Earthquake Research in China  2019, Vol. 33 Issue (4): 596-604     DOI: 10.19743/j.cnki.0891-4176.201904004
The Improvement of Earthquake Real-Time Monitoring System of Chinese National Digital Seismic Network
GUO Tielong, HUANG Zhibin, ZHAO Bo     
China Earthquake Networks Center, Beijing 100045, China
Abstract: The earthquake real-time monitoring system of the Chinese National Digital Seismic Network has been in operation since "the Ninth Five-year Plan" period,and the stability of the system has been well tested. In recent years,with the continuous improvement of monitoring technology and increase of public demands,the original real-time monitoring system needs to be upgraded and improved in terms of timeliness,stability,accuracy and ease of operation. Therefore,by accessing a total of more than 1,000 seismic stations,reducing the seismic trigger threshold of the monitoring system,eliminating the false trigger stations and optimizing the seismic waveform display interface,the current earthquake monitoring demands can be satisfied on the basis of ensuring the stable operation of the system.
Key words: Seismic monitoring     Earthquake location     Computer real-time processing    


The software for the Chinese National Digital Seismic Network Earthquake Real-time Monitoring System (Song Rui et al., 2001;Huang Zhibin et al., 2001) was developed during " the Ninth Five-year Plan" period and has been operated so far. The system architecture is clear and the operation is stable throughout the year. The system is an important part of the earthquake rapid report software of China Earthquake Networks Center (CENC). Independent of other provincial earthquake rapid wave processing software (such as MSDP), the National Digital Seismic Network seismic real-time monitoring system software is developed and operated by China Earthquake Networks Center. The system has served as important software for earthquake rapid report in China for many years and has fulfilled many important reports of major earthquakes, such as the 2008 Wenchuan MS8.0 earthquake, the 2010 Yushu MS7.0 earthquake, and the 2011 Japan MS9.0 earthquake. In these major earthquake reports, the system demonstrated its safety and stability characteristics. Moreover, in the use of the on-duty monitoring of the rapid report system, continuous optimization and improvement have been made based on the problems and deficiencies found in the actual work. With the continuous development of seismic monitoring technology and equipment, and the increasing demand for earthquake rapid reporting by the government, we have upgraded and improved the national digital seismic network real-time monitoring system software to adapt to the new requirements, such as mass data processing and the accuracy of automatic initial positioning.


The National Digital Seismic Network seismic real-time monitoring system uses a variety of programming techniques, featured with good real-time performance, high reliability, and system stability (Song Rui et al., 2001). The data receiving module, seismic waveform decompression module and seismic detection module programs of the system are written in C language, and the graphical interface is written by using Motif and X-Window graphical user interface. The operating environment of the system is Red Hat Enterprise 6.5 and CentOS 7, and the debugging environment of the system is Red Hat Enterprise 6.5, CentOS 7 and Ubuntu 16.04. The system has strong portability and no incompatibility in Linux versions.

During "the Ninth Five-year Plan" period, the National Digital Seismic Network earthquake real-time monitoring system collected and stored ground motion waveform data of 47 national digital seismic stations and 27 manned regional digital seismic stations in real time (Song Rui et al., 2001); at the end of 2007, the system received and stored the data of 145 national digital seismic stations in real time (Liu Ruifeng et al., 2008); after the completion of "the Tenth Five-year Plan", the system accesses real time the waveform data of 1, 069 seismic stations (including 166 national stations, 876 regional stations, and 27 stations of 3 arrays) to participate in earthquake triggering and location. In recent years, the increase of the number and the rational distribution of seismic stations have greatly improved seismic monitoring capabilities (Wang Yawen et al., 2017).

The National Digital Seismic Network seismic real-time monitoring system consists of a seismic data receiving module, a seismic data decompression and real-time waveform display module, and an automatic seismic detection module (Fig. 1). The system includes a data receiving shared area (rcvBuf) and a seismic data detection shared area (detBuf); the rcv process receives real-time waveform data of 1, 069 seismic stations nationwide, and stores it in the data receiving shared area (rcvBuf) in the mseed format; the sln process reads the seismic waveform data of the data receiving shared area and decompresses it into the waveform data of the evt format and stores it in the seismic data detection sharing area (detBuf); the earthquake automatic monitoring module (the detc process, the detL process and the detS process) reads the seismic data. The data in the shared area is detected for seismic waveform trigger detection. When the earthquake automatic detection module detects the earthquake event, the system automatically locates the earthquake event and triggers the alarm. At the same time, the waveform writing (wrte) process reads the seismic data stored in the data monitoring and shared area and writes it into the specified directory for data analysis in the artificial interaction system. The waveform display (waves) process reads the waveform data in real time and displays it on the screen.

