Radio Science

On the lithosphere-atmosphere coupling of seismo-electromagnetic signals



[1] Employing borehole and terrestrial antennas, vertical electric field components of the naturally occurring very low frequency electromagnetic emissions at the frequency of 3 kHz have been monitored simultaneously at Agra (geographic latitude 27.8°N, longitude 78°E), India, from 15 March 1999 to 30 September 1999. This period of observation included a major seismic swarm activity in the months of March and April which occurred in the Chamoli hills of north India, the next two months of May and June being quiet with respect to local lightning and spheric activities, and the rest of the three months from July to September being highly disturbed due to local lightning and thunderstorm activities. The abnormal electric field changes occurred in the form of noise bursts of varying amplitude and duration and included three kinds of data: (1) noise bursts observed by borehole antenna only, (2) noise bursts observed by terrestrial antenna only, and (3) noise bursts observed by both the antennas. We find that the occurrence of the first kind of data is positively correlated with major seismic activities in the region over the period of observations. The third kind of data indicates coupling between the two antennas dominated overall by atmospheric emissions. The long distance propagation of the seismo-electromagnetic emissions through the middle layer crust working as waveguide or through seismic faults is found to be associated with large attenuation ≈13 dB/km. Hence, the observation of the emissions at Agra, about 400 km from Chamoli, is interpreted in terms of leakage to the atmosphere through “windows” of low conductivity in the skin layer near the epicenter, possibly produced by some geophysical formations as discussed by other workers and then propagation in the Earth-ionosphere waveguide.

1. Introduction

[2] The association of electromagnetic emissions with impending earthquakes, first reported by Gokhberg et al. [1982], has evoked considerable interest in radio scientists and seismologists all over the globe, and a lot of work has been done in this field by ground- and satellite-based observations in various frequency bands [Warwick et al., 1982; Larkina et al., 1983; Parrot and Mogilevsky, 1989; Fujinawa and Takahashi, 1990; Molchanov et al., 1993, 1998; Hayakawa et al., 1996]. Such association has also been confirmed from laboratory experiments employing rock fracturing [Nitsan, 1977; Cress et al., 1987; Yamada et al., 1989]. Parrot [1995] and Hayakawa [1996] have reviewed the work done in this field thoroughly, and Hayakawa and Fujinawa [1994], Hayakawa [1999], and Hayakawa and Molchanov [2002] have presented recent work in three monographs.

[3] While the observations of ultra low frequency (ULF) emissions above the ground surface and in the ionosphere and magnetosphere have been supported by model calculations [Molchanov et al., 1995], it is still not clear how the emissions at higher frequencies (ELF/VLF to HF) are observed above the ground in the presence of heavy attenuation in the crustal region and skin effect. Some authors have suggested a mechanism related to redistribution of electrical charges in the Earth's atmosphere system which produced electrical discharges [Gokhberg et al., 1984; Schloessin, 1985], while others have suggested propagation through seismic faults in the presence of low attenuation in a manner similar to that of propagation of electromagnetic waves in Earth-ionosphere waveguide [Kingsley, 1989; Singh et al., 2000]. In order to resolve this problem we have conducted simultaneous measurements of vertical components of VLF electromagnetic emissions both below and above the ground surface by employing borehole and terrestrial antennas and present the results in this paper. We show that seismogenic emissions are mostly induced in the borehole antenna and lithosphere -atmosphere coupling is dominated by emissions of atmospheric sources. However, leakage of the seismogenic emissions to atmosphere is possible in especial conditions.

2. Experimental Setup

[4] The experimental setup employed by us is shown in Figure 1. It employs borehole and terrestrial antennas that are connected to amplifier and filter systems and the outputs are recorded on a two-channel D.C. ink chart recorder. The borehole antenna is naked copper wire of 120 m length and 4 mm diameter placed in a water- tight PVC pipe of 3.5 cm diameter with its lower end tightly fitted with an insulating black cork at the bottom. This is placed in another PVC pipe of 7.5 cm diameter which is open at both ends and is used to reduce the upward thrust (buoyancy) of the underground water on the pipe containing the antenna. Another electrode is placed 3 m down in contact with the ground to provide the Earth terminal. The natural potential between the two electrodes is found to be about 20 mV in noise-free environment. The vertical terrestrial antenna is also of similar copper wire but 20 m height. The amplifiers used are transistorized preamplifiers and main amplifiers of gain 40 dB each. The active band-pass filters provide peak frequency at 3 kHz with bandwidth ∼250 Hz. The D.C. ink chart recorder is similar to model A602C obtained from Esterline Angus USA. The chart speed is maintained at 0.5 cm/min. The chart recorder measures the current (0–5 mA) and its internal resistance is 65 ohms. However, we have modified it to measure the current between 0 and 10 mA.

