Earthquake Research in China  2019, Vol. 33 Issue (3): 367-390     DOI: 10.19743/j.cnki.0891-4176.201903008
A Preliminary Study on the Phenomenon of Microseismic Fluctuations before Impending Strong Earthquake
YANG Liming1,2, HAO Zhen1, WANG Jianjun1, ZHANG Shuzhen1, YAO Jiajun3, DONG Lei4
1. Lanzhou Institute of Seismology, CEA, Lanzhou 730000, China;
2. Qinghai Earthquake Agency, Xining 810001, China;
3. Geophysical Exploration Center, CEA, Zhengzhou 450002, China;
4. Chongqing Earthquake Agency, Chongqing 401147, China
Abstract: Based on repeated comparison studies of broadband digital seismic records before the Wenchuan MS8.0, Yushu MS7.1 and Qingchuan MS5.4 earthquakes, the possible microseismic fluctuations before impending earthquakes were preliminarily identified. In order to verify and test this phenomenon, a real-time tracking technical system was established by using continuous waveform records of more than 200 wide-band digital seismic stations in regional networks such as Gansu, Qinghai, Sichuan, Yunnan and Tibet. Through real-time tracking and dynamic monitoring of 24 earthquakes with M ≥ 5.0 occurring in the Qinghai-Tibetan block during the period of 2012-2014 and the observations of stations in some non-seismic areas, the reproducibility and objectivity of the impending earthquake phenomenon were verified. The main characteristics of the microseismic fluctuation phenomena immediately preceding the strong earthquakes are as follows:(1) the spectrum range is wider, the dominant frequency is 11-16Hz, and the spectrum shape is more regular; (2) it appears 6-24 days before the earthquake, averaging about 15 days; (3) it is possible to be recorded by the stations within the epicenter distance of 50km, and the stations with the epicenter distance of more than 50km generally cannot record it; (4) this phenomenon is directional, i.e. the direction in which the activity degree, N-value, varies significantly may be related to the location of the seismic source, the seismogenic fault and the distribution of aftershocks of the strong earthquake. The preliminary study shows that the impending-earthquakes microseismic phenomena may be related to the pre-activity, micro-vibration and micro-rupture in the source region in the imminent stage, or the micro-activity and micro-rupture associated with the active tectonics.
Key words: Impending microseisms     Dominant frequency     11-16Hz     Space-time characteristics     Pre-activity of source area

INTRODUCTION

The imminent earthquake prediction is the most difficult and challenging stage in the progressive earthquake prediction of long, medium, short, and impending terms. Research and exploration from the perspective of seismic waves is one of the main research directions of imminent earthquake prediction at home and abroad. As we all know, broadband digital seismic records contain a wealth of information, and to explore whether it contains information about the source or medium state changes associated with the seismogenic process, and try to identify and extract this information, are the research areas attracting increasing and continuous concern (Zhao Cuiping et al., 2004; Gao Yuan et al., 2004; Liu Jie et al., 2004; Feng Deyi et al., 1983, 1994; Guo Lücan et al., 1998; Yang Liming et al., 2006, 2018a, 2018b; Wang Yueqi, 2015).

1 MICRO-SEISMIC PHENOMENON IMPENDING STRONG EARTHQUAKES AND ITS CASES 1.1 Research Ideas and Methods

A normal wide-band digital micro-seism record is relatively stable, randomly undulating, characterized by random sequences, but the microseism record superimposed with known or unknown source information is not stable, there are obvious fluctuations in it. Fig. 1 shows the record of 120s recorded by the Baiyin station in Gansu Province on October 3, 2014 at 01:41:40. There are obviously some kinds of "interference event" superimposed on the normal micro-seism in the record. Generally speaking, such kinds of "interference event" may originated from typical geophysical events and environmental changes such as earthquake, blasting, collapse, typhoon, thunderstorm etc., or may originated from certain unknown or currently unrecognized processes, which may include processes such as pre-slip, micro-fracture, micro-seism originating from the source of the impending earthquake, and the micro-activity, micro-fracture, and fracture expansion of active tectonics associated with earthquake preparation when an earthquake is impending.

 Fig. 1 The microseism records superposed with "interference event" recorded at Baiyin station on October 3, 2014

Earthquakes, collapses, blaste and other events can be identified according to their time-domain and frequency-domain characteristics.The pre-activity, micro-seism, micro-fracture from the source region of the impending earthquake and the micro-activity, micro-fracture, and fracture propagation related to tectonic activity when the earthquake is impending in its preparation process are the goals of imminent earthquake prediction research. For such target events, the method of identification can only be studied by earthquake cases because the change characteristics are not yet known. Through repeated comparison and analysis of typical earthquake cases, the possible suspected "event" in imminent stage is identified preliminarily, and the reproducibility of the suspected "event" before different strong earthquakes is studied by a large number of earthquake examples and actual data so as to verify whether the "event" is related to the imminent stage of strong earthquakes. If there is a correlation, it can be further applied to the impending judgment of strong earthquakes.

Based on the above ideas, in order to explore the possible information in the imminent stage of an earthquake, we need to extract various "interference events" from the continuous microseismic waveform records at first, and then analyze and study the data in time-domain and frequency-domain. In general, the continuous microseism data should be scanned in full time based on the normal microseism records and the variance several times of the average of microseismic fluctuations. When the waveform record exceeds the variance control line in a certain period of time, the waveform record of a certain length of time is taken as the basis of the superimposed portion as the base data of the superimposition event, and then processed by digital signals such as FFT transform. The method performs spectrum analysis to achieve the purpose of identifying superimposed information and its spectral features.