Fig. 1 Modular structure diagram of the system

The automatic seismic detection module is the core component of the earthquake real-time monitoring system to detect and automatically locate seismic events, provide fast and reliable automatic positioning results for artificial earthquake rapid reports. Fig. 2 is a flow chart of the real-time monitoring program of the module. The module reads the real-time waveform data and determines whether the STA/LTA algorithm (Wu Zhitao et al., 2010) trigger threshold is met in stata.c. If the set algorithm threshold is exceeded, the AIC algorithm is used to accurately pick up the P-wave arrival time position near the trigger position of the data waveform (Akaike H., 1973; Wang Ji et al., 2006; Guo Tielong et al., 2017). In addition, Ltasta.c also records the periodic parameters of seismic body waves (P-wave and S-wave) and the maximum amplitude parameters of seismic S-wave (or Lg-wave), in preparation for the subsequent calculation of seismic automatic magnitude parameters. In Judge.c, it mainly identifies whether the seismic station can synthesize the seismic event with other stations within the effective time window after it is triggered (the trigger effective time window is 50 seconds). When the effectively triggered stations satisfy the condition, i.e. the triggered stations exceed 15 in an area, or the total number of triggered seismic stations is more than 40, it can be determined that synthesization of a seismic event occurs and it immediately enters the seismic positioning process. Otherwise, the seismic station exceeding the effective time window will return to the untriggered state and remains in the state to be detected and triggered in Ltasta.c. When the seismic event is synthesized, the earthquake trigger mark is turned on and the relevant seismic phase parameters recorded by the triggered seismic station are sorted in Locate.c; and the suspected false trigger station is removed according to the trigger station screening strategy. Finally, the seismic phase parameters of the stations participating in the positioning will generate a seismic phase file. Then, the seismic phase information is read by Sacloc.c for grid search. According to the residual value of each grid given by the travel time table (near earthquake travel time table, IASP91 travel time table), the position of the smallest residual value is selected as the epicenter location and the depth of the earthquake, and the magnitude formula is used to calculate the magnitude, after that, the calculated parameters are used to form an automatic location file for the seismic event. Finally, Wrte.c writes 120 waveform data of fixed stations with good waveform quality and even distribution, and then writes waveform data of seismic stations within 0.4° near the earthquake epicenter position obtained by automatic positioning, using regional stations. The combination with fixed stations ensures a rich seismic waveform data for human-computer interaction analysis.

Fig. 2 Flow chart of automatic earthquake detection module

The National Digital Seismic Network earthquake real-time monitoring system was compiled during "the Ninth Five-year Plan" period. The data access to the seismic stations was scarce, and the trigger threshold of the stations was high. It mainly monitors moderate-strong earthquakes in the Chinese mainland. In view of the weak seismic monitoring capability at that time, the requirements for earthquake rapid reporting were not high, and some of the drawbacks of the system were not clearly reflected. Nowaday, the magnitude of the earthquake rapid report is required to reduce to M2.0 for the densely populated areas in eastern China and M3.0 in the rest of China. The original system sometimes fails to detect M≤3.0 domestic earthquake events.

The automatic positioning result of the original system is seriously affected by the mis-triggered station, and mis-triggering of earthquake event (non-seismic event) and inaccurate automatic positioning would occur sometimes. When a seismic event is triggered or the seismic station evenly distributed around the epicenter of an earthquake, the individual mis-triggering stations have little influence on the positioning results, otherwise, mistriggered stations will cause fatal errors in the positioning results.

The instability of magnitude results produced by automatic positioning of the system has always been a common problem in the use of the system. The magnitude result automatically generated by the system isthat the S-wave (or Lg wave) amplitude of the seismic station is triggered by the record, and the average value of each seismic station is calculated by the magnitude formula as the automatic magnitude result of the earthquake event. The magnitude calculation is susceptible to stations with poor signal-to-noise ratio and the impact of mis-triggered stations, which does not guarantee a true reflection of the magnitude of the earthquake. In addition, in the system operation, we also encountered the problem of lower magnitude for the strong earthquakes measured by the automatic magnitude measurement. The reason is that the system has released the automatic magnitude result before measuring the maximum surface wave amplitude of strong earthquake.

