Earth Quake

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 6, NO. 1, JANUARY 2009

107

Experiment and Theoretical Study of Earthquake Detection Capability by Means of Microwave Passive Sensors on a Satellite Tadashi Takano, Fellow, IEEE, and Takashi Maeda, Member, IEEE

Abstract—This letter presents the possibility of detecting earthquakes (EQs) from microwaves emitted when rock fractures. The method is based on an experiment in which microwave emission was detected from rock fracturing in a laboratory for the first time in the world. First, the method of calibrating emitted microwave power from experimental data is presented. A model of microwave emission and propagation to a satellite is then proposed. An advantage of microwaves is that they penetrate the Earth’s ionosphere, unlike radiowaves of frequencies lower than several tens of megahertz. The power received by a satelliteborne receiver is estimated by assuming parameters of a radiometer currently operating in orbit. The result indicates that a satelliteborne receiver can detect microwave signals generated by an EQ. Based on this result, we attempted to detect some features associated with an actual EQ from the data of the Advanced Microwave Scanning Radiometer for Earth Observation System aboard the remote sensing satellite Aqua. Index Terms—Earthquake (EQ) detection, emitted microwave power, remote sensing satellite, rock-fracture experiment, signalto-noise (S/N ) ratio.

I. I NTRODUCTION

E

ARTHQUAKE (EQ) detection, whether preseismic (before EQ) or coseismic (simultaneously with EQ), is important for social well being. Much effort has been devoted to finding an effective means of EQ detection. Mechanical motion sensors, an electric sensor to measure the change of DC ground potential [1], very low frequency (several hertz to several kilohertz) detectors [2], and FM radio receivers were proposed for ground use to find anomalies of the ionosphere via radio wave propagation [3]. Additionally, synthetic aperture radars (SARs) may be applied to measure slight movements of the ground through interferograms. However, some means lack deterministic certainty, and even the signal power cannot be calibrated in them. Others are impractical at the current technology level. We were the first to find that microwave energy was emitted during rock fractures in a laboratory [4]. A detected signal is

Manuscript received July 6, 2008; revised August 8, 2008. First published December 12, 2008; current version published January 14, 2009. T. Takano is with the Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Sagamihara 229-8510, Japan (e-mail: [email protected]). T. Maeda is with the Earth Observation Research Center, Japan Aerospace Exploration Agency, Tsukuba 305-8505, Japan (e-mail: maeda.takashi@ jaxa.jp). Digital Object Identifier 10.1109/LGRS.2008.2005735

a sequence of pulses that include microwaves in the selected frequency bands of 300 MHz, 2 GHz, and 22 GHz. The features are the same as that of the signals in the case of a hypervelocity impact with several materials [5]. This fact suggested another means to detect an EQ, one associated with rock fractures or rock slips. The microwave energy associated with an EQ may be detected in several ways. A microwave receiver aboard a satellite in a low-Earth orbit is one candidate and has the following advantages. 1) Microwave energy is not blocked by the Earth’s ionosphere. 2) An antenna beam from a satellite can scan the Earth’s surface and obtain global data. 3) To investigate a detection method, we can first select a favorable EQ from the viewpoint of microwave detection and then perform a concentrated data analysis on it. A satellite system, however, has the disadvantage of a long propagation distance, resulting in a weak received signal. The weak signal should be distinguished from receiver and thermal noise emitted from the ground. Additionally, a large ambiguity exists in the underground propagation of microwaves due to unknown effects. This letter presents the result of calibrating the microwave signal received in the rock-fracture experiment. A satellite system is then proposed with a model of microwave emission and underground propagation. The received signal power will be calculated by assuming the parameters of the remote sensing satellite Aqua and the Advanced Microwave Scanning Radiometer for Earth Observation System (AMSR-E) carried aboard it to demonstrate the possibility of detecting an EQ. Based on the investigation result, we attempted to detect some features associated with an actual EQ from the data of AMSR-E. This letter finally presents the initial analysis result of the data of AMSR-E.

