Wavelet maxima curves of surface latent heat flux anomalies associated with Indian earthquakes

(1)

HAL Id: hal-00299125

https://hal.archives-ouvertes.fr/hal-00299125

Submitted on 17 Jan 2005

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Wavelet maxima curves of surface latent heat flux

anomalies associated with Indian earthquakes

G. Cervone, R. P. Singh, M. Kafatos, C. Yu

To cite this version:

G. Cervone, R. P. Singh, M. Kafatos, C. Yu. Wavelet maxima curves of surface latent heat flux

anomalies associated with Indian earthquakes. Natural Hazards and Earth System Science, Copernicus

Publications on behalf of the European Geosciences Union, 2005, 5 (1), pp.87-99. �hal-00299125�

(2)

Natural Hazards and Earth System Sciences (2005) 5: 87–99 SRef-ID: 1684-9981/nhess/2005-5-87

European Geosciences Union

Natural Hazards

and Earth

System Sciences

Wavelet maxima curves of surface latent heat flux anomalies

associated with Indian earthquakes

G. Cervone1, R. P. Singh1, 2, M. Kafatos1, and C. Yu1

1_{Center for Earth Observing and Space Research, School of Computational Sciences, George Mason University, Fairfax, VA}

22039, USA

2_{Department of Civil Engineering, Indian Institute of Technology, Kanpur 208 016, India}

Received: 6 May 2004 – Revised: 21 October 2004 – Accepted: 6 December 2004 – Published: 17 January 2005

Abstract. Wavelet maxima curves of surface latent heat flux

(SLHF) have been recently used to provide early warning information about coastal earthquakes. The present paper re-ports further validations of the spatial and temporal analysis of wavelet maxima curves associated with earthquakes oc-curred in different parts of the Indian sub-continent. Promi-nent anomalies that exhibit spatial and temporal continuity are found in case of coastal earthquakes, while no anomaly is detected in case of intraplate earthquakes. The precursory SLHF anomalies are found up to 2 weeks prior to the main earthquake event, with an extent area of up to 600 km.

1 Introduction

Significant changes in land, oceanic and atmospheric param-eters have been observed prior to large earthquakes using re-mote sensing data (Singh et al., 2001; Tramutoli et al., 2001; Tronin et al., 2002; Dey and Singh, 2003; Ouzonov and Fre-und, 2004; Cervone et al., 2004; Liu et al., 2004; Maekawa and Hayakawa, 2004; Pulinets and Boyarchuk, 2004). These changes suggest the existence of possible interaction be-tween the lithosphere and the atmosphere, and have opened up new possibilities to the use of satellite-based observations to study earthquakes precursors.

Cervone et al. (2004) have introduced a new data min-ing methodology based on wavelet transformations and sta-tistical analysis to detect precursory signals associated with earthquakes. This methodology was validated on two large earthquakes occurred in Greece using SLHF data. Promi-nent anomalies were detected about two weeks prior to both events, suggesting the possibility of developing an early warning system for impending earthquakes. Dey and Singh (2003) and Singh et al. (2004) have recently shown in the

Correspondence to: R. P. Singh

([email protected])

case of coastal earthquakes the consistent occurrence of anomalous SLHF peaks a few days prior to the main earth-quake event. They have concluded that the continuous mea-surements of SLHF can provide early warning information about an impending coastal earthquake. The magnitudes of the SLHF peaks are found to be variable, while SLHF tends to be higher over oceans and lower over land. The origin of the anomalous SLHF is likely to be related with the increase of surface temperature in the epicentral region which is as-sociated with the build up of stress and movements along the faults. It is believed that the temperature increases prior to an earthquake (Qiang, 1997). For example, changes in tempera-ture up to 5 K for the Gujarat earthquake of 26 January 2001 has been mapped using Infrared (IR) wavelength observa-tions by the Moderate Resolution Imaging Spectroradiome-ter (MODIS) sensor (http://amesnews.arc.nasa.gov/releases/ 2002/02images/quakes/earthquakes.html) and is likely due to frictional mechanism along the fault or due to movements of the fluids at depth (Ouzonov and Freund, 2004). Due to the heat conduction, the sea surface temperature increases which in turn is likely responsible for increasing the ocean evapora-tion giving rise to anomalous SLHF prior to the main earth-quake events (Dey and Singh, 2003).

In this paper, the general methodology discussed by Cer-vone et al. (2004) has been applied to five large earthquakes which occurred in different locations of the Indian sub-continent. The present paper further validates the method-ology on earthquakes occurring in different geographical re-gion, and to characterize the extent and the geometrical shape of the anomalies in order to discriminate from similar anoma-lies due to other atmospheric or oceanic phenomena. The earthquakes considered occurred in four different regions of India, which are characterized by different types of seismic events. Statistical t tests are used to determine the statistical significance of the detected anomalies.

