Hi-net
History of NIED borehole observations
Hi-net is a nationwide high-sensitivity seismograph network in Japan that was constructed primarily after the 1995 Kobe earthquake (Fig. 1 and Additional file 1: Figure S1). The observation style and technology to install velocity meters and tiltmeters at the borehole bottoms were basically established by the 1980s. In 1970, National Research Center for Disaster Prevention (NRCDP), which was reorganized into NIED in 1990, started a deep borehole observation project to investigate crustal activity in and around the Tokyo Metropolitan Area as part of a national earthquake prediction project. This densely populated and extremely prosperous region sits atop sedimentary layers that are between 2000- and 4000-m-thick. Therefore, sensors need to be installed at the bottom of deep boreholes located at the pre-tertiary basement to obtain high-quality data with good signal-to-noise ratios (S/Ns). To conduct deep borehole observations at such depths, it was first necessary to develop instruments capable of operating in extremely high temperature and pressure environments. The development and installation process took about 10 years, and NRCDP began observations at three sites, Iwatsuki, Shimohsa, and Fuchu, with sensor depths of 3510, 2300, and 2750 m, respectively, by 1981 (Takahashi 1982).
In addition to these deep borehole stations, a large-scale network for microearthquake and ground tilt observations was constructed in the Kanto and Tokai areas including the Tokyo Metropolitan Area during the period from the late 1970s to the early 1980s (Fig. 2). Most of these stations are composed of mainly 100-m boreholes enclosed in casing pipes with a diameter of 4 in. For this network, it was necessary to develop a small-sized, force-balanced pendulum tiltmeter. The measurement range of the tiltmeter is ± 2×10−4 radians and the resolution is 6 × 10−9 radians, and the long-term drift was found to be smaller than a few microradians per year in the half year of measurements taken after installation (Sato et al. 1980). 50 stations were constructed in the Kanto-Tokai area and have been operated along with the three deep borehole stations since 1983 (Hamada et al. 1982; Okada 1984). Additional 14 semi-deep borehole stations with sensor depths of 1000 to 2000 m and a deep borehole station, Koto, with a sensor depth of 3000 m were constructed in the 1990s (Okada et al. 2000). At the same time, an ocean bottom cable system was installed at six seismic and three pressure observation sites deployed as part of the Earthquake and Tsunami Monitoring Cable (ETMC) system in the Sagami Bay (Eguchi et al. 1998). Approximately 70 stations in operation to observe shallow microearthquakes in this area are collectively known as the Kanto and Tokai Crustal Activity Observation Network as shown in Fig. 2.
Hi-net as fundamental observation
After the 1995 Kobe earthquake, based on the HERP policy related to the fundamental survey and observation, Hi-net was established (Okada et al. 2004; Obara et al. 2005; Shiomi et al. 2009). The construction of this network was driven by the need to evaluate the largest possible earthquakes and the long-term occurrence potentials of earthquakes. For this purpose, we need to precisely evaluate the bottom of the seismogenic layer at depths of 10 to 20 km from hypocenters of shallow crustal earthquakes, because the largest possible crustal earthquakes are considered to be strongly related to the seismogenic layer thickness. To constrain the hypocentral depths to around 10 to 20 km, the nearest epicentral distances from the observation stations are required to be within 20 km; therefore, Hi-net station intervals were set to be approximately 20 km. Furthermore, to acquire much data for the evaluation of the seismogenic layer thickness, it is necessary to observe quite weak signals from microearthquakes contaminated by various types of noise. Therefore, Hi-net seismometers were installed at depths of 100 m or more.
HERP set a policy stipulating that data from fundamental observations would be widely shared and used for both earthquake mitigation and deepening our understanding of earthquake phenomena. Based on this policy, data exchange among NIED, JMA, universities, and institutes was promoted. JMA began analyzing the combined data as a unified seismic catalog that is widely distributed, and NIED, in its role as the national data center responsible for archiving and distribution, began storing waveform data and continuously transmitting the data to JMA, as well as to universities and institutes. All other interested researchers can acquire continuous and/or event-triggered waveform data directly from the Hi-net website.