FFT spectrum analysis is a commonly used signal processing method and one of the most mature digital signal processing theories and methods. This method was proposed by Cooley J. W. et al., (1965). The main idea is to express the general time function f(x) as a linear superposition of harmonic functions {ejωt, ωR} with different frequencies, thus transforming the study of the original time domain function to the study of harmonic functions of different frequency components. The continuous Fourier transform of the function f(x)∈L2(R) is defined as:

 $F(\omega)=\int_{-\infty}^{+\infty} f(\mathrm{x}) e^{j \omega t} \mathrm{d} x$ (1)
 $f(\mathrm{x})=\frac{1}{2 \pi} \int_{-\infty}^{+\infty} F(\omega) e^{-j \omega t} \mathrm{d} \omega$ (2)

The frequency of occurrence of different types of "interference events" in the microseismic record reflects the different levels of activity of such events. In order to quantitatively express this feature, this paper introduces the concept of event activity, that is, accumulates the frequency of occurrence of the same type of "interference event" at a certain time interval, and defines it as a quantitative indicator of the activity level of the event:

 ${N_i} = \sum\limits_{j = 1}^k n (i, j)$ (3)

Where i is time, i=1, 2, 3, …, in this study i refers to day, j is the order in which such events occur, n(i, j) means that certain types of events appear on the i-th day and j-th time, k is the total number of times that the event has occurred.

1.2 Basic Information

Taking the Wenchuan MS8.0 earthquake and the Yushu MS7.1 earthquake as the research objects, the continuous microseismic records of the Digital Seismic Networks in Sichuan, Qinghai and Gansu etc. are used as the basic data. The seismic networks of these areas have undergone digital transformation of the tenth "Five-year Plan", all of which have realized digital observation and network transmission, and generally use broadband seismic instruments such as BBVS-60, CMG-3ESPC, CTS-1 or wide-band seismographs for observation. The sampling rate of the instrument is 100 times/s and the maximum resolution frequency is 50Hz; the basement of the stations is good and the observation environment is stable so there are high qulity of the observed data.

1.3 The Imminent Microseismic Fluctuations and Earthquake Cases Study 1.3.1 The Imminent Microseismic Fluctuations before the Wenchuan MS8.0 Earthquake on May 12, 2008

There are 12 digital seismic stations within the radius of 150km around the epicenter of the Wenchuan earthquake (Fig. 2), and there are 23 within a radius of 250km, which can effectively monitor the possible impending earthquake anomalies in the source or near-source areas before the earthquake. The stations with its code and their epicentral distances are shown in Table 1.

 Fig. 2 Distribution of digital seismic stations within a radius of 150km around the epicenter of the Wenchuan earthquake

Table 1 Statistics of stations and their epicentral distances in different radius ranges around the epicenter of the Wenchuan earthquake

According to the above research ideas, the event waveforms of the Sichuan network before the Wenchuan earthquake were extracted one by one, and the FFT method was used for spectrum analysis. In the frequency range of 0-50Hz, through the repeated comparison study of a large number of event waveforms and the spectral characteristics of different frequency bands, the possible microseismic fluctuation events and the FFT spectrum appearing in the impending stage of Wenchuan earthquake are preliminarily identified (Fig. 3). Among them, Fig. 3(a) and Fig. 3(b) are the waveform records and their spectrum of N-S channel, with a 120s window length starting from 11:55:04 on May 10, 2008 in Dujiangyan station. As can be seen from Fig. 3(a) and Fig. 3(b), the waveform is recorded during the period of 11:55:30 to 11:56:00. On the basis of the relatively stable microseism recording, a significant event waveform is superimposed, and the duration is about 30s, the amplitude of the event exceeds the mean microseismic fluctuation by 5 times variance; when the frequency is 9-16Hz, the amplitude of the spectrum changes prominently, which is significantly different from other frequency bands. Fig. 3(c) and Fig. 3(d) are the waveform records and their spectra of the N-S chanel, with a 120s window length starting from 21:14:26 on May 10, 2008 in Dujiangyan station. As seen from Fig. 3(c) and Fig. 3(d), the waveform is recorded during the period from 21:14:50 to 21:15:20. On the basis of the relatively smooth microseismic recording, a significant event waveform is superimposed for a duration of approximately 30s, the amplitude of the event exceeds the microseismic fluctuation mean by a factor of 5, and its spectrum shows that the amplitude at the frequency of 11-16Hz has characteristics similar to those of Fig. 3(b).

 Fig. 3 The microseismic fluctuation before impending earthquake in Dujiangyan station

Further research shows that similar events and their spectrums in the records of Dujiangyan station were more common during the period from May 1 to 12, 2008, but rarely occurred in the period from January to April, 2008 for up to 4 months. The common feature is that the duration is about 30s, which is obvious with respect to the microseismic background variation. The corresponding spectrum is concentrated at the frequency of 11-16Hz, and the spectrum shape is relatively neat.

In order to further study whether these events were correlated with the imminent stage of Wenchuan earthquake, the continuous waveforms of 14 stations (including Jiangyou station (153) and Emei station (153)) in Table 1, 50km from the epicenter of Wenchuan earthquake, from January to May 2008, are analyzed and processed according to the above analysis method. The activity degree N-value of the event is calculated day by day (Fig. 4 (a)). From Fig. 4 (a), it can be seen that the N values of all stations fluctuated smoothly in a small range during the four-month period from January 1 to April 29, 2008. Since April 29, 2008, the N values of Dujiangyan station had a continuous high value change, which lasted until the Wenchuan earthquake occured on May 12, 2008. During this period, the N values of the remaining 13 stations such as Wenchuan Station kept stable fluctuations, and no similar changes occurred. In the same way, the time-varying comparison results of N values of 13 stations (including Huili station, Puge station, Garzê station, etc.) in the same radius range of 250km (except those selected in Fig. 4(a)) in Table 1 are given, which can be seen from Fig. 4(b)). It has a similar process of change with Fig. 4 (a).

 Fig. 4 (a) Variation of N-value of different stations over time before the Wenchuan earthquake from January 1 to May 12, 2008, within the epicenter distance of 150km. (b) Variation of N-value of different stations over time before the Wenchuan earthquake from January 1 to May 12, 2008; within the epicenter distance of 250km and part of the stations beyond this range

As can be seen from the comparison of Fig. 4(a) and Fig. 4(b), the microseismic fluctuations recorded by Dujiangyan station before the Wenchuan earthquake from April 29 to May 12 are unique. In the microseism records of 5 consecutive months of the stations within 250km around the epicenter of the Wenchuan earthquake and some stations with an epicenter distance larger than 400km, only this station observed the fluctuation phenomenon 14 days before the earthquake.

