multitaper coherence method for appraising the elastic thickness of the indonesian active...

8
Multitaper coherence method for appraising the elastic thickness of the Indonesian active continental margin Rajesh R. Nair * , Tanmay K. Maji, Tannishta Maiti, Suresh Ch. Kandpal, R.T. Ratheesh Kumar, Sharat Shekhar Indian Institute of Technology, Kharagpur 721 302, India article info Article history: Received 7 May 2009 Received in revised form 29 December 2009 Accepted 15 June 2010 Keywords: Indonesian margin Elastic thickness Hermite tapers abstract We estimate the mechanical strength of the Indonesian active continental margin. Values of effective elastic thickness (Te) were obtained for azimuthally averaged coherence measurements c 2 ðjkbetween Bouguer gravity and topography using a multitaper method with Hermite tapers. The T e estimates reveal uniform low strength (4–7 km) beneath the entire Indonesian continental margin. The 2D coherence function c 2 ðjkestimated with the help of MTM reveals varying anisotropy in the Indonesian margin and illustrates how the results correlate with the maximum horizontal stress orientations (SH max ), heat flow measurements, sediment thickness and shear wave speed measurements. The sediment thickness and heat flow in these regions do not appear to influence the strength of the Indonesian margin. We have inferred that low effective elastic thickness could have been fossilized on a much younger oceanic plate at the time of loading. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction The effective elastic thickness (T e ) of the lithosphere is defined as the thickness of an equivalent elastic plate, which would pro- duce the same deflection under the known tectonic loading struc- ture. Extensive studies on effective elastic thickness have been carried out in various parts of the world. However, much contro- versy still persists in the elastic thickness (T e ) characterization at the continental margins. Different techniques have been used to determine isostatic coherency in T e estimation which has resulted in major variations in the T e value and its relationship with the seismogenic thickness. A recent estimate of T e in the Irish–Atlantic margin (Daly et al., 2004) from wavelet and multitaper estimation yielded values of 6–18 km. Tiwari et al. (2003) conducted studies using spectral admittance in the Ninety-East Ridge adjacent to the Indonesian continental margin. They deduced that the north- ern (0–10°N) and the southern (20–30°S) parts of the ridge are flexurally compensated with an effective elastic thickness >15 km, whereas the central part (0–20°S) is locally compensated. T e values of 4 km (Cochran, 1979) and 10 km has been estimated in the East Pacific high and Mid-Atlantic Ridge respectively using admittance and mirrored periodogram method. Madsen et al. (1984) in the East Pacific Rise (0.7 km) have observed further, lower T e values. The variations in the method of T e have thus shown varying relationships between elastic thickness and seism- ogenic thickness. The Indonesian continental margin (Fig. 1) is a prototype of a complex subduction zone lying at the boundary of the Indo-Aus- tralian plate (IAP) (Schoffel and Das, 1999). The Sumatran Fault Zone (SFZ) is characterized by the presence of an active volcanic arc and several fore-arc structures. Normal subduction is observed beneath Java with associated occurrence of fore-arc basins. How- ever, the Sumatran region and other regions in the further north exhibit oblique subduction due to motion parallel to the arc accompanied by the dextral strike-slip displacement along the Sumatran Fault Zone (Newcomb and McCann, 1987). The increase in dip and depth of Wadati–Benioff Zone reflects the change in age and variation in lithospheric thickness going from west (Sumatra) to east (Java) (Newcomb and McCann, 1987). In the historical re- cords, great earthquakes have occurred more frequently in Suma- tra where the younger, more shallowly dipping seafloor enters the trench (Newcomb and McCann, 1987). The great 2004 Suma- tran earthquake had a focal depth of 30 km. The region is predom- inantly a thrust faulting one on a shallow (8°) dipping plane which has a strike value of 329° (Lay et al., 2005). Investigations from the Harvard CMT catalog for the Indonesian continental margin shows an average seismogenic thickness 40 km. The flexural rigidity of the lithosphere is directly estimated by the response of the plate to loading; wherever the load is known. An analysis of the correlation between the topography and the gravity data, with spectral methods, provides an alternative meth- od to T e estimation (Audet and Mareschal, 2004a,b). Spectral tech- niques can quantify the relationship between gravity and 1367-9120/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jseaes.2010.06.009 * Corresponding author. Present address: Indian Institute of Technology, Madras. Tel.: +91 03222 283362; fax: +91 03222 255303. E-mail address: [email protected] (R.R. Nair). Journal of Asian Earth Sciences 40 (2011) 326–333 Contents lists available at ScienceDirect Journal of Asian Earth Sciences journal homepage: www.elsevier.com/locate/jseaes

Upload: rajesh-r-nair

Post on 27-Oct-2016

215 views

Category:

Documents


1 download

TRANSCRIPT

Journal of Asian Earth Sciences 40 (2011) 326–333

Contents lists available at ScienceDirect

Journal of Asian Earth Sciences

journal homepage: www.elsevier .com/locate / jseaes

Multitaper coherence method for appraising the elastic thicknessof the Indonesian active continental margin

Rajesh R. Nair *, Tanmay K. Maji, Tannishta Maiti, Suresh Ch. Kandpal, R.T. Ratheesh Kumar, Sharat ShekharIndian Institute of Technology, Kharagpur 721 302, India

a r t i c l e i n f o a b s t r a c t

Article history:Received 7 May 2009Received in revised form 29 December 2009Accepted 15 June 2010

Keywords:Indonesian marginElastic thicknessHermite tapers

1367-9120/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.jseaes.2010.06.009

* Corresponding author. Present address: Indian InsTel.: +91 03222 283362; fax: +91 03222 255303.

