Radio Science

Reception of ELF transmitter signals at Moshiri, Japan, and their propagation characteristics



[1] Measurements are described of 82 Hz radio signals arriving from the Russian ELF transmitter located on the Kola Peninsula, Russia. We used two orthogonal calibrated horizontal magnetic field sensors and a vertical electric antenna at the Moshiri observatory, Hokkaido (geographic coordinates; 44.4°N, 142.2°E). Several propagation characteristics were studied: (1) signal amplitude and its variations on diurnal and seasonal scales, (2) phase difference between the two horizontal magnetic field components (wave polarization), and (3) wave arrival angle. The amplitude detected was compared with early published data, showing a good agreement. We estimated the source current moment of the transmitter by comparing our experimental amplitude with theoretical computations. This 82 Hz signal field is found to be linearly polarized, which allowed for the goniometric finding of source bearing. The arrival azimuth is consistent with the geometry of the experiment.

1. Introduction

[2] ELF (extremely low frequency) radio propagation in the Earth-ionosphere cavity can be studied either by using natural signals (continuous Schumann resonances signal or ELF transients) [Nickolaenko and Hayakawa, 2002] or by ELF transmissions [Fraser-Smith and Bannister, 1998]. Natural ELF waves, in particular those producing Schumann resonances, originate from pulsed sources and cover a wide frequency band. Their propagation characteristics depend on two factors: the distribution of lightning strokes in the world and properties of the Earth-ionosphere cavity. Sometimes it is difficult to distinguish between these two factors in analyzing the data. An advantage of artificial signals (although narrow-banded) is the known propagation path and transmitted power, which is extremely useful when estimating the subionospheric propagation parameters.

[3] This paper deals with the latter artificial ELF radio signals. Only two ELF transmitters are set in the world. The first one is the United States WTF/MTF (Wisconsin Transmitter Facility/Michigan Transmitter Facility) having two crossed transmitting antenna, and it operates at the center frequency of 76 Hz [Fraser-Smith and Bannister, 1998]. The second one is the Russian Kola Peninsula Transmitter (KPTF), working at 82 Hz [Belyaev et al., 2002; Fraser-Smith and Bannister, 1998]. We analyze the signals from this Russian KPTF transmitter.

[4] Fraser-Smith and Bannister [1998] examined signals from the Russian transmitter detected at a number of sites belonging to Stanford University, distributed around the globe. Unfortunately, the time interval of their analysis was relatively short. We treat the KPTF signal monitored during 1 year. The center of antenna of the KPTF is located at geographic coordinates 68.8°N and 34.5°E as seen in Figure 1. As described by Velikhov et al. [1996], the transmitter consists of two swept-frequency generators of sinusoidal voltage and two parallel horizontal grounded antennas, each about 60 km long. The generator provides 200–300 A currents in the antennas in the frequency range from 20 to 250 Hz.

Figure 1.

Relative location of the Russian ELF transmitter (KPTF) and Moshiri observatory.

[5] The ELF equipment was placed at the Moshiri field site to measure particular radio waves at the frequency of 82 Hz. We processed the long-term data covering the whole 2007 year. The long-term records enabled us to study the diurnal and seasonal variations (there are normally two transmissions per day). We address below the following issues: (1) signal amplitude, (2) wave polarization, and (3) wave arrival azimuth. The measured amplitude is compared with data published by Fraser-Smith and Bannister [1998]. Effects are discussed of ionospheric day/night asymmetry and seasonal changes. The linear polarization of the man-made radio wave is demonstrated, and the corresponding arrival azimuth is found by using the goniometer technique. These parameters have not been addressed in the literature before, and we present novel information.