Observation instruments
The primary Hi-net sensor is a short-period velocity meter manufactured by Akashi Co., Ltd. (now Mitutoyo Corporation) which is a moving coil-type sensor with a natural frequency of 1 Hz and a flat sensitivity of approximately 200 V/m/s above this frequency. To ensure the quality of velocity meters, NIED routinely examines the test coil responses of these instruments everyday using an electrical calibration mechanism and monitors their natural frequencies and damping constants (Obara et al. 2005; Ueno et al. 2015). In addition to the installed short-period three-component velocity seismometers, Hi-net also includes two-component tiltmeters. Analog voltage signals from the three-component velocity seismometers and two-component tiltmeters are digitized by an analog-to-digital (A/D) converter. This device has a resolution of 27 bits for 20 or 100 Hz sampling and a dynamic range of 130 dB based on oversampling and decimation filtering (Obara et al. 2005; Shiomi et al. 2005). NIED has been developing the Hi-net A/D converters for the last two decades. The latest A/D converter, which is manufactured by Keisokugiken Corporation, contains an energy-efficient decimation filter based on a field-programmable gate array (FPGA), and the A/D converter can continue operating for at least 3 days without an external power supply when a power outage. The A/D converter, batteries, and telephone-related equipment are installed in huts at each station.
Integrated observation networks and data
Borehole Hi-net stations, along with their huts, were constructed at approximately 700 sites during the period from 1997 to 2004. Most of these stations, which were constructed by NIED by 2000, began providing data in October 2000. In addition, NIED constructed 15 stations with 50 m boreholes without huts for use in an integrated observation research project for the Itoigawa-Shizuoka Tectonic Line active fault system during the period from 2005 to 2010 (Asano et al. 2010). Later, these 15 stations were included in Hi-net. In 2006, the existing A/D converters and telemetry system of the Kanto-Tokai Crustal Activity Observation Network (Kanto-Tokai Network) were replaced with the same system as that used throughout Hi-net, and these stations of the Kanto-Tokai Network were incorporated into Hi-net. Almost all of the approximately 800 stations (Fig. 1 and Additional file 1: Figure S1) equipped with short-period velocity meters are operated integrally as Hi-net.
Digitized data packets from these stations in WIN32 format (Urabe and Tsukada 1992; Shiomi et al. 2009) are continuously transmitted to the NIED DMC over Transmission Control Protocol/Internet Protocol (TCP/IP) through an IP-VPN based on telephone lines (Obara et al. 2005). NIED and NTT-Com developed this transmission system, which is now operated by NTT-Com, for EarthLAN. To supply A/D converted data to the NIED DMC, NTT-Com operates a COMBOX at each station and at the EarthLAN control center. The COMBOX at each station continuously receives and stores data from the A/D converter, and then sends the data to the control center. The control center integrates the data from Hi-net stations, and directly sends them to the NIED DMC and JMA in Tokyo and Osaka at a latency of typically less than 0.5 s. If there is packet loss, the COMBOX resends data according to a resending request from the control center. The COMBOX has sufficient storage capacity for 7-day worth of data and an uninterruptible power supply (UPS) that can operate for 24 h in the event of a power outage.
In 2002, a data exchange collaboration program was started among NIED, JMA, universities, and institutes. As a result, all Hi-net data are currently exchanged through the TDX installed in the EarthLAN control center, except for Hi-net data which are directly sent to the NIED DMC and JMA. Every day, the NIED DMC receives and stores approximately 70 GB of data in total from approximately 1500 stations belonging to Hi-net, JMA, universities, and institutes across Japan. The data can also be accessed on the Internet approximately 2 h after its acquisition. In the early 2000s, it was impossible to keep all the stored data available on searchable disk storage due to size limitations. Today, disk storage capable of holding several hundred terabytes of data is available, so Internet users can access all of the stored data on the Hi-net website.
Contributions
Acquired data are automatically analyzed for quality control and research utility at NIED. The approximately 300 to 900 earthquakes that are identified every day are shown on the Hi-net website. In the case of major earthquakes, the results and their interpretations are reported to the HERP Earthquake Research Committee.
Densely deployed stations and observations with high dynamic range in the quiet conditions at borehole bottoms are advantageous for detecting microearthquakes. Obara et al. (2005) showed that the ability to detect crustal microearthquakes with magnitudes of 1.5 to 0.9 in the Wakayama area in southwestern Japan was improved by Hi-net. Shiomi et al. (2009) showed similar improvements for the Chugoku area in southwestern Japan.