E-mail address: [email protected] (R.R.

We estimate the mechanical strength of the Indonesian active continental margin. Values of effectiveelastic thickness (Te) were obtained for azimuthally averaged coherence measurements c2ðjkjÞ betweenBouguer gravity and topography using a multitaper method with Hermite tapers. The Te estimates revealuniform low strength (4–7 km) beneath the entire Indonesian continental margin. The 2D coherencefunction c2ðjkjÞ estimated with the help of MTM reveals varying anisotropy in the Indonesian marginand illustrates how the results correlate with the maximum horizontal stress orientations (SHmax), heatflow measurements, sediment thickness and shear wave speed measurements. The sediment thicknessand heat flow in these regions do not appear to influence the strength of the Indonesian margin. We haveinferred that low effective elastic thickness could have been fossilized on a much younger oceanic plate atthe time of loading.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The effective elastic thickness (Te) of the lithosphere is definedas the thickness of an equivalent elastic plate, which would pro-duce the same deflection under the known tectonic loading struc-ture. Extensive studies on effective elastic thickness have beencarried out in various parts of the world. However, much contro-versy still persists in the elastic thickness (Te) characterization atthe continental margins. Different techniques have been used todetermine isostatic coherency in Te estimation which has resultedin major variations in the Te value and its relationship with theseismogenic thickness. A recent estimate of Te in the Irish–Atlanticmargin (Daly et al., 2004) from wavelet and multitaper estimationyielded values of 6–18 km. Tiwari et al. (2003) conducted studiesusing spectral admittance in the Ninety-East Ridge adjacent tothe Indonesian continental margin. They deduced that the north-ern (0–10�N) and the southern (20–30�S) parts of the ridge areflexurally compensated with an effective elastic thickness>15 km, whereas the central part (0–20�S) is locally compensated.Te values of 4 km (Cochran, 1979) and 10 km has been estimated inthe East Pacific high and Mid-Atlantic Ridge respectively usingadmittance and mirrored periodogram method. Madsen et al.(1984) in the East Pacific Rise (�0.7 km) have observed further,lower Te values. The variations in the method of Te have thus

ll rights reserved.

titute of Technology, Madras.

Nair).

shown varying relationships between elastic thickness and seism-ogenic thickness.

The Indonesian continental margin (Fig. 1) is a prototype of acomplex subduction zone lying at the boundary of the Indo-Aus-tralian plate (IAP) (Schoffel and Das, 1999). The Sumatran FaultZone (SFZ) is characterized by the presence of an active volcanicarc and several fore-arc structures. Normal subduction is observedbeneath Java with associated occurrence of fore-arc basins. How-ever, the Sumatran region and other regions in the further northexhibit oblique subduction due to motion parallel to the arcaccompanied by the dextral strike-slip displacement along theSumatran Fault Zone (Newcomb and McCann, 1987). The increasein dip and depth of Wadati–Benioff Zone reflects the change in ageand variation in lithospheric thickness going from west (Sumatra)to east (Java) (Newcomb and McCann, 1987). In the historical re-cords, great earthquakes have occurred more frequently in Suma-tra where the younger, more shallowly dipping seafloor entersthe trench (Newcomb and McCann, 1987). The great 2004 Suma-tran earthquake had a focal depth of 30 km. The region is predom-inantly a thrust faulting one on a shallow (8�) dipping plane whichhas a strike value of 329� (Lay et al., 2005). Investigations from theHarvard CMT catalog for the Indonesian continental margin showsan average seismogenic thickness �40 km.

The flexural rigidity of the lithosphere is directly estimated bythe response of the plate to loading; wherever the load is known.An analysis of the correlation between the topography and thegravity data, with spectral methods, provides an alternative meth-od to Te estimation (Audet and Mareschal, 2004a,b). Spectral tech-niques can quantify the relationship between gravity and

Fig. 1. Equivalent topography (m) map of the Indonesian continental marginshowing major tectonic features. Seismicity (M P 4),rSHmax after Zoback et al. (1989), averaged maximumstress(greenarrow)afterZoback(1992),ridgetorque(redarrow)afterRichardson(1992)andabsolutevelocity(bluearrow)afterRichardson(1992)isshown.RidgetorqueisthetorqueduetothetwoopposingforcesconsistingofRidgepushandslabpull.Absolutevelocityisthemotionofalithosphericplatewithrespecttoafixedframeofreference.Averagemaximumstressisthe stress due to which fracturing along a geological fault takes place. IAP-Indo-Australian plate, EP-Eurasian plate, BMT-Burma microplate, AS-Andaman Sea, JS-Java Sea, SMT-SumatraSFZ-Sumatran Fault Zone, MI-Mentawai Islands, MFZ-Mentawai Fault Zone, ST-Sunda Trench. W1–W8 are the eight windows of size 550� 550 km2 used for the estimation of elasticthickness.B1–B3arethreebiggerwindowsofsize880� 880 km2.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.

R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333 327

,

)

328 R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333

topography above the surface and subsurface load as a function ofits wavelength. The recovery of effective elastic thickness (Te) isfrom a wavelength at which the coherence drops to half of itslong-wavelength value defined as the transitional coherence wave-length (Audet and Mareschal, 2004a,b; Rajesh et al., 2003).

