,得到652个重定位事件,并与目录数据的结果进行了对比.本文采用了多个准则对走时差数据进行筛选,确保定位结果稳定可靠.得到MS5.7主震的震中为27.516°N,103.951°E,震源深度6.9km;MS5.6主震的震中为27.543°N,104.023°E,震源深度7.27km;重定位结果显示,地震序列紧缩为条带状并沿附近断裂走向分布,深度总体分布较重定位前变浅,集中分布在5~8km,地震群出现轻微倾斜.东西向、南北向、深度和发震时刻的平均相对误差分别为55.2 m,43.0 m,186.7 m和0.01s,走时残差16ms.研究表明:互相关数据的结果要优于目录数据;震源深度与速度模型存在较大的相关性;确定彝良—会泽断裂为本次彝良地震序列的发震构造.     

基于机器学习算法的银川台阵资料自动化震相分析和定位

近年来快速发展的机器学习算法显著提高了震相拾取的精度和效率。 采用卷积神经网络和递归神经网络的震相识别方法对银川台阵2019年6~7月的连续波形数据进行事件检测和P、 S震相拾取, 并通过快速震相关联和事件定位得到了银川地区较全的地震目录。 结果表明, 当震相数小于10时, 虽然可以检测出较多事件, 但分布呈弥散状, 与区域地震活动特征不符。 进一步对震相数≥10的事件进行了人工复核。 总体而言, 随着震相数量的增加, 事件的误检率逐步降低。 震相数16是该地区自动检测和定位结果准确性的拐点。 当震相数≥20时, 全部召回了地震目录中的13个地震事件, 二者平均定位差异4.27 km。 经过人工复核, 检测到的真实地震事件为区域内地震目录中事件数量的9倍。 本文使用的基于机器学习和快速震相关联和定位方法的流程可在确保准确率的基础上降低人工检测的难度, 提高地震检测的效率。     

Automatic Location of Swarm Earthquakes from Local Network Data

An advanced method of automated seismic phase picking and exact location and magnitude determination of swarm micro-earthquakes from local network data is presented. The phase picker is applied in two steps: first, S-wave groups are identified using a polarisation detector, and then corresponding P-wave groups are searched for. The times of maximum P- and S-amplitudes are then used as starting points for the determination of accurate P- and S-arrival times. The maximum S-wave amplitudes are utilised for determining local magnitudes. The whole procedure is checked by simultaneous preliminary hypocentre location providing estimates of local magnitudes and a compatibility check of the candidate P- and S-phases. The closest station to the earthquake cluster is used as a master, and the phase search at the remaining stations is governed by the P- and S-phases identified at the master station. Thanks to the use of apriori information on the approximate position of hypocentres, the procedure is also capable of picking the individual P- and S-phases of sequences of overlapping swarm events. The performance of the procedure was tested by comparison of the automatically and interactively created catalogues of the January 1997 NW-Bohemia micro-earthquake swarm. With stations located at epicentral distances between 0 and 20 km, the difference between hypocentre coordinates obtained by automatic and interactive processing did not exceed 80 m for 86% events. All events above magnitude 0.5 were identified, and the automatically determined polarity of first P-wave motion proved to be correct in 89% of them.     

地震自动速报综合触发系统产出结果评估

收集2015年1月—2020年8月地震自动速报综合触发系统的全部产出结果,选取国内天然地震事件结果共计1863条,与正式速报目录结果进行对比研究。研究结果显示,自动速报系统所产出的地震参数较正式速报目录发震时刻平均偏差约2s,震中位置平均偏差约8km,震源深度平均偏差约9km,震级绝对平均偏差约0.25,偏差均值为自动速报结果较正式速报结果偏大约0.1。到达产出震级标准的漏报地震事件176个,其主要原因为区域台站稀疏、台站分布不均匀、大震或前震尾波干扰等,未发生误报事件。发震时刻、震中位置和震源深度偏差较大事件的分布与漏报事件类似,多发生于台站密度较低、台站分布不均匀和速度模型与实际差异较大的地区。震级偏差主要与地震事件震级大小和震中所在位置有关,对于3.0≤M<5.0的地震事件,产出震级最为稳定,而对于M≥6.0的地震,震级结果可靠性较低;此外,震级偏差的大小与区域分布并无明显关系,但偏差方向具有区域特征。通过对自动速报系统运行情况的统计分析,提出了关于自动速报系统改进与完善的建议,有助于提高自动速报系统的应用效果。     