2. Measurements of ELF Signals at Moshiri, Japan

[6] The ELF radio waves are monitored at Moshiri, Hokkaido (geographic coordinates; 44.4°N, 142.2°E) since 1996 [Hobara et al., 2000]. The ELF measurement system was upgraded in 2005, and we mention only major parameters of the new system. Three electromagnetic field components (two horizontal magnetic fields BNS and BEW and the vertical electric field EZ) are measured by induction coil magnetometers and a capacitor (“ball”) antenna, respectively. The whole system was calibrated before installation, so that we can find the absolute amplitude of ELF waves. The sampling frequency is 4 kHz, and the ELF waveforms are continuously recorded on the hard disk of the PC. Further details of ELF equipment are given by Ando et al. [2005].

3. Analysis of Data

[7] The ELF antenna is often referred to as “Zeus,” and two horizontal wires are extended parallel with each other as the transmitting antenna [Belyaev et al., 2002], so that ELF radiation is directed eastward and westward. The Moshiri receiving point is located at the distance 6.13 Mm (1 Mm = 1000 km), and Figure 1 depicts in polar coordinates the relative position of KPTF transmitter, and the observatory. The great circle distance between the transmitter and receiver is schematically shown together with the wave arrival angle of 24.8° from north to west, which is in fact the source bearing seen from the observatory.

[8] We have used the Fourier analysis for the signal processing. The data length in the FFT procedure was 65.536 s. The relevant frequency resolution was 0.01526 Hz. The amplitude spectra of each field component were computed every 60 s, and thus we obtained dynamic spectra with temporal resolution of 1 min.

[9] Figure 2 illustrates a dynamic spectrum of the total magnetic field, equation image on a particular day of 8 January 2007. The abscissa of Figure 2 shows the universal time (UT) in hours, and the ordinate depicts the signal frequency (Hz) in the interval from 79 to 45 Hz. The amplitude is shown by black inking. One can identify the presence of a monochromatic ELF radio signal from KPTF transmitter at 82 Hz in the time intervals 0930–1030 h UT (the local time in Japan is 1830–1930) and 1730–1900 h UT (LT is 0230–0400). ELF signals are marked by an arrow on the right side of Figure 2. After deriving the amplitude spectra ∣HNS(f)∣ and ∣HEW(f)∣, we compute the amplitude spectrum ∣Hequation image(f)∣ ≡ equation image of the total horizontal magnetic field. Concurrently, the phase difference is found between the two horizontal magnetic fields Δ = θNS(f) − θEW(f) where θNS = arg[HNS(f)] and θEW = arg[HEW(f)] are correspondingly the arguments of the complex spectra HNS(f) and HEW(f). The phase difference Δ determines the ELF wave polarization.

Figure 2.

An example of ELF dynamic spectrum (total horizontal magnetic field) on a particular day of 8 January 2007. The abscissa shows the UT time in hours, and the ordinate is frequency in Hz. The signal from KPTF is clearly seen at 82 Hz (indicated by an arrow on the right).

[10] Figure 3 depicts the signal characteristics detected on 8 January 2007. Figure 3 (bottom) shows diurnal variations of the signal amplitude that reaches 0.2 μA/m. Intervals of artificial signal reception are marked by boxes in Figure 3. One may note that the maximal signal-to-noise ratio is about 5.

Figure 3.

Propagation characteristics of transmitter signals on 8 January 2007. (top) The direction of wave magnetic field (ϕH), (middle) the phase difference between the two horizontal magnetic fields (θEW − θNS), and (bottom) the spectra of total magnetic amplitude Hϕ = equation image. Time intervals in the boxes correspond to KPTF transmissions. We observe only noises outside the boxes.

[11] Points in Figure 3 (middle) show the dynamic spectrum of phase difference Δ. When the noise is usually detected at the frequency of 82 Hz, the phase difference is scattered in the interval from −180 to +180 degrees. When the KPTF signal appears, all points of Figure 3 (middle) fall down to near zero, thus indicating the linear polarization of radio waves. Therefore, we can use the goniometer technique [Hayakawa and Ohta, 2006] for establishing the source bearing.