Our paper is based on a novel approach of computing a robustcoherence method based on multitaper spectral analysis on over-lapping windows of equal size for Te estimation. The admittancemethod used earlier suffers from some obvious shortcomings suchas the free-air admittance will be a fiasco if the topography is re-moved by erosion. Banks et al. (2001) renders Bouguer coherencea more optimistic method than free-air admittance. The free-airgravity anomaly is perturbed more by leakage effects because ofthe relatively low power of the free-air gravity anomaly at longwavelengths. According to Forsyth (1985), a flexural isostatic modelmust include both the surface as well as subsurface loads in order toaccurately estimate Te. Bouguer gravity, which is less sensitive tothe ratio of subsurface to surface loading, is hence used. Pérez-Gussinyé et al. (2004) have argued that admittance method is moreprone to leakage at wavelengths where isostatic compensation oc-curs, because free-air gravity has low power at such wavelengths.

The main objective of this study is to reappraise the effectiveelastic thickness value of the Indonesian active continental margin.We have applied the spectral method of Coherence estimationusing orthonormalized Hermite functions as tapers on real datasets of continental margins. Our interpretation will take into ac-count, factors such as sediment thickness, surface heat flow andseismic wave speed anomaly.

2. Data

The GEBCO digital Atlas given at 1 min spacing (NOAA, 2003) isused for bathymetry of the region, which is scaled for the presentanalysis. The bathymetry model derived from satellite altimetryfree-air anomaly such as 2 min Mercator projected grid model (San-dwell and Smith, 1997) is not recommended for our analysis. San-dwell and Smith grid relies on machine interpolation ofconventional soundings to constrain wavelengths greater thanabout 160 km, because at longer wavelengths, the correlation ofgravity with bathymetry may be reduced by isostatic compensation(Marks and Smith, 2006). As the Sandwell and Smith (1997)bathymetry data already takes into account the local compensation,so we prefer GEBCO 1-min bathymetry which is prepared frombathymetric contours of the world’s oceans. According to Markand Smith, GEBCO has limited vertical resolution (i.e. abyssal hillsof 300 m will not be detected) as compared to Sandwell and Smithgrid. The Sandwell and Smith grid is also on a geographical grid,which is sometimes convenient. However, GEBCO grid is preferablefor shallow water and for displaying 500 m bathymetric contours.The Bouguer gravity anomaly, which is for the present analysis, isderived from DNSC07 1 min free-air anomaly map by applying Par-ker’s method. Parker’s method describes the use of fast Fouriertransforms (fft) for rapid calculation of gravity or magnetic anoma-lies in order to save computation time. In the earlier methods, thecrustal models were divided into a set of prisms or rectangularblocks and the resultant effect of the crustal models were obtainedby taking the sum of the contribution due to each of these blocks.Parker’s method is useful particularly when the topographic effectis not negligible as compared to the observation height, e.g., in caseof mid oceanic ridges or mountainous terrains. However this meth-od is not applicable when the data observed lies below the averageheight of topographic undulations (Parker, 1972).

The overall accuracy of the DNSC07 model is 2.78 mGal (Ander-son et al., 2008) when compared to KMS 02 (Anderson and Knud-sen, 1998), which is 4.99 mGal. Tiwari et al. (2003) used KMS 02

data, and computed elastic thickness over 90� East Ridge, adjacentto Indonesian margin. The equivalent topography is then com-puted from the GEBCO bathymetry data. The equivalent topogra-phy is the height or depth that the crust will assume in theabsence of ice or water present and under isostatic conditions gi-ven by hðxÞ ¼ qc � qw=qcð Þ � d.

Here qc, qw and d are the mean crustal density, density of sea-water and bathymetry, respectively. We restrict our analysis wellwithin the oceans, to avoid intricacies involved in the generationof satellite measurements from continental regions.

Audet and Mareschal (2004a,b) and Audet et al. (2007) have ar-gued the choice of window size being an indispensable limitationto Te estimations. High values of Te require a wide window to cor-rectly resolve large transitional wavelengths. If spatial Te variationsoccur at distances shorter than the width of actual window thensuch an effect may average out the elastic properties whereassmall window dimensions give high spatial resolution without suf-ficient resolution of long flexural wavelength. Therefore, we needto consider a tradeoff (Pérez-Gussinyé et al., 2004) between win-dow size and frequency resolution. To reduce such erroneous re-sults, Audet et al. (2007) have used a combination of windowsizes of 300 � 300 km2, 500 � 500 km2 and 800 � 800 km2 withan overlap of 70% between the adjacent windows for Te. Te esti-mated in the Canadian Shield by Audet and Mareschal (2004a,b)using maximum entropy method is as high as 140 km with a lowerlimit of 30 km. We have also used eight windows of size550 � 550 km2 for Te estimation with an overlap of 40–70%(Fig. 1) along with three bigger windows of size 880 � 880 km2

(Fig. 4). However, our tests with varying window sizes550 � 550 km2 and 880 � 880 km2 do not affect the Te results toany appreciable extent. However, our estimation is based on over-lapping windows so that small spatial changes in elastic thicknessare not evaded out completely. The flexural rigidity is related to theelastic thickness by the following equation

D ¼ ET3e

12ð1� r2Þ

where E is Young’s modulus, Te is the elastic thickness and r is Pois-son’s ratio.