1.3.2 The Imminent Microseismic Fluctuation before the April 14, 2010 Yushu MS7.1 Earthquake

There are 3 digital seismic stations in the area of 150km around the epicenter of Yushu earthquake, 4 digital seismic stations in the area of 250km, the distribution of the stations is sparse. The stations with different epicentral distance are shown in Table 2.

Table 2 Statistics of stations and their epicentral distances in different radius range around the epicenter of the Yushu MS7.1 earthquake

According to the same analysis idea as the Wenchuan earthquake, through repeated analysis of a large number of event waveforms and spectral characteristics of different frequency bands, it is found that there is a phenomenon of imminent microseismic fluctuations before the Yushu earthquake similar to that of the Wenchuan earthquake. Fig. 5(a) and Fig. 5(b) are the waveform records of the EW channel and their spectrums recorded by Yushu station, with the 120s window length starting from 11:11:18 on April 4, 2010. It can be seen from Fig. 5(a) and Fig. 5(b) that during the period from 11:11:48 to 11:12:08, on the basis of relatively stable microseismic recording, a significant event waveform is superimposed for a duration of about 30s. The amplitude exceeds the mean microseismic fluctuation by 5 times variance; when the frequency is 9-16Hz, the amplitude changes prominently, which is significantly different from other frequency bands. Fig. 5(c) and Fig. 5(d) are the waveform records and their spectrum on the E-W channel of Yushu station, with the 120s window length starting from 06:04:29 on April 3, 2010. It can be seen from Fig. 5(c) and Fig. 5(d) that during the period of 06:04:55 to 06:05:35, on the basis of relatively stable microseismic recording, a significant event waveform is superimposed for a duration of about 40s. The amplitude exceeds the mean microseismic fluctuation by a factor of 5; the amplitude at a frequency of 11Hz to 16Hz has the characteristics similar to those of Fig. 5(b).

 Fig. 5 Impending-earthquake microseismic fluctuations recorded by Yushu station

Further research shows that the events and their spectrum in the records of Yushu station were more common during the period from April 1 to 14, 2010, but rarely before the end of March. The common feature is that the duration of this kind of events is about 30-40s, which is obviously different from the microseismic background. The corresponding spectrum is concentrated at the frequency of 11-16Hz, and the spectrum shape is relatively neat.

In order to further study whether there is any correlation between this kind of event and the impending-earthquake stage of the Yushu earthquake, continuous waveforms of all the stations within 150km from the epicenter of the Yushu earthquake in Table 2 of the period from 6 months before the earthquake, that is, October 1, 2009 to April 14, 2010 are analyzed and processed according to the above analysis method, and the event activity degree N-value is calculated on a daily basis in the unit of day (Fig. 6(a)). It can be seen from Fig. 6(a) that during the period from October 1, 2009 to the end of March, 2010, the N values of all stations in the six-month period fluctuated smoothly within a small range, and beginning from April 1, 2010, a continuous high value of N-value appeared in Yushu station, which continued until the occurrence of Yushu earthquake on April 14. During this period, the N values of Zadoi station (130) and Qumalai station (140) remained stable and fluctuating, and no similar changes occurred. Based on the same idea and method, we obtained the comparison results of N-value variation with time for the stations within the radius of 250km in Table 2 in the same period (except the stations selected in Fig. 6(a)) and some of the far-distance stations (including Xining station (620), Qilian station (640), Xinghai station (390), Huatugou station (750) and Da Qaidam station (540), etc. (Fig. 6(b)), as can be seen from Fig. 6(b), it has a variation process similar to that of Fig. 6(a).

 Fig. 6 (a) Changes of N values of different stations with time before the Yushu earthquake from October 1, 2009 to April 20, 2010, within a distance of 150km from the epicenter. (b) Changes of N values of different stations with time before the Yushu earthquake from October 1, 2009 to April 20, 2010, within a distance of 250km from the epicenter and part of the stations farther away

It can be seen from Fig. 6(a) and Fig. 6(b) that the fluctuations recorded by Yushu station from April 1 to 14 is special. Among the stations within 250km around the epicenter of Yushu earthquake and some of the far-field stations, in the 6 consecutive months of microseism records, only Yushu station observed such fluctuations about 14 days before the earthquake.

Therefore, the phenomenon of imminent microseismic fluctuations recorded by Yushu station before the Yushu earthquake is similar to that recorded by Dujiangyan station before the Wenchuan earthquake. The phenomenon occurred in the imminent stage before the strong earthquakes, and can be recorded by stations with the epicenter distance less than 50km, while stations with epicenter distance more than 50km did not be recorded.

1.3.3 The Imminent Microseismic Fluctuations before the Qingchuan MS5.4 Earthquake on November 1, 2011

The stations are dense in the area around the epicenter of the November 1, 2011, Qingchuan MS5.4 earthquake. There are seven stations within 150km around the epicenter of this earthquake. The epicenter distance to Qingchuan station is about 7km. Stations with different epicentral distance and in different radius range are shown in Table 3.

Table 3 Statistics of stations and their epicentral distances in different radius ranges of the Qingchuan earthquake

According to the similar idea, the event waveforms of stations within 150km from the epicenter of the Qingchuan earthquake were extracted one by one, and the FFT method was used for spectrum analysis. Through the repeated comparison of a large number of event waveforms and the spectral characteristics of different frequency bands, it is found that there were imminent impending-earthquake fluctuation events before the earthquake similar to those before the Wenchuan earthquake and Yushu earthquake. Among them, Fig. 7(a) and Fig. 7(b) are the waveform records and its spectrum in the U-D channel of Qingchuan station, with a 120s window length starting at 17:37:21 on October 22, 2011. It can be seen from Fig. 7(a) and Fig. 7(b) that the waveform record contains a large number of fluctuation events with short duration; when the frequency is 11-16Hz, the amplitude of the spectrum changes sharply, which is significantly different from other frequency bands. Fig. 7(c) and Fig. 7(d) are the waveform records of a 120s window length and its spectrum in the U-D channel of Qingchuan station starting at 18:00:01 on October 22, 2011. It can be seen from Fig. 7(c) and Fig. 7(d) that on the basis of relatively stable microseism recording, a large number of event waveforms with short duration are superimposed, and the event amplitude exceeds the microseismic fluctuation mean by 5 times variance; the amplitude of the spectrum has characteristics similar to that of Fig. 7(b) at the frequency of 11-16Hz.