II. R OCK -F RACTURE E XPERIMENT Fig. 1 shows the waveforms detected in a rock-fracture experiment at an intermediate frequency of each microwave band [4]. In this case, the rock sample was quartzite with a volume of 3.4 × 10−5 m3 . The microwave signal in the 22-GHz band appears above the noise level for less than 0.1 ms after the rock fracture and is composed of many pulses in an envelope.

1545-598X/$25.00 © 2008 IEEE

Authorized licensed use limited to: BNM Institute of Technology. Downloaded on February 4, 2009 at 05:22 from IEEE Xplore. Restrictions apply.

108

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 6, NO. 1, JANUARY 2009

Fig. 1. Microwave signals obtained in a rock-fracture experiment (quartzite). Waveforms are composed of many pulses whose durations are at most several tens of nanoseconds. The entire emission time differs significantly with the kind of rock and frequency. (a) 2 GHz. (b) 22 GHz.

The signal in the 2-GHz band is also composed of many pulses, but lasts for a total measuring time of 4 ms. During the calibration, the receiving system is fed a microwave continuous wave (CW) of known power levels. The output voltage of the receiver is then calibrated with the input power level. The amplitude of the measured ith pulse is replaced with the calibrated CW power level (Pi ) with the specific period (τi ). The received power (Pr ), which is averaged over time (τave ), is expressed by n Pi τi Pr = i=1 (1) τave where n is the number of pulses in τave . In Fig. 1(a), τi values are 6–10 ns. τave is taken as 3 ms, so that Pr = 1.1 × 10−3 pW. Pr is a part of the total emitted power from the crushed sample (Pexp ) and is expressed by the Friis equation as follows: Pr = Pexp Lf Gr .

(2)

Here, Lf is the free-space loss represented by Lf = (λ/4πl)2 , where λ is the microwave signal wavelength, l is the distance from the wave source to the receiving antenna, and Gr is the gain of the receiving antenna. In the experiment, l = 0.48 m and Gr = 126; therefore, Pexp is calibrated to be 32 pW at 2 GHz and 1.6 pW at 22 GHz. In addition to quartzite, rock samples of gabbro, granite, and basalt were tested, and signals were detected at 2 GHz but not detected at 22 GHz. The power level emitted from gabbro was high, almost the same as that from quartzite. This fact indicates that the emitted power level does not depend on piezoelectricity alone. III. S ATELLITE S YSTEM FOR EQ D ETECTION It is inferred that rock fractures or slips occur around the same time as an EQ. Therefore, we can conceive a satellite system for detecting microwaves associated with an EQ, as shown in Fig. 2. The microwave signal emitted by rock fractures

Fig. 2. Satellite system for EQ detection. a: Radius of crushed-rock sphere. d: Distance from the hypocenter to the epicenter of an EQ. h: Distance from the epicenter to a satellite in orbit. PEQ : Microwave signal power generated by rock fractures in an EQ. LG : Propagation loss under the ground. LF : Freespace loss. PREQ : Power associated with crushing rock in an EQ received by a satelliteborne receiver. The signal path from the hypocenter to the satellite is included.

in an EQ (PEQ ) may be extrapolated from the experimental data of Pexp as PEQ = Pexp ×

V v

(3)

where v (V ) is the volume of the rock sample fractured in the experiment (area where rocks are crushed due to an EQ). Since the area where rocks are crushed is assumed to be a sphere whose radius is a, V = 4πa3 /3. Microwave energy generated by rock fractures propagates under the ground to the ground surface and then through space to a satellite in orbit. The propagation loss (LG ) includes ohmic and scattering losses in the ground and the reflection loss at the ground surface, which are all significantly affected by the ground conditions. LG may be separated from the free-space loss (LF ). By another Friis equation, the microwave signal power received by a satelliteborne receiver (PREQ ) is expressed by PREQ = PEQ LG LF GR