Elastic model parameters used in the analysis are: r (Poisson’sratio) = 0.25, E (Young’s modulus) = 1011 N/m2, qc (mean crustaldensity) = 2.85 kg m�3, qm (mean mantle density) = 3.35 kg m�3,qw (density of sea water) = 1.03 kg m�3, tc (average crustal thick-ness) = 16 km (W1 and W2) and 20 km (for all other windows)(Curray et al., 1982; Grevemeyer et al., 2001).

3. Estimating elastic thickness

3.1. Coherence

The isostatic coherence response is the isostatic response of thelithosphere estimated from the coherence between Bouguer grav-ity and topography. The coherence method for estimating Te isbased on the assumption that there is no correlation between sur-face and internal loads. At long wavelengths, the Bouguer anomalyis coherent with the topography as the surface (or internal) loadsare fully compensated. For shorter wavelengths, the Bougueranomaly and the topography are incoherent because the loadsare supported by the strength of the lithosphere. The transitionwavelength from low to high coherence is used to estimate therigidity of the plate (Audet and Mareschal, 2004a,b). For two non-stationary random processes {X} (gravity) and {Y} (topography),defined on r in the spatial domain and on k in the Fourier domain,the coherence-square function relating both fields, c2

XY is definedas the ratio of their cross-spectral density, SXY, normalized by the

k (rad/km)

W 8 Gravity Equi. Topo

k (rad/km)

W 4 Gravity Equi. Topo

0 0.02 0.04 0.060 0.02 0.04 0.060 0.02 0.04 0.06k (rad/km)

2

3

4

5

6

log

(Pow

er)

W 1 Gravity Equi. Topo

k (km-1)

0

0.4

0.8

γ2(|k

|)

W1

10-2 10-1

------ Te= 5 km------ Te= 7 km------ Te=10 km

R=2 R=3 R=4

k (km-1)

W4

10-2 10-1

------ Te= 0 km------ Te= 5 km------ Te=10 km

R=2 R=3 R=4

k (km-1)

W8

10-2 10-1

------ Te= 0 km------ Te= 4 km------ Te=10 km

R=2 R=3 R=4

Fig. 2. Coherence between Bouguer gravity and topography for three windows shown in Fig. 1. From top to bottom: the radially averaged power spectrum of gravity andtopography (Spector and Grant, 1970) for these blocks, coherence-square function for a set of values of R (calculated) and Te (predicted).

R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333 329

individual power spectral densities, SXX and SYY (Bendat and Piersol,2000)

c2XYðr; kÞ ¼

SXY ðr; kÞj j2

SXXðr; kÞSYYðr; kÞ¼ jEf~Xðr; kÞ~Y�ðr; kÞgj2

Ef~Xðr; kÞ~X�ðr; kÞgEf~Yðr; kÞ~Y�ðr; kÞgð1Þ

Here E denotes an expectation operator, tildes refer to the Fourier-transformed signal, and the asterisk refers to the complex conju-gate. The periodogram ~X~X� is a direct spectral estimator of X,although not a particularly accurate one.

In Forsyth’s method (Forsyth, 1985), the theoretical coherenceis considered to be a function of wavelength and flexural rigidity,assuming that surface and subsurface loadings are statisticallyindependent. If surface and subsurface loading have equal ampli-tude (i.e. when F = 1) the theoretical coherence of the mantle Bou-guer anomaly with bathymetry can be computed from,

c2 ¼ ð1þ ðF=nÞ2uÞ2

ð1þ ðF=nÞ2Þð1þ F2u2Þð2Þ

where u = 1 + D(2pk)4/(qc � qw)g, n = 1 + D(2pk)4/(qm � qw)g and kis the wave number.

3.2. Spatiospectral localization properties

3.2.1. Multitaper spectral analysis methodThe multitaper method is a nonparametric technique for opti-

mal spectral estimation. This method consists of calculating thespectra with multiple orthogonal windows used as data tapersand averaging over different (approximately) independent subsetsof the data (Percival and Walden, 1993; Slepian, 1978). The averag-ing of the orthogonal tapers minimizes the spectral leakage as wellas reduces the variance of the estimate thereby leading to optimalspectral estimation.

The multitaper cross-spectral estimator between two channels land m is the average of a number of direct cross spectral estimates

between the two channels l and m. Mathematically, this can berepresented as:

Slmk ðf Þ ¼

1K

XK�1

k¼0

Slmk ðf Þ

where Slmk ¼ 1

NDt Jlkðf Þ

h i� Jm

k ðf Þ� �

is the kth direct cross-spectral esti-mator between channel l and m. Dt and N are the sampling intervaland the number of samples, respectively.

3.2.2. Hermite tapersWe have used orthonormal Hermite functions as data tapers in-

stead of Slepian functions (Simons et al., 2000, 2003, 2006). Theorthonormal Hermite functions are the eigenfunctions of operatorsconcentrating in a disc shaped time frequency domaint2 þ ð2pf Þ2 6 R2.

Hermite functions hj(t) can be represented as the Hermite poly-nomials Hj modulated by a Gaussian function

hjðtÞ ¼HjðtÞe�t2=2

p1=4

ffiffiffiffiffiffiffi2jj!

q ð3Þ

The Hermite polynomials can be calculated using the recurrencerelation

Hnþ1ðtÞ ¼ 2tHnðtÞ � 2nHn�1ðtÞ ð4Þ

starting from H0(t) = 1 and H1(t) = 2t.One advantage of using the Hermite functions as data tapers is

that they are computationally very fast. This is due to the fact thatthe eigenfunctions of the concentration operator are independentof R whereas the eigenvalues are functions of R.

