Low-Frequency Electromagnetic Signals Observed before Strong Earthquakes

We consider two kinds of signals preceding earthquake (EQ): intensification of internal electromagnetic (EM) field – lithosphere emission (LE) and change of the Earth interior response function (RF). Several cases of LE before strong EQs were reviewed and analyzed, and preliminary portrait of LE precursor was compiled. LE can appear several times with lead time month(s), weeks, days, and hours and can attain amplitude of several hundreds of nT which not uniformly decreases with increasing distance from the source. Typical LE frequency content/maximum is 0.01–0.5 Hz. Data of 19 Japanese geomagnetic observatories for 20 years preceding the Tohoku EQ on March 11, 2011 were analyzed, and RFs (mainly induction vector) were calculated. At six observatories in 2008–2010, anomalous variations of RF were separated which can be identified as middle-term precursors. Applying the original method developed in Ukraine, a short-term two-month-long precursor of bay-like form was separated by phase data of observatory KNZ in the Boso peninsula where electrical conductivity anomaly was also discovered. Hypothetical explanation based on tectonic data is advanced: Boso anomaly connects two large-scale conductors—Pacific seawater and deep magma reservoir beneath a volcanic belt. Between two so different conductors, an unstable transition zone sensitive to changes of stress before strong EQs can be expected.


Introduction
One of the long lasting challenges for the Earth sciences is earthquake (EQ ) prediction. EQ precursors deliver unique information which is necessary for the solution of two interconnected fundamental problems-EQ prediction (humanitarian practical aspect) and support to the development of EQ theory (scientific aspect).
The history of the EQ precursor study goes back to antiquity. But even now their study remains purely empirical, and any precursor even recorded with perfect instrument can be treated as not related with seismicity (skepticism to prediction widely spread now), and it is difficult to prove that it is genuine EQ precursor because physics of EQ preparation process is still not well understood. The causes of such situation are (1) the complexity of the real Earth and processes in it, (2) the absence of direct information from the place of EQ preparation, nucleation, and occurrence in the Earth interior. Nevertheless, consider the unique case of the successful EQ prediction-Haicheng prediction.

Haicheng EQM7.3
The Haicheng EQM7.3 occurred on February 4, 1975 at 19:35 local time in northeast China. After 1965, activation of seismicity occurred in an area of 120 km to SSW from Beijing with several destructive EQsM > 6. After this long-term prediction, the Chinese government greatly strengthened EQ study and precursor monitoring (telluric currents, well water, animal behavior, and other phenomena related with EQ ) attaching to observation experts and also amateurs and scholars. Many precursors were observed in 1973 and 1974 in a large area of 200 × 300 km (middle-term precursors) which in December 1974-January 1975 concentrated in a smaller area. In January 1975, quiescence of seismicity was observed but anomalies of groundwater, telluric currents, radon, tilt, animal behavior, etc. increased till January 23 (shortterm precursors), then slightly decreased, and since February 1 rouse in hundreds times. From 16:00, February 3, to 18:30, February 4, 500 EQs with M up to 4.2 occurred in the area between Yingkou and Haicheng cities. They were interpreted as foreshocks of strong EQ. Emergency evacuation was ordered by authorities, and law-abiding Chinese left their houses. At 19:36, a devastating EQ struck dozens of cities including Haicheng with almost 1 million inhabitants. Thousands of buildings collapsed, but hundred thousand lives have been saved by the well-timed prediction. Instructive lessons from successful Haicheng prediction are (1) continuous transition from long−/intermediate-term prediction to short/immediate ones with consistent improvement of expected EQ time, place, and magnitude; (2) close cooperation of authorities, scientists, amateurs, and mass media; and (3) the use of all available precursors, including "nonscientific" ones as animal behavior [1].
Multiparameter monitoring and strong scientific efforts of the last decades reveal some unexpected features of precursors: (1) Long-distance appearance up to thousands km from EQ epicenter. (2) Spatial selectivity: EQ precursors can be observed in some sensitive zones (usually fault zones) and be not observable in vast territories even not far from impending EQ. (3) Spatial-temporal migration of precursors: initially it appears in one locality, and then it appears in the next locality, usually with changed parameters. Such features did not find explanation in the framework of simple dominant ideas in the middle of the twentieth century about geological media. These features evidenced the complexity of geological media, and in the second half of the twentieth century, several new concepts have appeared to explain new data (Sadovsky MA, Varotsos P, Gufeld IL and many others).
For effective EQ prediction, we need automatic system of monitoring, processing, and analysis of all observed precursory parameters, their cross-correlation analysis to estimate probability of expected EQ (taking into account all previous global, regional, and local analyzes). High-level scientific team must keep contact with decision-making authorities for providing public announcement of the EQ prediction and plan of emergency measures. Such system needs great funds. Some elements of such system are created in few regions (California, China, ex-USSR countries).