(3)

88 G. Cervone et al.: Wavelet maxima curves of surface latent heat flux anomalies

2 Data and Methodology

The SLHF data used in the present study cover the pe-riod from 1 January 1998 to 28 March 2004 for the region bounded by latitudes 33 N to 45 N and longitudes 14 E to 28 E. The SLHF data have been downloaded from the web-site of the Scientific Computing Division of the National Center for Atmospheric Research (NCAR) (http://ingrid. ldeo.columbia.edu/SOURCES/NOAA/NCEP-NCAR/). The plate boundary data have been downloaded from http:// www-geology.ucdavis.edu/GEL102/demets.html, and con-sist of the best fitting Euler vectors, closure fitting Euler vec-tors and the global model NUVEL-1 to describe geologically current plate motions between 12 assumed rigid plates. The details of this dataset are described by DeMets et al. (1990). The methodology employed uses wavelet transformations as data mining tools by computing the wavelet maxima that propagate from coarser to finer scales (Cervone et al., 2004). Those maxima are used to identify strong anomalies in the data; however only those anomalies that show continuity in both time and space are assumed as possible precursors for earthquakes. Time continuity means that the detected anomalies occur at the same time or with a short delay with respect to each other, while space continuity means that the detected anomalies are distributed in space according to a precise geometry conforming to the relevant geological set-tings of the region. Details of the SLHF data used in the present study and methodology are discussed in detail by Cervone et al. (2004).

The wavelet analysis is also paired with statistics to de-termine the significance of the singularities found. The first step consists in computing the 30 days average and its stan-dard deviation using several years of prior data. In the present work, five years of prior data are used due to the unreliability of data with longer temporal coverage. For each year prior to the earthquake, a 365-days long time series is transformed into 12 data points, each being the average of 30 days, except for the last that is the average of either 35 or 36 days. The av-erage and standard deviation are computed using the 12 data points for all previous years, and are later expanded to the original size of 365 days, by assigning the value of each data point to a sequence of 30 consecutive points, except the last that has 35. As a result the average and standard deviation are two step functions with a length of 365. A spline interpo-lation is used to smooth both the computed average and the standard deviation.

The following t test is performed on the singularities de-tected by the wavelet transformation to determine the statis-tical significance of the peaks. The t test is commonly used in statistics to determine if the mean of the population and the mean of the sample are statistically different Miller et al. (1997).

Let

X =P Xi

n (1)

where X is the unbiased estimator of the population mean, n is the number of points used in the computation of the aver-age and standard deviation (in the present work, n=30*5=150 because we use the 30 days average for five years of data) and Xi is the ith point of the time series for the year of the earthquake, and S2= P Xi −X 2 n −1 (2)

where S2 is the unbiased estimator of the population vari-ance, Xi is the ith point of the time series for the year of the earthquake, and X is the mean previously computed.

Based on the Central Limit Theorem

X ∼ N µi,

σ_i2 n

!

(3)

where N represents the normal distribution, with mean µi, estimated by X, and variance σ_i2estimated by S2. In other words, we expect the sample mean to asymptotically come from a normal distribution.

A new mean is created

µ0_i =µi +b

σi

√

n (4)

where µi and σ are the mean and the standard deviation of the normal distribution, and b is a constant used to adjust the allowed level of oscillations in the data. In the experiments b is always set equal to 1. Hence

µ0_i ∼N µi+b σi √ n, σ_i2 n ! (5)

A new data point yi is tested by comparing µ0_i with yi, using the two following hypotheses:

H0:yi =µ0i (6)

H1:yi > µ0i (7)

An hypothesis is accepted or rejected according to the t test t = yi−µ 0 i σi √ n (8)

In the present work, the estimate of the test statistics, which is t , is computed by ˆ t = yi−X − b√S_n S √ n (9)

If ˆt is greater than a predefined quantile value (at α 0.05 significance level, the corresponding quantile is ˆt1−α,n or

1.65) then we reject H0and conclude that the data point yiis statistically significant. 1−α is used because the methodol-ogy is for a one-sided t test.

(4)

G. Cervone et al.: Wavelet maxima curves of surface latent heat flux anomalies 89 6 1Gujarat Mw 7.6_{26 Jan 2001} 2, 3 Andaman MW 6.5 15 Sep 2002 Andaman MW 5.4 10 Apr 2004 4 Pondicherry 5.4 25 Sep 2001 5 Jabalpur Mw 6.0 22 May 2002 66˚E 66˚E 72˚E 72˚E 78˚E 78˚E 84˚E 84˚E 90˚E 90˚E 6˚N 6˚N 12˚N 12˚N 18˚N 18˚N 24˚N 24˚N 30˚N 30˚N 36˚N 36˚N

Fig. 1. Map of India indicating the location of the Earthquakes. The epicenters are marked with a star, and annotated with the location, time and magnitude of the earthquake. The plate boundary is also indicated

Fig. 1. Map of India indicating the location of the Earthquakes. The epicenters are marked with a star, and annotated with the location, time and magnitude of the earthquake. The plate boundary is also indicated

Table 1. Details of the earthquakes used in the experiments.

