You are on page 1of 13

Frequency-Dependent Scattering Observed

in P- and Surface-Wave Arrivals From


South India

A.K.Rai

Pure and Applied Geophysics


pageoph

ISSN 0033-4553
Volume 174
Number 3

Pure Appl. Geophys. (2017) 174:895-905


DOI 10.1007/s00024-016-1426-5

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
International Publishing. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com.

1 23
Author's personal copy
Pure Appl. Geophys. 174 (2017), 895905
 2016 Springer International Publishing
DOI 10.1007/s00024-016-1426-5 Pure and Applied Geophysics

Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India


A. K. RAI1

AbstractAnomalies in polarization angles of teleseismic during its geologic history and a major tectono-thermal
waves have been used to understand effect of scattered arrivals
from subsurface heterogeneities. Seismological data recorded in
stress at the ProterozoicPaleozoic boundary
southern India show polarization anomalies up to 5 for several (*600 Ma). These are recognized by isotopic re-setting
stations. These anomalies are most pronounced for earthquakes of minerals and whole rock assemblages of older pro-
from western and southern azimuths. Furthermore, stations located
toliths, and lithospheric deformations evident from
near the boundary of Dharwar craton and southern Granulites are
more affected by scattered waves. Considering that many of the seismological observations (Rai et al. 2009).
nearby stations show similar patterns of polarization anomalies, it The presence of anisotropy and scattering in seis-
is likely that the source of scattered energy is located at shallower mic data has been reported from various parts of India
depths. The non-stationary nature of seismic arrivals warrants
determination of frequency-dependent polarization. Result and elsewhere (Rai et al. 2008; Deshpande and Mohan
obtained using multi-taper spectral analysis method indicates that 2016). While seismic anisotropy produces azimuthal
data are contaminated at frequencies greater than *2 Hz for most variation of seismic arrivals, shear-wave birefrin-
of the stations. Furthermore, surface-wave records also indicate off-
azimuth arrivals, and quasi-Love waves indicating heterogeneities
gence, and apparent discrepancy between Love and
or anisotropy in the subsurface. These small-scale heterogeneities Rayleigh waves (Anderson 1961; Backus 1962),
that may be located in crust may be important for studies using effects of scattering due to small-scale heterogeneities
converted phases and ground motion prediction studies.
are usually random on seismic arrivals. Analysis of
Key words: Heterogeneity, scattering, multi-taper spectral polarization of first arrivals in three component seismic
analysis, India. waveform can be used to determine dominant orien-
tation of arriving waves. Estimates of deviations of
arrival azimuths from expected directions are often
interpreted as due to contamination of seismic records
1. Introduction from off-azimuth arrivals. These off-azimuth arrivals
are primarily caused by scattering due to the presence
The peninsular India is an amalgamation of a num- of structural features, such as sedimentary basins and
ber of Archaean cratons, Proterozoic mobile belts, intra- crustal faults (Crampin et al. 1982; Hu and Menke
cratonic sedimentary basins (e.g., Cuddapah) and large 1994; Farra and Begat 1995; Bokelmann 1995;
igneous flood basalt province, such as Deccan volcanics Morozov and Din 2008; Kawahara 2011). Thus,
(Fig. 1). The grade of metamorphism gradually information on polarization of teleseismic body and
increases from low-grade granite-gneiss greenstone surface waves provides important information about
(*3.4 Ga) rocks of Dharwar craton in the north, to seismic characteristics in the vicinity of the recording
*2.5 Ga-old high-grade granulites in the southern part stations (receiver-side effects).
(Ramakrishnan and Vaidyanathan 2008). Various ter- One of the qualitative ways to determine orienta-
ranes of the Indian shield are traversed by a number of tion of dominant arrival of seismic phases is by visual
shear zones and lineaments in the lithosphere. The inspection of the particle motion plotted for two pairs
shield region has experienced episodic deformation of the three component seismograms. The linearity of
the motion indicates whether the incident arrival rep-
resents a plane polarized, or a pure state (Samson
1
Indian Institute of Technology Bhubaneswar, Bhubaneswar 1983a), whereas the inclination of motion allows
751007, India. E-mail: akrai@iitbbs.ac.in estimation of the azimuth and inclination of that
Author's personal copy
896 A. K. Rai Pure Appl. Geophys.

