Regional Seismograph Network Research Articles

Graph‐Partitioning Based Convolutional Neural Network for Earthquake Detection Using a Seismic Array

AbstractWe present a deep‐learning approach for earthquake detection using waveforms from a seismic array consisting of multiple seismographs. Although automated, deep‐learning earthquake detection techniques have recently been developed at the single‐station level, they have potential difficulty in reducing false detections owing to the presence of local noise inherent to each station. Here, we propose a deep‐learning‐based approach to efficiently analyze the waveforms observed by a seismic array, whereby we employ convolutional neural networks in conjunction with graph partitioning to group the waveforms from seismic stations within the array. We then apply the proposed method to waveform data recorded by a dense, local seismic array in the regional seismograph network around the Tokyo metropolitan area, Japan. Our method detects more than 97% of the local seismicity catalog, with less than 4% false positive rate, based on an optimal threshold value of the output earthquake probability of 0.61. A comparison with conventional deep‐learning‐based detectors demonstrates that our method yields fewer false detections for a given true earthquake detection rate. Furthermore, the current method exhibits the robustness to poor‐quality data and/or data that are missing at several stations within the array. Numerical experiments using subsampled data demonstrate that the present method has the potential to detect earthquakes even when half of the normally available seismic data are missing. We apply the proposed method to analyze 1‐h‐long continuous waveforms and identify new seismic events with extremely low signal‐to‐noise ratios that are not listed in existing catalogs. We also show the potential portability of the proposed method by applying it to seismic array data not used for the training.

Large‐Scale Fracture Systems Are Permeable Pathways for Fault Activation During Hydraulic Fracturing

AbstractInduced seismicity due to fluid injection, including hydraulic fracturing, is an increasingly common phenomenon worldwide; yet, the mechanisms by which hydraulic fracturing causes fault activation remain unclear. Here we show that preexisting fracture networks are instrumental in transferring fluid pressures to larger faults on which dynamic rupture occurs. Studies of hydraulic fracturing‐induced seismicity in North America have often used observations from regional seismograph networks at distances of 10s of km, and as such lack the resolution to answer some of the key questions about triggering mechanisms. To carry out a more detailed analysis of the mechanisms of fault activation, we use data from a dense sensor array located at a hydraulic‐fracturing site in Alberta, Canada. The spatiotemporal distribution of event hypocenters, coupled with measurements of seismic anisotropy, reveal the presence of preexisting fracture corridors that allowed communication of fluid‐pressure perturbations to larger faults, over distances of 1 km or more. The presence of preexisting permeable fracture networks can significantly increase the volume of rock affected by the pore‐pressure increase, thereby increasing the probability of induced seismicity. This study demonstrates the importance of understanding the connectivity of preexisting natural fractures for assessing potential seismic hazards associated with hydraulic fracturing of shale formations and offers a detailed case exposition of induced seismicity due to hydraulic fracturing.

Recent seismicity in northern European Russia

A revised comprehensive catalog has been made for earthquakes that occurred in northern European Russia for the period between 2005 and 2017. The earthquake parameters were determined to greater accuracy using the same velocity model (BARENTS), the same location method (based on generalized beamforming), combining data from catalogs, bulletins, and (in part) from waveform supplied by regional seismograph networks in Russia and the Scandinavian countries. The resulting unified catalog formed a basis to assess the recent seismicity in northern European Russia, which occurs as low magnitude earthquakes. The distribution of earthquake epicenters is not uniform. The Fennoscandian Shield shows the highest activity (by seismicity rate and epicenter density), the next to follow are the northeastern Russian plate and the northern Urals Fold-Thrust Region. All earthquakes typically occurred at crustal depths.