Earthquake Event Date Location Focal Depth Magnitude

Gujarat 26 January 2001 23.399 N 70.316 E 17 km 7.6

Andaman 1 13 September 2002 13.04 N 93.07 E 21 km 6.5

Andaman 2 10 April 2004 13.04 N 93.25 E 33 km 5.4

Pondicherry 25 September 2001 11.94 N 80.22 E 19 km 5.4

Jabalpur 22 May 1997 23.08 N 80.04 E 35 km 6.0

3 Results and Discussion

Detailed analysis has been performed using SLHF data for five earthquakes which occurred in different parts of India, namely Gujarat (1), Andaman (2 and 3), Pondicherry (4), and Jabalpur (5) (Fig. 1). Details of these earthquakes are given in Table 1.

A 1-D wavelet transformation has been applied on sev-eral grids adjacent to the epicentral region. The goal is to identify singularities in the data, defined as abrupt changes in the first derivative of the time-series. The maxima of the wavelet coefficients at different scales have been

com-puted by generating maxima curves which propagate from the coarser to the finest scales. All the singularities corre-sponding to maxima curves which propagate from the finest to a fraction of the total number of scale, are considered to be significant. Such technique has the advantage of effectively filtering out seasonal components, which show primarily at coarser scales (lower frequency), and also high frequency noise which does not propagate through the wavelet coef-ficients. Additionally, only the anomalies with statistical sig-nificance are kept, namely those whose magnitude is above the noise level for the region, computed using five years of prior data. This wavelet transformations is carried out not

(5)

Table 2. Parameters of the numerical real wavelet transformation (RWT) used in the experiments.

Function RWT

Mother Wavelet Sombrero

Nvoice 10

Scale 10

Octave 2

Propagation Factor 1₂

only for the grid comprising the epicenter of the earthquake, but also for several adjacent region. The wavelet analysis has been performed using 365 SLHF data points, corresponding to one year of data. The parameters used in the wavelet anal-ysis are given in Table 2.

The second step is to discriminate anomalies associated with earthquakes from those due to other atmospheric or oceanic phenomena by using the concept of spatial and time continuity. Time continuity means that the detected anoma-lies occur at the same time or with a short delay of each other, while space continuity means that the detected anomalies are distributed in space according to a precise geometry con-forming to the geological settings of the region. Anomalies associated with earthquakes are caused by a large scale event, and thus their extent is not confined only to the epicentral area. Additionally, such anomalies have a peculiar charac-teristic geometrical shape, which can be used to discriminate from other anomalies. The identifications of anomalies asso-ciated with earthquakes is carried out by selecting anomalies with a characteristic geometrical shape and occurring all at the same time or within a short time period (usually 1 or 2 days). The geometrical shape of the anomalies are likely to be related to the geological characteristics of the region, such as continental boundaries or fault lines. The results of the wavelet analysis are given in three parts:

1. The first part shows the time series for the original sig-nal, the 30 days average for the previous five years 1998–2002, and the 1 and 2 standard deviations (sigma) line for the 30 days average. The average has been in-terpolated using a spline based algorithm to generate a smooth line.

2. The second part is a graphical representation which shows the time when significant wavelet maxima are de-tected. The color indicates the propagation depth in or-der to emphasize those maxima which propagate to the finer scales.

3. The third and last part consists of a graphical represen-tation of the wavelet coefficients, and the corresponding maxima lines.

The results of the space and time continuity are shown in Figs. 2b, 6b and 8b. In the figures, the x-axis represents time expressed in days, and the y-axis represents the grid of the

grid path. The strength of the anomaly is color coded. The parameters used to compute the spatial and temporal conti-nuity are given on top of each figure.

3.1 Gujarat

The Gujarat earthquake (Mw=7.6), occurred on 26 January, 2001, at 03:16 UTC, and its epicenter was calculated at lo-cation 23.399 N 70.316 E, with a focal depth of 17 km. The earthquake occurred along an approximately east-west trend-ing thrust. This is an intraplate earthquake, since it occurred far away from the continental boundary. Historically, this re-gion is hit by a major earthquake every 100 years. During the period of quiescence, none or very little seismic activ-ity is recorded. In particular, in the 6 months prior to the 2001 earthquake, there was no seismic activity recorded by the global seismic network.

The 26 January, 2001, earthquake is one of the most deadly earthquakes that occurred in the Indian subcontinent. The large magnitude of the event and the lack of safety stan-dards as well as a low level of preparedness of the population caused great losses of lives and properties. Official Indian Government figures reported about 20 000 deaths and about 170 000 injured. The entire region suffered with great de-pression causing damages between 1 and 5 billion dollars, and affecting the life of several million people. The general destruction of building caused large amount of dust, which increased the aerosol concentration and affected the local cli-mate (Okada et al., 2004).

