Abstract—Fractal analyses of harmonic tremors recorded at

Sakurajima volcano were carried out to investigate the dynamical system regarding to their generating mechanism. We applied delay embedding theorem to reconstruct the phase space which describes the evolution behavior of a nonlinear system. We analyzed 2 types of harmonic tremor at the volcano into 2 groups: Harmonic Tremors occurred after B-type earthquake swarms (HTB) and Harmonic Tremor which occurred immediately after Eruption (HTE). The delay time used in the reconstruction was chosen after examining the first minimum of Mutual Information (MI) of the seismogram. In both harmonic tremors it was found that the method yielded a delay time in the range of 0.050-0.095 s for HTB and 0.060-0.125 s for HTE. The sufficient embedding dimension was estimated based on Cao’s method which had value in the range 3-5 for both events. Based on embedding parameters it was possible to calculate the correlation dimension of the resulting attractor. The values of the correlation dimension both of HTB and HTE are distributed in the range of 1.1–2.5. Keywords—Fractal, Sakurajima volcano, harmonic tremor, correlation dimension

.

I. INTRODUCTION HIS document is a template for Word (doc) versions. If you are reading a paper version of this document, so you can Harmonic tremor is one of the most elusive seismic

signals recorded in a volcanic environment. It is a continuous ground shaking characterized by peaks of spectra with regular frequency interval, composed of fundamental frequency and its multiple integers. Harmonic tremors have been observed at volcanoes with long-term eruptive activity, such as Sakurajima, Japan, Langila, Papua New Guinea, Arenal, Costa Rica, Mt. Erebus, Antarctica, Lascar, Chile and Mt. Semeru volcanoes, Indonesia. Various models of sources have been proposed to explain the mechanism of harmonic tremor based on their spectra analysis. For example, Chouet [3] proposed a model linear resonance of organ-pipe modes of a conduit triggered by sudden sustained disturbance. Schlindwein et al. [4] reported that spectra of harmonic tremor at Mt. Semeru could be triggered by repetition of sources. Benoit and McNutt

Sukir Maryanto, is with the Physics Department of Brawijaya University, Malang, East Java, Indonesia (corresponding author to provide phone: +62-34-575833; fax: +62-341-554403; e-mail: sukir@ub.ac.id ).

Masato Iguchi is a Assc. Professor at Sakurajima Volcano Research Center (SVRC), Disaster Prevention Research Institute (DPRI), Kyoto University, Japan. (e-mail: iguchi@svo.dpri.kyoto-u.ac.jp).

Takeshi Tameguri is a researcher from Sakurajima Volcano Research Center (SVRC), Disaster Prevention Research Institute (DPRI), Kyoto University, Japan. (e-mail: tamekuri@svo.dpri.kyoto-u.ac.jp ).

[5] proposed a linear resonance of a 1D vertical conduit for source model of harmonic tremor observed at Arenal volcano. Three physical source models, those are, eddy shedding, slug flow, and soda bottle have been proposed and discussed relationships of fundamental frequency and power spectrum to fluid dynamics variables, such as conduit size, kinematics viscosity of fluid and flow velocity [6]. Other model, however, suggests that site or path effects may be also responsible, in some cases, for the apparent harmonic tremor [7].

Most of the previous models mentioned above are mainly based on the characteristics of spectra which mostly triggered by linear process, however little intent has been given on nonlinear process. Julian [8] suggested that volcanic tremor generated by some kind of nonlinear processes involved in the source mechanism modeling. The methods based on the discipline of nonlinear dynamics have been rarely applied to the volcanic tremor recorded at some volcanoes. The first investigation of fractal properties of volcanic tremor conducted by Chouet and Shaw [9] for the tremor and gas piston events recorded at Kilauea volcano, Hawaii. Their results revealed a fractal dimension of the tremor attractor in the range of 3.1-4.1 with the average value of 3.75. This was interpreted that the source of tremor is not controlled by a stochastic process (where fractal dimension should be infinite), therefore it can be described by only a few degrees of freedom. Some studies of tremor and low-frequency events reported similar estimates of the fractal dimensions confirming the low-dimensional nature of the phenomena [9-13]. Knowledge of the dimension of an attractor is importance in modeling the dynamics of complex harmonic tremor source because it offers of the glimpse of the degree of self-organization, or dimensional simplification, present in these dynamics.