Goal and scope of the chapter
Multitude of EQ precursors is the unique base for EQ prediction. Complexity of the earth and poor understanding of seismicity process enforce us to use as many kinds of precursors as possible. Two approaches are perspective for the fishing of the precursory signals from geomagnetic data: (1) Direct observation of the lithosphere emission (LE) of the internal electromagnetic fields arising in the course of EQ preparation and nucleation process and (2) transformation of the time series of the observed three components of geomagnetic field into time series of response functions of the Earth's interior conductivity.
In the next sub-section, we review few case stories of the most reliable records of ultralow-frequency (ULF) LE.
In Sections 3 and 4, we apply RF method to the Japanese geomagnetic observatories data in the attempt to separate the precursors of great Tohoku EQM9 11.03.2011, wherein obtained quite reliable result on the Boso conductivity anomaly and advance its tectonic interpretation.
In Section 5, we discuss the results obtained. In Section 6, we summarize the results and give the recommendations for the improvement of the low-frequency EM precursor study.

Case stories of LE records before strong EQs
There are many reports on LE registration, in particular before EQs. Consider fortunate cases when EM observations turned out to be located in the places where LE field was well above magnetotelluric (MT) field+noise background and can be easily identified.

Great Alaska EQM9.2 on March 28, 1964
Geomagnetic observatory in the city of Kodiak was located in the distance of 440 km from the epicenter of the EQ and only in 30 km from the fault zone along which displacement occurred. The full field proton magnetometer recorded several magnetic disturbances. The strongest one with intensity 100nT appeared 1 h 06 min before the EQ [6].

Loma Prieta EQM7.1 on October 18, 1989
One of the most convincing cases of LE precursors was recorded before that EQ [7]. The monitoring system of Stanford University (created for traffic noise study) operated since October 1987 at the distance of 7 km from the future EQ epicenter. The system included induction coils and special computer which calculated half-hourly averages of the magnetic field power spectra in each of nine narrow frequency bands covering the overall range 0.01-10 Hz.
During 23 months record was normal with low noise. After September 12, 1989, anomalous signal began to appear in two frequency bands: 0.05-0.1 and 0.1-0.2 Hz and grew up to 1.5 nT. In October 5, a large increase of amplitude appeared at all frequencies with the strongest one at the lowest frequency 0.01 Hz, where it reached 30 times the normal level. On the last several days before EQ , anomaly gradually diminishes (a quiescence!), and 3 h before the EQ , very large amplitude appeared only at periods longer 0.5 Hz, exceptionally large at frequency 0.01 Hz. We must emphasize that instrumentation of Stanford University which allowed to get results every half an hour is very good for LE monitoring. Unfortunately, it did not continue the operation for EQ prediction.