The radius of concentration R defines the radius of the circulararea of signal space, which contains an appreciable amount ofthe signal. A time signal X(t) is considered to be concentrated with-in an interval [�T, T] if the fraction of energy concentrated withinthe interval is close to unity, i.e. if kWXk2

kXk2 tends to unity.See Simons et al. (2003) for a description of the MTM applied to

flexural studies. Fig. 2 shows calculated coherence value for R = 2–

0

0.4

0.8

γ2(|k

|)λ (km)

1001000

7 km

W1

λ (km)

1001000

6 km

W2

0

0.4

0.8

γ2(|k

|)

6 km

W3

5 km

W4

k (km-1)

0

0.4

0.8

γ2(| k

|)

5 km

W7

10-2 10-1

0

4

8

12

RM

SE

W1 W2

0

4

8

12

RM

SE

W3

W8

Te (km)0 10 20 300 10 20 30

0

4

8

12

RM

SE

W7

Te (km)

W4

0

4

8

12

RM

SE

W5 W6

k (km-1)

4 km

W8

10-2 10-1

0

0.4

0.8

γ2(| k

|)

4 km

W5

6 km

W6

Fig. 3. Left: Te inversions based on multitaper coherence analysis for data windows. Predicted coherence is indicated by solid lines and observed by symbols. Right: modelmisfit in the inversion of Te for windows. Plotted is a root mean square error (RMSE) calculation based on the misfit between observed and predicted.

330 R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333

4 and predicted coherence for a set of Te values. The number of ta-pers used for each value of R is equal to R2 in each dimension. Wehave noticed that coherence calculated for R = 2 and 4 lie on eitherside of that of for R = 3. The error bars in coherence curve are smal-ler for R = 3 compared to R = 2 and 4. The relative magnitudes of er-ror bars in coherence curve vary from 0.001 to 0.08. This analysismakes us to take R = 3 and nine tapers in each dimension (total81 different Hermite tapers) for our further coherence study.

We applied the Hermite multispectrogram method to charac-terize the coherence between the topography and Bouguer data.Window sizes of 550 � 550 km2 are extracted from the full extentof the data. The data was analyzed using 81 different Hermite ta-pers (the outer products of nine tapers in each dimension) and aconcentration region of R = 3.

4. Results

The distribution of Te estimates along with the error plot isshown in Fig. 3. We have compared the observed and predictedcoherence as a function of wavelength. The red dots symbolize cal-culated coherence and the solid lines are the best-fit predictedcoherence. The calculation of root mean square error (RMSE) isbased on the misfit between observed and predicted coherencenormalized by the standard deviation of the coherence measure-ment. A best fit is considered whenever RMSE drops below the va-lue of minimum with the lowest value of Te. All the Te values aremade at minimum estimates. The misfit curves show distinct min-

ima for Te in all the windows. The Te inversion in the spectral do-main for the eight windows W1–W8 yield values in the range of�4–7 km for MTM with Hermite tapers. In the spectral domain,the coherence function c2ðjkjÞ is characterized in the transitionalwavelength range. Low consistent values of Te have been observedfor all the eight windows in the Indonesian continental margin.Tests, with window size as large as 880 � 880 km2 also give similarlow Te estimates (Fig. 4). This low range is in agreement with sim-ilar low Te values obtained in the Irish–Atlantic margin (Daly et al.,2004). Further, our Te estimate is contained well within the seism-ogenic thickness of 40 km in this region.

The flexural anisotropy derived from isostatic coherence esti-mation and its comparison with observed stress indicators (themaximum horizontal stress SHmax) provides one basis for evaluat-ing the lithospheric weakness direction.

All the estimated flexural anisotropy results are shown in Fig. 1aand some examples of determination of isotropic coherence areshown in Fig. 6.

In the region adjacent to 90�E Ridge near to 0� latitude SHmax isoriented approximately in the east–west direction while the flex-ural anisotropy is oriented at an angle �45� with the horizontaland perpendicular to the tectonic faults (Fig. 1a).

In the region with latitude around 50�N and longitude around97�E, SHmax is oriented approximately in the NW–SE directionand the flexural anisotropy in the adjacent region correspondingto the windows W3, W4, B1, and B2 is at angles 90�, 45�, 48� and90�, respectively.

k (km-1)

0

0.4

0.8

γ2(| k

|)

λ (km)

10-2 10-1

6 km

B1

k (km-1)

λ (km)

1001000

10-2 10-1

7 km

B2

1001000

B2

Te (km)

0

4

8

12

RM

SE

B1

Te (km)0 10 20 300 10 20 30 0 10 20 30

B3

Te (km)

k (km-1)

λ (km)

10-2 10-1

4 km

B3

1001000

Fig. 4. Te inversions based on multitaper coherence analysis for bigger windows of size 880 � 880 km2. Predicted coherence is indicated by solid lines and observed bysymbols. Bigger windows (red) are shown in Fig. 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. (a) Sediment thickness (m) of Indonesian continental margin after Divins (2008), (b) heat flow (mW m�2) ranges in the Indonesian continental margin after Pollacket al. (1993) and (c) upper mantle shear wave perturbation of the average velocity (km/s) relative to the average global velocity at the depth of (Moho + 20) km after Barminet al. (2001), Ritzwoller et al. (2002) and Shapiro and Ritzwoller (2002).