东祁连山北缘断裂带基于深度学习的密集台阵地震事件快速检测与定位研究

为监测东祁连山北缘断裂带附近的地震活动性,布设包含240台短周期地震仪的面状密集台阵,进行约30 d的连续观测。首先使用基于深度学习的多台站地震事件检测算法（CNNDetector）进行地震事件检测,然后使用震相拾取网络（PhaseNet）对地震事件进行P波和S波到时拾取,其次使用震相关联算法（REAL）进行震相关联及初定位,最后使用双差定位（hypoDD）进行地震重定位,最终的精定位地震目录中共有517个地震。在密集台阵观测期间,中国地震台网正式地震目录中共有39个位于台阵内的地震事件,相比而言,密集台阵检测到大量小于0级的地震。因此通过布设密集台阵,可提高活动断裂微地震活动性的监测能力。与历史地震空间分布相比,密集台阵地震精定位分布具有较好的一致性,表现出更明显的线性分布特征。基于地震分布,发现研究区域存在与地表断层迹线走向不同的隐伏活跃断裂。     

基于恢复地震数据获取震级、震源机制及破裂过程的评价——以2013年四川芦山MW6.6地震为例

区域地震波形对于震源研究非常重要,但限幅问题限制了区域地震台网数据的运用,并影响到震源参数测定的准确度.本文利用恢复后的芦山地震区域地震波形,研究了芦山地震的震级、点源机制解以及破裂过程.基于震中距99~300 km恢复前与恢复后地震数据获取的面波震级分别为7.01与7.06级.分别利用7个震中距150~250 km宽频带台站的恢复前和恢复后的数据反演点源机制解,与参考机制解相比,滑动角偏差自13°减小到了4°.基于7个震中距81~134 km的区域地震波形联合远场数据获得的震源破裂过程结果,其主要参数（如滑动分布、破裂速度等）与强地面运动波形联合远场数据得到的结果具有很好的一致性.研究结果表明,本文所采用的数据恢复方法具有较高的可靠性,有效提高了震源参数测定的准确度.     

General seismic wave and phase detection software driven by deep learning

We developed an automatic seismic wave and phase detection software based on PhaseNet, an efficient and highly generalized deep learning neural network for P- and S-wave phase picking. The software organically combines multiple modules including application terminal interface, docker container, data visualization, SSH protocol data transmission and other auxiliary modules. Characterized by a series of technologically powerful functions, the software is highly convenient for all users. To obtain the P- and S-wave picks, one only needs to prepare three-component seismic data as input and customize some parameters in the interface. In particular, the software can automatically identify complex waveforms (i.e. continuous or truncated waves) and support multiple types of input data such as SAC, MSEED, NumPy array, etc. A test on the dataset of the Wenchuan aftershocks shows the generalization ability and detection accuracy of the software. The software is expected to increase the efficiency and subjectivity in the manual processing of large amounts of seismic data, thereby providing convenience to regional network monitoring staffs and researchers in the study of Earth's interior.     