Caucasus
Kopytenko et al. [8,9] developed three-component magnetometers for frequency range 0.005-10 Hz. The first instrument started the record 23 days before destroying Spitak EQM6.9 at 7.41 UT, on December 7, 1988, in geomagnetic observatory at Dusheti, 129 km to the north from the EQ epicenter. It recorded intensive B LE which started 4 h prior the EQ (Figure 1a, b).
In the time interval November 14, 1988 to March 5, 1989 in frequency range 0.1-1 Hz 59 unusual noise-like bursts of LE with an amplitude well higher the background noise (0.03 nT) and the duration ranging from several minutes to several hours (mean duration ≈30 min) were recorded mainly before the strong aftershocks. Decrease in aftershock activities and B LE activity was quite synchronous.
The next strong event was Racha EQM6.9 at 9.13 UT, on April 29, 1991, occurred at the epicentral distance of 90 km from Dusheti where no pronounced B LE were recorded. It means that Dusheti is not a sensitive place for the EQs in Racha region (effect of spatial selectivity). Kopytenko's team organized B LE observations in two field sites Nik and Oni for June-July 1991-the time of Racha aftershock activities. Forty-seven B LE with intensity up to 2 nT and duration ranging from several minutes to several hours were observed at both observatories, 23 of them were recorded 1-4 days before the strong aftershock M6.2 on June 15,1991. Figure 1c presents B LE bursts observed before aftershock M4.0. The distance from the epicenter to Oni was two times smaller than to Nik, but intensity of LE signal at Nik was considerably larger: for H component in 10-20 times, for D in 1.3-2 times, and for Z in 5-7 times. It means strong spatial selectivity of the LE signals preceding EQ [10,11].  [8,9]. Low-Frequency Electromagnetic Signals Observed before Strong Earthquakes DOI: http://dx.doi.org /10.5772/intechopen.88522 In the conclusion we like to emphasize that magnetometers in described studies can register variations in the frequency band of 0.005-10 Hz, but they recorded most intensive B LE at the frequencies 0.1-1 Hz that differs from the result in Loma Prieta, where the maximum tends to their lowest frequency 0.01 Hz.

Taiwan
The island-wide geomagnetic network consisted of three-component geomagnetic observatory LP in the seismically quiet area and seven full field stations equipped by proton magnetometers with 0.1 nT sensitivity and sampling rate of 10 min distributed in areas of high seismicity [12]. Chi-Chi EQM7.6 occurred on September 21, 1999, in the middle of Taiwan. Stations LY turned out to be just near to surface rupture of the EQ along Chelungpu fault and recorded the strongest LE which clearly separated from comparison with records of remote observatory LP. LE begun more than a month before the EQ and attained in maximum 200 nT, and then its amplitudes gradually weakened, and the disturbance level reduced to that of a quiet period almost right after the second strong Chia-Yi EQM6.4 that occurred near the southern end of the Chelungpu fault on October 22, 1999 [12].

Greece
In Greece in the early 1980s for the registration of the LE electric components, the so-called seismic electric signals (SES) special network was created by Prof.  [11]. Amplitudes are given in relative units. P. Varotsos [11,13]. The network consists of 10-15 stations. Each station included several (from 6 up to 100) grounded electrical dipoles with the length from 50 m up to 20 km that allows to study spatial characteristics of observed field and separate SES from MT field and noise. In the course of continuous monitoring for more than 35 years, Prof. Varotsos and collaborators identified (as the result of a posteriori analysis) many SES before the following EQs and studied their regularities [11], for example, the selectivity effect: SES can be observed in some sensitive zones and be not observable in vast territories even not far from impending EQ. Prof. Varotsos made a number of correct EQ predictions registered officially before the event. We show interesting case of joint registration SES and horizontal magnetic components recorded on April 19, 1995, 25 days before Grevena-Kozani EQM6.6 on May 13, 1995 (Figure 2). Magnetic components look as derivative of electrical impulses that are clearly seen in the lower graph (b) with expanded time scale. In the latter years, Prof. Varotsos' group develops deeper insight in the physics of LE: entropy and natural time analysis for the better understanding of the EQ preparation process and for the distinguishing LE signals from similarly looking variations of MT and noise origin [14,15].

Varying geomagnetic field
Varying geomagnetic field B(t) = B x e x + B y e y + B z e z (where e x , e y , and e z are unit vectors directed to north, east, and downward correspondingly) continuously exists in and around the Earth and is recorded nowadays digitally with a time reading interval Δt (1 min or 1 s in our study). All components are functions of time t which we skip out for brevity.