The waveform display interface is used to display real-time waveform data of station. The original system has a single display function and does not have the function of displaying a seismic event trigger station. When an earthquake is triggered, the original systems waveform display function does not have the function of quickly determining the location of the earthquake by observing the position of the station that originally triggered the earthquake event, and the function of determining the magnitude of the earthquake by the number of seismic stations that subsequently trigger the earthquake event. The staff on the fast report cannot quickly understand the preliminary information of the location of the earthquake and the size of the earthquake, which indirectly affects the speed of earthquake fast reporting.

With the increasing requirements for earthquake rapid reporting, the original system is unable to meet the present seismic monitoring needs. Therefore, in the process of using the system software, the problems in the optimization of the system are also debugged. On the premise of ensuring the stable operation of the system, the above problems are gradually resolved to meet the requirements of rapid earthquake reporting, while improving the accuracy and convenience of the automatic location parameters of earthquake events of the system.

3 MEASURES AND EFFECTS OF SYSTEM IMPROVEMENT 3.1 Improvement of Waveform Data Access in Seismic Stations

Since the construction in "the Ninth Five-year Plan" and completion in "the Tenth Five-year Plan" of the system, the system has accessed in real time the waveform data of 1069 seismic stations to participate in the automatic earthquake triggering and location. The data receiving module of the original system had only 100 stations receiving seismic data, and the problem of memory space consuming did not emerge. However, the existing system receives the real-time waveform of massive seismic stations and caches them into the computer data receiving shared area (rcvBuf), which consumes memory. Therefore, by increasing the shared memory capacity of the software system, the static array of the original system is changed to the dynamic allocation array of the received waveform data, thereby improving the repeated utilization of the limited storage resources, avoiding the waste of memory resources by the static array of the data receiving module, solving the problem of lack of buffering space caused by the access to a large number of seismic stations in the program upgrading, optimizing the memory resource usage of the station waveform receiving, and getting prepared for accessing more seismic station data in the future.

3.2 Trigger Threshold Adjustment

The system trigger threshold indicates the sensitivity of the seismic real-time waveform data to the detection of the initial arrival time of the seismic P-wave. The advantage of high triggering threshold is that it is not easy to trigger by mistake and the pick-up degree is high for medium and large-size earthquakes; the disadvantage is that the position of P-wave first arrival of microearthquake is not picked up, and a missingearthquake event will occur. For the original system, due to the low requirements for earthquake rapid reporting at that time, the seismic stations were scarce, and the high earthquake trigger threshold had little effect on the seismic monitoring capability at that time. However, at present, the seismic monitoring capability is more stringent, so it is urgent to reduce the trigger threshold.

The STA/LTA algorithm of the National Digital Seismic Network earthquake real-time monitoring system picks up the default value of position of the P-wave first arrival: STA (signal short-time window average) is 1.5s, LTA (signal long-term window average) is 20s, trigger threshold 6.5, and optimizes and improves the parameter value: keeping the short and long time window unchanged, the trigger threshold is modified as 6.0. After improvement, on the premise that the number of seismic stations has increased by more than 7 times, and by lowering the trigger threshold, we can ensure that more subtle abrupt changes in the amplitude of waveforms triggered by seismic stations can be detected, which enhances the ability of monitoring small earthquakes and ensures the sensitivity of seismological stations to pick up the first arrival position of P-waves. At the same time, we also use the strategy of screening false trigger stations. The problem of false alarm of earthquake events caused by wrong triggering of seismic stations is effectively prevented.

3.3 Improvement of Processing of the False Triggering Stations

The real-time seismic monitoring system of the national digital seismic network firstly improves sorting of the first arrivals of P-wave detected by triggering seismic stations, and then screens out the seismic stations whose arrival time is more than 6 seconds different from that of any other triggering stations according to the sorting results and eliminates them. Then, according to the sequence of the first arrivals, the amplitudes of the seismic body waves (P-wave and S-wave) are compared, and the waveform amplitudes are arranged in descending order. The system will automatically select and delete the triggering stations whose amplitude of seismic waveform is 1.6 times larger than that of a waveform on the sequence or less than one-half of the amplitude of subsequent seismic waveform, and judge them to be false triggering stations and no longer participate in seismic positioning. Finally, if the distance of individual stations to themajority of stations is calculated to exceed 2° (the station distances in Tibet and Xinjiang are set to be more than 5°), these stations will also be judged as false triggering stations and be eliminated. The above method effectively improves the accuracy of automatic positioning and reduces the impact of false triggering stations on the automatic positioning results of the real-time seismic monitoring system of the national digital seismic network. After system optimization, the problem of false triggering of earthquake events in the original system is solved. The number of false triggering earthquake events of the improved system is 0, and the recognition and eliminating rate of false triggering seismic stations is 92%.