The SLHF data are analyzed over several grids shown in Fig. 2a. Each grid is annotated with a unique number on the bottom left, and with the day of when the anomaly was de-tected, and the offset from the day of the earthquake. The annotation “23 Jan -3”, means that the anomaly was detected on 23 January 2001, which is 3 days prior to the earthquake event. When grids have more than one anomaly, they are listed consecutively in chronological order. This grid path is chosen following the coastline, and it also reflects the lo-cations where thermal anomalies associated with this earth-quake have been recently found from MODIS data1. The MODIS data has shown an increase up to 50K in the sur-face temperature over the epicentral region prior to the earth-quake event (Ouzonov and Freund, 2004). The increase in thermal temperature is found to be associated with the tec-tonic settings of the region. In fact, one of the crucial parts of this experiment was the selection of the grids in the ab-sence of distinct tectonic features, such as continental bound-aries. This region contains very complicated faults rather short in length which, combined with the coarse resolution of the SLHF data, make the selection of the grid path a non-trivial task. In this particular case, the thermal anomalies observed using MODIS are important for the selection of the appropriate grid path.

1_{http://amesnews.arc.nasa.gov/releases/2002/02images/quakes/}

(6)

G. Cervone et al.: Wavelet maxima curves of surface latent heat flux anomalies 91

7

34 35

26 27 28

19 20

09Jan -17

09Jan -17 09Jan -17 09Jan -17

09Jan -17 12Jan -14 13Jan -13

14Jan -12

14Jan -12 14Jan -12 15Jan -11

15Jan -11 21Jan -5

23Jan -3 23Jan -3 23Jan -3

23Jan -3 23Jan -3 24Jan -2 24Jan -2

26Jan 0

64˚E 66˚E 68˚E 70˚E 72˚E

18˚N 20˚N 22˚N 24˚N 26˚N

(a) Extent of the SLHF anomalies

W_M^2 0 50 100 150 200

02 Dec 31 Dec 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 01 Dec Years 2000/2001 Avg 1999..2003 1 sigma 2 sigma

Surface Latent Heat Flux : Years 2000/2001

02 Dec 31 Dec 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 01 Dec

Wavelet Maxima Log2(scale) Day 5.1 4.6 4.1 3.6 3.1 2.6 2.1 1.5 1 0.5 0

02 Dec 31 Dec 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 01 Dec

Wavelet Coefficients

0 0.1 0.20.30.40.5 0.60.70.8 0.9 1

Color Scale

Event Date = 26 Jan 2001 Data analyzed =02 Dec 2000..01 Dec 2001 Wavelet Library = WaveLab Function = RWT (X, 10 , Sombrero , 1 , 10 ) Propagation Factor = 1 / 2

Location 23.81N 71.25E Grid 37

(b) Wavelet analysis using SLHF data

Grid Number

02 Dec 01 Jan 31 Jan 02 Mar 01 Apr 01 May31 May 30 Jun 30 Jul 29 Aug 28 Sep 28 Oct 01 Dec

34 35 26 27 28 19 20

Continuity along the Grids

Max Disc=2, Min Len=7, Pen Disc=2, Tol=90%

1 2 3 4 1 2 3 4 5 6 Magnitude of Anomaly Anomaly Discontinuity Earthquake Weak Signals Strong Signals

(c) Time and space analysis of SLHF data

24 Jan 24 Jan 23 Jan 23 Jan 23 Jan 23 Jan 23 Jan Anomaly Grid Number 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 34 35 26 27 28 19 20

Anomalies for Signal 1

Values Indicate the Day of the Anomaly

30 Jan 30 Jan 30 Jan 31 Jan 31 Jan 01 Feb 31 Jan Anomaly Grid Number 1.0 1.5 2.0 2.5 34 35 26 27 28 19 20

Anomalies for Signal 2

Values Indicate the Day of the Anomaly

(d) Magnitude of the SLHF anomalies

(7)

8

100 200 300 400 500 600

01Jan01 12Jan01 23Jan01 03Feb01 14Feb01 25Feb01 Day

W_M^2

SLHF for 23.81N 65.62E Grid 34

−1 0 1 2 3 4 5

Quantile

Test for 23.81N 65.62E Grid 34

100

200

300

400

500

W_M^2

SLHF for 23.81N 67.5E Grid 35

−1 0 1 2 3 4

Quantile

Test for 23.81N 67.5E Grid 35

100

200

300

400

W_M^2

SLHF for 21.9N 65.62E Grid 26

−1

0

1

2

3

Quantile

Test for 21.9N 65.62E Grid 26

Fig. 3. Statistical test of the SLHF anomalies for the Gujarat earthquake of 26 January 2001

Fig. 3. Statistical test of the SLHF anomalies for the Gujarat earthquake of 26 January 2001.