Real-time passive monitoring with a fibre-optic ocean bottom array

To achieve high recovery rates, modern-day production management can benefit from not only snapshot images of the state of the reservoir at regular time intervals, but also continuous monitoring of the dynamic processes induced by pressure changes and fluid movement during production. Production management using time-lapse 4D snapshots is reactive, i.e., adjustments addressing the sweep efficiency or reservoir integrity can only be instigated once the next snapshot image is available after acquisition, processing and interpretation, often years later. For a more proactive reservoir management, it is important to have dynamic reservoir information in real time between the seismic time-lapse snapshots. Such information is contained in microseismic monitoring data and in surface or borehole deformation measurements. If sensors are permanently installed, this information comes at a negligible additional cost, provided that the data can be transferred to shore in real-time and processed automatically. Time-lapse 4D snapshot images are typically obtained over a period of years and are inadequately sampled for capturing dynamic reservoir changes taking place over much shorter time intervals, from hours to days. Such changes can include variations in the permeability caused by scaling or compaction (Barkved and Kristiansen, 2005), changes of the fluid phase owing to pressure variations (e.g., Osdal et al., 2006), unintended alterations of the flow paths owing to out-of-zone injections and fault reactivation (e.g., Schinelli et al., 2015), or movement in the overburden, potentially compromising the integrity of infrastructure in the form of casing failures or seafloor subsidence (e.g., Yudovich et al., 1989; Hatchell et al., 2017). While 4D seismic data can capture the cumulative effect of such processes by evaluating differences in still images every few years, they provide little information about when exactly the associated dynamic changes occurred and how they relate to changes in flow rate and pressure that may have been captured through continuous measurements in the wellbores accessing the reservoir. Microseismic events from within or around a producing reservoir can be indicative of reservoir fluid pathways and sub-seismic reservoir compartmentalization (e.g., Maxwell and Urbancic, 2001), or stress changes and associated production- related deformations in the vicinity (e.g., Teanby et al., 2004; Zoback and Zinke, 2002; Wuestefeld et al., 2011, 2013). Continuous monitoring of seismicity can also help in assessing deformation-related risks to infrastructure over the life of a field. Combined with pressure and flow rate, such data can provide the necessary information to capture dynamic processes in the reservoir right when they happen. In conjunction with geomechanical flow modelling, production optimization strategies can thus be validated and adjustments can be properly planned at an early stage. The result will be an improved sweep efficiency with a further increased recovery factor, as well as better risk assessment with the avoidance of potentially costly mitigation actions. A better understanding of, and continuous information about, the reservoir dynamics may even help to plan a 4D seismic strategy better. This can include better definition of suitable intervals for the acquisition of time-lapse images. These intervals could be irregular, depending on the state of reservoir development and type of recovery method. Continuous monitoring may also provide a means to high-grade areas of the reservoir for partial 4D imaging at lower cost and faster turnaround in between ‘full’ time-lapse surveys (Hatchell et al., 2013). In this paper we discuss the general ability to monitor microseismic events in an offshore setting and presents results from a real-time monitoring pilot in Norway. We validate the concept of continuous real-time monitoring from a fibre-optic deep-water installation by comparing our automatic detections with data from a regional seismograph network.

Temporary seismological observations in the area of the 2012–2013 Tolbachik Fissure Eruption: Results

The seismicity that accompanied the Tolbachik Fissure Eruption was recorded by additional seismic stations that were installed in the southern Klyuchevskoi Volcanic Cluster area in January to October 2013. We used broadband (0.033–50 Hz) three-component digital Guralp CMG-6TD seismometers. This temporary network provided seismicity data at a lower energy level than can be done using the regional seismograph network of Kamchatka. The processing of the resulting digital records supplied data for compiling a catalog of over 700 M L = 0–3.5 (K S = 1.5–8.5) earthquakes, which is an order of magnitude greater than the number of events located by the regional network for the same period of time. The seismicity in the area of Ploskii Tolbachik Volcano was found to concentrate mostly in spatially isolated areas during the eruption. The main isolated clusters of earthquakes were identified both in the eruption area itself and along the periphery of Ploskii Tolbachik Volcano, in the area of the Zimina volcanic massif, and in the Tolud epicenter zone; the eruption zone was not dominant in the seismicity. The region of a shallow seismicity increase beneath Ploskii Tolbachik before the eruption was not found to exhibit any increased activity during the time the temporary seismograph network was operated, which means that a seismicity inversion took place at the beginning of the eruption. We discuss the question of what the earthquake-generating features are that we have identified.