 Fig. 7 The impending-earthquake microseismic fluctuation before the Qingchuan earthquake recorded at Qingchuan station

With the same analysis method, further research shows that the spectrum of similar events in the records Qingchuan station is more common during the period from October 18 to 22, 2011, but rarely before. The characteristics of the spectrum change are similar to those of the Wenchuan earthquake and the Yushu earthquake. The spectrum is roughly concentrated at 11-16Hz, and the spectrum shape is relatively neat. The N values of the remaining stations remain stable for up to 5 months. The specific results are shown in Fig. 8(a) and Fig. 8(b).

 Fig. 8 (a) Changes in N values of different stations over time before the Qingchuan earthquake from May 1 to November 1, 2011, within an epicenter distance of 150km. (b) Changes in N values of different stations over time before the Qingchuan earthquake from May 1 to November 1, 2011, within the epicenter distance of 250km and partly of the stations farther away

Therefore, the fluctuations recorded at Qingchuan station before the Qingchuan earthquake are similar in the frequency domain to the phenomenon recorded in the Dujiangyan station before the Wenchuan earthquake and the Yushu station before the Yushu earthquake, and all occur in the imminent stage before strong earthquake, the epicentral distances are less than 50km, so that can be reproducible.

2 MAIN FEATURES OF THE IMPENDING-EARTHQUAKE MICROSEISMIC FLUCTUATION PHENOMENON

According to the long-term and large-scale scanning calculations and analysis of the microseismic databefore impending strong earthquakes such as the Wenchuan earthquake, Yushu earthquake and Qingchuan earthquake, it can be seen that there are similar impending microseismic fluctuation phenomena in the three earthquake cases, which seems to have the characteristic of reproducibility. Its main features are as follows:

(1) Spectral characteristics: The frequency range is wide, the dominant frequency is concentrated around 11-16Hz, and the spectrum form is relatively neat.

(2) Time characteristics: It may occur 12 to 14 days before the earthquake, and it seems to have the characteristics that the longer the duration is, the larger the earthquake magnitude will be, and the shorter the duration, the smaller the earthquake magnitude will be. Specifically, the microseismic fluctuation occurred 14 days before the Wenchuan earthquake and lasted for 14 days until the main shock. The phenomenon occured 12 days before the Yushu earthquake, lasted for 8 days, and returned to normal 4 days before the main shock. The phenomenon occurred 13 days before the Qingchuan earthquake, lasted for 5 days, and returned to normal 8 days before the main shock.

(3) Spatial characteristics: Stations within 50km from the epicenter may record the phenomenon before impending earthquakes, while stations with epicenter distance greater than 50km generally cannot record the impending-earthquake microseismic fluctuations. Specifically, the impending-earthquake microseismic fluctuations were recorded by Dujiangyan station (the epicenter distance 19km), Yushu station (the epicenter distance 46km), and Qingchuan station (the epicenter distance 7km), while the other stations with the epicenter distance larger than 50km did not record the phenomenon.

(4) Directional characteristics of the activity degree N-value: The impending-earthquake phenomenon may be directional, and the change of direction of activity degree, N-value, may be related to the seismogenic fault, aftershock distribution and focal position of strong earthquake. Specifically, the N-value variation before the Wenchuan earthquake was significant on the N-S channel, while it remained relatively stable on the E-W and U-D channel. This is consistent with the strike of the seismogenic fault of Wenchuan earthquake and the spatial distribution of the aftershocks, which is in N-S, with N-S as the dominant direction. The variation of N-value of the Yushu earthquake is significant in the E-W direction, and the activity of the N-S and U-D channels is relatively stable. This is consistent with the strike of the Yushu earthquake seismogenic fault and the spatial distribution of aftershock, which is in near E-W direction, with E-W as the dominant direction. The change of N-value before the Qingchuan earthquake is relatively significant on U-D channel, and the variation on the N-S and E-W channels remain stable. Because Qingchuan station is only 7km away from the earthquake, the earthquake is a nearly right-beneath-station earthquake, so the U-D channel is dominant.

(5) The variation of microseismic activity degree N-value: The maximum N-value of Wenchuan earthquake reaches 18, the maximum N-value of Yushu earthquake is 8 and of Qingchuan earthquake is 5, showing the characteristic of larger magnitude corresponding to higher activity degree and smaller magnitude to lower activity degree.

3 TRACKING RESEARCH AND VERIFICATION OF THE PHENOMENON OF IMPENDING-EARTHQUAKE MICROSEISMIC FLUCTUATIONS

In order to further verify and test the impending-earthquake microseismic fluctuations and its main features, and considering that strong earthquakes are frequent in the Qinghai-Tibetan Plateau, which facilitates the verification, the author built a real-time earthquake dynamic monitoring system, tracking and Analysis Technical System for Qinghai-Tibetan Plateau in Lanzhou in June, 2011 (referred to as "Lanzhou Microseism Real-time Monitoring Technology System"). The system integrates the microseismic waveform data of about 200 broadband digital seismic stations in Gansu, Sichuan, Yunnan, Tibet, Qinghai and other regional networks to Lanzhou in real time, and uses 6 servers to form a technical support platform to realize the real-time, dynamic monitoring and tracking analysis of microseisms in Qinghai-Tibetan Plateau. The tracking and verification of impending-earthquake microseismic fluctuations and their main features were carried out through the whole-course tracking of the observations of 24 earthquakes with M≥5.0 during the period from 2012 to 2014, and the earthquake monitoring indicators and criteria were further refined.

3.1 Overview of Seismic Activity of the Qinghai-Tibetan Block from 2012 to 2014

Due to the huge amount of data, it has high requirement on data flow and technical system to guarantee a stable and continuous real-time calculation. In the 3-years continuous real-time monitoring, there are interruptions of recording in some periods of time, that is, the data in this period is incomplete and the earthquakes occurring in these periods cannot be analyzed and processed. Specifically, they are the Ninglang MS5.7 earthquake on June 2, 2012, the Baiyu MS5.4 earthquake on January 18, 2013, the Zadoi MS5.1 earthquake on January 30, 2013, and the Minxian-Zhangxian MS6.6 earthquake on July 22, 2013. In a short time before these 4 earthquakes, the station or the "Lanzhou microseism real-time monitoring technology system" was in a state of breakdown or records lacking.