Figure 1
Location of stations (left) shown by red triangles. The locations of earthquakes recorded by the stations (small circles) are shown on right. The
two circles denote distances of 30 and 90 from of the center (star). DVP Deccan Volcanic Province, EGMB Eastern Ghat Mobile Belt, WDC
& EDC Western and Eastern Dharwar Craton, SGT Southern Granulite Terrane

particular arrival. If the seismic instruments are prop- Institute, University of Cambridge, UK and few other
erly aligned, the main reasons for changes in wave institutes at sites distributed in the Indian shield region
polarization (or propagation directions) are refractions (Rai et al. 2008, 2009). Data were recorded at sampling
and reflections in the subsurface from randomly dis- frequency of 20 Hz; therefore, signals up to 10 Hz are
tributed geological features. These scattering effects or preserved without aliasing effect. A large number of
anomalies are mostly interpreted to be due to receiver- earthquakes that were recorded by the network were
side effects, because polarization anomalies due to either from Andaman-Sumatra subduction zone in
source-side scattering are small because of large dis- south-east direction or from the Japan trench in the
tances. Furthermore, different dimensions of north-east. Earthquakes from western and southern azi-
heterogeneities would affect different frequencies. muths are recorded less frequently. About 1000 station-
Therefore, it is important to remove contaminating event pairs that were within an epicentral distance of 30o
signal for obtaining results of high accuracy in obser- and 90o were analyzed in this study. The time-domain
vational seismology, such as study of structure of the principal component analysis or eigenvalue analysis is
lithosphere using Ps and Sp converted phases. Such one method to determine polarization of seismic waves
contaminations can be easily determined by polariza- and gain insights into the nature of the seismic arrivals
tion analysis and frequency-domain multi-taper and their interactions within the subsurface.
spectral analysis (Thomson 1982; Samson 1983b; Park Using a wide frequency signal to determine
1987), as described in the following sections. polarization may result in estimates that are domi-
nated by peak amplitudes in a specific frequency
band. Multi-taper spectral analysis method has been
2. Data and Analysis used to determine the polarization characteristics of
seismic waves as a continuous function of frequency.
Earthquake records from seismic stations located in This provides an important constraint on scale of
southern part of India were used in this work (Fig. 1). heterogeneity in the subsurface and information that
The seismographs were temporarily installed and oper- is independent from those obtained from travel times,
ated jointly by the National Geophysical Research shear-wave splitting, and coda analysis.
Author's personal copy
Vol. 174, (2017) Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India 897

 
2.1. The Principal Component Analysis k2 k3
Rectilinearity 1  : 3
In one-dimensional, radially stratified, isotropic, 2k1
and homogeneous Earth, primary or P waves are Rectilinearity takes the value 1.0 when there is only
expected to be polarized in the direction of wave one non-zero eigenvalue as for pure P wave, the value
propagation. However, the real earth is significantly is 0.5 when k1 = k2, k3 = 0 (i.e., circular polariza-
different from these ideal cases, and seismic records tion), and it is 0 when k1 = k2 = k3 (i.e., spherical
are contaminated with back-ground noise, scattered polarization). A schematic representation of the polar-
waves from small-scale heterogeneity, refracted, ization ellipsoid associated with a seismic arrival
reflected, converted phases that immediately follow exhibiting elliptical motion is presented in Fig. 2.
direct arrivals and contaminate the same. The Here, the azimuth of approach is measured clockwise
discrepancy between the predicted and observed from north and the angle of incidence is measured from
polarization of teleseismic P waves can be used to vertical.
study effect of inhomogeneity in subsurface and There are several sources of errors that may affect
contamination of direct arrivals with scattered calculation of the azimuthal anomalies. For example,
energy. The predicted azimuth of arrival can be uncertainties in the epicentral locations will lead to
determined by the known station-earthquake loca- uncertainties in the calculated azimuth and are
tions. The eigenvalue analysis of first few seconds relatively small. The reliability of polarization mea-
(one cycle) of the three component P waveform surements also depends on analysis window length,
provides a quantitative estimate of the degree of since multiple reflected and converted phases gener-
linearity of the particle motion. The direction of ated in the near surface can affect the results. In
approach (azimuth) and angle of incidence (inclina- practice, only the first cycle of the direct P wave is
tion) can be estimated by the analysis of the windowed. An example of a typical onset window is
covariance matrix of the three components of ground shown in Fig. 2. The degree of linearity is also a
motion (Kanasewich 1975; Jurkevics 1988). guiding factor in determining window length for the
For a three component seismic time series (Z, N, analysis. However, in most of the cases, the window
E), the covariance matrix (S) is given by of analysis extends one cycle of dominant period of
0 1 motion both before and after the P-wave onset. For
VarZ  CovZ; N  CovZ; E the typical seismogram and window length shown in
S @ CovN; Z  VarN  CovN; E A: 1
Fig. 2, the rectilinearity is 0.98. The P waveforms
CovE; Z  CovE; N  VarE
were manually picked considering above factors. An
If V1, V2, and V3 are three eigenvectors corre- example of P-wave polarization measurements for
sponding to the three eigenvalues k1, k2, and k3, such some of the stations is shown in Fig. 3. The data are
that k1  k2 and k3, then the largest eigenvalue (k1) plotted at two (0.060.2 and 0.332.0 Hz) frequen-
corresponds to the dominant direction of incident cies and as a function of azimuth of arrivals to the
energy. Then, the azimuth of approach (Baz) and station. The figure gives an indication that southern
angle of incidence (Inc) are and western directions show higher misfit between
the observed and predicted back azimuths. It is
  