A long-history hypocenter catalog can be a useful tool for background seismicity investigations. Figure 3 shows the distribution of shallow earthquakes that have occurred in central Japan over the last 19 years. We can identify many earthquake clusters, several of which correspond to recent moderate earthquakes and their aftershocks, while others are related to historical earthquakes such as the 1891 Nobi earthquake, or active faults such as the Itoigawa-Shizuoka Tectonic Line active fault system (The Headquarters for Earthquake Research Promotion 2015). The 1891 Nobi earthquake is one of the largest crustal earthquakes in recorded Japanese history. The time decay of aftershock activity for this earthquake was investigated by Utsu (1961). According to his empirical equation of time-dependent aftershock decay, the recent seismicity of felt aftershocks is expected to be approximately four events per year. Assuming that the minimum magnitude of a felt earthquake is approximately Mj 3.0 and has a b-value of 1.0, the seismicity of microearthquakes with magnitudes larger than 1.0 is expected to be approximately 400 events per year. Because recent observed seismicity is consistent with model-predicted aftershock activity, the seismicity of this cluster can be considered to be the aftershocks of the 1891 Nobi earthquake.
More detailed hypocenter distributions have been investigated using the double-difference relocation technique (Waldhauser and Ellsworth 2000). For example, Yano et al. (2017) relocated 1.1 million shallow earthquakes that occurred in inland areas all over Japan and provided their results as the Japan Unified hI-resolution relocated Catalog for Earthquakes (JUICE) on the Hi-net website (www.hinet.bosai.go.jp/topics/JUICE/?LANG=en, Accessed 30 Apr 2020). They also showed that the seismogenic depth inferred from this catalog agrees with a recent moderate earthquake in central Japan.
K-NET and KiK-net
K-NET and KiK-net overview
NIED currently operates two large strong motion networks, K-NET and KiK-net (Kinoshita 1998; Aoi et al. 2004, 2011; Fujiwara et al. 2007; Kunugi et al. 2009; Fig. 1 and Additional file 1: Figure S2). After the 1995 Kobe earthquake, K-NET was planned as a nationwide strong motion seismograph network research project of NIED. The “K” in K-NET is an abbreviation for the Japanese word “kyoshin”, which means strong ground motion. The K-NET construction plan is based on the Basic Plan for Nationwide Deployment of Strong Motion Seismographs (Strong-Motion Earthquake Observation Council 1988) formulated by the Strong-Motion Earthquake Observation Council in 1988. This plan called for the uniform deployment of several thousand strong motion seismographs throughout Japan.
Later, when HERP established policies regarding the fundamental survey and observation for earthquake research, it was recommended that a strong motion seismometer and a high-sensitivity seismometer be installed together at each station to prevent the saturation of seismic records due to strong motions. Following that recommendation, NIED installed accelerometers on the ground surface and at the bottom of each borehole as a vertical array, in addition to high-sensitivity seismometers, at newly constructed Hi-net stations. The strong motion seismograph network, which consists of stations with these surface and downhole accelerometer pairs, operates independently as KiK-net. The name “KiK-net” refers to the strong motion network planned under the fundamental survey and observation for earthquake research because the Japanese word “kiban” means fundamental.
K-NET began operations in June 1996, just one and half years after the 1995 Kobe earthquake. The network initially consisted of 1000 stations, separated by an average distance of about 20 km. Later, more stations with three-component strong motion accelerometers were added, including stations installed before 1995 and six accelerometer-equipped seafloor stations of the ETMC system in the Sagami Bay (Eguchi et al. 1998). Currently, the number of stations is 1037. Because K-NET was designed to record strong motions in residential areas, most stations are located on the grounds of public spaces such as government office buildings, schools, and parks. Each K-NET station is equipped with standardized observation facilities that are installed on a site 3 m × 3 m square. A typical station consists of a lightweight hut made of fiber-reinforced plastic (FRP), a concrete base on which a set of accelerometers and recorders is installed, facilities for electric power and communications, and a protective fence. The hut is designed to withstand 4 m of snow. In places where the temperature falls below -20 degrees Celsius, the base on which the seismograph is installed extends from the ground surface to approximately 80 cm.