Figure 2b shows the wavelet transformation for the epi-central region using SLHF data. The earthquake signal is not seen in this figure because the vertical scale is too small due the presence of the monsoon in the summer months. However, the SLHF signal is found to be anomalous 3 days prior and 5 days after the earthquake event, with a mag-nitude of about 3 times the average for this period of the year. Figure 2 shows the time and space continuity computed

using the defined grid path (c) and the magnitudes of the de-tected anomalies for the signals associated with the earth-quake event (d). During a one year period, only four signals are found which follow this precise geometrical path, out of which one occurs about 3 days prior and one about 5 days after the Gujarat earthquake. The grid path was chosen se-lecting grids over the ocean adjacent to the epicentral region. The highest anomalies occur in grid 35, which is one of the

(8)

9

100 200 300 400

W_M^2

SLHF for 21.9N 67.5E Grid 27

−1

0

1

2

3

Quantile

Test for 21.9N 67.5E Grid 27

100

200

300

400

W_M^2

SLHF for 21.9N 69.38E Grid 28

−1

0

1

2

3

Quantile

Test for 21.9N 69.38E Grid 28

50 100 150 200 250 300 350

W_M^2

SLHF for 20N 69.38E Grid 20

−1

0

1

2

Quantile

Test for 20N 69.38E Grid 20

Fig. 4. Statistical test of the SLHF anomalies for the Gujarat earthquake of 26 January 2001

Fig. 4. Statistical test of the SLHF anomalies for the Gujarat earthquake of 26 January 2001.

grids closest to the epicenter. The statistical test described in Sect. 2 was performed to determine which anomalies in the grid path are statistically significant, and the results are pre-sented in Figs. 3 and 4. Each figure is composed of pairs of images lined up horizontally. The left image shows the origi-nal sigorigi-nal for the months of January and February 2001, plus the 30 days average, one sigma and two sigmas. The right image shows the normalized signal for the same time period,

and the horizontal line indicating the minimum quantile re-quired to pass the test. If a peak is higher than the horizontal line, it is considered to be statistically significant. All grids used in the study show significant peaks 3 days before and 5 days after the earthquake.

Grids over the land were not chosen because their SLHF value for this time of the year is too small, around 10 W/m2. Figure 5 shows a sequence of the detected SLHF anomalies.

(9)

10 ‘

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Anomaly

Surface Latent Heat Flux Anomaly - 07 Jan 2001

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

62˚E 64˚E 66˚E 68˚E 70˚E 72˚E 74˚E 76˚E 78˚E 14˚N 16˚N 18˚N 20˚N 22˚N 24˚N 26˚N 28˚N 30˚N 32˚N

(a) SLHF anomaly for January 7 2001

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Anomaly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

(b) SLHF anomaly for January 8 2001

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Anomaly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

(c) SLHF anomaly for January 18 2001

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Anomaly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

(d) SLHF anomaly for January 24 2001

Fig. 5. Sequence of SLHF anomalies for the Gujarat earthquake of 26 January 2001

Fig. 5. Sequence of SLHF anomalies for the Gujarat earthquake of 26 January 2001.

(10)

12

3 4 5 12 20 28 29 37 45 53 31Aug -13 31Aug -13 31Aug -13 07Sep -6 09Sep -4 09Sep -4 09Sep -4 09Sep -4 09Sep -4 09Sep -4 10Sep -3 10Sep -3 10Sep -3 13Sep 0

88˚E 90˚E 92˚E 94˚E 96˚E 4˚N 6˚N 8˚N 10˚N 12˚N 14˚N 16˚N 18˚N 20˚N

(a) Map of the region of the first Andaman Earthquake. W_M^2 50 150 250 350

01 Jan 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 26 Nov 31 Dec Year 2002 Avg 1999..2003 1 sigma 2 sigma

Surface Latent Heat Flux : Year 2002

01 Jan 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 26 Nov 31 Dec

0 0.10.20.30.40.50.60.70.80.9 1

Color Scale

Event Date = 13 Sep 2002 Data analyzed =01 Jan 2002..31 Dec 2002 Wavelet Library = WaveLab Function = RWT (X, 10 , Sombrero , 1 , 10 ) Propagation Factor = 1 / 2

Location 6.67N 90E Grid 3

(b) Wavelet analysis for the first Andaman Earthquake.

Grid Number

01 Jan 31 Jan 02 Mar 01 Apr01 May 31 May 30 Jun 30 Jul 29 Aug28 Sep 28 Oct 27 Nov 31 Dec

3 4 5 12 20 28 29 37 45 53

Max Disc=1, Min Len=6, Pen Disc=0.5, Tol=1%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 Size of Anomaly Anomaly Discontinuity Earthquake Weak Signals Strong Signals

(c) Result of the time and space analysis for the first Andaman Earthquake.

(11)

13

4.0 18.8 33.6 48.4 63.2 78.0 92.8 107.6 122.4 137.2 152.0 SLHF Surface Latent Heat Flux - 02 Apr 2004

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

84˚E 86˚E 88˚E 90˚E 92˚E 94˚E 96˚E 98˚E 100˚E 102˚E 4˚N 6˚N 8˚N 10˚N 12˚N 14˚N 16˚N 18˚N 20˚N 22˚N

(a) Extent of the signal for the second Andaman Earthquake.