Performance Evaluation of the Regional Seismograph Network in Northeast British Columbia, Canada, for Monitoring of Induced Seismicity

An important factor in mitigation of seismic hazard from induced seismicity is properly established seismic networks suitable for consistent identification of small‐to‐moderate events (magnitudes less than four). Here, we evaluate the performance of the newly established regional broadband seismic network in northeast British Columbia, Canada. The seismic network was designed for monitoring of induced seismicity due to oil and gas operations related to hydraulic fracturing and fluid injection in the region. We use regional and local earthquake catalogs for the period 1985–2015 to analyze magnitude of completeness and epicentral uncertainty. We also perform a theoretical assessment of minimum detectable magnitude across the study region based on analysis of ambient noise and simulated ground motions. From the frequency–magnitude distribution of the reported events in the regional earthquake catalog, the magnitude of completeness has decreased ∼1 magnitude unit from ∼3 in the periods 1985–2013 to ∼2 in the period 2013–2015 as a result of the establishment of new stations. The minimum detectable magnitude in the region is 1.6–2.6 based on the signal‐to‐noise ratio (SNR) of 10 and higher at four or more stations. By comparing the regional and dense array catalogs, we determine that the error in epicentral location in well‐constrained areas by seismic stations is below 3 km in both east–west and north–south directions. However, location uncertainties can be up to 10 km in the east–west direction in areas where the current regional network is sparse. The magnitude detection threshold in the Montney Play, where most of the current oil and gas activities are taking place, can be further reduced by up to 1 magnitude unit with the addition of four new stations.

Automatic classification of seismic events within a regional seismograph network

This paper presents a fully automatic method for seismic event classification within a sparse regional seismograph network. The method is based on a supervised pattern recognition technique called the Support Vector Machine (SVM). The classification relies on differences in signal energy distribution between natural and artificial seismic sources. We filtered seismic records via 20 narrow band-pass filters and divided them into four phase windows: P, P coda, S, and S coda. We then computed a short-term average (STA) value for each filter channel and phase window. The 80 discrimination parameters served as a training model for the SVM. We calculated station specific SVM models for 19 on-line seismic stations in Finland. The training data set included 918 positive (earthquake) and 3469 negative (non-earthquake) examples. An independent test period determined method and rules for integrating station-specific classification results into network results. Finally, we applied the network classification rules to independent evaluation data comprising 5435 fully automatic event determinations, 5404 of which had been manually identified as explosions or noise, and 31 as earthquakes. The SVM method correctly identified 94% of the non-earthquakes and all but one of the earthquakes.The result implies that the SVM tool can identify and filter out blasts and spurious events from fully automatic event solutions with a high level of accuracy. The tool helps to reduce the work-load and costs of manual seismic analysis by leaving only a small fraction of automatic event determinations, the probable earthquakes, for more detailed seismological analysis. The self-learning approach presented here is flexible and easily adjustable to the requirements of a denser or wider high-frequency network.

The shear-wave splitting in the crust and the upper mantle around the Bohai Sea, North China

In order to infer the distribution of local stress and the deep geodynamic process in North China, this study detects seismic anisotropy in the crust and upper mantle beneath the Bohai Sea area. A total of 535 local shear-wave and 721 XKS (including SKS, PKS and SKKS phases) splitting measurements were obtained from stations in permanent regional seismograph networks and a temporary seismic network called ZBnet-E. The dominant fast polarization orientation of local shear-waves in the crust is nearly East-West, suggesting an East-West direction of local maximum compressive stress in the area. Nearly North-South fast orientation was obtained at some stations in the Tan-Lu fault belt and the Zhang-Bo seismic belt. The average fast orientation from XKS splitting analysis is 87.4° measured clockwise from the North. The average time-delays of XKS splitting are range from 0.54s to 1.92s, corresponding to a 60–210km thick layer of anisotropy. The measured results indicate that upper mantle anisotropy beneath Bohai Sea area, even the eastern part of North China, is mainly from asthenospheric mantle flow from the subduction of the Pacific plate. From the complicated anisotropic characteristics in this study, we infer that there might be multiple mechanisms in the crust and upper mantle around the Bohai Sea area that led to the observed anisotropy.

Crustal thickness and vP/vS ratio in Shanxi Graben, China

Shanxi Graben is in the middle part of the North China Craton, from south to north. With the teleseismic data recorded by Regional Seismograph Networks and the temporary ZBnet-W Seismic Array around east part of Shanxi Graben, we measured the crustal thickness and vP/vS ratio beneath each station using theH-κ stack of receiver functions. The observed crustal thickness shows obvious lateral variation, increasing gradually from east to west in the Shanxi Graben. Beneath the Shanxi Graben the crust is relatively thicker than both sides of the south and the north. In addition, the vP/vS ratio in the north of study zone is higher than that in the south. The highest vP/vS ratio exists in the crust of the Xinding basin and the Datong basin. Our study also suggests that high velocity ratio might result from the strong activities of the magmation and volcanism.