[12] Figure 3 (top) presents the orientation of total magnetic field vector HϕH). The angle ϕH ≈ 65° found from the orthogonal components HNS and HNS is connected with the wave arrival angle ϕk by the relation ϕk = ϕH ± 90°. Hence, direction to the source is ϕk ≈ −25° (in the definition of positive for north to east), which is very close to the purely geometrical estimate. One may observe that the arrival angle usually fluctuates owing to the presence of natural ELF radiation from global thunderstorms. When we detect the man-made signal, the arrival azimuth becomes rather stable, as seen by the boxes in Figure 3.

[13] Figure 4 demonstrates the effect of temporal data length used in particular DFT (discrete Fourier transform) procedure. The ELF transmissions were processed from the “morning” time interval, i.e., from 0830 to 1030 UT, on 8 January 2007. Figure 4 (left) corresponds to the data fragments 100 s long. Figure 4 (right) presents the results obtained with DFT elements 1500 s long. Figure 4 (bottom) depicts the amplitude of 82 Hz radio signal detected in the total magnetic field vector. Figure 4 (middle) demonstrates the phase difference between horizontal magnetic field components, and Figure 4 (top) shows the wave arrival angle (ϕH). Individual spectral estimates were obtained with the time step of 1 min, while the length of data processed by DFT was 100 or 1500 s. As one may observe, the longer fragments provide more stable results. This fact follows from the properties of Fourier transform: a longer duration of the time domain realization gives a higher-frequency resolution, and hence, the higher signal-to-noise ratio, provided that the background noise is distributed over the frequencies. Concurrently, a “reduction of the filter bandwidth” results in a slower time domain reaction. This is why Figure 4 (right, especially the signal amplitude) is stable and has slow increase/decrease patterns.

Figure 4.

Characteristics of 82 Hz signal radiated in the “morning” time interval (0830–1030 UT) on 8 January 2007. The data length varies from (left) T = 100 s to (right) T = 1500 s. (top) The source bearing, (middle) the phase difference, and (bottom) the amplitude of the magnetic field.

[14] Similar analysis was applied to ELF records collected during 1 year from January to December 2007. We have found that similar to Figure 2, the man-made ELF radio signals were observed in two UT intervals: 0800–1100 and 1700–2000 h. The duration of ELF transmission was normally about 1 h, but an exact time of transmission onset varied in a range from 8 to 11 h and in 17–20 h intervals.

[15] Figure 5 compares the time of KPTF transmission (experimental data are shown as bluish composite bars) with the computed sunrise-sunset times at the both ends of propagation path. Sideway “jumps” in these bars probably correspond to the transition from wintertime to summertime in Russia. Moments of sunrise and sunset were found for the ground surface, not for the ionosphere [see Nickolaenko and Hayakawa, 2002]. It is seen from Figure 5 that the whole propagation path is in the night hemisphere for December, during both transmission intervals. The “evening” ∼19 h UT transmissions correspond to the night propagation path from January to March. Simultaneously, the “morning” ∼10 h UT transmissions are associated with the day-to-night propagation during the same months. Transmissions correspond to daytime propagation from May to August.

Figure 5.

UT times of ELF transmissions and the seasonal variations of sunrise and sunset times at the transmitter (KPTF) and observatory (Moshiri).

[16] Figure 6 shows seasonal variations of the amplitude at the two UT times, ∣Hϕ (0800–1100 h)∣ (Figure 6, top) and ∣Hϕ (1700–2000 h)∣ (Figure 6, bottom) measured in dB. Of course, the processing suggests that the power transmitted remained constant. There are gaps in Figure 6; for example, no data were received from September to the end of November. The breaks probably indicate the periods when the KPTF transmitter did not work. This fact is clearly seen in Figure 5. The amplitude shown in Figure 5 is rather unstable, and its overall average value is 0.29 dB indicated by the horizontal line. It is interesting to compare the amplitude between day and night, to be compared with our later theoretical analysis. As already mentioned before, the paths in December for both UT times are in full night conditions, while the interval of UT = 0800–1100 h (Figure 6, top) in May–August is found to be in the daylight condition. So we try to compare the amplitudes between the daylight and nighttime propagation paths. It seems that there is no significant difference in amplitude between these day and night propagation conditions.