W_M^2

100

200

300

400

02 May 31 May 30 Jun 30 Jul 29 Aug 28 Sep 28 Oct 27 Nov 27 Dec 26 Jan 25 Feb 26 Mar 30 Apr Years 2003/2004 Avg 1998..2002 1 sigma 2 sigma

Surface Latent Heat Flux : Years 2003/2004

02 May 31 May 30 Jun 30 Jul 29 Aug 28 Sep 28 Oct 27 Nov 27 Dec 26 Jan 25 Feb 26 Mar 30 Apr

0 0.10.20.3 0.40.50.6 0.70.80.9 1

Color Scale

Event Date = 10 Apr 2004 Data analyzed =02 May 2003..30 Apr 2004 Wavelet Library = WaveLab Function = RWT (X, 10 , Sombrero , 1 , 10 ) Propagation Factor = 1 / 2

(b) Wavelet analysis for the second Andaman Earthquake.

Fig. 7. Results for the Andaman Earthquake of 10 April 2004.

spatial analysis of SLHF data to discriminate anomalies

associated with earthquakes from other anomalies. The

results presented in this paper show that prominent

SLHF anomalies are found prior to Indian coastal

earthquakes, which follow continuity both in space and

time. Such SLHF anomalous peaks were not found in

the case of the earthquake occurring far away from the

coast (Jabalpur earthquake). The results presented here

further validate the hypothesis that SLHF data contains

precursory information only for coastal earthquakes, and

set the basis for establishing a relationship between the

magnitude and the extent of the SLHF anomaly and the

magnitude of the earthquake.

5 Acknowledgments

This research was partially supported by NASA’s office

of Earth Science Enterprise under grant NAG12-01009,

NAG13-02054 and NAG13-03019, VAccess/MAGIC

projects

References

Cervone, G., Kafatos, M., Napoletani, D., and Singh, R. P., Wavelet maxima curves associated with two recent Greek earthquakes,

Natural Hazards and Earth System Sciences, 4, 359–374, 2004.

DeMets, C., Gordon, R., Argus, D., and Stein, S., Current plate motions, International Journal of Geophysics, 101, 425–478, 1990. Dey, S. and Singh, R. P., Surface latent heat flux as an earthquake

precursor, Natural Hazards and Earth System Science, 3, 749–755, 2003.

Liu, J., Chuo, Y., Shan, S., Tsai, Y., Chen, Y., Pulinets, S., and Yu, S., Pre-earthquake ionospheric anomalies registered by continuous gps tec measurements, Annales Geophysicae, 2, 1585–1593, 2004. Maekawa, S. and Hayakawa, M., A study on the association of

ionospheric perturbations with the earthquakes in Japan, as based on the JG2AS signal observation at Moshiri (Hokkaido) and at Kamchatka (Russia), in Geophysical Research Abstracts, vol. 6, European Geosciences Union, 2004.

Miller, R. G., Miller, R., and Brown, B. W., Beyond ANOVA: Basics of

Applied Statistics, Chapman & Hall/CRC, 1997.

Okada, Y., Mukai, S., and Singh, R., Changes in atmospheric aerosol parameters after Gujarat earthquake of January 26, 2001, Advances

in Space Research, 33, 254–258, 2004.

Ouzonov, D. and Freund, D., Mid-infrared emission prior to strong Fig. 7. Results for the Andaman Earthquake of 10 April 2004.

The higher anomalies found are over the ocean and closer to the epicentral region.

3.2 Andaman 1

The first Andaman earthquake occurred on 13 September 2002 at 22:28 UTC, at location 13.04 N 93.07 E, with a focal depth of 21 km and magnitude 6.5. This earthquake occurred on one of the most active seismic regions. The epicenter is ly-ing within few kilometers of the continental boundary where the Indian plate meets the Eurasian plate. Numerous earth-quakes and micro-earthearth-quakes have been recorded since the last three years, Fig. 6a shows the grids considered for the analysis of SLHF. Unlike Gujarat, Pondicherry and Jabalpur earthquakes, whose epicenter were located away from the continental boundary, the Andaman island lies exactly over the continental boundary. The result of the wavelet analysis also shows the anomalous SLHF peak and the wavelet max-ima curve associated with this earthquake event (Fig. 6b). The computation of the time and space continuity of the SLHF anomalies were chosen by selecting the grids lying ex-actly over the continental boundary. The maximum amount of stress occurs over the continental boundary, and this is the location where the largest anomaly was found. The results of the space and time continuity are shown in Fig. 6c. The

results show that throughout one year period, there is only a single signal which propagates through all the chosen grids, and this occurs about 4 days prior to the earthquake event. The other anomalous signals do not propagate through the continental boundary. Other significant SLHF anomalies are seen about 12 days prior to the earthquake event, but such anomalies are found to be localized only over the epicentral region, and are not found to propagate throughout the con-tinental boundaries. The other anomalies are not found to be associated with earthquake events due to their smaller ex-tent, and smaller statistical significance. The selection of the grids related to this earthquake is found to be easier than in the case of the Gujarat earthquake, because the epicenter is in the proximity of the continental boundary. We emphasize that knowledge of the relevant geological structure and tec-tonics of the region is crucial for the selection of grid paths. 3.3 Andaman 2