In this study, another approach of the harmonic tremor analysis has been applied based on the non linear theory. This paper describes the results of fractal analyses applied to harmonic tremors recorded at Sakurajima volcano, Japan. In our analysis we take full advantage of our data and investigate the fractal properties of harmonic tremor at various location and depths to determine how these properties vary in space as well as in time. First, we briefly describe the data obtained from Sakurajima Volcano Research Center, Kyoto University. We then present a method to reconstruct and calculate the dimension of the harmonic tremor attractor. We follow with an application of the method to our data and terminate with a discussion of the implications of these results for driving mechanism of harmonic tremor.

Sukir Maryanto, Masato Iguchi, and Takeshi Tameguri

Fractal Analyses of Harmonic Tremors at Sakurajima Volcano, Japan

T

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

318

II. OBSERVATION Data used in this study obtained from the permanent

seismic stations at Sakurajima volcano operated by Sakurajima Volcano Research Center (SVRC), Kyoto University [1]. We selected data from five of them as shown in Fig. 1. The 4 boreholes stations, ARI, HAR, KAB and KOM are distributed 2.7 - 4.4 km apart from the active crater, Minamidake and are installed at depths of 85 – 290 m below the ground surface with the deepest station, HAR. The ground based station, HIK, is installed at 1.7 km the closest station from the crater. A three-component and 1 Hz short-period seismometer was installed at each station with the sensors of two horizontal components are oriented to the direction to the crater (L : longitudinal component) and perpendicular to the crater (T : transverse component). Before May 2001, the recording seismic signals from the stations are transmitted to SVRC via telephone lines or radio waves and were recorded on analog magnetic tapes. The seismic signals have been digitized at a rate of 200 Hz and 22 bit resolution at the stations and 1s packet data are transferred to SVRC by UDP protocol since May 2001. For analysis, a sampling rate of 200 Hz and 12 bit resolution were applied to digitize the analog records.

Sakurajima

N

Minamidakecrater

Fig. 1 Location of seismic stations used in this study. Squares and circle denote borehole and ground-based seismometers, respectively. Triangle shows location of the summit crater of Minamidake. SVRC: Sakurajima Volcano Research Center to register seismic records

The characteristics of spectral and particle motion of

harmonic tremors at Sakurajima as well as classification of them have been previously published by Maryanto et. al. [1], therefore only a summary of these aspects will be presented here. They classified harmonic tremor recorded at Sakurajima volcano into two types; HTB and HTE. HTB follows B-type earthquake swarm and precedes explosive eruption several hours before, while HTE occurs a few minutes after an eruption. Seismic records before and after HTB at 14h – 15h on July 20, 1990 (JST=9h+UTC) are shown in Fig. 2a. A swarm of B-type earthquakes began at 2h on July 19, 1990 and continued to 19h. Nineteen hours after the B-type earthquake swarm, HTB began to occur at 14h06m on July 20. The HTB continued for 2 hours. An example of waveform of HTE is shown in Fig. 2b. Explosive eruption occurred on 11h15m27s on October 11, 2002 and volcanic tremor

succeeded. The waveform of the tremor seemed to become harmonic at 11h19m, three minutes after the beginning of the eruption. The HTE continued for 12 minutes.

0 10 20 30 40 50

15:06

HTB

15:05

15:04

15:03

15:02

15:01

14:59

July 20, 1990

0 10 20 30 40 50 60(s

11:24

11:23

11:22

11:21

11:20

11:19

11:18

11:17

11:16

October 11, 200211:15

HTE

Explosionearthquake

160µ

m/s

Fig. 2 Two types of harmonic tremors. The seismograms were observed by a vertical component seismometer at station HIK. (a) Apart of records (13h14m - 13h16m on July 19, 1990) of BL-type earthquake swarm which began at 02h35m, HTB from 14h59m - 15h07m on July 20, 1990 and the following explosion earthquake at 20h47m. (b) Explosion earthquake and the following HTE records from 11h15m - 11h25m on October 11, 2002.