Figure 1.

Experimental arrangement for subsurface and atmospheric measurement of vertical electric field emissions using borehole and terrestrial antennas.

[5] Observations were taken round the clock except for a couple of breaks two hours in the morning between 0700–0900 hours and two hours in the evening between 1700–1900 hours at Bichpuri, a rural area, about 12 km west of Agra (geographic latitude 27.8°N, longitude 78°E), where local electric and electromagnetic noises are very low.

3. Earthquake Data

[6] The earthquake data have been obtained from United States Geological Survey through India Meteorological Department, New Delhi. For the purpose of the present study, we have considered only high magnitude earthquakes (Ms ≥ 4.5) that occurred between 15 March and 30 September 1999 over seismic zones in northern India and around in the latitude-longitude ranges of 25°–36°N and 69°–100°E. However, the earthquakes whose epicenters are outside India in neighboring countries of Pakistan and Afghanistan are only those whose magnitudes are ≥5.0 because the effects of such earthquakes are observed in India also. In Figure 2, we show the locations of these earthquakes in a map by solid circles. An interesting feature of the Indian earthquakes is that they included a severe seismic swarm which occurred in the Chamoli hills of Himalayan region (geographic latitude 30.51°N, longitude 79.4°E) of which the main shock (Ms ≥ 6.6) and six others of high magnitudes occurred on the same day of 29 March 1999. The seismic activity continued with aftershocks up to 18 April 1999 and included five shocks of large magnitudes on the same day of 7 April 1999. All these earthquakes occurred at the same place within ±50 km from the epicenter of the main shock and their locations are encircled in the map of Figure 2 and marked by CHM. Other details of the figure will be presented later.

Figure 2.

Map showing the locations of high magnitude earthquakes (Ms ≥ 4.5) by solid circles considered in this paper during the period of observations. The encircled earthquakes are a seismic swarm which occurred in Chamoli region of Himalaya. The thick curve is a main boundary fault (MBF). The observing station Agra (AGR) and Indian capital Delhi (DLH) are indicated by big solid circles. The thick line indicates transverse conductive channels near Delhi-Haridwar ridge (DHR) shown by parallel lines.

4. Results and Discussion

[7] The subsurface measurements of vertical components of electric field emissions at the frequency of 3 kHz was started at our station from 1 February 1998 employing a borehole antenna initially. The frequency of 3 kHz was chosen as a trade-off between the unwanted noise contamination caused by lower frequencies of power line harmonic radiations and ELF atmospherics and increasing attenuation at frequencies above 3 kHz. The anomalous electric field changes appeared in the form of signal bursts of varying amplitudes and durations. For the sake of convenience, we call them as noise bursts. A majority of these noise bursts were produced by known sources identified as local lightning, spherics, local radio transmissions, and magnetospheric VLF radiations etc. A detailed description of these bursts and their implication in earthquake prediction research has been given by Singh et al. [2000].

[8] The simultaneous operation of the terrestrial antenna in conjunction with the borehole antenna was started from 15 March 1999 and continued till 30 September 1999. This period of observation is significant in the sense that the first two months of March and April included a major seismic swarm activity that occurred in the Chamoli hills of north India about 400 km north of the observing station at Agra (described in detail in the preceding section), the next two months of May and June were quiet months in respect of lightning and spheric activities in our Agra region, and the remaining period from July to September was full of local lightning and thunderstorm activities in this region. Hence, the data corresponding to six and half months of observations are highly useful for the purpose of present study.