KiK-net is a strong motion seismograph network consisting of pairs of accelerometers installed at the ground surface and at the bottom of a borehole, and each KiK-net station shares a borehole and observation facilities with a Hi-net station. Currently, KiK-net consists of about 700 stations separated by an average distance of about 20 km. The network covers almost all of the Japan Islands, except for Okinawa Prefecture and most remote islands. Most KiK-net stations are located on rock or mostly rock sites covered with a thin weathered layer because they share facilities with Hi-net stations that are designed for microearthquake observations. Each KiK-net station has an observation borehole drilled to a depth of 100 m or more. 13 stations have boreholes deeper than 2000 m, 12 of which are in the Kanto region and the other (OSKH02) is located in Konohana, Osaka Prefecture. The deepest KiK-net borehole (SITH01), which is 3510 m deep, is located in Iwatsuki, Saitama Prefecture, and it is currently the deepest observation facility for strong ground motion in Japan.
Observation instruments
All of the strong motion seismographs installed at K-NET stations, K-NET95, K-NET02, K-NET02A, and K-NET11 listed in order of development, are specially designed for use by this network. At this point in time, all K-NET95 and K-NET02 have been replaced by K-NET02A and K-NET11.
K-NET95, the first-generation seismograph developed for K-NET, is manufactured by Akashi Co., Ltd. and consists of a three-component accelerometer and a data logger equipped with a 24-bit A/D converter with a sampling frequency of 100 Hz. The maximum measurable acceleration is ± 2000 cm/s/s and the effective dynamic range is 114 dB or more. A V403BT accelerometer with a nominal full-scale range of ± 3g, manufactured by Akashi Co., Ltd., is used. The internal clock of the recording system, which has a precision of one ppm, is calibrated by the Global Positioning System (GPS) signal every hour to provide an accuracy of five ms. During a power outage, K-NET95 can function for about 20 h using its built-in battery. K-NET95 is equipped with 8 MB of flash memory and can store a total of about 2.5 h of three-component data digitized at a sampling frequency of 100 Hz. Strong motion data are stored only when the ground acceleration exceeds the pre-set trigger condition. The data stored in K-NET95 can be obtained over dial-up connection from the NIED DMC.
K-NET02, the second-generation K-NET seismograph developed to replace K-NET95, is manufactured by OYO Corporation and began operating in 2003. K-NET02 is improved to measure a maximum acceleration of ± 4000 cm/s/s and has an effective dynamic range of more than 132 dB. K-NET02 is equipped with 512 MB of flash memory and a 24-bit A/D converter with a sampling frequency of 100 Hz. The timing accuracy is improved to be 0.1 ms. K-NET02 is equipped with an EpiSensor FBA-ES-deck, manufactured by Kinemetrics, Inc., which has a nominal full-scale range of ± 4g. One of the advantages of K-NET02 and later seismograph models is that they use the Linux operating system (OS), which means that they can be easily programmed to perform a wide variety of functions. K-NET02, K-NET02A, and K-NET11 can share most programs because they use the same OS. K-NET02 automatically establishes a connection over a telephone line to the NIED DMC within a few seconds after being triggered by an event, and can send waveform data even while recording. This feature significantly reduces the time required to collect data and telephone line congestion, which often occurs immediately after a major earthquake. K-NET02 is officially approved by JMA as a seismic intensity meter, and seismic intensity is automatically measured and sent to the NIED DMC within 1 min after triggering. The seismic intensities recorded by K-NET are broadcast within one and half to two minutes on television, radio, and the Internet through JMA. This information contributes to the decision-making and actions of national and local governments and private sector post-earthquake responses. During a power outage, K-NET02 can record strong motion data for 7 days using its backup battery and send seismic intensity data to the NIED DMC for 1 day.
K-NET02A is a modified version of K-NET02 that uses a different accelerometer JA-40GA04, which has a nominal full-scale range of ± 4g and is manufactured by Japan Aviation Electronics Industry, Ltd. Because the hinge supporting the proof mass of this accelerometer is made of quartz, the generation of step-wise noise at the output is greatly reduced compared to that for accelerometers with metal hinges, such as those used in K-NET95 and K-NET02. This is important because even though the amplitude of step-wise noise is very small on a raw accelerogram, it can have a considerable effect on the velocity or displacement seismograms calculated from the integration of accelerogram data. Hence, a significant advantage of K-NET02A is its ability to reduce instability.