R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333 331

Again another SHmax observation at latitude 80�N and longitude90�E shows a NE–SW direction which matches with the flexuralanisotropy direction estimated within the window B1 coveringthe adjacent region.

Finally the SHmax observed at latitude 13�N longitude 96�E indi-cates a NW–SE direction whereas the flexural anisotropy estimatedfor window B3 covering the adjacent region shows a N–S direction.

5. Discussion

Te is a wavelength dependent parameter and hence its accuracyis controlled basically by the spatiospectral resolution of the tech-nique applied. In our present discussion, we have applied the win-

dow based method and have used the Hermite tapers for differentwindow sizes. As already discussed, we use the Te values computedwith R = 3 to be a plausible case in showing elastic thickness trendin the subduction zone. Fig. 2 shows typical examples of the deter-mination of Te in different parts along the subduction region. Thedeterministic window illustrates that it is difficult to determine areliable Te with different radii of concentrations. Thus, we empha-size on the relative variation of Te values rather than absolute ones.The windows near to the Sumatran Fault Zone and Mentawai FaultZone shows a Te value of 6–7 km and is consistently decreasing to-wards south east up to a value of 4 km. The Sumatran Fault Zone ischaracterized by fore-arc and active volcanoes but with a low heatflow of 10–50 mW m�2. The entire subduction zone plate strengthis nominal, varying from 4 to 7 km. To support our findings further,

Fig. 6. 2D coherence for the windows B1, B2, B3 and W1. The white dotted linesindicate the direction of the coherence anisotropy.

Table 1List of sediment thickness, heat flow and estimated effective elastic thickness valuesfor analyzed windows respectively.

Windowno.

Sediment thickness(km)

Heat flow(mW m�2)

Elastic thickness(km)

W1 0.5–3.5 17–80 7W2 0.1–0.2 20–155 6W3 0.1–0.2 37–155 6W4 0.1–0.6 29–167 5W5 0.1–0.6 27–164 4W6 0.1–0.8 24–199 6W7 0.6–0.8 20–78 5W8 0.3–0.7 20–78 4B1 0–1.0 12–167 6B2 0–0.6 27–167 7B3 0–0.8 24–199 4

332 R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333

elastic thickness computed with larger windows of size880 � 880 km2 show low Te values of 4–7 km.

Results from our multitaper spectral analysis approach in theIndonesian continental margin reveal uniformly low strength. Ear-lier works by Karner (1991) attributes the presence of low Te valuedue to thermal blanketing or load of large sedimentary material.This theory is further supported by the Lavier and Steckler(1997) who argued that the presence of huge sedimentary loadinevitably reduces the thermal conductivity resulting in weaklithospheric strength. Sediment layer of thickness 3–5 km, effec-tively reduces Te value by 25%. The vast deposition in the Bengalsubmarine fans that lie adjacent to the Indonesian continentalmargin is likely to contribute to sedimentation in this region otherthan the tectonic plate marginal deposits. There is a thick pile ofsediment, extending from 20�N as far as 7�S with a maximumwidth of 1000 km at 15�N. The sediment thickness gradually de-creases from 2 km (Curray, 1991) to few 100 m in the Southernpart. Kopp (2002), by means of seismic reflection studies correlateda high sedimentation rate (>0.4 km/Ma) along the Indonesian mar-gin. Our present tectonic system has Te value of 4–7 km and lowersediment thickness as shown in Fig. 5a, which may not be sup-ported by the sediment decoupling model as it cannot explainTe < 15 km (Lowry and Zhong, 2003). The Indonesian region in-cludes Island arcs and collision zones, with a record of Mesozoicand Cenozoic subduction of Indian and Pacific lithosphere as wellas history of continental collision at its margins (Hamilton, 1979).

Te may be attributed to the mechanism of mantle decoupling(Courtney and Beaumont, 1983) due to high mantle heat flow withsubsequent reduction in the competency of mantle. In case of verylow effective viscosity of the lower crust, a flow occurs in the lowercrust and the topographic compensation of load is restricted to thelower crust. However, such low viscosities are seen in young lith-osphere and volcanic regions (Burov and Diament, 1995). Fig. 5cshows the upper mantle velocity in the Indonesian continentalmargin, which shows high positive upper mantle velocities, and

rules out the possibility of mantle decoupling, reducing themechanical strength. Subduction beneath the Indonesia resumedat 45 Ma, after subduction ceased in the Late Cretaceous as Austra-lia began to move rapidly, northwards. As a result of subduction,high heat flow occurs in the region beyond the vicinity of volcanicarcs. In our present study, we are unable to find any direct relationbetween heat flow data and Te values, which are uniformly low.

The sediment thickness, heat flow values and the Te values cor-responding to the different windows are provided in a tabular formin Table 1. Our Te estimates contradict the explanation that heatflow and elastic thickness are inversely correlated. We have ob-tained constant Te values for varying heat flow values within thedifferent windows. Hence, we infer that Te may not be attributedto heat flow alone (Table 1).

Flexural anisotropy is often used to compare with seismicanisotropy (Audet and Mareschal, 2004a,b; Simons et al., 2003; Ra-jesh et al., 2003). In the absence of seismic anisotropy in the Indo-nesian margin, we have used flexural anisotropy to explain thefindings. If deformation has operated in a vertically-coherentway, the seismic anisotropy direction should either be parallel orperpendicular to the surface geological trends (Silver and Chan,1988). Our observations from flexural anisotropy as well as thestress indicators reveal a combination of coherent and incoherentdeformation indicating the Indonesian margin to be truly aniso-tropic (Figs. 1a and 6). These results highlight the existence of sev-eral directions of weakness of the plate resulting in predominantlowering of the elastic strength of the plate. The varying state ofstress and flexural anisotropy might be a consequence of the sub-duction zone where the dipping cold slab exists beneath it.