[9] From a careful examination of the chart records corresponding to the period of observations, we have noticed three kinds of data in general, i.e. (1) noise bursts recorded by borehole antenna only, (2) noise bursts recorded by terrestrial antenna only, and (3) noise bursts recorded by both the borehole and terrestrial antennas. Examples of the three kinds of data are shown in Figure 3. The top panel of this figure shows the electric field changes recorded by borehole antenna only on 28 March 1999 about 21 hours before the occurrence of main shock in Chamoli, the middle panel shows the noise bursts recorded by terrestrial antenna at 2148 hours on 24 May 1999 and the bottom panel shows the electric field changes recorded by both the antennas in the early morning hours of 18 March 1999. We have separated each kind of data and calculated their occurrence number in each month between March and September 1999. Here it may be made clear that a noise burst varying in duration from few minutes to few hours with amplitude enhancement greater than 1 dB from its normal DC level is considered as one event (the scale is converted into dB by the relation dB = 20 log10 (enhancement)). Smaller and unmeasurable enhancements of less than 1 dB are not included in the analysis. The first and second kind of noise bursts, for example, were observed on nine and five occasions, respectively, in the month of March during simultaneous operations by the two antennas.

Figure 3.

Examples of noise bursts observed by borehole antenna (top), terrestrial antenna (middle), and both the antennas simultaneously (bottom).

[10] In the top panel of Figure 4, we show the occurrence number of such noise bursts recorded by borehole and terrestrial antennas in each month between March and September 1999 by open and dark histograms, respectively. Although the data in the month of March are available for 15 days only, they are included here for the reason that seismic swarm activity started during the later period of this month and before 15 March there were two cases of earthquakes, only one of which occurred far away from the Chamoli region outside India and the other occurred at a large depth of 105 km also far away from Chamoli. These earthquakes have not produced significant noise burst activity at Agra [Singh et al., 2000]. In the cases where the amplitude enhancements occurred in both the channels simultaneously corresponding to borehole and terrestrial antenna measurements, we measured the amplitude enhancements from the graph and deduced the ratio Eb/Et. Here, Eb and Et are the amplitude enhancements produced by borehole and terrestrial antennas, respectively. Then we counted in each month the number of cases in which Eb/Et was greater than one or less than one. In the middle panel of Figure 4, we show their occurrence number by open and dark histograms, respectively. In the bottom panel, we show the variation in the number of earthquakes considered in the present study over the period of observations.

Figure 4.

Occurrence number of noise bursts observed by borehole and terrestrial antennas separately in each month during observations (top), occurrence number of noise bursts with amplitude ratio Eb/Et > 1 and Eb/Et < 1 (middle), and variation in occurrence number of large magnitude earthquakes (Ms ≥ 4.5) during each month of observations.

[11] From a close examination of the variation in the occurrence number of noise bursts observed by borehole antenna shown in the top panel in relation to variation of earthquakes shown in the bottom panel we find that there is a positive correlation between the two. Further, the occurrence number of the noise bursts observed by borehole antenna is larger than that observed by terrestrial antenna in the month of March when seismic swarm activity started with devastating main shock of Ms = 6.6 followed by six aftershocks of high magnitudes (4.5 < M < 6.6) on the same day of 29 March 1999. The occurrence number increases in the month of April following the large number of aftershocks and then reduces in the month of May. Minor increase in the occurrence number in the month of June is due to increase in number of earthquakes in this month. Then it reduces following the reduction in number of earthquakes in the rest of the months. The domination of the occurrence number of noise bursts measured by the terrestrial antenna between July to September is well understood in view of enhanced local lightning and thunderstorm activities in these months.

[12] The coupling between the two antennas is shown in the middle figure in terms of variations of Eb/Et > 1 and Eb/Et < 1. Here, it may be noted that Eb and Et are measured independently by the two antennas and not that the same signal is coupled between them. The results are dominated mostly by Eb/Et < 1 which corresponds to enhancement in the number of noise bursts recorded by terrestrial antenna due to emission sources in the atmosphere except between May to July when atmospheric emissions are at lower level. Since the seismic activity in the month of September is very low, a large enhancement in Eb/Et > 1 in this month may be as a result of combined effect of seismicity and enhanced local lightning and thunderstorm activities. Further, the earthquakes of small magnitudes (Ms < 4.5) may also contribute to enhanced Eb/Et > 1 during all the months.

[13] In order to strengthen our result that there is a positive correlation between earthquakes and electromagnetic data shown in Figure 4 we have made some statistical analysis and tested the Null hypothesis. The results are summarized in Table 1.