V1E V rad observed that stations located on different tectonic
Baz arctan N ; Inc arctan 1Z 2
V1 V1 units (e.g., Dharwar craton, granulite terrane, etc)
show similar patterns of polarization anomalies.
where VE1 , VN Z rad
1 , V1 , and V1 are components of the
eigenvector corresponding to the largest eigenvalues
in the east, north, vertical, and radial directions, 2.2. Frequency-Domain Multi-Taper Analysis
respectively. The component Vrad 1 can be derived once The amplitudes of seismic signals vary strongly
back azimuth (Baz) is known from Eq. 2. Similarly, with frequency due to scattered waves of different
rectilinearity of the motion is defined as wavelengths from structural heterogeneities of
Author's personal copy
898 A. K. Rai Pure Appl. Geophys.

Figure 2
Polarization ellipsoid (left) indicating directions of back azimuth (Baz) and angle of incidence (Inc). On the right panel, polarization analysis
window (top) and corresponding particle motions in three orthogonal directions (bottom) for bandpass filtered data (515 s) is shown

various dimensions (Park et al. 1987; Park 1987). Figure 3 c


Back azimuths for stations BGL, GBA, and KOD. Frequency band
Therefore, it is important to understand the effect of used in principal component analysis is given adjacent to figures.
scattering as a function of frequency. The angle of The circles denote distances of 30, 60, 90, and 120. The blue
polarization as a function of frequency can be and red lines are back-azimuth directions computed from known
event-station locations, and principal component analysis of P
determined by singular value decomposition of the
wave, respectively
matrix of eigenspectra. Samson (1983a) suggested
estimate of the spectral density matrix S(f) having
components: samples the time series in a different manner while
j k being resistant to spectral leakage. The data dis-
Sjk f y f :y f where j; k 1; 2; 3
carded by the first taper are partially recovered by
and the second, and higher order taper and the infor-
1 XN 1 mation discarded by the first two tapers are
y j f xn x j nT ei2PfnT 4 partially retrieved by the third higher order tapers
NT n0
and so on. Thus, using several tapers (Slepian
as the Fourier transform of jth component of 1978), a spectral estimate can be computed which
the time-series X(t), and the chosen data taper is not significantly hampered by the trade-off
(xn). Moving average is applied in frequency between spectral leakage and variance that affects
domain to smooth the matrix S(f), and eigenvectors single taper estimation.
and eigenvalues of smoothed matrix are calculated. The eigentapers are defined by their order and
The largest eigenvalue and the corresponding time-bandwidth product which governs the width of
eigenvector represent the direction of polarization the frequency band about a center frequency outside
at frequency f. To separate effects of later arrivals which the spectral leakage is minimized and the
that may travel through different paths, polariza- resolution available in the frequency domain. For
tion of short-time windows around seismic phases example, a taper with time-bandwidth product equal
is computed. The taper (xn) applied to data before to 4 will minimize spectral leakage at frequency f0
its transformation. In brief, data are multiplied from outside the band jf  f 0j  4=NT, where NT is
with several leakage-resistant tapers. The tapers length of the time series. This method allows us to
are constructed in such a way that each taper identify the frequencies at which plane P arrivals
Author's personal copy
Vol. 174, (2017) Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India 899
Author's personal copy
900 A. K. Rai Pure Appl. Geophys.

Table 1
Summary of P-wave polarization analysis at low- and high-frequency bands

Station name Low freq. observations High freq. observations

No. (N) Baz gap (deg.) MD No. (N) Baz-gap (deg.) MD

BGL 58 165 2.8 54 165 2.4


GBA 87 030 3.3 67 030 3.3
KOD 55 180 2.2 54 180 2.2
SLM 54 050 2.2 54 050 2.1
MBN 93 030 2.2 115 030 2.2
KIL 64 175 2.2 59 175 3.2
BKR 8 170 4.0 9 170 3.5
BOM 10 190 5.5 10 190 5.0
CRP 29 120 3.3 29 120 3.8
CUD 45 120 1.5 49 120 1.5
DHR 22 250 1.2 22 250 1.4
DVG 8 200 1.1 8 200 1.3
GDP 23 115 2.3 24 115 3.4
GOA 8 265 3.0 8 265 3.3
GRR 136 060 3.6 137 060 4.2
KBC 13 150 1.0 14 150 2.1
KDM 44 110 3.3 45 110 4.2
KSL 40 190 4.2 41 190 4.2
KRD 25 210 1.3 26 210 1.3
KOL 15 120 1.1 16 120 1.3
LTV 83 090 2.2 71 090 3.1
MDR 19 090 3.3 20 090 3.3
MTP 27 090 2.3 28 090 2.4
MYS 24 170 1.4 25 170 1.4
NTR 10 260 1.4 11 260 1.4
PCH 11 140 2.3 12 140 2.5
PLK 26 120 2.1 27 120 2.0
PUN 35 090 1.1 36 090 1.3
TMK 43 060 1.4 44 060 1.3
TPT 135 060 1.2 137 060 1.4
TRV 31 175 4.8 31 175 4.3
N number of observations, MD mean deviation of computed back azimuths from predicted back azimuths