According to the China Seismic Network, there are 28 independent earthquakes with M≥5.0 occurring in the monitoring area in the period of 2012-2014. After deducting the above four earthquakes, the specific parameters of the remaining 24 earthquakes are shown in Table 4.

Table 4 Verification of earthquake cases with M≥5.0 in the monitoring area from 2012 to 2014
3.2 Tracking and Verifying the Impending-earthquake Microseismic Fluctuation Phenomenon and Its Main Features

In order to verifiy the impending-earthquake microseismic fluctuation phenomenon and its main features, the author conducted a full space-time real-time tracking study of the 24 earthquakes of M≥5.0 listed in Table 4. At the same time, for the local areas with continuous observations but no earthquakes in the monitoring area, the impending-earthquake microseismic phenomenon is examined from the perspective of non-seismic inspection.

3.2.1 Study on the Earthquake Cases in the Monitoring Area from 2012 to 2014

The amount of data for real-time tracking analysis of microseisms is very large. Although the actual tracking process includes continuous real-time observations of more than 200 stations in the period from 2012 to 2014, the following discussion only introduces the tracking results of stations within the epicenter distance of 100km. The event activity degree N-value of different stations also shows only the data of 2 months before the earthquake. The tracking results of other stations and other periods are not described here. Because the analysis ideas are consistent, the representative earthquake is selected year by year as an example.

(1) The Yiliang MS5.6 Earthquake on September 7, 2012

The epicenter of the Yiliang MS5.6 earthquake on September 7, 2012 is located at 34km northeast of Zhaotong station, 85km from the Bingyidi station, 71km from Yanjin station, 96km from Leibo station, and 83km from Junlian station. During the period from August 1 to September 17 before the mainshock, the change of the activity degree of the microseisms is shown in Fig. 9. There was an obvious abnormal change in Zhaotong station in the period from August 21 to 27, 16 days before the earthquake occurred; the abnormal amplitude was 21, and the anomaly in N-S channel was prominent.

 Fig. 9 Changes in N values of different stations before the MS5.6 earthquake in Yiliang on September 7, 2012

(2) The Eryuan MS5.5 Earthquake on March 3, 2013

The epicenter of the Eryuan MS5.5 earthquake is located at 35km southwest of Eryuan station, 33km from Yunlong station, 65km from Tuanshan station, 85km from Luoshui station, and 86km from Heqing station. From February 1 to March 3 before the earthquake, the change of the activity degree of impending-earthquake microseismic fluctuation events is shown in Fig. 10. During February 9 to 15, there was a significant abnormal change in the Eryuan station, 24 days before the earthquake occurred. The abnormal amplitude reached to 6, and the anomaly in N-S channel was prominent.

 Fig. 10 Changes in N values of different stations before the Eryuan MS5.5 earthquake on March 3, 2013

(3) The Lushan MS7.0 Earthquake on April 20, 2013

The epicenter of the Lushan MS7.0 earthquake on April 20, 2013 is located on 24km north of Ya'an station, 82km from Guzan station, 91km from Emei station, and 99km from Xiaojin station. The changes of the activity degree of the impending-earthquake microseismic events from March 1 to April 20 before the earthquake are shown in Fig. 11. From April 15-19, there was an obvious abnormal change in Ya'an station, 6 days before the earthquake occurred; the maximum amplitude was 8 and the anomaly in N-S channel was prominent.

 Fig. 11 Changes in N values of different stations before the Lushan MS7.0 earthquake on April 20, 2013

(4) The Zhaotong MS6.5 Earthquake on August 3, 2014

The epicenter of the August 3, 2014, Zhaotong MS6.5 earthquake is located at 48km southwest of Zhaotong station, 41km from Qiaojia station, 81km from Puge station, 60km from Shimenkan station, and 80km from Bingyidi station. The change of activity degree of the impending-earthquake microseismic events from June 3 to August 3 before the earthquake is shown in Fig. 12. Among them, from July 20-25, there was a sudden change in the microseismic fluctuation at Zhaotong station, which was 15 days before the earthquake occurred; the maximum abnormal amplitude was 8, and the anomaly on N-S channel was prominent.

 Fig. 12 Changes in N values of different stations before the Zhaotong MS6.5 earthquake on August 3, 2014

(5) The Kangding MS6.3 Earthquake on November 22, 2014

The epicenter of the November 22, 2014, Kangding MS6.3 earthquake is located 50km west of Guzan station, 73km from Yajiang station, 96km from Daofu station and 100km from Xiaojin station. From September 22 to November 22 before the earthquake, the change of the activity degree, N-value, of the impending-earhtquake microseismic events is shown in Fig. 13. Among them, there were significant abnormal changes in Guzan station from November 8-18, 14 days before the earthquake occurred; the maximum abnormal amplitude was 6, and the anomaly in E-W channel was prominent.

 Fig. 13 Changes in N values of different stations before the Kangding MS6.3 earthquake on November 22, 2014

According to the above ideas, the other 20 earthquakes occurring in 2012-2014 were studied one by one. The basic results are summarized in Table 4. It can be seen from Table 4 that:

(1) Among the 24 earthquake cases, there are 14 earthquakes around which there are seismic stations within a radius of 50km from the epicenters. Between them, the Yiliang MS5.7 earthquake in 2012, the Eryuan MS5.5 earthquake in 2013, the Lushan MS7.0 earthquake in 2013, the Zhaotong MS6.5 earthquake in 2014 and the Kangding MS6.3 earthquake in 2014, similar phenomena of impending-earthquake microseismic fluctuations as that before the Wenchuan and Yushu earthquakes were observed.

(2) In the above 5 earthquake cases, in which the impending-earthquake microseismic fluctuations were observed, the longest duration from the occurrence of the microseismic fluctuation to the main shock is 24 days, the shortest is 6 days, and the dominant duration is concentrated in 14-16 days; In terms of spatial characteristics, stations within the epicenter distance of 50km recorded the impending-earthquake microseismic fluctuations before the 5 earthquakes, while stations with the epicenter distance larger than 50km failed to record them; in terms of directional characteristics, the direction, along which the activity degree N-value changes significantly, is basically consistent with the seismogenic fault, aftershock distribution and source location. Therefore, the results from the five earthquake cases are basically consistent with that of the Wenchuan, Yushu and Qingchuan earthquakes.