UNIFIED EARTHQUAKE CATALOG FOR CHINA’S SEAS AND ADJACENT REGIONS AND ITS COMPLETENESS ANALYSIS

China’s seas and adjacent regions are affected by interactions among the Eurasian plate, the western Pacific plate, and the Philippine Sea plate. Both intraplate and plate-edge earthquakes have occurred in these regions and the seismic activities are frequent. The coastal areas of China are economically developed and densely populated. With the development and utilization of marine energy and resources along with the development of national economy, the types and quantity of construction projects in the marine and coastal areas have increased, once an earthquake happens, it will cause huge damage and loss to these areas, therefore, the earthquake-related research for these sea areas cannot be ignored and the need for study on these areas is increasingly urgent. One type of essential basic data for marine seismic research is a complete, unified earthquake catalog, which is an important database for seismotectonics, seismic zoning, earthquake prediction, earthquake prevention, and disaster reduction. Completeness and reliability analysis of an earthquake catalog is one of the fundamental research topics in seismology.
At present, four editions of earthquake catalogs have been officially published in China, as well as the earthquake catalogue compiled in the national fifth-generation earthquake parameter zoning map, these catalogs are based on historical data, seismic survey investigations, and various instrumental observations. However, these catalogs have earlier data deadlines and contain the earthquake records for only the offshore regions of China, which are extensions of coastal land. Distant sea regions, subduction zones, and adjacent sea regions have not been included in these catalogs. Secondly, there were no cross-border areas involved in the compilation of earthquake catalogs in the past. It was not required to use magnitudes measured by other countries’ seismic networks and observation agencies to develop an earthquake catalog with a uniform magnitude scale, moreover, there was no formula suitable for the conversion of magnitude scale in China’s seas areas and adjacent regions. Little research has been conducted to compile and analyze the completeness of a unified earthquake catalog for China’s seas and adjacent regions. Therefore, in this study, we compiled earthquake data from the seismic networks of China and other countries for China’s seas and adjacent regions. The earthquake-monitoring capabilities of different sea areas at different time periods were evaluated, and the temporal and spatial distribution characteristics of epicentral location accuracy for China’s seas and adjacent regions were analyzed. We used the orthogonal regression method to obtain conversion relationships between the surface wave magnitude, body wave magnitude, and moment magnitude for China’s seas and adjacent regions, and established magnitude conversion formulae between the China Seismic Network and the M_L magnitude of the Taiwan Seismic Network and the M_S magnitude of the Philippine Seismic Network. Finally, we developed an earthquake catalog with uniform magnitude scales for China’s seas and adjacent regions.
On the basis of the frequency-magnitude distribution obtained from the magnitude-cumulative frequency relationship (N-T) and the Gutenberg-Richter(GR)law, we conducted a completeness analysis of the unified earthquake catalog for China’s seas and adjacent regions, Then, we identified the beginning years of each magnitude interval at different focal depth ranges and different seismic zones in the earthquake catalog.
This study marks the first time that a unified earthquake catalog has been compiled for China’s seas and adjacent regions, based on the characteristics of seismicity in the surrounding sea regions, which fills the gap in the compilation of the earthquake catalogue of China’s seas and adjacent areas. The resulting earthquake catalog provides a basis for seismotectonics, seismicity study, and seismic hazard analysis for China’s seas and adjacent regions. The catalog also provides technical support for the preparation of seismic zoning maps as well as for earthquake prevention and disaster reduction in project planning and engineering construction in the sea regions. In addition, by evaluating the earthquake-monitoring capability of the seismic networks in China’s seas and adjacent regions and analyzing the completeness of the compiled unified earthquake catalog, this study provides a scientific reference to improve the earthquake-monitoring capability and optimizing the distribution of the seismic networks in these regions.     

由地震分布丛集性给出断层参数的一种新方法

由于大范围内地质构造的复杂性和介质的非均匀性,发震断层面的几何形态一般十分复杂.如果大地震的破裂过程涉及多个断层的活动,则发震断层并非是单一断层平面,而是多个断层面的组合.利用地震空间位置分布丛集性,即震源点成丛位于断层面附近的假设,结合稳健扩充算法和主成分分析给出一种可以重构活断层网络三维空间结构的新方法.该方法每次从震源点集中处开始,利用假设检验扩充子断层面,并得到多个子断层面.接着按震源点属于最近断层面的准则把各子断层面内的震源点进行竞争,并根据一定假设合并和删除一些子断层面,最后用主成分分析确定每个子断层面参数.于是可根据地震事件目录给出一组矩形区域来描述断层面网络结构,其中每个矩形断层面由其位置、走向和倾角确定.通过计算机模拟发现,新方法可成功地重建模拟地震目录的断层面,最后用于南加州1992年6月28日发生的Landers地震部分余震目录中,得到各个子断层面参数与已知地质破裂或隐伏断层相当一致.     

Universality of the Seismic Moment-frequency Relation

—We analyze the seismic moment-frequency relation in various depth ranges and for different seismic regions, using Flinn-Engdahl's regionalization of global seismicity. Three earthquake lists of centroid-moment tensor data have been used the Harvard catalog, the USGS catalog, and the Huang et al. (1997) catalog of deep earthquakes. The results confirm the universality of the β-values and the maximum moment for shallow earthquakes in continental regions, as well as at and near continental boundaries. Moreover, we show that although fluctuations in earthquake size distribution increase with depth, the β-values for earthquakes in the depth range of 0–500 km exhibit no statistically significant regional variations. The regional variations are significant only for deep events near the 660 km boundary. For declustered shallow earthquake catalogs and deeper events, we show that the worldwide β-values have the same value of 0.60 ± 0.02. This finding suggests that the β-value is a universal constant. We investigate the statistical correlations between the numbers of seismic events in different depth ranges and the correlation of the tectonic deformation rate and seismic activity (the number of earthquakes above a certain threshold level per year). The high level of these correlations suggests that seismic activity indicates tectonic deformation rate in subduction zones. Combined with the universality of the β-value, this finding implies little if any variation in maximum earthquake seismic moment among various subduction zones. If we assume that earthquakes of maximum size are similar in different depth ranges and the seismic efficiency coefficient, χ, is close to 100% for shallow seismicity, then we can estimate χ for deeper earthquakes for intermediate earthquakes χ≈ 5%, and χ≈ 1% for deep events. These results may lead to new theoretical understanding of the earthquake process and better estimates of seismic hazard.     