Main sources of observed geomagnetic field
where B MT = B MT ext + B MT intn + B MT inta is the magnetotelluric field. B MT ext is the external primary magnetic field of the currents in magnetosphere and ionosphere of the Earth. B MT intn is the normal internal secondary magnetic field of the currents induced by B MT ext in the Earth's interior having a hypothetical "normal" horizontally layered conductivity structure. The horizontal components of B MT intn add together with B MT ext increasing observed horizontal MT-field up to doubling over hypothetical ideally conducting Earth. On the contrary, the vertical component of B MT intn subtracts from B MT ext diminishing observed normal vertical MT field up to 0 (phase neglected).
Thus, B MT n = B MT ext + B MT intn has the horizontal component much greater than the vertical one and embraces great territory equal to external source field dimension.
B MT inta is the anomalous internal secondary field arising on local/regional conductivity anomalies as a result of redistribution of the currents responsible for B MT intn.
Such subdividing of secondary internal field is rather artificial, but it is used in geoelectromagnetic methods: B MT intn in sounding methods, magnetotelluric sounding (MTS) and geomagnetic deep sounding (GDS); B MT inta in profiling one, magnetic variation profiling (MVP) [2][3][4][5]. Bnoise is the ever-present noise, both man-made and natural. B LE is the purely internal field of lithosphere emission, it is usually local and transient, and the ratio of vertical component to the horizontal one usually differs from the same ratio for MT field.
After the conventional processing using Fourier transform, a B(t) record of duration t 2 −t 1 is transformed into a superposition of harmonic components with periods T 1 , T 2 …T n .

Response function
Response function is the term widely used in natural sciences and mathematics. In the geoelectromagnetic studies of electrical conductivity σ (x, y, z) of the Earth's interior [2][3][4], the RFs are supposed to be some functions derived from the Earth's electromagnetic (EM) data that provides us with a possibility to determine the conductivity structure of the Earth. EM RFs are usually frequency/(period T) dependent, and then they are complex functions having real (index u) and imaginary (index v) parts. We use two of these functions.

Induction vector
Induction vector C = Ae x + Be y (A and B are determined from the linear equation: B z = AB x + BB y ). Real induction vectors C u = A u e x + B u e y possess an important property: in the notation of Wiese used here, they are directed away from a good conductor.

Anomalous horizontal magnetic variation tensor
Anomalous horizontal magnetic variation tensor [M] is determined from the linear system of equations B x (r i ) = M xx B x (r 0 ) + M xy B y (r 0 ) and B y (r i ) = M yx B x (r 0 ) + M yy B y (r 0 ), where r 0 and r i are coordinates of base (reference) and some other observation place. Tensor [M] reflects the change in geoelectric structures between two places (if the source field used is uniform).

The processing of observed geomagnetic field
The processing of observed geomagnetic field B(t) for the monitoring of geodynamic and other environmental processes implies transformation of three components of geomagnetic field time series with time reading interval Δt (1 min or 1 s in our study) into a variety of time series with temporal resolution Δτ (Δτ»Δt) of different RF components: Re and Im and x and y at the set of periods T 1 , T 2 …T n of received harmonics (Δt « T i « Δτ).

The theory of geoelectromagnetic methods
The theory of geoelectromagnetic methods [2][3][4] is developed for natural source field in the form of vertically incident plane wave (Tikhonov-Cagniard (T-C) model), which usually holds for an external source field of magnetosphereionosphere origin (named as magnetotelluric field) for the periods less than 10 4 s. Ideally RF depends only on the Earth's conductivity distribution which is sensitive to the stress variations and therefore to geodynamic processes including the earthquake preparation.

Variations of geomagnetic response functions (mainly induction vector) before the 2011 Tohoku earthquake
RFs and their variations, especially in relation with EQs preparation, were studied in Japan for many years and were described in many works among which we cite only few [16][17][18]. In the last two decades, the RF approach became less used for EQ studies because of strong noise at Japanese observatories. After the catastrophic Tohoku EQ on March 11, 2011, near Japan, we analyzed the available geomagnetic data to obtain some EQ precursors using the RF method. Some results were presented in Russian [19][20][21], which together with the latest results of our study will be summarized below.
Data used: We obtained data of 17 Japanese geomagnetic observatories (Figure 3, Table 1) for time interval of 12-20 years with temporal reading every 1 min (and few observatories with 1 s reading). For the conversion of geomagnetic field B time series into RF time series, we used the advanced multi-window remote reference (rr) robust programs with coherency control [22,23]. After processing we got values of four components of induction vector A u , B u , A v , B v for each day for five period intervals centered at: 225, 450, 900, 1800, and 3600 s. To reduce the great scattering, everyday values were smoothed with moving windows and/or were found average or median values for some interval (usually 1 month).