Table 1. Statistical Results for Testing Null Hypothesis
Noise Burst ActivityCorrelation Coefficient; RTest Te Statistic Computed for Null Hypothesis; TProbability for R = o
Eb/Et > 10.3440.8160.449
Eb/Et < 10.5531.4000.218

[14] From the above results it is seen that probability that R = 0 for borehole data is 0.137. This is certainly higher than 0.05 (the level of significance) but not too bad for a geophysical problem. In other cases of the data the probability is much higher suggesting that there is no useful correlation with the earthquakes. From these results we find that seismic events of larger magnitude do induce electromagnetic signals in deep boreholes.

[15] In Figure 5 we examine the diurnal occurrence pattern of the first two kinds of data in three different cases including a seismic swarm activity (top), an isolated earthquake (middle), and seismically quiet period (bottom). For this purpose, we show the occurrence of the two kinds of data along with their amplitude enhancements ten days before to ten days after the main shock of 29 March 1999 during the seismic swarm activity. In the bottom two figures the data is shown for five days before to five days after. Here the open histograms show the occurrence days and the amplitude of the noise bursts recorded by the borehole antenna and hatched histograms show the same for the noise bursts recorded by the terrestrial antenna. The amplitude enhancements are shown in mA in order to include the enhancements as low as 1 mA. From this figure it is clearly seen that while the noise bursts recorded by the terrestrial antenna are independent of seismic activity there is a clear association of borehole data with seismic activity. During swarm activity of March/April such noise bursts occurred five days before the main shock of 29 March 1999 and continued even after. In the case of isolated earthquake of 28 July 1999 the borehole noise burst appeared a little less than one day before. In contrast to the above, there are no such data visible during quiet period around 19 September 1999. In the top figure, we show the two kinds of data appearing in some cases on the same day. This is possible because both kinds of data are collected independently from sources lying above and below the crust and their observations on the same day is quite likely. From these results it is clear that during seismic swarms electromagnetic precursors may appear much before than those during isolated earthquakes. Further, small magnitude earthquakes possibly do not generate strong signals to be picked up by the borehole antenna. This is very well seen from the figure at the bottom.

Figure 5.

Occurrence pattern of noise bursts of first and second kind within ±10 days from the main shock of 29 March 1999 during the swarm activity (top), within ±5 days from an isolated earthquake of 28 July 1999 (middle), and within ±5 days corresponding to nonseismic days between 14 and 24 September 1999 (bottom). Here blank and shaded histograms show the noise bursts recorded by borehole and terrestrial antennas, respectively.

[16] Since the results in Figure 4 suggest long distance propagation of electromagnetic signals in crustal region and possible leakage to atmosphere also, the question arises how such emissions can be propagated in view of high attenuation at such frequencies and skin effect. Further, since Chamoli is located about 400 km away from our observing station and epicenters of other earthquakes considered in this study are at large distances more than 1000 km in neighboring countries of Pakistan and Afghanistan, the question arises how such emissions can be propagated to such long distances to be observed at Agra. There are three possibilities in which this question can be examined. One possibility is that the signals generated at the source are propagated through seismic faults in a manner similar to that in Earth-ionosphere waveguide and reached our observing station at Agra. The possibility of propagation through seismic faults was suggested by Yoshino and Tomizawa [1988] and Kingsley [1989]. Incidentally, there exists a big seismic fault at southern boundary of Himalayan region between northeast India to most seismically active zones in Afghanistan. This fault has been created by four major earthquakes (Ms > 8) and many other large magnitude earthquakes that occurred during the last 100 years. This is known as main boundary fault (MBF) and is shown by thick curve in Figure 2. The locations of four major earthquakes (Ms > 8) and years of their occurrences are indicated by stars. There also exist transverse conducting channels near Delhi-Haridwar ridge (DHR) across the main boundary fault [Arora and Reddy, 1995]. The transverse conducting channels and Delhi-Haridwar ridge (DHR) are shown by thick and parallel lines, respectively. The observing station Agra (AGR) and Indian capital Delhi (DLH) are indicated by big solid circles. The main boundry fault and transverse conductive channels provide ideal conditions for the propagation of seismogenic emissions to our station at Agra, if the propagation through fault is taken for granted. The second possibility has been suggested by Tsarev and Sasaki [1999] following Keller [1989] in which the signals generated in the shallow region may be propagated to long distances through middle layer crust of low conductivity similar to that in a waveguide and on the way they may find some windows where rock conductivity can be much lower than on average in the upper layer through which they may appear at the Earth surface. Considering the conductivity of the middle layer to be of the order of 10−4–10−6 S/m, Tsarev and Sasaki [1999] have shown quantitatively that ELF emissions (ω = 102 Hz) can be propagated to a distance of 102–103 km which reduces to 101–102 km at ω = 1 kHz. The third possibility is that such a window existed near the site of earthquake itself through which emissions emerged at the Earth surface and then propagated to long distances through Earth-ionosphere waveguide without much attenuation.