Shear-wave splitting beneath Yunnan area of Southwest China

Systematic analyses of seismic data recorded by the Yunnan regional seismograph network reveal significant crustal and upper mantle anisotropy. Splitting of the S phase of local earthquakes and teleseismic SKS, PKS, and SKKS phases indicates time-delays from 1.60 ms/km to 2.30 ms/km in the crust, and from 0.55 s to 1.65 s in the upper mantle which corresponds to an anisotropic layer with a thickness about between 55–165 km. The polarization orientations of fast shear waves in the crust are complicated with a predominantly north-south direction, and the mantle anisotropy has a nearly west-east direction. Our results show different deformation styles and mechanisms exist between the crust and upper mantle.

Shear wave splitting in the crust in North China: stress, faults and tectonic implications

SUMMARY This paper uses data from the Capital Area Seismograph Network in North China (NC), currently the largest regional seismograph network in China to investigate the compressional stress orientations from the polarizations of faster shear waves (PFS), a parameter of shear wave velocity anisotropy. Data are restricted to earthquakes deeper than 5 km to reduce the influence of the heterogeneous uppermost few kilometres of the crust. The results show that the regional compressional stress direction is ENE–WSW to nearly EW in NC, which agrees with seismic mechanisms and drilling and GPS data. Predominantly, PFS are also nearly EW in NC and are influenced by faults, geology and tectonics. PFS orientations at stations on seismically active faults are consistent with the strike of strike-slip faults and indicate large seismically active faults breaking, or nearly breaking, the free surface. PFS orientations are influenced by both regional compressional stress and by nearby faults. It is an effective way to study detailed spatial distribution of crustal compressional stress by shear wave anisotropy with dense seismograph network. Shear wave splitting may also be used to indicate hitherto unknown faults.

Time-lapse slip variation associated with a medium-size earthquake revealed by "repeating" micro-earthquakes: the 1999 Xiuyan, Liaoning, <i>M</i><sub>S</sub> = 5.4 earthquake

Abstract. We obtained the time-lapse cumulative slip before and after the 29 November 1999, Xiuyan, Liaoning, China, Ms = 5.4 earthquake by using "repeating events" defined by waveform cross-correlation. We used the seismic waveform data from the Liaoning Regional Seismograph Network from June 1999 to December 2006. Two "multiplets" located near the seismogenic fault of the 1999 Xiuyan earthquake and the 4 February 1975, Haicheng Ms = 7.3 earthquake, respectively, were investigated. For the "multiplet" that occurred before and after the 1999 Xiuyan earthquake, apparent pre-shock accelerating-like slip behavior, clear immediate-post-seismic change, and relaxation-like post-seismic change can be observed. As a comparison, for the "multiplet" near the 1975 Haicheng earthquake which occurred a quarter century ago, the cumulative slip appears linear with a much smaller slip rate.

Structure of the upper mantle and transition zone beneath Southeast Asia from traveltime tomography

Tomographic images of the mantle beneath East Asia were obtained from the inversion of traveltime data from global and regional seismograph networks and from temporary arrays on and around the Tibetan plateau. Our results are consistent with previous studies but the unprecedented resolution of mantle heterogeneity provides new insight into the large‐scale tectonic framework of the continental India‐Asia collision in the western part of the study region and subduction of the oceanic lithosphere in the east. In the realm of continental collision, west of ∼100°E, a relatively slow P‐wave speed characterizes the upper mantle beneath much of the Tibetan plateau but the wave speed is high beneath cratonic India, the southern and western part of the Tibetan plateau, Hindu‐Kush, and the Tian Shan. In the subduction realm, east of ∼110°E, the main structures are (i) pronounced low‐wave‐speed anomalies at a depth of between 100 and 400 km beneath Asia's southeastern seaboard and the back‐arc regions of ongoing subduction; (ii) narrow, fast anomalies in the upper mantle beneath major subduction zones; and (iii) widespread fast anomalies at a depth of 500–700 km beneath the Sea of Japan, the northern part of the Philippine Sea plate, and southeastern China. If the latter anomalies represent stagnant slabs, their fragmented nature and large lateral extent suggest that they are produced by different episodes of subduction beneath western Pacific island arcs, along the old SE margin of Asia, or during the Mesozoic collision of cratonic units in Southeast Asia. Attribution to ancient subduction systems implies that slab fragments can reside in the transition zone for (at least) several tens of millions of years. Shallow, slow anomalies beneath the Red River fault region connect to deep anomalies beneath the South China fold belt and South China Sea, suggesting a causal relationship between the evolution of the continental lithosphere of SW China and deeper mantle processes. Between the collision and the subduction realms, tomography reveals high‐wave‐speed continental roots beneath the western part of the North China craton (Ordos block) and the South China, or Yangtze, craton (Sichuan Basin) to a depth of ∼300 km.