The second Andaman earthquake occurred on 10 April 2004 at 14:05 UTC, at location 13.04 N 93.25 E, with a focal depth of 33 km and magnitude 5.4. This earthquake occurred extremely close to the epicenter of the earthquake of 13 September 2002. Although of much lesser intensity, this earthquake provides a good test case to study the shape and

(12)

15

44 37 30 31 32 09Sep -16 09Sep -16 13Sep -12 14Sep -11 16Sep -9 18Sep -7 19Sep -6 19Sep -6 19Sep -6 19Sep -6 22Sep -3 22Sep -3 23Sep -2

76˚E 78˚E 80˚E 82˚E 84˚E 86˚E 88˚E 8˚N

10˚N 12˚N 14˚N 16˚N

(a) Map of the region of the first Pondicherry Earthquake.

W_M^2

50

100

200

300

01 Jan 30 Jan 01 Mar 31 Mar 30 Apr 30 May 29 Jun 29 Jul 28 Aug 27 Sep 27 Oct 26 Nov 31 Dec Year 2001 Avg 1999..2003 1 sigma 2 sigma

Surface Latent Heat Flux : Year 2001

0 0.10.2 0.30.4 0.5 0.60.7 0.80.9 1

Color Scale

Event Date = 25 Sep 2001 Data analyzed =01 Jan 2001..31 Dec 2001 Wavelet Library = WaveLab Function = RWT (X, 10 , Sombrero , 1 , 10 ) Propagation Factor = 1 / 2

(b) Wavelet analysis for the Pondicherry Earthquake.

Grid Number

01 Jan 31 Jan 02 Mar 01 Apr 01 May 31 May 30 Jun 30 Jul 29 Aug 28 Sep 28 Oct 27 Nov 31 Dec

44

37

30

31

32

Max Disc=1, Min Len=5, Pen Disc=0.5, Tol=50%

1 2 3 4 1 2 3 4 5 6 Size of Anomaly Anomaly Discontinuity Earthquake Weak Signals Strong Signals

(c) Result of the time and space analysis for the Pondicherry Earthquake.

Fig. 8. Results for the Pondicherry earthquake of 25 September 2001. Fig. 8. Results for the Pondicherry earthquake of 25 September 2001.

(13)

16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 02May -20 02May -20 02May -20 02May -20 02May -20 02May -20 03May -19 04May -18 04May -18 05May -17 05May -17 05May -17 05May -17 05May -17 05May -17 06May -16 06May -16 06May -16 07May -15 07May -15 07May -15 07May -15 08May -14 10May -12 10May -12 11May -11 11May -11 11May -11 11May -11

11May -11 12May -10 12May -10 12May -10 12May -10 13May -9 14May -8 15May -7 15May -7 16May -6 18May -4 19May -3 19May -3 19May -3 19May -3 20May -2 21May -1 21May -1

72˚E 74˚E 76˚E 78˚E 80˚E 82˚E 84˚E 86˚E 88˚E

14˚N 16˚N 18˚N 20˚N 22˚N 24˚N 26˚N 28˚N 30˚N 32˚N

Fig. 9. Map of the region of the Jabalpur Earthquake of 22 May 1997. The grids used in the study are marked, labeled, and annotated with the date

of the detected anomaly and the offset from the day of the earthquake. The epicenter is marked with a star.

Fig. 9. Map of the region of the Jabalpur Earthquake of 22 May 1997. The grids used in the study are marked, labeled, and annotated with the date of the detected anomaly and the offset from the day of the earthquake. The epicenter is marked with a star.

the extent of the anomaly for this region. Figure 7a shows the extent of the SLHF anomaly of this earthquake. The anomaly does not propagate through the continental boundary as ob-served for the other larger Andaman earthquake, but is more confined to the epicentral area. It is believed that the smaller extent of the SLHF anomaly is associated with the smaller in-tensity of the earthquake. The results of the wavelet analysis for the epicentral region shows the peak and the wavelet max-ima curve associated with this earthquake event (Fig. 7b). 3.4 Pondicherry

The Pondicherry earthquake occurred on 25 September 2001 at 14:54 UTC, at location 11.94 N 80.22 E, with a focal depth of 19 km and magnitude 5.4. This earthquake also occurred far away from the continental boundary, but close to the ocean. A single prominent SLHF anomaly was detected

about one week prior to the earthquake event that satisfies the space and time continuity requirements (Fig. 8a). Fig-ure 8b shows the results for the wavelet analysis for the epi-central region. The other signals which satisfy the space and time continuity are likely to be associated with aftershocks (Fig. 8c).