(4)

where GR is the gain of the satelliteborne receiver’s antenna. We will now estimate the detection capability of the proposed satellite system. The ground is assumed to be a pure rock of quartzite. The electrical characteristics of quartzite are the most important factors in the system and are assumed to be those of granite [6], [7]. However, the relative permittivity (r ) and loss tangent (tan δ) differ widely in the literature. Daniels [6] states that r is 5, and the attenuation constant is from 0.5 to 5 dB/m at 100 MHz for granite. Therefore, the electric conductivity (σ) varies from 10−6 to 10−2 S/m. In contrast, Ulaby et al. [7] state that r and tan δ at 450 MHz and 35 GHz, respectively. σ is calculated to be much larger by these r and tan δ values. Consequently, it is, at least, clear that σ depends heavily on water content, composition, and frequency. Here, we assume that r = 5 and σ = 10−6 S/m. Satellite parameters such as altitude above ground, receiver antenna gain, and receiver noise figure are taken from those of Aqua and AMSR-E [8].

Authorized licensed use limited to: BNM Institute of Technology. Downloaded on February 4, 2009 at 05:22 from IEEE Xplore. Restrictions apply.

TAKANO AND MAEDA: EXPERIMENT AND THEORETICAL STUDY OF EARTHQUAKE DETECTION CAPABILITY

109

TABLE I ESTIMATION OF THE S/N OF THE EQ-GENERATED MICROWAVE SIGNAL

Fig. 3. Epicenter of the target EQ supposed by IAG (35.142◦ N, 3.997◦ W) and the terrain feature around it. The circle at the origin represents the epicenter, and black lines represent faults and fissures.

Table I lists the estimation results. Based on the experiment result at 22 GHz, PEQ is estimated to be 18.7 GHz to fit with the characteristics of AMSR-E. Therefore, Pexp at 18.7 GHz is assumed to be the same as Pexp at 22 GHz, and the receiver’s bandwidth is assumed to be 200 MHz (AMSR-E’s bandwidth at 18.7 GHz), while that of the receiver used in the experiment is 500 MHz. PEQ , LG , and LF are derived in the Appendix. The hypocenter depth is assumed to 10 km. Table I indicates that we can distinguish microwave signals generated by rock fractures from receiver noise with a signal-to-noise (S/N ) ratio of 8.8 when rock is crushed in a spherical volume with a = 1000 m. The microwave signal level received by a satelliteborne radiometer (PR ) is obtained by averaging brightness temperatures in an instantaneous field of view (IFOV) (Tb ) as follows: PR = kTb B.

(5)

Here, k is Boltzmann constant, and B is the receiver bandwidth. Let Tb0  and PR0 denote Tb  and PR when an EQ is not considered. At this time, Tb  is almost dominated by the average land surface temperature (LST) of the IFOV (Tg ); therefore, Tb0   Tg . Therefore, the following is obtained from (5): PR0 = kTg B.

(6)

When an EQ is considered, PR is the superposition of PR0 and PREQ . If Tb , at this time, is denoted by TbEQ , the following is obtained from (5): PR0 + PREQ = kTbEQ B.

(7)

Therefore, taking the ratio of (6) and (7), we obtain PR0 + PREQ TbEQ  . = PR0 Tg 

(8)

For Tg  = 290 K and B = 200 MHz, we obtain PR0 = 8.0 × 10−13 W. As PREQ = 3.7 × 10−12 W, we obtain TbEQ  = 1630 K, assuming that T0  = 290 K for simplicity. TbEQ  is large enough to be recognized in the distribution of Tb  from the background of the LST level from this perspective as well.