(3) In terms of seismic intensity, all 5 earthquake cases preceded with impending-earthquake microseism fluctuations are with magnitude greater than 5.5. However, in 7 out of 9 earthquake cases with magnitude 5.0-5.4, the phenomenon of impending-earthquake microseismic fluctuations was not recorded by stations within a radius of 50km from the epicenters. That is to say, the appearance of the impending earthquake phenomenon seems to be related to the earthquake magnitude. The larger the magnitude, the more likely the continuous impending-earthquake microseismic fluctuations would occur. Table 5 gives the statistical results of the above 14 earthquakes in different magnitude intervals. It can be seen from Table 5 that in the 7 earthquakes of magnitude 5.0-5.4, there were no imminent microseismic fluctuation phenomena before these earthquakes; in the 3 earthquakes of magnitude 5.5-5.9, 2 earthquakes were preceded with the phenomena, with the earthquake-reflecting rate of 67%; of the four earthquakes of magnitude 6.0 and above, three were preceded with pre-earthquake fluctuations, and the earthquake-reflecting rate is 75%. In fact, if only earthquakes of magnitude 7.0 or higher are considered, the Wenchuan, Yushu, Lushan earthquakes that occurred in the monitoring area since 2008 were all preceded with the phenomenon of impending-earthquake microseismic fluctuations, the earthquake-reflecting rate is 100%. Therefore, the impending earthquake phenomenon may be related to the magnitude of the earthquake. The larger the magnitude is, the greater the possibility of the continuous microseismic fluctuation before the earthquake will be.

Table 5 Earthquake-reflecting rate of different magnitude
3.2.2 The Seismic Tracking of Impending-earthquake Microseismic Fluctuation

There is a phenomenon that deserves attention in the 3-year real-time continuous observation of the impending-quake microseismic fluctuations. That is to say, during the three-year continuous observation from January 2012 to December 2014, there are very few stations which recorded the imminent microseismic fluctuations, and even if the fluctuations occur, the anomalous duration hardly reached or exceeded 4 days. To illustrate this phenomenon, stations of Tibet Digital Seismic Network and some stations in the Gansu-Qinghai-Sichuan border area are taken as examples.

The Tibet Digital Seismic Network consists of 13 stations. During the period from January 2012 to December 2014, none of the epicenters of earthquakes with M≥5.0 in Tibet were within the radius of 50km away from the stations. The variation process of N-value of 13 stations is shown in Fig. 14. As can be seen from Fig. 14, during the three-year continuous observation period, the N values of all stations are in a state of random fluctuation, and none of the stations has experienced the abnormal state similar to the cases of the Wenchuan, Yushu and Qingchuan earthquakes.

 Fig. 14 N-value change process of Tibetan seismic stations from January 2012 to December 2014

Similarly, there was no earthquake of M≥5.0 occurring in the period from January 2012 to December 2014 in the border area of Gansu-Qinghai-Sichuan. The N-value change process of stations in this area such as MQT, DBT, LTT, REG station are shown in Fig. 15. It can be seen from Fig. 15 that the N-value change process is similar to the characteristics of the Tibet network stations, that is, no abnormality, as those appearing before the Wenchuan, Yushu and Qingchuan earthquakes, occurred in any of the stations in three years.

 Fig. 15 The change of N-value of some stations at the junction of Gansu-Qinghai-Sichuan from January 2012 to December 2014

Therefore, the fact that stations of Tibet Digital Seismic Network and some stations in the border area of Gansu-Qinghai-Sichuan did not observe the phenomenon of the impending-earthquake microseismic fluctuation and that no earthquake occurred verifies the validity of the phenomenon of impending-earthquake microseismic fluctuation from another perspective.

4 CONCLUSIONS AND DISCUSSION 4.1 The Impending-earthquake Microseismic Fluctuation Phenomena and the Main Features

On the basis of case studies on the Wenchuan earthquake, Yushu earthquake, Qingchuan earthquake and other earthquake cases, and in combination with the actual tracking results of the above 24 earthquake cases, the following preliminary conclusions are drawn:

(1) The main characteristics of the phenomenon of microseismic fluctuation before impending earthquakes are as follows:① the spectrum range is wide, the dominant frequency is 11-16Hz, and the spectrum shape is relatively neat; ② it appears 6 to 24 days before the earthquake, with an average of about 15 days; ③ it may be recorded by stations with an epicenter distance of 50km. Stations with an epicenter distance greater than 50km generally cannot record it; ④ the anomaly may be directional, the direction of significant change of the activity degree N-value may be related with the source location, the seismogenic fault and the aftershocks distribution of the strong earthquake.

(2) The phenomenon of impending-earthquake microseismic fluctuation is reproducible. This phenomenon occurred not only before the Wenchuan, Yushu, Qingchuan earthquakes, but also before the earthquakes occured in Yiliang, Eryuan, Lushan, Kangding, etc.; and the three consecutive years observations and the seismic activity of Tibet and Gansu-Qinghai-Sichuan border areas have examined the objectivity of the impending-quake microseismic fluctuation from another perspective. This means that this phenomenon is a tracking indicator, and further exploration on it has practical value for imminent earthquake prediction.

(3) The phenomenon of impending-quake microseismic fluctuation may be related to the magnitude of earthquake. The larger the magnitude, the more likely it is that the earthquake will continue for several days before the earthquake. For earthquakes below magnitude 5.5, almost no impending-earthquake microseismic fluctuations are observed, except for earthquakes near the stations, such as the Qingchuan MS5.4 earthquake on November 1, 2011, the epicenter distance to Qingchuan station is only 7km.

4.2 Preliminary Refinement of Earthquake Tracking Indexes Based on the Impending-earthquake Microseismic Fluctuation

Based on the earthquake cases such as the Wenchuan earthquake, Yushu earthquake, Qingchuan earthquake and the actual tracking results of 24 earthquake cases from 2012 to 2014, the criteria for monitoring and tracking the earthquake situation can be preliminarily refined as a basis for further research and verification.