Results of processing for separation of middle-term precursors
Analyzing large material of processed data for 15 years from 2001 till 2015, we found that aperiodic variations (or enhancement of annual variation) of induction vectors were observed at periods 225, 450, and 900 s during 3-5 years before the Tohoku EQ at stations: HAR, KAK, OTA, KNZ, and TTK, most clearly at period 450 s presented in Figure 4. We should emphasize that no such aperiodic variations were observed at other stations including ESA and MIZ, which are the nearest to the EQ epicenter. The best correlation of middle-term anomalies is observed between the two most remote (620 km) from each other stations HAR and TTK: at both we see strong synchronous variations of induction vectors with maxima in the end of 2008, the end of 2009, with several maxima in 2010, in the beginning of 2011, and  return to the previous level after the Tohoku EQ. We may suppose that these aperiodic variations can be the middle-term precursors of the Tohoku EQ. These observatories are located not at the shortest distance from the EQ , which is in agreement with well-known phenomenon of spatial selectivity of EQ precursors known during the centuries for hydrological precursors and recently proven for LE registered in the form of seismic electric signal [11,13].
Having 1 min time series, we can analyze only geomagnetic variations with period T > 3 min, and the most interesting shorter part of ULF spectra (0.01-10 Hz), where strongest B LE have been observed [7][8][9], is left not resolved. So, we get 1 s data for observatories KAK, KNZ, ESA, MIZ, and for short time intervals for UCU and KYS to analyze RF for periods T > 5 s.

Boso conductivity anomaly
Processing of records from 18 observatories (16 of them are shown in Figure 3 and KYS and UCU in Figure 5) for the determination of horizontal tensors [M] with KAK as the base station yields the absence of noticeable horizontal tensor anomalies in ESA, MIZ, HAR, TTK, and MUR but reveals their existence in KNZ, UCU, OTA, and KYS (Figure 4a). In KNZ and OTA the enhancement of real tensor components M xx and M yy equals to ≈40% and ≈30% correspondingly at periods T < 500 s decreasing at longer periods. This result was supported by direct visual measurements described below. At closely located observatories KNZ and OTA, the latitudinal (E-W) component of induction vector at period 450 s and shorter increased (in 2011 comparatively to 2001) in opposite directions: westward in KNZ and eastward in OTA (see Figure 3b). It means that between these two observatories, an additional current (of geodynamic origin) appeared in 2011.

Visual analysis of geomagnetic records
Considering the geomagnetic field synchronous records (Figure 3a), we noticed that magnetotelluric field appears synchronously at all observatories, while noise appears locally at each one. The stations most contaminated by noise are UCU and KNZ which are the nearest to DC railways. But during the after-midnight time interval from ≈1:30 to ≈4:30 LT (16:30-19:30 UT) the strong noise from DC railways almost disappears.
Direct measurements of the strong MT amplitude variations in each component provide a check (not precise but very reliable) of the results obtained by processing. So, the enhancement of B x at KNZ and OTA at approximately 30-40% exists, and it can be interpreted only by the electrical conductivity anomaly under the observatories, i.e. under the central part of the Boso peninsula.