K-NET11, the third-generation K-NET seismograph, is currently the latest model in use. It is manufactured by OYO Corp. and its installation began in 2012. K-NET11 has a maximum measurable acceleration of ± 8000 cm/s/s, 512 MB of flash memory, and a 24-bit A/D converter with a sampling frequency of 100 Hz. The accelerometer is JA-40GA08, which has a nominal full-scale range of ± 8 g. During a power outage, K-NET11 can record strong motion data and send seismic intensity data to the NIED DMC for up to seven days. In addition, K-NET11 is equipped with a redundant accelerometer aligned on the skew axis, which is not parallel to any axis of the remaining three orthogonal accelerometers. This redundancy provided by observing three-component ground motion using a four-component sensor can be used for detecting accelerometer or A/D converter errors.
All of the strong motion seismographs installed in KiK-net stations, SMAC-MDK, KiK-net06, and KiK-net11 listed in order of development, are specifically designed for use by this network. As of this time, all SMAC-MDK have been replaced, and KiK-net06 and KiK-net11 are currently in operation. The instruments used in K-NET and KiK-net are based on the same technology.
SMAC-MDK, the first-generation KiK-net seismograph, is manufactured by Akashi Co., Ltd., and includes a three-component V404BT accelerometer with a nominal full-scale range of ± 3g for surface observations and a data logger equipped with a 24-bit A/D converter with a sampling frequency of 200 Hz. The six-channel A/D converter uses three channels to digitize the three-component accelerometers at the surface and the other three channels to digitize the three-component accelerometers in the borehole. The acceleration resolution, maximum measurable acceleration, and effective dynamic range are 1 μm/s/s, ± 2000 cm/s/s, and 114 dB, respectively, which match the K-NET95 specifications. SMAC-MDK is equipped with 85 MB of flash memory, which allows it to store a total of about 6.5 h of six-component data digitized at a sampling frequency of 200 Hz.
The accelerometers currently installed in KiK-net boreholes are V404BT, V410BT, and JA-40GA04. The accelerometer employed at the time of construction of the Hi-net and KiK-net stations was V404BT. At some stations, V404BT has been replaced by V410BT, which is the successor to V404BT, or JA-40GA04. These accelerometers and the Hi-net velocity meters are mounted in a pressure-resistant vessel. Here, it should be noted that accelerometer replacement in station boreholes is performed independently from the seismographs at the ground surface.
KiK-net06, the second-generation KiK-net seismograph, is developed based on K-NET02A, and it has a 24-bit A/D converter with a sampling frequency of 100 Hz and is equipped with a JA-40GA04 accelerometer.
KiK-net11, the third-generation KiK-net seismograph developed based on K-NET11, is currently in operation, and the K-NET11 and KiK-net11 have many components in common. KiK-net11 began operating in 2012. KiK-net11 has a 24-bit A/D converter with a sampling frequency of 100 Hz and uses JA-40GA08 for surface observations. Both KiK-net06 and KiK-net11 share bandwidth with EarthLAN to continuously transmit strong motion indices such as real-time seismic intensity (Kunugi et al. 2008, 2013), long-period ground motion intensity (Aoi et al. 2020a), peak ground acceleration (PGA), and response spectrum each second or at an optional timing. Furthermore, at some stations located in the Kanto area, KiK-net06 and KiK-net11 estimate earthquake epicentral distance and magnitude using the algorithm of Tsukada et al. (2004) and then transmit this information to JMA through EarthLAN for use in earthquake early warning.
Data collection, archiving, and dissemination
Previously, strong motion data recorded by K-NET95 and SMAC-MDK were collected over dial-up connections from the NIED DMC where the operation was automatically processed. However, data collection was often delayed due to telephone line congestion after a large earthquake. To solve this problem, a new data collection system was introduced in 2003 and has since been updated. In the current system, when a seismograph is triggered by ground motion exceeding a threshold that is set prior to the event, it automatically sends waveforms and seismic intensity data to the DMC at Tsukuba using more than 1000 telephone lines. The seismographs also send seismic intensity values of three or greater in parallel to the sub DMC located in Miki, Hyogo Prefecture, for robustness. The received seismic intensity data are transferred to JMA independently from not only the DMC in Tsukuba, but also the sub DMC in Miki within one and half minutes after an earthquake.