(1) Tracking indicators: With the target of the impending-earthquake microseismic fluctuations and its activity degrees N, the change of the earthquake situation is dynamically tracked. If the impending microseismic activity degree is randomly fluctuating within a certain range, it is normal activity and does not have strong earthquake risk; if the impending earthquake dynamic activity exceeds the variation range of random fluctuation and the duration reaches or exceeds 4 days, then this means that there may be a strong earthquake risk.

(2) Time criterion: When the N-value of the impending-earthquake microseismic dynamic activity degree changes abnormally, it is predicted that there may be a strong earthquake risk in the next 6 to 24 days or in an average of 15 days, calculated from the date of occurrence of the anomaly.

(3) Location criteria: There may be three criterias for sequential approximation:①The area with a radius of 50km around the station which records the variation of N-value may be the location where the earthquake occurs in the future. ②The predominant direction of impending-earthquake microseismic activity degree anomaly is possibly the direction of occurrence of strong earthquake in the future. ③ The active fault section existing in the area and direction indicated by criterion ① and criterion ② is possibly the earthquake risk section.

(4) Strength criterion: The earthquake magnitude predicted by the impending earthquake microseismic phenomenon should generally be above MS5.5, and the earthquake below MS5.5 is generally not predictive, unless it occurs near the station.

4.3 Discussion of the Related Issues

The relationship between the anomalous amplitude of the impending-earthquake microseismic activity degree, N-value, and the magnitude

In the cases of Wenchuan earthquake, Yushu earthquake, Qingchuan earthquake, the magnitude is large, so the N-value is large and has a long duration. and the magnitude is small, the N-value is small and the duration is short. However, this feature does not seem obvious in the earthquake cases tracking for the period between 2012 and 2014. Therefore, further research and verification are needed on the relationship between the abnormal magnitude, duration and magnitude of the N-value of the impending-earthquake microseismic activity degree.

Although this paper only gives the tracking results of the impending-earthquake microseismic fluctuations from 2012 to 2014, in fact, the "Lanzhou Microseism Real-time Monitoring Technology System" has been running continuously for 6 years. Concerning the typical earthquake cases in the period from 2015 to 2017, the results of analysis on the differences between spectral characteristics of typical geophysical events and environmental changes, and impending-quake microseismic fluctuations are subject to further discussion in a separate paper.

The Jinggu station, which is 27km away from the epicenter, did not record the impending-quake microseismic fluctuations before the Jinggu MS6.6 earthquake on October 7, 2014. Further investigation on the situation is needed.

4.4 Preliminary Discussion on the Mechanism of the Phenomenon of the Impending-earthquake Microseismic Fluctuation

Many studies have been published on the focal activities of strong earthquakes in the impending-earthquake stage, such as possible acceleration of fault creep (Chen Yuntai et al., 1979), and pre-slip before the earthquake (Guo Zengjian et al., 2000; Feng Deyi et al., 1984; Zhang Guomin et al., 1990; Sun Jizhu et al., 1996). The existing research results suggest that if micro activities such as impending-quake fault creep, micro fracture, pre-slip, etc. do exist, these pre-activities in the source regions can be recorded by broadband digital seismic observation instruments, then identifying these imminent pre-activities and further application to earthquake prediction will be a very valuable work.

As mentioned earlier, broadband digital seismic records contain a wealth of information, in which "interference events" superimposed on the normal microseisms may originate not only from the typical geophysical events and environmental changes such as earthquakes, blasts, collapse, typhoons, thunderstorms etc. but also from unknown or currently unrecognized processes, which may include pre-activity, micro-fracture and micro-vibration originating from the impending earthquake source, or micro-movement, micro-fracture, crack expansion of the active structure associated with the impending earthquake stage in an earthquake preparation process. To this end, Yang Liming(2009, 2014), Wang Yueqi (2015) have studied the spectrum of typical geophysical events and environmental changes such as blasting, collapse, typhoon, thunderstorm, etc., and initially believe that the spectral characteristics of the impending-earthquake microseismic fluctuations are different from the above events.

It is diffcult to determine that the impending-earthquake microseism fluctuations are related with the pre-activity, micro-fracture of the impending earthquake source area or the micro-movement and micro-fracture of the active structure when the earthquake preparation process enters into the imminent stage. However, the characteristics of reproducibility, directionality, impendency, near-source, and weakness of the impending-earthquake microseismic fluctuations appearing in the three earthquake cases of the Wenchuan earthquake, the Yushu earthquake, and the Qingchuan earthquake prompted the authors consider that the impending-earthquake microseismic fluctuations may be related to the pre-activity, micro-vibration, micro-fracture of the source area in the imminent stage, or micro-activity, micro-fracture and fissure expansion of the related active structure in the imminent stage. The continuous tracking study for 3 years from 2012 to 2014 tested and verified the phenomenon of impending earthquake microseismic fluctuation. The increase of relevant earthquake cases further strengthens the understanding of the reproducibility and objectivity of the phenomenon. Of course, this understanding requires more case studies to verify.

In fact, according to existing research (Ma Jin et al., 2012; Ma Jin, 2016), earthquakes are structural deformations that appear in the form of bursting, and there is a physical process that changes from slow to fast. Pre-slip, tremor and other phenomena just describe the process of structural deformation from slow to fast. At the same time, the occurrence of an earthquake is a mechanical process. When the fault is in the meta-stable stress state after the stress peak intensity, it enters an irreversible deformation stage, which indicates that the occurrence of the earthquake is inevitable. According to this, it is not difficult to consider that the impending microseismic fluctuation does not exist in the observations for years till the imminent stage, and it has characters of reproducibility, directionality, impendency, near-source, weakness etc. Whether this characteristic is the manifestation showing the source stress state entering into meta-stable stress state, and whether it can be used as a sign indicating that the earthquake preparation has entered into meta-stable stress state, it is worthy of further tracking research.

ACKNOWLEDGEMENT

Special thanks are extended for the valuable guidance, inspiration and encouragement from Academician Ma Jin, Professors Lu Yuanzhong, Zhang Xiaodong, Jiang Zaisen, Xu Ping, Nie Yong'an, Wu Yun, Che Shi, Ma Hongsheng, Zhou Longquan et al., and also to the Seismic Network Data Backup Center at Institute of Geophysics, China Earthquake Administration for providing the specific waveform data, and to Research Professor Zheng Xiufen who has offered a lot of helps.