The characteristics of particle motions and spectra are summarized as follows;

1. Particle motion of HTB and HTE are dominated by Rayleigh; and mixed by Loves waves. Spatial distribution pattern of Rayleigh and Love waves are similar to each other. Dominant surface waves and similar distribution pattern suggest that HTB and HTE are generated by a similar source at a shallow depth.

2. Although HTB and HTE have similar pattern of spectra, having fundamental peak and higher mode peak at frequencies of multiple-integers of the fundamental frequency, the temporal characteristics are different. Peak frequency of HTB is nearly constant. In contrast, peak frequencies of HTE gradually increase

III. RECONSTRUCTION OF THE PHASE SPACE The complexity of the spectra and particle motion

associated with the harmonic tremor at Sakurajima volcano, suggests that the standard approach on spectral and particle motion analysis, while applicable, is not sufficient to describe the dynamical system chaotic behavior of their generating mechanism. Here we applied another approach to obtain a

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

319

description of harmonic tremor dynamic system by reconstructing the phase space from the time series of a single variable represented by a seismogram.

The problem of reconstruction of the phase space from a scalar time-series (such as a seismogram) is of great practical importance, since it is the starting point of all non-linear time-series analysis methods. Anytime series resulting from a non-linear process can be considered as the projection on the real axis of a higher-dimensional geometrical object that describes the behavior of the system under study. Takens [14], showed that it is possible to recover this object from a series of scalar measurements x(t) in an m-dimensional Euclidean space using points y with coordinates

( ) ( ) ( )ττ )1(,...,, −++= mtxtxtxy (1) where τ is called the delay time and for a digitized time-series is a multiple of the sampling interval used. The dimension m of the reconstructed space is considered as the sufficient dimension for recovering the object without distorting any of its topological properties, thus it may be different from the true dimension of the space where this object lies. This procedure of phase space reconstruction is termed embedding and the formulation of Taken’s is called the delay embedding theorem, with m being the embedding dimension. In practical applications both the delay time and the embedding dimension have to be determined from the time-series itself. A.1 Selection of the delay time

In a phase space reconstruction procedure, we must ensure that the points in each dimension (coordinate) are independent of each other. Therefore, time delay τ must be chosen so as to result in points that are not correlated to previously generated points. One proposed way of choosing the delay time τ for phase space reconstruction is by calculating the Mutual Information of the time series data and choosing τ as the time of the first local minimum [2]. The Mutual Information ( )YXMI , of time series xi{t=1, 2, …, M} is defined:

( ) ( ) ( )( ) ( )jpip

jipjipYXMI

yx

xy

jxy

i

,log,, ∑∑= (2)

Where in the first step, a sequence Ｘ i{i=1, 2, …, M} of length M is considered in the time series Xi. Let the probability distribution in Xi be px(i) and the probability distribution in the sequence Yi{j=α, α+1, …, α+M} obtained by shifting Xi by α be py(j). By varying the time shift α, let the time delay, for which first takes the local minimum, be τ.

Fig. 3 shows the mutual information of the HTB. The first minimum value of the mutual information is approximately 0.05 sec. A.2 Estimation of the embedding dimension

The next step in reconstructing phase space is to recover the appropriate number of coordinates m of the phase space. The idea of a number of coordinates m is a dimension in which the geometrical structure of the phase space is completely unfolded.

0 20 40 60 80 1000

2

4

6

8

time lag (multiple of the sampling interval)

MI

L T V

Fig. 3 Mutual information of HTB calculated for delay times 1–100s. Solid, dash and dot lines represent component L, T and V, respectively.

The basic method in determining the embedding

dimension in phase-space reconstruction is the False Nearest Neighbor method. Suppose the vector y

NN is a false neighbor

of y, having arrived in its neighborhood by projection from a higher dimension, because the present dimension m does not unfold the attractor, then by going to the next dimension m+1, we may move this false neighbor out of the neighborhood of y. Thus, if the additional distance is large compared to the distance in dimension d between nearest neighbors, we have a false neighbor. Otherwise, we have a true neighbor. In order to have a straightforward representation of the minimum embedding dimension, Cao et al., [15] defined the mean value of E1, which generally represents the relative Euclidean distance between y

NN and y

NN in two consecutive dimensions.

Cao’s number E1 consequently will stop changing when the dimension m is greater than the minimum embedding dimension m0. Fig. 4 depicts the Cao number as a function of embedding dimension. It can be observed that E1 approaches a constant value for a dimension higher than four. Thus, we can conclude that the minimum dimension that will totally unfold the phase space is 5.