3.4 Optimization of Magnitude Release

When a triggering seismic event happens, the original system simply records the maximum amplitude of S-wave (or Lg wave) of the triggered seismic station and then uses magnitude formula to calculate and generate the magnitude result of the automatically located earthquake. As a result, the instability of magnitude of earthquake event often happens. In order to solve this problem, the S-wave (or Lg wave) or surface wave amplitudes identified by the system are arranged in descending order, and then the singular value amplitudes at both ends of the sequence are removed according to the Gauss distribution characteristics. Finally, the magnitude results are published after calculating with the magnitude formula. In the trial operation stage, the catalogue of three-month manual rapid-report earthquakes was extracted and compared. The maximum deviation between the improved automatic positioning magnitude and the manual quick-report magnitude was 0.4, and the magnitude deviation was obviously smaller than that before the improvement.

During the automatic earthquake rapid reporting process, the speed of releasing seismic parameters was overemphasized and the accuracy was ignored. When the parameters were released, the S waves (or Lg waves) did not arrive or did not arrive completely at some stations, resulting in a smaller average magnitude from the calculation (Liang Jianhong et al., 2015). Affected by the large difference between the magnitude results of the automatic positioning of the Lushan earthquake in 2013 and the manually revised magnitude, the software system has been optimized accordingly: when the initially measured magnitude by the system exceeds M4.5, the system will keep waiting for magnitude updating, and the magnitude will be released after the program finishes searching for the maximum surface wave amplitude, thus avoiding the large deviation between automatic location magnitude and the actual magnitude. After the optimization, the system adjusts the strategy of magnitude determination of strong earthquakes, and the output time is slightly delayed, but the reliability of magnitude is effectively improved. The magnitude of strong earthquakes is similar to the result of manual rapid-report. After putting into operation, the maximum deviation between automatic location magnitude and the formal rapid-report magnitude of subsequent strong earthquakes is 0.5, and the average deviation is 0.23.

3.5 Enhancement of Real-time Waveform Display Function

The basic function of the real-time seismic waveform display is to provide earthquake real-time waveform display for the earthquake rapid-report personnel to check whether the real-time waveform of the provincial networks is abnormal. Through the improvement of the real-time waveform display module, the module code is modified, the trigger function of the seismic station is increased, the dimension of the seismic waveform data array is adjusted, and the data structure of the module is unified with the system program interface. After the improvement, besides the above basic functions, the triggering state of seismic stations displayed by the system is expressed by color when an earthquake is triggered. This improvement not only enables the on-duty staff to quickly observe the approximate epicenter position of the earthquake event, but also estimates the magnitude of the earthquake by the scale of the triggered stations. For example, at 11:44:12 on February 18, 2018, an MS4.1 earthquake occurred in Qingchuan County, Guangyuan City in Sichuan Province (Fig. 3), the color of the triggered stations gradually changed outward from the epicenter.

Fig. 3 Trigger Map of Qingchuan Earthquake Event in Sichuan, China

In Fig. 3, the red seismic stations are the triggered stations near the epicenter of a synthetic earthquake event. By observing the area of the red seismic stations, it can be roughly understood that the earthquake epicenter is located in the territory of Sichuan. The process of color change from red to yellow to pink simulates the direction of propagation when seismic stations are triggered by an earthquake event (white seismic station represents the state where the station is not triggered by an earthquake event). Based on the experience, the earthquake rapid-report attendants can roughly estimate the size of the earthquake to be about M4.0 from the trigger map, therefore, it is convenient for quickly characterizing the earthquake event and providing a location reference for the artificial interaction result, and is conducive to saving the operation time of earthquake rapid-reporting.


The software of the real-time seismic monitoring system of the National Digital Seismic Network runs uninterruptedly throughout the year. Each module runs steadily and has strong ability to monitor moderate and strong earthquakes. With the development of the Internet, public-oriented seismic information service is more rapid and comprehensive, which requires the real-time seismic monitoring system to respond better to small and medium-sized earthquakes. Therefore, the system module is improved in order to solve the shortage of system software and improve the monitoring ability and convenience of the software system, so as to ensure more efficient and accurate automatic determination of earthquake events.