John Clark Lahr (1944-2009)

![][1] John Clark Lahr (1944–2009) John Clark Lahr, 64, passed away on 17 March 2009 at his home in Corvallis, Oregon, from malignant brain cancer. John was a curious, resourceful, hands-on seismologist who increased understanding of earthquakes, tectonics, and volcanic processes, particularly in Alaska. He inspired teachers and youngsters to listen to and explore the Earth, and his enthusiasm and generosity touched many lives. Born in 1944 in Indianapolis, John lived most of his school years in Buffalo, New York, where he loved taking things apart to understand how they worked and then how they could be fixed or improved. This quality endured throughout his life. John earned a B.S. degree in physics and math in 1966 from Rensselaer Polytechnic Institute in Troy, New York, and then pursued graduate work at Columbia University and Lamont Geological Observatory during the exciting years of the plate tectonics revolution. While earning an M.S. degree, awarded in 1970, he was introduced to Alaskan tectonics and earthquakes through survey campaigns to determine if major strike-slip faults in Alaska were creeping. John joined the U.S. Geological Survey in 1971 in Menlo Park, California, where he developed software to locate subduction-zone earthquakes recorded by a sparse regional seismograph network. He applied these methods … [1]: /embed/graphic-1.gif

Source mechanism of strong aftershocks (M s⩾5.6) of the 2008/05/12 Wenchuan earthquake and the implication for seismotectonics

Dozens of >M5, hundreds of >M4, and much more >M3 aftershocks occurred after the 2008/05/12 Wenchuan earthquake, which were well recorded by permanent and portable seismic stations. After relocated with P arrival, the >M3 aftershocks show two trends of distribution, with most of the aftershocks located along the north-east strike consistent with Longmenshan fault system, yet there is a north-west trend around the epicenter. It seems that substantially more aftershocks occur in regions with crystalline bedrocks. Then we collected waveform data from National Digital Seismograph Network and regional seismograph network of China, and employed “Cut and Paste” method to obtain focal mechanisms and depths of the big aftershocks (M⩾5.6). While most of those aftershocks show thrust mechanism, there are some strike slip earthquakes in the northern-most end of the rupture. Focal mechanisms show that the events located on the southern part of central Beichuan-Yingxiu Fault (BY) are mainly thrust earthquakes, which is consistent with initial mechanism of the main shock rupture. In the north part the aftershocks along the BY are also dominated by thrust slip, which is quite different from the right slip rupture of the main shock. Around Qingchuan-Pingwu Fault, the focal mechanisms are dominated by right-slip rupture with large depths (∼18 km). So we suspected that in the north part the main shock might rupture on two faults: Beichuan Fault and Qingchuan-Pingwu Fault. The complex pattern of aftershock mechanisms argues for presence of a complicated fault system in the Longmenshan area.

Compressive stress field in the crust deduced from shear-wave anisotropy: an example in capital area of China

The rocks in the crust are pervaded by stress-aligned fluid-saturated microcracks, and the complex fault tectonics and stress control the configuration of the microcracks, however shear-wave splitting could indicate this kind of characteristics. In this paper, Capital Area Seismograph Network (CASN), the widest scope and highest density of regional seismograph network presently in China, is adopted to deduce the principal compressive stress field distribution pattern from polarizations of fast shear-waves, based on shear-wave splitting analysis. The principal compressive stress in capital area of China is at NE85.7°±41.0° in this study. Compared with the results of principal compressive stress field in North China obtained from other methods, the results in this study are reliable in the principal compressive stress field distribution in capital area. The results show that it is an effective way, although it is the first time to directly obtain crustal stress field from seismic anisotropy. It is effectively applied to the zones with dense seismograph stations.