Comparison with geology and tectonic evidence
The relation between M xx and M yy anomalies in KNZ defines WNW-ESE strike of the Boso conductivity anomaly. Geological data [25,26] presented in Figure 4b-c  Figure 6a. It means that the direction of anomalous currents is also different under each observatory. Calculations show that at least 50% of anomalous currents should be located near the surface in the sediments of the Kanto plain to fit the received data.
On the other hand, the plate tectonic evidences that the Boso anomaly is located over the Sagami trough, structure at the depth 15-20 km in the complex junction of three lithosphere plates (Figure 3a). Strike of the trough is the same, WNW-ESE, so some part of the anomalous conductor can be located in the Sagami trough.
The eastern part of both conductors (shallow sediments and deep trough) has contact with seawater, while the western one can contact with a magma reservoir. In such a circuit it can be some unstable area(s) with conductivity strongly dependent on stress and sensitive to stress changes related with EQs.  (15 h UT). Magnetic activity was very low: Kp =0+. DC railway traffic is almost stopped, but the noise-like variations were not terminated in the Boso peninsula. Maybe it is a precursory LE which according to [24] started on February 22, 2011. In Figure 6b, vectors are shown for a period of about an hour (4000 s), at which industrial noise is practically absent and vectors adequately reflect the heterogeneity of geoelectric structure. In Figure 6c, vectors at the period 25 s are built with dominated noise field, which is greater than MT field on four observatories considered. Real and imaginary vectors at periods 16,25,50, and partly at 200 s (Figure 6d) are directed to the source of noises-the nearest railway. To reduce influence of the noise, night records and remote reference technique were used (Figure 6f). Received corrected vectors appeared still very scattered (Figure 6e) and for monitoring of geodynamic processes can be used cautiously. However, vectors averaged over a long period of time can be used for clarifying of the geoelectric structure. Corrected real vector in KNZ at the shortest periods directs to north.  [25,26] and induction vectors for period ≈1 h, named observatories with real and imaginary vectors-is our processing, the other six real vectors are taken from [16]. It means that the most conductive part of Boso electrical conductivity anomaly is located to south of KNZ, apparently near the southern side of the asymmetric "N.8 half graben fills" of the Miura Group sediments [26].
Results of the single station and nighttime records processing at KNZ are given in Figure 6f. We see that full-day and nighttime results significantly differ from each other only for B u and partly B v components because a railway is located to the west from KNZ and brings the distortions mainly in the eastern component. The northern component is less affected by noise at periods 100 s and more that opens the possibility to use it for separation of EQ precursor that will be used in the next section.

Introductory remarks
The induction vector derived from very noisy records, practically from noise field, has small stable phase. Therefore, if some other magnetic fields, which are usually not so stable (let it be a precursor field), are superimposed on the field of noise, exactly the phase of induction vector will be the most sensitive component for a precursor separation.
Below we apply a new approach developed by Tregubenko [27], who used it for processing the data of seismo-prognosis monitoring network in Ukraine. He separated precursors before few strongest (M ≈ 4) Crimean EQs occurred during 15 years, in particular before the Sudak, Crimea EQM3.9 on January 24, 2005 [27]. We applied this approach to KNZ, KAK, and ESA 1-s data, but the precursor was found only in KNZ. We can explain this by the spatial selectivity of the precursors: high sensitivity of KNZ place is quite natural in virtue of Boso electrical conductivity anomaly located just under KNZ observatory.

Processing of KNZ data
The processing was made with the use of Varentsov's [22] program. Coherences were used as weight estimates for averaging the results. To minimize the effect of noise, the estimates with multiple coherences less than 0.7 were ignored that allowed us to obtain minimally shifted estimates of induction vector's components. Maximum anomalous effect before the Tohoku EQ was observed for the phase of the induction vector northern component-arg(A) for periods 100-200 s. For the longer periods, the anomaly gradually decreases, so that we now present the results for T = 100 s. The processing and analysis were made in two steps: Step 1. A 7-year-long (2005-2011) geomagnetic field time series with every 1 s reading was processed for every month as a single unit. Arg(A) time series with every month reading were received as in Figure 7a and analyzed by a polynomial approximation approach. The most significant first-, second-, and seventh (quasiannual)-order polynomials were extracted from the rough data, and we obtained Figure 7b which is more perspective for comparison with EQs. But 1-month temporal resolution of RF is not sufficient for such an analysis. As for the Tohoku EQ precursors, we see a 9-month-long negative anomaly in 2010 approximately 1 year before the main event.
Step 2. A 2-year-long (2010-2011) 1 s time series were processed for every day as single unit. Large scatter of everyday results was reduced by averaging with moving window of 5 days long with 1-day shift. From these curves, i.e., arg(A) time series with everyday reading, we extracted first-, second-, and seventh (quasi-annual)order polynomials determined in step 1. The result is shown by the gray rough curve