Preliminary analysis of earthquake probability based on the synthetic seismic catalog

The analysis of seismic hazards relies on the statistical analysis of historical seismic data and the instrumental seismic catalog to obtain the regional earthquake recurrence interval and earthquake probability. The accuracy of analysis thus depends strongly on the completeness of the seismic data used. However, available seismic catalogs are too short or incomplete for the reliable analysis of the statistical characteristics of earthquakes. If a long-term synthetic seismic catalog can be generated using a physics-based numerical simulation, and the simulation results match the crustal deformation, seismicity, and other observations,then such a synthetic catalog helps us to further understand the characteristics of seismic activity and analyze the regional seismic hazard. In this paper, taking the northeastern Tibetan Plateau as a case study, we establish a three-dimensional visco-elastoplastic finite-element model to simulate earthquake cycles and the spatiotemporal evolution of earthquakes on the model fault system and obtain a seismic catalog on a time scale of tens of thousands of years. On the basis that the model satisfies the regional geodynamics of the northeastern Tibetan Plateau, we analyze seismicity on the northeastern Tibetan Plateau using the simulated synthetic earthquake catalog. The characteristics of earthquake recurrence at different locations and different magnitudes, and the long-term average probability of earthquake occurrence within the fault system on the northeastern Tibetan plateau are studied. The results are a reference for regional seismic hazard assessment and provide a basis for the physics-based numerical prediction of earthquakes.     

五指山形变台VP型宽频带倾斜仪记录地震特征

以五指山形变台为圆心,选取2016—2020年5 000 km范围内M_S≥2.0地震资料,分析该台VP型宽频带倾斜仪记录地震特征,得出倾斜仪对震中距5 000 km范围内地震的记录能力。结果表明,该倾斜仪对海南及周边中强以上地震同震响应明显。     

基于Datist的地震目录资料自动下载与数据存储

地震目录在地震预报和震害分析等工作中具有重要的科学研究价值。从网站上手动下载各个地震目录资料时效率比较低下,此外由于下载的地震目录格式固定,对开展更广泛的数据服务具有一定的局限性。为了提高地震目录资料的下载效率,采用节点流程的方法对地震目录文本进行自动下载,并将其解析后存储到数据库,实现了高效率的地震目录资料自动下载和数据存储功能,方法简单灵活,可为科研预报及相关地震数据应用服务提供相应的数据支持。     

USGS地震目录中4～5级震源深度异常地震可靠性初步研究:以南北地震带若干地震为例

在南北地震带地区,USGS全球地震目录中存在一些震源深度大于30km的地震.这些地震的震源深度是否可靠,对于研究这一地区的孕震机制、岩石圈强度和构造演化等科学问题具有重要意义.本文以南北地震带2012年发生的5个4~5级地震为例,利用区域地震台网的波形数据,基于sPL深度震相、短周期瑞利面波以及CAP等独立方法测定了其震源深度.结果表明:sPL深度震相和CAP方法给出的震源深度比较一致,差别小于2~3km,能够得到比较可靠的震源深度;短周期瑞利面波及其与P波振幅比也确定了地震震源深度较浅的特征.本文研究结果显示:宁夏会宁4.7级、云南富民4.8级和四川会东4.7级地震的震源深度约为8~12km左右,仍为发生于上地壳的地震,USGS地震目录给出的30km甚至更深的震源深度存在明显偏差;对于四川隆昌4.6和4.9级地震,本文给出的震源深度为1~2km,属于极浅源地震,USGS地震目录给出的10km和35km的震源深度结果尚需进一步改进.     

四川地震预警网外玛多7.4级地震预警分析研究

2021年5月22日2时4分青海省玛多县发生M7.4地震。四川地震预警台网成功地处理并产出了本次地震预警结果。本文针对四川地震预警系统产出的8次地震预警处理结果及预警减灾效能进行了分析。此次地震震中距四川地震预警网55km,首台触发后12.19s发布首次预警处理结果,前5次预警处理结果用时距离首台触发时间均小于35s。与正式地震目录对比,第1次预警处理结果震中位置偏差为17.1km,震级偏差为-1.8,盲区半径为121km。随着参与计算台站数量增多,震中位置偏差和震级偏差不断减小,盲区半径不断增大。由于地震破裂时间长和台站限幅等原因,8次预警震级均偏小。本次地震预警的有效获益区为震中距121～324km的环形区域,此区域内减灾效能显著。本次地震预警结果表明,四川地震预警系统在本次地震中预警产出效果较好,对于网外大地震预警有重要参考意义。