IV. EQ D ETECTION F ROM AMSR-E’ S D ATA A. Overview Based on the investigation result described in Section III, we attempted to detect features associated with an actual EQ from AMSR-E’s brightness temperature data of vertically and horizontally polarized signals at 18.7 GHz (TV and TH ). Since Aqua (with AMSR-E aboard) is traveling in a sun-synchronous subrecurrent orbit, all points on the Earth’s surface are observed at least twice a day, at local daytime (nighttime) when Aqua is ascending (descending). We used only the data of TV and TH obtained on descending tracks in order to evaluate microwave emissions from the land surface without the influence of reflecting sunlight. We focused on an EQ that occurred at 02:27:46 on February 24, 2004 (UT) in Morocco. According to an International Seismological Centre On-line Bulletin (http://www. isc.ac.uk/), the location, depth, and magnitude of the seismic center were supposed 35.142◦ N, 3.997◦ W, 14.0 km, and 6.3, respectively, by the Instituto Andaluz de Geofisica (IAG, Spain). Fig. 3 shows the epicenter supposed by IAG and the terrain feature around it. According to [9], the area in the south side of the epicenter is composed of quartzite. As described in Section II, a microwave emission at 22 GHz was experimentally confirmed in connection with quartzite’s fracture. Additionally, it was confirmed by interferograms formed by the data of Advanced SAR aboard the European Space Agency’s satellite Envisat that the terrain around the epicenter was displaced in association with the EQ [10]. In Section III, we assumed that a microwave signal associated with rock fractures is emitted from a seismic center under the ground. However, a deformation of the terrain on the land surface is likely to be accompanied by rock fractures. Microwave energy generated by rock fractures near the ground surface is more likely to be detected by AMSR-E. B. Method We first extracted the data of TV and TH obtained on descending tracks from June 1, 2002 (observation start) to December 31, 2007 and clipped out the scene of the area shown in Fig. 3 for each observation. Considering the sampling interval of AMSR-E (9 × 10 km) [8], each scene was restructured

Authorized licensed use limited to: BNM Institute of Technology. Downloaded on February 4, 2009 at 05:22 from IEEE Xplore. Restrictions apply.

110

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 6, NO. 1, JANUARY 2009

at 0.05◦ ( 5 km) intervals by a nearest neighbor algorithm. Since it is natural to assume that a microwave signal due to rock fractures is weakly dependent on polarization, we sought an area where TV and TH increased at the same time around the epicenter in the period of the EQ. If such an area was found, we then obtained time variations of TV and TH from June 1, 2002 (observation start) to December 31, 2007 with respect to any point in the area and verified whether the increases of TV and TH were particular to the period of the EQ. However, since each scene was restructured by a nearest neighbor algorithm, there was a constraint when a time variation of TV (TH ) was obtained. Since AMSR-E scans the Earth’s surface conically, each scanning line draws an arc, and observation points are located on it. According to an observation, the position on the scanning line of the epicenter and, eventually, the distribution of observation points around the epicenter vary. Therefore, strict locations of observation points regarded as the same by the nearest neighbor algorithm are distributed within the range of the sampling interval of AMSR-E. Considering that the IFOV of AMSR-E at 18.7 GHz is 16 × 27 km centered on an observation point [8], TV (TH ) values obtained from observation points whose IFOVs do not overlap can be treated as those at the same point. Such TV (TH ) values normally cannot be compared as the data at the same point. However, since Aqua is traveling in a sun-synchronous subrecurrent orbit, if the track number is the same, observation points are similarly distributed. Consequently, a time variation of TV (TH ) consisted of only the data obtained in observations with the same track number. With this constraint, the resolution of a time variation of TV (TH ) became 16 days which is the recurrent period of Aqua [8]. The time variation of TV (TH ) at each point had an inherent shape according to its location, but generally became high in summer and low in winter. Therefore, we calculated a difference of TV and TH (ΔTV and ΔTH ) at two points with as short a distance as possible and attempted to detect a short-span increase of TV (TH ). Considering that the nearest neighbor algorithm is applied to each scene, the distance between these two points (d) was set to 0.1◦ ( 10 km) since d should be more than the sampling interval of AMSR-E in order to always calculate ΔTV and ΔTH . Additionally, the following value S was introduced in order to more simply determine that ΔTV and ΔTH increased at the same time:  ΔTV 2 + ΔTH 2 , (ΔTV > 0, ΔTH > 0) (9) S≡ 0, (otherwise). S becomes positive only when both ΔTV and ΔTH are positive at the same time. Eventually, we evaluated a time variation of S from June 1, 2002 (the observation start) to December 31, 2007 with respect to a combination of two points that we focused on and verified whether S values became maximum in the period of the EQ. C. Results We could find an area where TV and TH increased simultaneously around the epicenter on February 22, 2004 (two days

Fig. 4. Distributions of TV and TH around the epicenter on February 22, 2004 (two days before the EQ).