Repeating earthquakes recorded by Liaoning Regional Seismograph Network

In the list of ‘repeating pairs’ or ‘doublets’ of earthquakes in China identified by Schaff and Richards using teleseismic waveform cross-correlation, there were 23 repeating pairs located in Liaoning Province. In this study the waveforms of these events were cross-correlated using records from Liaoning Regional Seismograph Network (LRSN), and the ‘repeating events’ in the sense of regional waveform cross-correlation were obtained. The result was compared with that of Schaff and Richards and was used for the assessment of the seismic phase picking and event location practice of LRSN. The result shows that ‘repeating events’ in the sense of teleseismic waveform cross-correlation and those in the sense of regional waveform cross-correlation have significant difference, although with some overlap. However, the overall assessment of the location accuracy and the phase pick errors of LRSN by using these two sets of ‘repeating events’, respectively, provides similar results, while ‘repeating events’ in the sense of regional waveform cross-correlation seem to be better performing in such an assessment. With the assumption that the separation between the ‘repeaters’ be less than 1 km, the uncertainty in routine earthquake location of LRSN is estimated to be below 5 km, with the average of 2 km. In the observational bulletins of LRSN the time error in phase picking is estimated to be within ±1s for 94% Pg readings and for 88% Sg readings.

Sequence of strong intraplate earthquakes in the Kodiak Island Region, Alaska in 1999–2001

A sequence of strong earthquakes was registered in 1999–2001 in the Kodiak Island region of the Alaska‐Aleutian subduction zone. Two Mw 7 earthquakes occurred in December, 1999 and January, 2001 and an Mw 6.5 event occurred in July, 2000. These events and their aftershocks recorded by the regional seismograph network were relocated using a Joint Hypocenter Determination (JHD) method. Regional broadband data were used to obtain seismic moment tensors for the main shocks and their largest aftershocks. Relocation and moment tensor inversion results indicate that these events originated inside the subducting Pacific plate. The focal mechanisms indicate down‐dip tension with the fault planes being nearly vertical and parallel to the strike direction of the subducting plate. The 1999 and 2000 events were located down‐dip of the locked portion of the megathrust, while the 2001 event was located directly beneath it.

Discrepancies in earthquake location between ISC and other agencies

The different operating requirements of the International Seismological Centre (ISC) from those of the National Earthquake Information Service of the US Geological Survey(NEIS), and of the prototype International Data Center to monitor the Comprehensive Test Ban Treaty (pIDC), result in some discrepancies between earthquake locations computed by the three agencies. For larger events recorded by many stations the differences are small, but for some smaller events differences in location of up to 20° may occur. The largest discrepancies are found for small events in areas where later analysis by ISC has the benefit of additional readings from regional seismograph networks and where ISC has made a different interpretation of the station readingsavailable to pIDC and NEIS. We identify regions where such discrepancies occur most frequently, and give some examples for which the augmented data set of ISC has resulted in significant improvement for specific earthquakes. NEIS and, particularly, pIDC produce their results more speedily than ISC, and these form a valuable starting point for the later, more complete ISC analysis, which is commonly considered the most definitive compilation of global earthquake information.

Stress pattern in the double seismic zone beneath Cook Inlet, south-central Alaska

Relocation of south-central Alaskan earthquakes recorded by the regional seismograph network, using the joint hypocenter determination method, revealed a double seismic zone in the underthrusting Pacific plate beneath the Cook Inlet area. The lower zone is separated from the upper one by a decrease in seismic activity within a 5–10 km wide band, and extends from a depth of 50 km to about 90 km. Stress tensor inversion for earthquakes in the upper and lower zones shows different stress regimes. In the upper zone minimum principal stress is oriented in the down-dip direction and maximum principal stress is normal to the Wadati-Benioff zone (WBZ). In the lower zone both maximum and minimum principal stresses are horizontal and rotated by 40–60° from the down-dip direction of the WBZ. Below 80–90 km depth, stress tensor inversion results show down-dip orientation of the minimum principal stress and the maximum stress aligned in the direction of convergence of Pacific plate.