REFERENCES
 Chen Yuntai, Huang Liren, Lin Banghui, Liu Miaolong, Wang XinhuaChen Yuntai, Huang Liren, Lin Banghui, Liu Miaolong, Wang Xinhua. A dislocation model of the Tangshan earthquake of 1976 from the inversion of geodetic data[J]. Acta Geophysica Sinica, 1979, 22(3): 201-217 (in Chinese with English abstract). Cooley J.W., Tukey J.W.Cooley J.W., Tukey J.W. An algorithm for the machine calculation of complex Fourier series[J]. Mathematics of Computation, 1965, 19(90): 297-301. DOI:10.1090/S0025-5718-1965-0178586-1 Feng Deyi, Sheng GuoyingFeng Deyi, Sheng Guoying. Premonitory abnormal characteristics of short-period Rayleigh surface wave before strong earthquakes[J]. Acta Geophysica Sinica, 1983, 26(3): 288-294 (in Chinese with English abstract). Feng Deyi, Pan Qinlong, Zheng Sihua, Xue Feng, Min XiangyiFeng Deyi, Pan Qinlong, Zheng Sihua, Xue Feng, Min Xiangyi. Long-period deformational waves and short-term and imminent earthquake precursors[J]. Acta Seismologica Sinica, 1984, 6(1): 41-57 (in Chinese with English abstract). Feng Deyi, Chen Huaran, Ding WeiguoFeng Deyi, Chen Huaran, Ding Weiguo. Study of anomalous characteristics of seismic wave spectra before large earthqukes[J]. Journal of Seismological Research, 1994, 17(4): 319-329 (in Chinese with English abstract). Gao Yuan, Liang Wei, Ding Xiang, Xue Yan, Cai Mingjun, Liu Xiqiang, Su Youjin, Peng LiguoGao Yuan, Liang Wei, Ding Xiang, Xue Yan, Cai Mingjun, Liu Xiqiang, Su Youjin, Peng Liguo. Variational characteristics of shear-wave splitting on the Shidian earthquakes in Yunnan, China[J]. Acta Seismologica Sinica, 2004, 26(6): 576-582 (in Chinese with English abstract). Guo Lücan, Yang Dongmei, Hu Changxi, Wu JianpingGuo Lücan, Yang Dongmei, Hu Changxi, Wu Jianping. On the research of foreshock and foreshock sequences by using of digital and simulated seismic records[J]. Earthquake Research in Shanxi, 1998(3/4): 11-16 (in Chinese with English abstract). Guo Zengjian, Qin BaoyanGuo Zengjian, Qin Baoyan. Discussion on some problems of earthquake source[J]. Earthquake, 2000, 20(1): 103-109 (in Chinese with English abstract). Liu Jie, Zheng Sihua, Kang Ying, Chou YongqingLiu Jie, Zheng Sihua, Kang Ying, Chou Yongqing. The focal mechanism determinations of moderate-small earthquakes using the first motion and amplitude ratio of P and S wave[J]. Earthquake, 2004, 24(1): 19-26 (in Chinese with English abstract). Ma Jin, Sherman S.I., Guo YanshuangMa Jin, Sherman S.I., Guo Yanshuang. Identification of meta-instable stress state based on experimental study of evolution of the temperature field during stick-slip instability on a 5° bending fault[J]. Science China Earth Sciences, 2012, 55(6): 869-881. DOI:10.1007/s11430-012-4423-2 Ma JinMa Jin. On "whether earthquake precursors help for prediction do exist"[J]. Chinese Science Bulletin, 2016, 61(4/5): 409-414. Sun Jizhu, Tang Chun'anSun Jizhu, Tang Chun'an. Ellipse area and precursor law of fault earthquakes[J]. Earthquake, 1996, 16(4): 355-362 (in Chinese with English abstract). Wang Yueqi. Study on Frequency Spectrum Characteristics of Different Type Events Record by Broadband Seismograph[D].Lanzhou: Lanzhou Institute of Seismology, 2015 (in Chinese with English abstract). Yang Liming, Liu XiaofengYang Liming, Liu Xiaofeng. A compositive scheme of earthquake short-term prediction in Qilianshan seismic belt[J]. Northwestern Seismological Journal, 2006, 28(3): 193-203, 209 (in Chinese with English abstract). Yang LimingYang Liming. Preliminary study on the tremors with special frequency recorded by seismograph before Wenchuan earthquake and its Charaters[J]. Recent Developments in World Seismology, 2009(1): 14-19 (in Chinese with English abstract). Yang Liming, Hao Zhen, Wang Jianjun, Zhang Shuzhen, Yao Jiajun, Dong LeiYang Liming, Hao Zhen, Wang Jianjun, Zhang Shuzhen, Yao Jiajun, Dong Lei. Preliminary study on phenomenon of microwave fluctuation impending strong earthquake (1)[J]. Earthquake Research in China, 2018a, 34(2): 219-233 (in Chinese with English abstract). Yang Liming, Hao Zhen, Wang Jianjun, Zhang Shuzhen, Yao Jiajun, Dong LeiYang Liming, Hao Zhen, Wang Jianjun, Zhang Shuzhen, Yao Jiajun, Dong Lei. Preliminary study on phenomenon of microwave fluctuation impending strong earthquake (2)[J]. Earthquake Research in China, 2018b, 34(2): 234-243 (in Chinese with English abstract). Zhang Guomin, Fu ZhengxiangZhang Guomin, Fu Zhengxiang. Discussions on complexity of earthquake precursors from a point of rock instability[J]. Journal of Seismological Research, 1990, 13(3): 215-222 (in Chinese with English abstract). Zhao Cuiping, Zhang Zhiqiang, Xia Aiguo, Liu JieZhao Cuiping, Zhang Zhiqiang, Xia Aiguo, Liu Jie. Study of attenuation characteristics of the central and eastern Tianshan, Xinjiang using digital seismic waves[J]. Journal of Disaster Prevention and Mitigation Engineering, 2004, 24(3): 300-305 (in Chinese with English abstract).