Fig. 5. Time variation of S for the combination of ϕ and ρ from June 1, 2002 (the observation start) to December 31, 2007.

before the EQ). Fig. 4 shows the distributions of TV and TH on February 22, 2004. The origin of each figure represents the epicenter. In particular, we focused on a point ϕ indicated by a diamond in Fig. 4. ϕ is closest to the epicenter of all points where TV and TH increased simultaneously in this observation and corresponds to meshes formed of columns F to G and row g in Fig. 3. We then defined a point ρ 0.1◦ south of ϕ represented by a triangle in Fig. 4. The coordinates of ϕ (ρ) are 35.1◦ N, 4.0◦ W (35.0◦ N, 4.0◦ W). We calculated ΔTV and ΔTH by subtracting the TV and TH at ρ from those at ϕ, and finally obtained a time variation of S with respect to the combination of ϕ and ρ. Fig. 5 shows the time variation of S with respect to the combination of ϕ and ρ from June 1, 2002 (the observation start) to December 31, 2007. The track number of the observation on February 22, 2004 is 205, and by the aforementioned constraint of the nearest neighbor algorithm, this time variation of S consists of only 124 data points obtained in observations whose track number is 205. According to Fig. 5, S becomes maximum on February 22, 2004. Therefore, we concluded that the increases of TV and TH at ϕ for those at ρ were particular to the two days before the EQ, at least in observations with a track number of 205. From Fig. 1, it seems difficult to consider that these increases of TV and TH at ϕ on February 22, 2004 were generated by rock fractures in association with the EQ since microwave energy was observed after rock was crushed but not before. However, the latest seismological studies indicate that a preseismic slip of faults which is called preslip is likely to induce a main shock [11]. A slowly occurring preslip may not be

Authorized licensed use limited to: BNM Institute of Technology. Downloaded on February 4, 2009 at 05:22 from IEEE Xplore. Restrictions apply.

TAKANO AND MAEDA: EXPERIMENT AND THEORETICAL STUDY OF EARTHQUAKE DETECTION CAPABILITY

detected, even by a seismograph. Of course, this is just our initial finding. V. C ONCLUSION A satellite system to detect microwave signals generated by rock fractures in an EQ was proposed. The emitted microwave signal in the rock-fracture experiment was successfully calibrated by comparison with CWs. The microwave signal power to be received by a satelliteborne receiver was estimated by extrapolating the experimental results and based on the parameters of Aqua and AMSR-E. The investigation result indicates that an EQ can be detected from brightness temperatures under favorable conditions. Based on this investigation result, we analyzed the brightness temperature data obtained by AMSR-E with respect to an EQ that occurred on February 24, 2004 in Morocco and attempted to detect features associated with an actual EQ. As a result, we indicated the possibility that microwave energy was specifically emitted at 18.7 GHz around the epicenter two days before the Al Hoceima EQ. However, this analysis result is just the first step in the process to verify the possibility of EQ detection by a microwave sensor. For further verification, we obviously should have made fundamental improvements to the data analysis method. We were motivated by the analysis result indicated in this letter, so that we developed a specialized data analysis method to extract local and faint changes from the brightness temperature data of AMSR-E [12]. We then applied this analysis method to the Al Hoceima EQ and, finally, could strongly conclude that microwave energy was highly likely to be emitted two days before the Al Hoceima EQ. This topic will soon be presented in another paper. A PPENDIX Here, we will derive PEQ , LG , and LF . PEQ is calculated by PEQ = Pexp ×

200 [MHz] V × . v 500 [MHz]

(10)

For dielectrics, the attenuation constant (α m−1 ) is approximated as follows with the relative permittivity (r ) of quartzite set to 5 and its conductivity (σ) set to 10−6 S/m:  σ μr μ0 . (11) α 2 r 0 Here, 0 F/m (μ0 H/m) is the permittivity (permeability) in vacuum, and μr is the relative permeability. Assuming that μr = 1, we obtain α = 2.7 × 10−5 for 0 = 8.85 × 10−11 F/m and μ0 = 1.26 × 10−6 H/m. Based on this, LG is calculated as LG = e−2αd

(12)