[17] However, in all the three cases mentioned above emissions have to propagate certain distances in crustal region and one is required to show that they do not suffer much attenuation during this propagation. For this purpose, we make a quantitative analysis in which we calculate attenuation suffered by seismo-electromagnetic emissions in a suitable conductivity model.

[18] It is well known that conductivity of the upper layer of the Earth's crust (so-called skin layer) is very high of the order of 10−1–10−2 S/m and signals generated under ground cannot be propagated to the surface through this layer in normal conditions. Molchanov et al. [1995] have assumed a conductivity of the order of 10−4–10−5 S/m near the source while studying the propagation characteristics of ULF signals through the lithosphere, atmosphere, and ionosphere. As mentioned above, Tsarev and Sasaki [1999] have also considered similar conductivities for the middle layer of shallow crust in which most of devastating earthquakes occur. Here, keeping these results in view, we consider a conductivity of 10−4 S/m and calculate the attenuation at various frequencies between ULF-VLF range (1–104 Hz) and present the results in Figure 6. We find that the attenuation is very low in ULF range but increases steeply in VLF range. At 3 kHz, the attenuation is around 13 dB/km which is very high for signals to be propagated to long distances. Under these circumstances the first two possibilities in which the signals are expected to travel long distances in the crustal region do not seem to be tenable. Hence, the only possibility left is that the signals generated at the source are propagated upward towards the Earth surface. In the case of Chamoli earthquakes which occurred in shallow region mostly at a depth of 10 km, the signals at 3 kHz have to suffer an attenuation of 130 dB to come to the surface provided there is no skin layer. This attenuation is not much considering the amount of energy liberated at the source. Assuming that only 0.1% of the strain energy is liberated for 103 to 104 s before the main shock, Sumitomo [1994] has made a rough estimate of the liberated energy and found it to be ∼6 × 107–8 watts. However, the main problem is the skin layer of much higher conductivity. In order to overcome this problem some authors [Tsarev and Sasaki, 1999] have assumed “windows” of very low conductivities in the skin layer which may be produced by some special geophysical formations, one of them being the emergence of some basement rocks, through which the signals may be propagated to Earth surface. However, the problem of propagation of seismo-electromagnetic signals to Earth surface is still in question and it requires great attention. Once the emissions emerged at the Earth surface, they can be propagated to long distances in the Earth-ionosphere waveguide where attenuation at VLF is much low as compared to that in the Earth's crust.

Figure 6.

Attenuation of seismo-electromagnetic emissions computed at various frequencies in ULF-VLF range.

[19] The long distance precursory signals have been observed by other workers also who conducted observations at different frequencies and measured electric and magnetic fields of the seismogenic emissions. For example, Oike and Ogawa [1986] have recorded anomalous low frequency noises before and after the earthquakes at distances of about 1000 km. Warwick et al. [1982] have observed electromagnetic emissions at 18 MHz six days prior to the great Chilean earthquake of 22 May 1960 at distance of about 10,000 km from the epicenter. Yoshino et al. [1992] have considered seismo-electromagnetic emissions in frequency range from 36 Hz to 82 kHz generated from a number of locations lying between the vicinity of the observing station to about 1000 km in their statistical analysis to obtain a clear explanation of source mechanism. It has also been suggested that depending upon the conductivity and the characteristics of the rocks, signals generated due to mechanical distortion may be propagated from hundreds to thousands of kilometers away from the epicenter [Rikitake and Yamazaki, 1967].


[20] The authors are thankful to the Department of Science and Technology, Government of India, New Delhi, for financial support in the form of a major research project. Thanks are also due to the Director General, India Meteorological Department, New Delhi, for supplying the earthquake data. The valuable help extended by our research colleague Mr. Vinod Kumar Kushwah during preparation of the manuscript and figures is gratefully acknowledged.