Figure 6.

Seasonal variation of the amplitudes at the two UT intervals: (top) ∣Hϕ (0800–1100 h)∣ and (bottom) ∣Hϕ (1700–2000 h)∣ measured in dB. Data gaps correspond to silent transmitter. The average value is indicated by a horizontal dotted line in each plot.

[17] Let us return to the signal amplitude detected. The 82 Hz signal amplitude depicted in Figures 2 and 3 had the average value of 2.0 × 10−7 A/m or −134 dBA/m. Model computations (see below) with an account for experiment geometry and antenna angular pattern predict the signal at Moshiri of −137 dBA/m level.

[18] We plot our experimental value in Figure 7 as a black star at the distance of 6.13 Mm (Moshiri), together with the published values of amplitude at different distances from KPTF [Fraser-Smith and Bannister, 1998]. The legend for notations is given in the caption of Figure 7. Points in Figure 7 show that amplitude variations were detected at definite distances: the daytime values deviate from those in the night in the work of Fraser-Smith and Bannister [1998].

Figure 7.

Wave amplitude versus distance computed for the “whole day” (red curve) and the “whole night” (blue curve) models. Amplitude at Moshiri is shown by a black star. The results of Fraser-Smith and Bannister [1998] are plotted as well. SS, Sondrestromfjord, Greenland; DU, Dunedin, New Zealand; AH, Arrival Heights, Antarctica; CO, Connecticut, USA; HW, Hawaii; KB, King's Bay, Georgia, USA. Letters “d” and “n” denote day or night models, respectively.

[19] We plot two model curves in Figure 7. Computations were based on the uniform cavity model [Nickolaenko and Hayakawa, 2002] with purely daytime and purely nighttime ionosphere models. The blue curve corresponds to the “whole nighttime” model [Nickolaenko, 2008], i.e., when the global ionosphere parameters correspond to the night [Bannister, 1999]. The red curve corresponds to the “whole day” ionosphere: the globally uniform ionosphere with the daytime profile. The following formulas are used in the computations of ELF fields from a horizontal electric dipole antenna grounded at its ends (or equivalent horizontal magnetic dipole) [Nickolaenko and Hayakawa, 2002; Nickolaenko, 2008]:

equation image
equation image
equation image

Here, a is the Earth's radius, M(ω) is the source current moment of the loop antenna in A/m2 which is the product of the current and loop effective area, the factors of cosη and sinη account for the radiating antenna angular pattern, the angle η is counted from the direction to the east at the antenna center; ɛ0 is the dielectric constant of the vacuum, θ is the angular distance from the antenna center to the observer, Pν[cos(π − θ)] is the Legendre function, the simple prime indicates the first-order derivative with θ and the double prime indicates the second-order derivative, ν(f) is the complex propagation constant of ELF radio wave, and hE and hM are correspondingly the “electric” and “magnetic” ionosphere heights [see, e.g., Bannister, 1999].

[20] We accept that the source current moment is M(ω) = 1.0 × 1011 A/m2 as in the work of Fraser-Smith and Bannister [1998]. The daytime ionosphere heights were hED = 55.1 km and hMD = 75.6 km. And the nighttime heights were hEN = 75.6 km and hMN = 91.5 km. These values are taken from Bannister [1999], though they are smaller than those by Füllekrug et al. [2002]. The propagation constant for the particular model was found from the “standard” equation.

equation image

Here the wave attenuation was αD = 1.33 dB/Mm for the day and αN = 0.82 dB/Mm for the night, and the wave phase velocity (V) was characterized by (c/V)D = 1.245 for the day and (c/V)N = 1.12 for the night.

[21] The values computed appeared to be 3 dB lower than the measured amplitudes. In order to fit with the experimental value on amplitudes, we had to enhance the source current moment up to M = 1.42 × 1011 A/m2. Thus the effective source moment was found to be corresponding to measurements at Moshiri in 2007.