3.5 Jabalpur

The Jabalpur earthquake occurred on 22 May 1997 at 22.51 UTC, at location 23.08 N 80.04 E with focal depth of 35 km and magnitude 6.0. This earthquake occurred in a re-gion with very low seismicity, with the only other similar event recorded in 1938 in Satpura with magnitude 6.3. Anal-ysis of SLHF data have been carried out over a large area of about 2000 km2 (Fig. 9). Since this earthquake occurred far away from the coast, no SLHF anomaly was found to

(14)

G. Cervone et al.: Wavelet maxima curves of surface latent heat flux anomalies 99 be associated. It is believed that in the case of earthquakes

occurring away from the ocean, the land-ocean-atmosphere coupling is not significant enough to provide anomalous changes in atmospheric parameters, and thus SLHF data do not provide characteristic anomalies which can be used as a precursor (Dey and Singh, 2003).

4 Conclusion

This paper presents an application of the study of SLHF anomalies associated with five recent Indian earthquakes. The methodology is based on temporal and spatial analysis of SLHF data to discriminate anomalies associated with earth-quakes from other anomalies. The results presented in this paper show that prominent SLHF anomalies are found prior to Indian coastal earthquakes, which follow continuity both in space and time. Such SLHF anomalous peaks were not found in the case of the earthquake occurring far away from the coast (Jabalpur earthquake). The results presented here further validate the hypothesis that SLHF data contains pre-cursory information only for coastal earthquakes, and set the basis for establishing a relationship between the magnitude and the extent of the SLHF anomaly and the magnitude of the earthquake.

Acknowledgements. This research was partially supported by

NASA’s office of Earth Science Enterprise under grant

NAG12-01009, NAG13-02054 and NAG13-03019, VAccess/MAGIC

projects.

Edited by: M. Contadakis Reviewed by: two referees

References

Cervone, G., Kafatos, M., Napoletani, D., and Singh, R. P.: Wavelet maxima curves associated with two recent Greek earthquakes, Nat. Haz. Earth Sys. Sci., 4, 359–374, 2004,

SRef-ID: 1684-9981/nhess/2004-4-359.

DeMets, C., Gordon, R., Argus, D., and Stein, S.: Current plate motions, Intern. J. Geophys., 101, 425–478, 1990.

Dey, S. and Singh, R. P.: Surface latent heat flux as an earthquake precursor, Nat. Haz. Earth Sys. Sci., 3, 749–755, 2003,

SRef-ID: 1684-9981/nhess/2003-3-749.

Liu, J., Chuo, Y., Shan, S., Tsai, Y., Chen, Y., Pulinets, S., and Yu, S.: Pre-earthquake ionospheric anomalies registered by con-tinuous gps tec measurements, Ann. Geophys., 22, 1585–1593, 2004,

SRef-ID: 1432-0576/ag/2004-22-1585.

Maekawa, S. and Hayakawa, M.: A study on the association of ionospheric perturbations with the earthquakes in Japan, as based on the JG2AS signal observation at Moshiri (Hokkaido) and at Kamchatka (Russia), in: Geophys. Res. Abstr., Vol. 6, European Geosciences Union, 2004.

Miller, R. G., Miller, R., and Brown, B. W.: Beyond ANOVA: Ba-sics of Applied Statistics, Chapman & Hall/CRC, 1997. Okada, Y., Mukai, S., and Singh, R.: Changes in atmospheric

aerosol parameters after Gujarat earthquake of January 26, 2001, Adv. Space Res., 33, 254–258, 2004.

Ouzonov, D. and Freund, D.: Mid-infrared emission prior to strong earthquakes analyzed by remote sensing data, Adv. Space Res., 33, 268–273, 2004.

Pulinets, S. A. and Boyarchuk, K. A.: Ionospheric Precursors of Earthquakes, Springer Verlag, Berlin, 2004.

Qiang, Z.: Thermal infrared anomoly precursor of impending earth-quakes, Pure Appl. Geophys., 149, 159–17, 1997.

Singh, R. P., Sahoo, A. K., Bhoi, S., Kumar, M. G., and Bhuiyan, C. S.: Ground deformation of the Gujarat earthquake of January 26, 2001, J. Geological Society of India, 58, 209–214, 2001. Singh, R. P., Dey, S., Singh, V. P., Cervone, G., Sarkar, S., and

Kafatos, M.: Prediction of coastal earthquakes using surface la-tent heat flux retrieved from satellite data, in: Proceedings of the World Congress on Natural Disaster Mitigation, Vol. 2, pp. 129– 134, World Federation of Engineering Organisation, 2004. Tramutoli, V., Bello, G. D., Pergola, N., and Piscitelli, S.: Robust

satellite techniques for remote sensing of seismically active ar-eas, Annali di Geofisica, 44, 295–312, 2001.

Tronin, A., Hayakawa, M., and Molchanov, O.: Thermal IR satellite data application for earthquake research in Japan and China, J. Geodynamics, 33, 519–534, 2002.