0 2 4 6 8 100

1

2

L T V

E 1(m)

Dimension (m)

Fig. 4 Minimum embedding dimension for HTE calculated using CAO’s method. Solid, dash and dot lines represent component L, T and V, respectively.

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

320

xx

x

x+0.05sx+0.045s

x+0.04s

x

x+0.035s

Fig. 5 Four projections of the HTB event for delay time 0.035 s, 0.040 s, 0.045 and 0.050, respectively.

Fig. 5 shows an attractor of HTB for several delay time.

We used 1024, 2048, 4096 points to define the harmonic tremor attractor. Chouet and Shaw [9] used 1000-8000 points to determine the dimension of strange attractor of volcanic tremor and gas piston event at a fixed sampling rate of 200 samples/s.

B. Estimation of Correlation Dimension

The attractor resulting from the embedding procedure is a

fractal object and the estimation of its dimension forms a part of every successful nonlinear time series analysis. The importance of the fractal dimension stems mainly from two reasons: (1) it gives a measure of the effective degrees of freedom that are present in the physical system under study; (2) it is a quantity that does not vary under smooth transformations of the coordinate system (i.e., an ‘invariant measure’). Even though there exists a number of definitions for the dimension of a fractal object (Box-counting dimension, Information dimension, etc.), the correlation dimension definition suggested by Grassberger and Procaccia [16] was found to be the most efficient for practical applications.

For an N number of points attractor ny resulting from the embedding and for distance values r, the correlation sum definition as [14]

( ) ( ) ( )∑−

≠=∞→

−−−

=1

,0,11lim

N

jijijiNn yyrH

NNrC

(3)

where H(x) is the Heaviside step function, with H(x)=0 if x < 0 and H(x)=1 for x ≥ 0. And . is suitable norm, chosen here as the Euclidian measure of distance (square root of the sum of the squares components). When the r can be scaled like a power law Cn(r) ∼ rD and then correlation dimension d2 is defined as:

r

rCd n

ln)(ln

2 = (4)

We calculated the correlation sum for our dataset using the delay

times and embedding dimensions determined in the previous section. Fig. 6 shows the correlation dimension of component V, L and T of HTB at station KOM.

KOM-Ld2=1.7

(a)

KOM-Td2=1.8

(b)

KOM-Vd2=1.8

(c)

Fig. 6 Correlation dimension of HTB recorded at station KOM. a) Longitudinal component, b) Transversal component and c) Vertical component.

IV. RESULTS AND DISCUSSIONS In our study of the fractal properties of harmonic tremor,

we investigated the spatio-temporal evolution of the attractor by computing the fractal dimension for 5-s, 10-s and 20-s windows of the HTB and HTE of five stations and three components. A comparison of the values of the correlation dimension of the two groups shows that the HTE events possess small larger correlation dimensions than the HTB events, which suggests that HTB have more stationary process than HTE.

Each component in each group have no significant different of correlation dimensions. For time windows 5 s, the correlation dimensions of HTB fluctuate in the range of 1.2-1.8, 1.3-1.9 and 1.2-1.9 for longitudinal, transversal and vertical components, respectively. For HTE, they fluctuate in the range of 1.8-2.3, 1.6-2.3 and 2.1-2.5 for longitudinal, transversal and vertical components, respectively. The similarity range appears in the time windows of 10 s and 20 s for all components of both types.

In addition, the spatio-temporal distributions of fractal dimension associated with the wave fields of HTB and HTE recorded at five stations are given in Figs. 7, 8 and 9. There

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

321

are no significant differences among five stations and between borehole and ground based stations. These facts above suggest that the two types are governed by lightly different physical mechanism.

Fig. 7 Temporal variation of correlation dimension of the HTB component V recorded at station HIK.

Maryanto et al. [17] and Tameguri et al.[18], suggest that based on the spectra analysis, moment tensor analysis, visual observations, and ground deformation associated with HTB and HTE, the following facts should be considered as constraints for modeling of their generation mechanism; 1) Spectrum peaks appear at the frequencies of multiple

integers of the fundamental frequencies for HTB and HTE.