111

where the exponential part is not −αd but −2αd since LG is the propagation loss of electric power. LF is calculated as  2 λ LF = (13) 4π(d + h) where λ is 16.0 mm since the microwave signal associated with rock fractures is considered to be emitted at 18.7 GHz. ACKNOWLEDGMENT The authors would like to thank Dr. K. Maki of the Institute of Physical and Chemical Research, Dr. E. Soma of the Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency (JAXA), and Dr. S. Yoshida and Dr. M. Nakatani of the Earthquake Research Institute, University of Tokyo, for performing the rock-fracture experiments, and Mr. K. Imaoka of the Earth Observation Research Center, JAXA, for arranging the AMSR-E data as well. R EFERENCES [1] P. Varotsos, K. Alexopoulos, and M. Lazaridou, “Latest aspects of earthquake prediction in Greece based on seismic electric signals II,” Tectonophysics, vol. 224, no. 1–3, pp. 1–37, Aug. 1993. [2] M. Tsutsui, “Identification of earthquake epicenter from measurements of electromagnetic pulses in the Earth,” Geophys. Res. Lett., vol. 32, no. 20, p. L20 303, Oct. 2005. DOI:10.1029/2005GL023691. [3] T. Yoshida and M. Nishi, “Observation of co-seismic electromagnetic phenomena in VHF associated with the Tottori-ken Seibu earthquake in 2000 and the Geiyo earthquake in 2001,” J. Seismol. Soc. Jpn., vol. 55, no. 2, pp. 107–118, 2002. [4] K. Maki, T. Takano, E. Soma, S. Yoshida, and M. Nakatani, “An experimental study of microwave emissions from compression failure of rocks,” J. Seismol. Soc. Jpn., vol. 58, no. 4, pp. 375–384, 2006. [5] T. Takano, Y. Murotani, K. Maki, T. Toda, A. Fujiwara, S. Hasegawa, A. Yamori, and H. Yano, “Microwave emission due to hypervelocity impacts and its correlation with mechanical destruction,” J. Appl. Phys, vol. 92, no. 9, pp. 5550–5554, Nov. 2002. [6] D. Daniels, Ed., Ground Penetrating Radar, 2nd ed. London, U.K.: IEE, 2004. [7] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing—Active and Passive, vol. III. New York: Addison-Wesley, 1982, pp. 2083–2084. [8] T. Kawanishi, T. Sezai, Y. Ito, K. Imaoka, T. Takeshima, Y. Ishido, A. Shibata, M. Miura, H. Inahata, and R. W. Spencer, “The advanced microwave scanning radiometer for the Earth observing system (AMSR-E), NASDA’s contribution to the EOS for global energy and water cycle studies,” IEEE Trans. Geosci. Remote Sens., vol. 41, no. 2, pp. 184– 194, Feb. 2003. [9] L. Ait Brahim, C. Nakhcha, B. Tadili, A. El Mrabet, and N. Jabour, “Structural analysis and interpretation of the surface deformations of the February 24th, 2004 Al Hoceima earthquake,” EMSC-Newslett., vol. 21, pp. 10–12, 2004. [10] Z. Cakir, M. Meghraoui, A. M. Akoglu, N. Jabour, S. Belabbes, and L. Ait-Brahim, “Surface deformation associated with the Mw 6.4, 24 February 2004 Al Hoceima, Morocco, earthquake deduced from InSAR: Implications for the active tectonics along North Africa,” Bull. Seismol. Soc. Am., vol. 96, no. 1, pp. 59–68, Feb. 2006. [11] T. Matsuzawa, N. Uchida, T. Igrashi, T. Okada, and A. Hasegawa, “Repeating earthquakes and quasi-static slip on the plate boundary east off northern Honshu, Japan,” Earth Planets Space, vol. 56, no. 8, pp. 803–811, 2004. [12] T. Maeda and T. Takano, “Discrimination of local and faint changes from satellite-borne microwave radiometer data,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 9, pp. 2684–2691, Sep. 2008.

Authorized licensed use limited to: BNM Institute of Technology. Downloaded on February 4, 2009 at 05:22 from IEEE Xplore. Restrictions apply.