Tiwari et al. (2003) have discussed that large values of subsur-face loading can cause buoyancy uplift and would result in a largetopography (�1 km compared to other parts) as observed in thecentral part of Ninety-East Ridge. A low value of EET, large topog-raphy, and large subsurface loading can be explained by presenceof a hot spot. Tiwari et al. (2003) have also reported that the centralpart of the Ridge is locally compensated (Te = 0 km) which is nearto our study area. We have demonstrated low rigidity of the plateat the present time. Low effective elastic thickness could have been‘‘frozen in” on a much younger oceanic plate at the time of loadingsimilar to that of the central part of Ninety-East Ridge.

6. Conclusions

The Bouguer gravity and topography coherence analysis of datawindows over the Indonesian continental margin provides elasticthickness in the range 4–7 km using MTM. The Te values obtainedrepresent the lower limit from free-air admittance computations.Results from multitaper coherence estimates are contained withinthe seismogenic thickness, reflecting the nature of predominantinfluence of the tectonic regime. The Te results obtained are found

R.R. Nair et al. / Journal of Asian Earth Sciences 40 (2011) 326–333 333

to be consistent with similar lowering of Te coinciding with thelower limit of values 6–18 km, obtained in the Irish margin (Dalyet al., 2004), using the multitaper technique and wavelet ap-proaches. The sediment thickness and heat in these regions arenot sufficient enough to cause significant lowering of Te as deducedfrom our estimated values. We have demonstrated low rigidity ofthe Indonesian active continental margin at the present time. Weconclude that the low effective elastic thickness could have been‘‘frozen in” on a much younger oceanic plate at the time of loading.

Acknowledgments

The authors sincerely thank two anonymous reviewers for theiruseful suggestions that significantly improved the manuscript.They also thank INCOIS and DST fast track project for necessaryfunding.

References

Anderson, O., Knudsen, P., 1998. Global marine gravity field from the ERS-GEOSATgeodetic mission altimetry. J. Geophys. Res. 103, 8129–8137.

Anderson et al., 2008. The DNSC07 Ocean wide altimetry derived gravity field. In:Presented EGU 2008, Session G1, General Assembly, Vienna, Austria, April 14th–18th.

Audet, P., Mareschal, J.C., 2004a. Anisotropy of the flexural response of thelithosphere in the Canadian Shield. Geophys. Res. Lett. 31. doi:10.1029/2004GL021080.

Audet, P., Mareschal, J.C., 2004b. Variation in elastic thickness in the CanadianShield. Earth Planet. Sci. Lett. 226, 17–31.

Audet, P., Jellinek, A.M., Uno, H., 2007. Mechanical controls on the deformation ofcontinents at convergent margins. Earth Planet. Sci. Lett. 264, 151–166.

Banks, R.J., Francis, S.C., Hipkin, R.G., 2001. Effects of loads in the upper crust onestimates of the elastic thickness of the lithosphere. Geophys. J. Int. 145, 291–299.

Barmin, M.P., Ritzwoller, M.H., Levshin, A.L., 2001. A fast an reliable method forsurface wave tomography. PAGEOPH 158, 1351–1375.

Bendat, J.S., Piersol, A.G., 2000. Random Data: Analysis and MeasurementProcedures, third ed. John Wiley, New York.

Burov, E., Diament, M., 1995. The effective elastic thickness (Te) of the continentallithosphere: what does it rally mean? J. Geophys. Res. 100, 3905–3927.

Cochran, J.R., 1979. An analysis of isostasy in the world‘s oceans 2. Midocean Ridgecrests. J. Geophys. Res. 84, 4713–4729.

Courtney, R., Beaumont, C., 1983. Thermally-activated creep and flexure of theoceanic lithosphere. Nature 305, 201–204.

Curray, Joseph.R., 1991. Possible greenschist metamorphism at the base of a 22 kmsediment section, Bay of Bengal. Geology 19, 1097–1100.

Curray, J.R., Frans, J.E., Moore, D.G., Ram, R.W., 1982. The Indian Ocean: aseismicridges, spreading centers, and oceanic basins. In: Narin, A.E.M., Stehli, F.G. (Eds.),The Ocean Basins and Margins. The Indian Ocean, vol. 6. Plenum, New York.

Daly, E., Brown, C., Stark, C.P., Ebinger, C.J., 2004. Wavelet and multitaper coherencemethods for the assessing the elastic thickness of the Irish Atlantic margin.Geophys. J. Int. 159, 445–459.

Divins, D.L., 2008. NGDC Total Sediment Thickness of the World’s Oceans &Marginal Seas. <http://www.ngdc.noaa.gov/mgg/sedthick/sedthick.html>.

Forsyth, D.W., 1985. Subsurface loading and estimates of the flexural rigidity ofcontinental lithosphere. J. Geophys. Res. 90, 12623–12632.

Grevemeyer, I., Flueh, E.R., Reichert, C., Bialas, J., Klaschen, D., Kopp, C., 2001. Crustalarchitecture and deep structure of the Ninetyeast Ridge hot spot trail fromactive-source ocean bottom seismology. Geophys. J. Int., 414–431.