2) Peak frequencies of HTB are rather stable, while those of HTE shift toward higher frequency.

3) Source depths of HTB and HTE coincide with the gas pocket at uppermost part the conduit.

4) HTB and HTE have the similar source mechanism to each other, involving more than 50 % of isotropic component and may be generated by repetition of expansion and contraction of horizontal crack.

0 60 120 180 240 300 360 420 480 540 600

0 60 120 180 240 300 360 420 4800

4

8

12

0 60 120 180 240 300 360 420 4800

2

4

6(s)

80 µm

/s

(s)

(s)

Time (s)

Frac

tal d

imen

sion

f1 f2f3f4Fr

eque

ncy

(Hz)

Fig. 8 Temporal variation of correlation dimension of the HTE component V recorded at station HIK.

The suggestion that volcanic tremor is the result of non-linear source processes involving one or several different kinds of magmatic activity is not only supported by theoretical considerations Julian [7], but also by certain characteristics observed in tremor signals, which are

believed to be common among systems exhibiting aperiodic, chaotic behavior.

Fig. 9 Spatial trend of correlation dimension of HTB and HTE in the component of Longitudinal, Transversal and Vertical recorded at station HIK, ARI, HAR, KAB and KOM.

In this study, the low values of the correlation dimension both of HTB and HTE (1.1–2.5) suggest that a second-order nonlinear differential equation may be enough to describe the Sakurajima harmonic tremor source. Qualitatively, this fact can be considered as constraint in source modeling of harmonic tremor. Similar low values of correlation dimension of tremor have been reported at Sangay volcano [12]. The similarities between the properties of the Sangay and Sakurajima tremor as demonstrated by the appearance of the attractors and the upper bounds of their fractal dimensions, point to the possibility of similar characteristics of tremor source processes.

Even though the results of the application of fractal analysis methods to harmonic tremor events can tell us what kind of oscillatory behavior, they cannot provide a clue as to what the physical mechanism of this oscillation is. Other methods, that can extract information concerning the physical properties and geometrical configuration of the rock-fluid system from seismic and acoustic data combined with visual observations, are needed in order to accomplish this task.

Future volcano monitoring and prediction efforts should, therefore, rely on a multidisciplinary approach both on observational and theoretical when trying to study the nature of harmonic tremor sources.

V. CONCLUSION We have analyzed two kinds of harmonic tremor, HTB

and HTE, recorded at Sakurajima volcano, Japan. The HTE events possess small larger correlation dimensions than the HTB events, which is suggests that HTB have more stationary process than HTE, which suggest that the two types are related to lightly differences process. The fractal properties are summarized as follows; 1. Each component in each group have no significant

different of correlation dimensions. 2. The correlation dimension is lower for vertical

component than horizontal components at station、suggesting a lesser degrees of path-dependent

0 60 120 180 240 300 360 420 480

0 60 120 180 240 300 360 420 4800

2

4

6

8

0 60 120 180 240 300 360 420 4800

1

2

3

4

38 µm

/sFr

actal

dim

ensio

n

f1 f2f3f4Fr

eque

ncy

(Hz)

Time (s)

0 1 2 3 4 50

1

2

30 1 2 3 4 5

0

1

2

3

KOM

KAB

HAR

ARIHIK HTE L HTE T HTE V

Distance (km)

FD

KOM

KAB

HAR

ARI

FD

HTB L HTB T HTB V

HIK

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

322

complexity in vertical than horizontal motion. Some questions still open regarded to the physical mechanism of the nonlinear oscillation of source.

ACKNOWLEDGMENT The author greatly appreciate Prof. Kazuhiro Ishihara of

Disaster Prevention Research Institute, Kyoto University, for their encouragement and valuable helps, and also thanks to all staffs of Sakurajima Volcano Research Center, DPRI, Kyoto University for supporting the data.

REFERENCES [1] S Maryanto, M Iguchi, and T Tameguri, “Spatio-Temporal

Characteristics on Spectra and Particle Motion of Harmonic Tremors at Sakurajima Volcano, Japan,” Ann. Disast. Prev. Res. Inst., Kyoto University, vol. 48B, pp. 329-339, Apr. 2005

[2] M Nakagawa, “Chaos and Fractal in Engineering,” Singapore: World Scientific, 1999, ch. 3.