In this paper, various modules of the system software have been optimized and improved to varying degrees. The data receiving module uses dynamic memory allocation to receive and store real-time waveform data of thousands of seismic stations, thus makes more reasonable use of limited memory resources and lays a good foundation for increasing the use of memory resources for waveform data of seismic stations in the future. In the seismic waveform display module, the receiving data structure of thewaveform display module is adjusted, the display function of seismic event triggering station is added, the graphical display content of the software system is enriched, and the earthquake process is presented more intuitively on the screen of the seismic monitoring system. Seismic automatic detection module is the core of the system software. Firstly, the strategy of eliminating false trigger stations is optimized to ensure the accuracy of automatic determination of earthquake events. Then, the trigger threshold is optimized to ensure the sensitivity of seismic stations to detect small earthquakes. Finally, the deviation between the automatic output magnitude of the system and the rapid-report magnitude is found and solved in the practice of rapid reporting.

Many improvement schemes in this paper are summarized and refined on the basis of the encountered problems, proposed ideas and accumulated experience in the use of the software by the on-duty staff of earthquake rapid reporting. With the improvement of earthquake monitoring capabilities and the rapid dissemination of earthquake information, the improvement of earthquake real-time monitoring system will be an ongoing process, so as to lay the foundation for providing faster and more accurate seismic information for the public.

Akaike H. Information theory and an extension of the maximum likelihood principle. In: Proceedings of the 2nd International Symposium on Information Theory. Budapest: Academiai Kiado, 1973: 267-281.
Guo Tielong, Zhang Xuemei, Zou LiyeGuo Tielong, Zhang Xuemei, Zou Liye. Influence of the STA/LTA-AIC algorithm on the stability of P-wave pickups[J]. Seismological and Geomagnetic Observation and Research, 2017, 38(3): 13-17 (in Chinese with English abstract).
Huang Zhibin, Song Rui, Gu Xiaohong, Xue Feng, Wang Yongli, Zuo Rubin, Huang Jinli, Zheng SihuaHuang Zhibin, Song Rui, Gu Xiaohong, Xue Feng, Wang Yongli, Zuo Rubin, Huang Jinli, Zheng Sihua. The technological system of China Center of Digital Seismic Network[J]. Earthquake, 2001, 21(4): 7-11 (in Chinese with English abstract).
Liang Jianhong, Sun Li, Xu Zhiguo, Liu JieLiang Jianhong, Sun Li, Xu Zhiguo, Liu Jie. Analysis on magnitude deviation of the 2013 Lushan, Sichuan, MS7.0 earthquake from automatic earthquake location and report system and its improvement[J]. Acta Seismologica Sinica, 2015, 37(6): 983-996 (in Chinese with English abstract).
Liu Ruifeng, Gao Jingchun, Chen Yuntai, Wu Zhongliang, Huang Zhibin, Xu Zhiguo, Sun LiLiu Ruifeng, Gao Jingchun, Chen Yuntai, Wu Zhongliang, Huang Zhibin, Xu Zhiguo, Sun Li. Construction and development of digital seismograph networks in China[J]. Acta Seismologica Sinica, 2008, 30(5): 533-539 (in Chinese with English abstract).
Song Rui, Gu Xiaohong, Wang Yongli, Zhao Yong, Wang Bin, Xie XiangweiSong Rui, Gu Xiaohong, Wang Yongli, Zhao Yong, Wang Bin, Xie Xiangwei. The software system for real-time processing and large earthquake rapid determination of NCDSN[J]. Earthquake, 2001, 21(4): 47-59 (in Chinese with English abstract).
Wang Ji, Chen Jiuhui, Liu Qiyuan, Li Shuncheng, Guo BiaoWang Ji, Chen Jiuhui, Liu Qiyuan, Li Shuncheng, Guo Biao. Automatic onset phase picking for portable seismic array observation[J]. Acta Seismologica Sinica, 2006, 19(1): 44-53. DOI:10.1007/s11589-001-0044-0
Wang Yawen, Jiang Changsheng, Liu Fang, Bi JinmengWang Yawen, Jiang Changsheng, Liu Fang, Bi Jinmeng. Assessment of earthquake monitoring capability and score of seismic station detection capability in China seismic network (2008-2015)[J]. Chinese Journal of Geophysics, 2017, 60(7): 2767-2778 (in Chinese with English abstract).
Wu Zhitao, Li ShixiongWu Zhitao, Li Shixiong. Comparison of STA/LTA P-pickers for micro seismic monitoring[J]. Progress in Geophysics, 2010, 25(5): 1577-1582 (in Chinese with English abstract).