The collected waveform data are archived and provided on the K-NET and KiK-net website. To download data, users can select specific events or stations based on key parameter combinations. Users can also browse, select, and obtain various types of information such as acceleration distribution maps and station maps. The soil condition data surveyed at K-NET stations and the geological and geophysical data derived from drilled boreholes at KiK-net stations are also available. All of the archived waveform data are manually verified, and operators exclude inappropriate data such as traffic noise and link strong motion data to earthquake events.
F-net
F-net overview
NIED began construction of a broadband seismic observation network as part of the Fundamental Research on Earthquakes and Earth’s Interior Anomaly (FREESIA) project in 1994 (Fukuyama et al. 1996). The aim of this project is to gain a better understanding of the earthquake source process, detailed structures of the crust and upper mantle, and deeper parts of the Earth’s structure. A total of 20 FREESIA stations had been constructed by March 1997. After that, based on the HERP policy, NIED constructed a broadband observation network covering the Japan Islands with station intervals of about 100 km, and the FREESIA network was integrated into the network in April 2001. The integrated broadband seismograph network was named Full Range Seismograph Network of Japan (F-net) in April 2002. The total number of F-net stations reached 73 (Matsumoto et al. 2009; Fig. 1 and Additional file 1: Figure S3). The network now covers almost all of Japan, including remote Izu, Ogasawara, and Ryukyu Islands. Because broadband seismographs are significantly affected by temperature changes, the seismographs at most F-net stations are installed at the end of dead-end tunnels with a length of 20 to 60 m, where the sensitive instruments are isolated from outside weather effects by several boundaries. At 25 stations, vault-type seismic and/or geodetic observatories owned by various universities are operated as F-net stations under joint research agreements between NIED and the universities. At some observatories, NIED uses vaults constructed for purposes other than broadband seismic observations such as a railway tunnel of dead tracks and a tunnel at a dam. Nine and three stations have vaults longer than 60 m and shorter than 20 m, respectively, and seismometers at the Sapporo station are installed in a vertical shaft.
Observation instruments
For broadband seismic observations, two types of sensors are installed at each F-net station. As broadband sensors, STS-1 (Wieland and Streckeisen 1982) or STS-2, manufactured by G. Streckeisen AG, whose responses are flat to ground velocity from 0.1 to 360 s and from 0.1 to 120 s, respectively, are installed. In addition, CMG-1T and CMG-3T sensors manufactured by Guralp Systems Ltd. were used at some F-net stations until 2002. STS-2.5 manufactured by Kinemetrics Inc. has been used as a successor to STS-2 at some stations since 2014. The numbers of stations with STS-1, STS-2, and STS-2.5 sensors are 16, 39, and 18, respectively. These broadband sensors have a clip level of about ± 10 mm/s and cannot observe ground motions caused by regional large earthquakes around Japan. Therefore, a velocity-type strong motion sensor is also installed at each F-net station to extend the dynamic range of the observations. The VSE-355G3 sensor manufactured by Tokyo Sokushin Co. Ltd., which has a flat response to ground velocity from 0.008 to 70 Hz and a clip level of ± 2 m/s, along with the TSM-1 sensor manufactured by Tokyo Keiki Inc., which has a flat response to ground velocity from 0.01 to 80 Hz and the clip level of ± 3 m/s, are installed at 62 and 11 stations, respectively.
Until 2003, Quanterra Q680 was used as a data logger. Observation data were digitized with 24-bit A/D converters that had sampling frequencies of 80, 20, 1, 0.1, and 0.01 Hz, and were time-stamped based on GPS time information. Later, to facilitate more rational operation of all NIED observation networks, data logger systems that were compatible with those used by Hi-net were installed at F-net stations. At present, data loggers with 27-bit A/D converters and sampling frequencies of 100 Hz are installed at all F-net stations.