[22] Figures 2 and 3 indicate that during the ELF transmissions the phase difference between the two horizontal magnetic field components are concentrated around 0°. Hence, the artificial radio signal is found to have the linear polarization, so that we can apply the goniometric direction finding technique based on the amplitudes of orthogonal magnetic fields [Hayakawa et al., 1995; Hayakawa and Ohta, 2006]. In particular, we used the following equation for the wave arrival angle β:

equation image

Our measurements showed that β = 25° (see Figure 3, top), which is very close to geometrical considerations.

4. Discussion

[23] Our estimate for the source current moment is found to deviate from the previously published data. By using the distance dependence of signal amplitude acquired at various positions and then performing a fitting with the simple expression by Bannister [1999], Fraser-Smith and Bannister [1998] found that the current moment of the KPTF antenna was equal to M ≅ 200(A) × 55(km) × 10(km) = 1.1 × 105 (A · km2). However, our results indicate that the source moment must be M ≅ 1.42 × 105 (A · km2). This estimate is about 2 dB higher than that found by Fraser-Smith and Bannister [1998]. This deviation might be caused by different reasons. First, the time of measurements was very different: the 1990s in the work of Fraser-Smith and Bannister [1998] versus 2007 in our analysis, and the power of transmitter could be enhanced in 17 years. Second, we used more exact computations. In particular, the zonal harmonic series representation with accelerated convergence was applied for the Legendre functions, and we have used also the most advanced model of propagation parameters that incorporated all ELF data [Bannister, 1999]. All of these features might play its role; however, we cannot determine the particular cause of deviations at the moment.

[24] An important feature of Figure 7 is that in our model the “day” and “night” curves meet around a critical distance of ∼6 Mm. Fraser-Smith and Bannister [1998] found a ∼3 Mm critical distance. The physical explanation of why a critical distance exists is rather simple. The wave attenuation is higher in the day propagation, and relevant amplitude decreases with distance faster than in the ambient night conditions. Simultaneously, the excitation factor of a source is greater in ambient day conditions. Hence, the rapidly decreasing “day” curve starts from a higher initial amplitude, and therefore it meets the nighttime curve at some critical distance. Our computations show that this crossover occurs around 6 Mm (apart from oscillations caused by interference of direct and antipodal waves at larger distances). This theoretical prediction on the absence of day/night propagation amplitude at the distance of Moshiri, seems to be supported by our previous experimental property as seen in Figures 5 and 6.

[25] The position of Moshiri observatory corresponds to a minor impact of the ionosphere day/night asymmetry on the measured amplitude. The absence of regular daily variations will facilitate detection of localized disturbances associated, say, with seismic activity.

5. Conclusion

[26] The 82 Hz CW transmission was successfully received in two time intervals per day and an analysis was performed of the 1 year data recorded at Moshiri, Hokkaido, Japan. The source of radio waves was the Russian KPTF transmitter located at the Kola Peninsula, Russia 6 Mm away from the observatory.

[27] Detailed spectral analysis of the 1 year data showed that the amplitude in two time intervals was nearly the same showing no daily and seasonal variations. The field amplitude was −132.4 dBA/m. We have found that the source moment was M = 1.42 × 1011 A/m2, which is about 2 dB higher than that estimated earlier by Fraser-Smith and Bannister [1998].

[28] Observation data also indicate that the artificial signal from the KPTF is linearly polarized: the phase difference between the two orthogonal magnetic field components was close to 0°. This linear polarization enabled us to perform the goniometric direction finding technique. The wave arrival angle for the 82 Hz radio signal was 25° from north to west, which is in excellent agreement with the experiment geometry (24.8° from north to west).


[29] The authors thank E. Titova from Polar Geophysical Institute, Russia, for her useful discussions. The present work was made under the financial support of NiCT in the framework of a research and development promotion scheme funding international joint research.