[3] B Chouet, “Excitation of a buried magmatic pipe: a seismic source model for volcanic tremor,” J. of Geophys. Res., vol. 90, pp 1881-1893, 1985.

[4] V Schlindwein, J Wassermann，and F Scherbaum. Spectral analysis of harmonic tremor signals at Mt. Semeru volcano, Indonesia. Geophys. Res. Lett., vol. 22, no. 13, pp 1685-1688, Jul. 1995.

[5] J Benoit and SR McNutt, “New constraints on source processes of volcanic tremor at Arenal volcano, Costa Rica, using broadband seismic data” Geophys. Res. Lett., vol. 24, no. 4, pp. 449–452, Feb. 1997.

[6] M Hellweg, “Physical model for the source of Lascar`s harmonic tremor,” J. Volcanol. Geotherm. Res., vol 101, pp. 183-198, 2000.

[7] S Kedar, S Bradford, and H Kanamori, “The origin of harmonic tremor at Old Faithful geyser,” Nature, vol. 379, no. 22, pp. 708-711, Feb. 1996.

[8] BR Julian, “Volcanic tremor: non-linear excitation by fluid flow,” J. Geophys. Res., vol. 99, pp 11859-11877, 1994.

[9] B Chouet and HR Shaw, “Fractal Properties of Tremor and Gas-piston Events Observed at Kilauea Volcano, Hawaii,” J. of Geophysics Research, vol. 96, no B6, pp. 10177-10189, Jun. 1991.

[10] C Godano, C Cardaci, and E Privitera, “Intermittent Behaviour of Volcanic Tremor at Mt. Etna,” Pure Appl. Geophysics, vol. 147, no. 4, pp. 729–744, 1996.

[11] C Godano and P Capuano, “Source Characterisation of Low Frequency Events at Stromboli and Vulcano Islands (Isole Eolie Italy),” Journal of Seismology, vol. 3, pp. 393-408, 1999.

[12] KI Konstantinou, “Deterministic Nonlinear Source Processes of Volcanic Tremor Signals Accompanying the 1996 Vatnajökull Eruption, Central Iceland,” Geophys. J. Int., vol. 148, pp 663-675, 2002.

[13] KI Konstantinou and Lin CH, “Nonlinear Time Series Analysis of Volcanic Tremor Events Recorded at Sangay Volcano, Ecuador. Pure Appl. Geophysics, vol. 161, pp. 145–163, 2004.

[14] F Takens, “Detecting Strange Attractors in Turbulence,” Lecture notes in Math. New York : Springer; 1981.

[15] L Cao, “Practical method for determining the minimum embedding dimension of scalar time series,” Physica D, vol. 110, pp. 43-50, 1997.

[16] P Grassberger and I Procaccia, “Characterisation of Strange Attractors,” Phys. Rev. Lett., vol. 50, pp 346-349, 1983.

[17] S Maryanto, T Tameguri, and M Iguchi, “Constraints for source mechanism of harmonic tremor based on seismological, ground deformation and visual observations at Sakurajima Volcano, Japan,” J. Volcanol. Geotherm. Res. (accepted for publication), Oct. 2007.

[18] T Tameguri, S Maryanto, and M Iguchi, “Moment tensor analysis of harmonic tremors at Sakurajima Volcano,” Ann. Disast. Prev. Res. Inst. Kyoto University, vol. 48B, pp. 323-328, Apr. 2005.

Sukir Maryanto was graduated from Physics Department, Brawijaya University in 1995, Master of Science in Geophysics of Gadjah Mada University, Yogyakarta, Indonesia in 2000, and obtained Doctor of Science from Kyoto University, Japan in 2007. Current research interests in Physical Volcanology at Geophysics laboratory, Brawijaya University, Malang, East Java, Indonesia. Masato Iguchi is a Assc. Professor at Sakurajima Volcano Research Center (SVRC), Disaster Prevention Research Institute (DPRI), Kyoto University, Japan. Takeshi Tameguri is a researcher from Sakurajima Volcano Research Center (SVRC), Disaster Prevention Research Institute (DPRI), Kyoto University, Japan.

International Conference on Chemical, Biological and Environment Sciences (ICCEBS'2011) Bangkok Dec., 2011

323