Data management system and data distribution
In the FREESIA network, observed real-time data are continuously transmitted from each station to the NIED DMC over a 64 Kbps dedicated telephone line. At present, WIN32 format waveform data with 100 Hz sampling are directly transmitted through EarthLAN to both the NIED DMC and JMA. Moreover, international real-time data exchange is performed with Korea Meteorological Administration (KMA), Korea Institute of Geoscience and Mineral Resources (KIGAM), and Institute of Earth Sciences, Academia Sinica, Taiwan. The observed waveform data acquired at the NIED DMC are provided to the public on the NIED websites. For the convenience of researchers that use broadband waveforms, the 100 Hz data are routinely converted at the NIED DMC into 20 and 1 Hz sampling data and made available on the F-net website.
Quality control of observed waveform data
Most of the STS-1 and STS-2 seismometers have been in operation for about 20 years and it is thus possible that these sensors have degraded in terms of their response to ground motion and noise levels. To determine the conditions of the instruments, NIED developed a system for monitoring the quality of waveform data that consists of two subsystems. One monitors the instrument response to ground motion (Kimura et al. 2015) and the other monitors the background noise in the waveform data (McNamara and Buland 2004). These subsystems enable us to monitor the instrument conditions without inputting a calibration signal not to disturb the continuous waveform data. The output of these systems is used to detect instrument errors in F-net operation and is provided to users on the F-net website.
For response monitoring, Kimura et al. (2015) developed a method to compare waveform data at a target station with periods of 50–200 s, which result from large shallow teleseismic events, with those at several nearby stations. Because the long-period surface waveform observed at the target station will be almost identical to those observed at nearby stations, after the observed waveforms have been corrected with the reported instrumental responses, the differences among the waveforms can be used to evaluate any errors in the responses. This method has enabled us to confirm the normality of the responses for most of the stations for the period from 1995 to 2013 and has revealed errors at some stations, especially those equipped with STS-1 seismometers. In the monitoring system, this method is routinely applied to teleseismic observation data, and is used to detect temporal changes in the instrument responses. In the other system, the power spectral densities (PSDs) of continuous waveform data at each station are routinely calculated using the method devised by McNamara and Buland (2004). The probability density function of the calculated PSDs enables us to monitor the characteristics of background noise in waveform data, as well as their temporal changes.
Moment tensor analysis using F-net waveform data
Using the F-net waveform data, automated moment tensor (MT) analyses of regional earthquakes (Dreger and Helmberger 1993) have been performed by NIED since 1997 (Fukuyama et al. 1998; Kubo et al. 2002). In this system, earthquake information from JMA triggers an automated inversion analysis when an earthquake is Mj 3.5 or larger, and an inversion analysis is performed using the F-net waveform data with a period of 20 to 200 s at three or fewer stations. The analyzed MT results are automatically displayed on the F-net website within approximately 7 min after the origin time. The automatically determined MT solutions are then reviewed manually to improve their accuracy and reliability, and the F-net website information is revised if necessary. The total number of events in the F-net MT catalog currently reaches 38,000.
Matsumura et al. (2006) developed another automated source-parameter estimation system called the Accurate and QUick Analysis System for Source Parameters (AQUA). In operation, the AQUA system first performs earthquake detection, and estimates the hypocenter location and magnitude using three algorithms: the real-time earthquake information system (Horiuchi et al. 2005; Nakamura et al. 2009), the wavefront estimation method, and the conventional hypocenter location method based on the Hi-net data. The resulting estimated parameters are used as the initial values of the latter two-moment tensor analyses, with the horizontal location fixed to the initial value (AQUA-MT), and with hypocenter searching (AQUA-CMT) where F-net data from at most 20 stations are used. From the time of earthquake detection, it takes less than 30 s for the initial hypocenter information estimation, 2 to 5 min for the AQUA-MT analysis, and 4 to 8 min for the AQUA-CMT analysis. The MT and CMT solutions as well as the hypocenter information estimated by the AQUA system are posted on the Hi-net website immediately after an earthquake (www.hinet.bosai.go.jp/AQUA/aqua_catalogue.php?LANG=en, Accessed 30 Apr 2020). The combined use of F-net broadband and strong motion data has recently improved the accuracy of magnitude estimation for M9-class large earthquakes by the AQUA analyses (e.g., Kimura et al. 2020) such as those of the 2011 Tohoku earthquake.