Discussion of short term precursor
The variations of arg(A) given by the bold solid line in Figure 8b are cleaned from periodic variations of annual and monthly periods, which make it more convenient to identify EQ precursors. Indeed, we see a strong bay-like variation that begins almost 2 months before the Tohoku EQ and has approximately 2-month duration. This is in good agreement with the finding of Prof. Varotsos' team [14,15]. In particular, Varotsos et al. showed that the initiation of an SES activity in Japan appears almost simultaneously with the minimum of the fluctuations of the order parameter of seismicity analyzed in natural time, and such a minimum was clearly observed [15] on January 5, 2011, that is, almost 2 months before the Tohoku EQ occurrence.
Time of beginning of a bay-like precursory variation and its duration depends on the magnitude of an expected EQ. This time is equal to approximately 2 weeks for the processed Crimean EQs with magnitude approximately 4 [27] and 2 months for Tohoku EQ with magnitude 9 according to our study.  Kopytenko et al. [24] compared nighttime records of KAK and UCU observatories (see Figures 3, 6) in frequency range of 0.33-0.01 Hz for the interval of 21 days before the Tohoku EQ , that is, since February 2, 2011 till March 3, 2011. They found the appearance of anomalous changes on February 22, 2011 (18 days before EQ ): decrease of the correlation coefficients between geomagnetic components of KAK and UCU observatories and rise of B z component in sub-diapason 0.033-0.01 Hz. It was interpreted as appearance LE. Now let us see Figure 5b. It presents nighttime records on February 24, 2011, 16 days before the Tohoku EQ , in geomagnetically quite interval with rather good temporal and amplitude resolution. MT variations in the horizontal components are almost the same at all presented observatories distributed at the territory 2000 km long. In the records of UCU and KNZ (separated by 17 km) we see strong varitations with frequencies of 0.002-0.1 Hz and amplitude of 0.2 and 0.5 nT, respectively, and even more strong variations in B z component. All of this is in good agreement with the results of [24]. Variations in KNZ and UCU cover approximately the same frequency diapason as in [24], slightly correlate one with the other, and are not observed at other observatories. All signs of LE! But we cannot exclude that there are some remainders of the daytime noise from DC traffic. We need several more observatories (as SES network in Greece) for more definite conclusions.

Discussion of LE features
To use the LE for EQ prediction, one needs to know its lead time, amplitude, frequency characteristic, and expected distribution of sensitive places in the Earth surface. This knowledge can be obtained now only empirically. We can extract the necessary properties from the data presented in Section 1.3 supplementing them with other published data. An attempt of such extraction is presented in Table 2.
The results depend of the conditions of observation. So, sampling rate of 10 min and compressed time scale in [12] describing two EQs in Taiwan exclude frequency content estimation. Important parameter -lead time ∆t is properly determined only before Loma Prieta EQ when signal-to-noise ratio was large during long time that allowed separate three stages of the precursor appearance. Spatial selectivity complicates the formulation of the LE spatial regularities. Thus, we are in the very beginning of LE phenomenon study and use.

Conclusions
We have calculated induction vectors using data from Japanese observatories for many years preceding the 2011 Tohoku EQ. In 2008-2010 at six observatories, we found anomalous variations of induction vectors, which are regarded as middleterm EQ precursors. Those observatories are located not at the shortest distances from the EQ epicenter, which is in general agreement with the well-known phenomenon of spatial selectivity of EQ precursors. The analysis of horizontal tensors reveals a conductivity anomaly under the central part of the Boso peninsula with a WNW-ESE strike coinciding both with the Sagami trough strike and well conducting 3-km-thick sediment strike. A joint analysis of geoelectric and tectonic data leads to a preliminary conclusion that the Boso conductivity anomaly connects two large-scale conductors: Pacific seawater and a deep magma reservoir beneath a volcanic belt. Similar anomaly was found earlier in Kamchatka [21]. Then, applying original data analysis with the elimination of annual and monthly variations, we separated two-month-long short-term EQ precursor of the Tohoku EQ.
Several cases of lithosphere emission LE before strong EQs were reviewed and analyzed, and preliminary portrait of LE precursor was compiled: LE can appear several times with lead time a month(s), weeks, days, hours, and minutes and can attain amplitude several hundreds of nT which rapidly and not uniformly diminishes with moving away from the source. Typical frequency content/maximum is 0.01-0.5 Hz. As it is widely accepted [5], LE is generated by the process of