Hamilton, W., 1979. Tectonics of the Indonesian region: US Geological Survey Prof.Paper 1078.

Karner, G.D., 1991. Sediment blanketing and flexural strength of extendedcontinental lithosphere. Basin Res. 3, 177–185.

Kopp, H., 2002. BSR occurrence along the Sunda margin: evidence from seismicdata. Earth Planet. Sci. Lett. 197, 225–235.

Lavier, L.L., Steckler, M.S., 1997. The effect of sedimentary cover on the flexuralstrength of the continental lithosphere. Nature 389, 476–479.

Lay, T., Kanamori, H., Ammon, C.J., Nettles, M., Ward, S.N., Aster, R.C., Beck, S.L., Bilek,S.L., Brudziski, M.R., Buttler, R., Deshon, H.R., Ekstrom, G., Satake, K., Sipkin, S.,2005. The great Sumatra–Andaman earthquake of 26th December, 2004.Science 308, 1127–1133.

Lowry, A.R., Zhong, S., 2003. Surface versus internal loading of the Tharsis rise Mars.J. Geophys. Res.. doi:10.1029/2003JE002111.

Madsen, J.A., Forsyth, D.W., Detrick, R.S., 1984. A new isostatic model for the eastPacific rise crest. J. Geophys. Res. 89, 9997–10016.

Marks, K.M., Smith, W.H.F., 2006. An evaluation of publicly available globalbathymetry grids. Mar. Geophys. Res. 27, 19–34. doi:10.1007/s1001-005-2095-4.

National Oceanic and Atmospheric Administration, 2003. General BathymetricChart of the Oceans. <www.ngdc.noaa.gov/mgg/gebco/grid/development.pdf>.

Newcomb, K.R., McCann, W.R., 1987. Seismic history and seismotectonics of theSunda Arc. J. Geophys. Res. 92 (B1), 421–439.

Parker, R.L., 1972. The rapid calculation of potential anomalies. Geophys. J. Roy. Astr.S. 31, 447–455.

Percival, D.B., Walden, A.T., 1993. Spectral Analysis for Physical Applications,Multitaper and Conventional Univariate Techniques. Cambridge Univ. Press,New York.

Pérez-Gussinyé, M. et al., 2004. On the recovery of effective elastic thickness usingspectral methods: examples from synthetic data and from the FennoscandianShield. J. Geophys. Res. 109, B10409.

Pollack, H.N., Hurter, S.J., Johnson, J.R., 1993. Heat flow from the earth’s interior:analysis of the global data set. Rev. Geophys. 31 (3), 267–280.

Rajesh, R.S., Stephan, J., Mishra, D.C., 2003. Isostatic response and anisotropy of theEastern Himalayan–Tibetan Plateau: a reappraisal using multitaper spectralanalysis. Geophys. Res. Lett. 30 (2), 1060. doi:10.1029/2002GL016194.

Richardson, R.M., 1992. Ridge force: absolute plate motions and the intraplate stressfields. J. Geophys. Res. 97, 11739–11748.

Ritzwoller, M.H., Shapiro, N.M., Barmin, M.P., Levshin, A.L., 2002. Global surfacewave diffraction tomography. J. Geophys. Res. 107 (B12), 2335.

Sandwell, D.T., Smith, W.H.F., 1997. Marine gravity anomaly from Geosat and ERS-1satellite altimetry. J. Geophys. Res. 102, 10039–10054.

Schoffel, H.J., Das, S., 1999. Fine details of the Wadati–Benioff zone under Indonesiaand its geodynamic implications. J. Geophys. Res. 104 (B6), 13101–13114.

Shapiro, N.M., Ritzwoller, M.H., 2002. Monte-Carlo inversion for a global shearvelocity model of the crust and upper mantle. Geophys. J. Int. 151, 88–105.

Silver, P.G., Chan, W.W., 1988. Implications for continental structure and evolutionfrom seismic anisotropy. Nature 335, 34–39.

Simons, F.J., Zuber, M.T., Korenaga, J., 2000. Isostatic response of the AustralianLithosphere: estimation of effective elastic thickness and anisotropy usingmultitaper spectral analysis. J. Geophys. Res. 105 (19), 163–184.

Simons, F.J., van der Hilst, R.D., Zuber, M.T., 2003. Spatiospectral localization ofisostatic coherence anisotropy in Australia and its relation to seismicanisotropy: implication for lithospheres deformation. J. Geophys. Res. l08(B5), 2250. doi:10.1029/2001JB000704.

Simons, F.J., Dahlen, F.A., Wieczorek, M.A., 2006. Spatiospectral concentration on asphere. SIAM Rev. 48 (3), 504–536.

Slepian, D., 1978. Prolate spheroidal wave function, Fourier analysis anduncertainty, the discrete case. Bell Syst. Tech. 27, 1371–1429.

Spector, A., Grant, F.S., 1970. Statistical models for interpreting aeromagnetic data.Geophysics 35, 293–302.

Tiwari, V.M., Diament, M., Singh, S.C., 2003. Analysis of satellite gravity andBathymetry data over Ninety-East Ridge: variation in the compensationmechanism and implication for emplacement process. J. Geophys. Res. l08(B2), 515–645.

Zoback, M.L. et al., 1989. Global patterns of tectonic stress. Nature 341, 291–298.Zoback, M.L., 1992. First and second order patterns of stress in the Lithosphere: the

world stress map project. J. Geophys. Res. 97, 11703–11728.