become contaminated by scattered or converted azimuths are more pronounced from the NW and
phases. SW directions, generally less than 5 for most of
the station-event pairs. Furthermore, deviations are
more at the higher frequencies. Stations located on
3. Results and Discussions Dharwar craton (e.g., GBA) indicate deviations in
polarization directions for events recorded from
The analysis results of polarizations of the P southern and western azimuths. The data from
waveforms for some of the stations are shown in seismic stations located on high-grade granulite
Fig. 3. Calculated and predicted back azimuths terrane (KOD and PALK) indicate mostly on-az-
obtained from polarization analysis and station- imuth arrivals and least sign of deviations. In
event locations, respectively, are illustrated as a summary, polarization anomalies are relatively
function of frequency, azimuth, and epicentral more for stations located on Dharwar craton (EDC
distance. The results indicate that the deviations of and WDC), Cuddapah basin (CB), and Eastern
computed azimuths from those of the predicted Ghats (EGMB) than those of southern Granulites
Author's personal copy
Vol. 174, (2017) Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India 901

Figure 4
Multi-taper analysis of a record at station BGL. Signal versus noise amplitude spectra (black and red lines, respectively) is shown on the top
for vertical, northsouth, and eastwest components. Eigenvalues, angles of incidence and back azimuths are shown at bottom with
corresponding error bars

Table 2
Multi-taper analysis results for some of the analyzed earthquakes

Station name Evt. name (YYYJDAYHHMMSS) Dist. (deg.) Baz. (deg.) Freq. band

BGL 1998313053014 54.7 108.23 0.01.75


BKR 2001016132501 32.0 131.50 0.01.70
KIL 1999019033533 78.8 098.60 0.01.75
PLK 2001115144006 53.8 054.48 0.02.00
BOM 2001055072349 55.0 102.01 0.01.65
CRP 2001342202937 51.7 064.26 0.01.50
CUD 1999132175922 62.0 048.81 0.02.45
DHR 2001225201123 63.0 051.56 0.01.75
DVG 2002064211609 48.0 094.83 0.01.25
GDP 2001352040258 45.0 067.98 0.03.00
GRR 2001045133639 72.1 042.94 0.03.00
GBA 1998271133430 40.9 119.86 0.04.00
LTV 1999339220034 47.5 063.90 0.02.00
MBN 1999077175543 60.6 052.02 0.03.00
SLM 2000162182329 40.4 072.05 0.01.75
The frequency band within which the computed and expected polarizations are same is given
Author's personal copy
902 A. K. Rai Pure Appl. Geophys.

Figure 5
Three component (transverse, radial, and vertical components), low-pass filtered (100 s), seismic record of station KOD (Evt.
1998333141031). The small signal (shown by asterisk mark) on vertical component shows the quasi-love waveform anomaly

(SGT) and Deccan Volcanics (DVP). The results Figure 6 c


Long period surface-wave data (radial, transverse, and vertical
are summarized in Table 1. components) (Evt. 1998148211144, station: KOD) (top panel).
Multi-taper analysis has provided further insights Vertical lines at A and F denote time window chosen for multi-
into deviations of arrivals azimuths and incidence angles taper spectral analysis. a Three eigenvalues and b ellipticity of the
particle motion obtained by multi-taper surface-wave polarization
as a function of frequency. An example of multi-taper
analysis of the data, c azimuth of arrivals, and d tilt of the particle
measurements of back azimuth and angle of incidence motion
for stations Bangalore (BGL) is shown in Fig. 4. Results
indicate that at frequencies, above 2 Hz noise deterio-
rates the data quality as evident from signal and noise The decrease in signal-to-noise ratio of the data above
amplitude spectra (Fig. 4, top). Therefore, the eigen- *2.0 Hz could be due to scattered energy arising from
values are seen to converge after *2.0 Hz (Fig. 4, subsurface heterogeneities present in the vicinity of the
bottom). Obtained values of back azimuths remarkably station. Therefore, earthquake records from south India
match the expected values calculated from station-event having frequencies greater than *2.0 Hz should be
locations only up to frequency of *2.0 Hz. A summary used carefully.
of the multi-taper measurement for other stations is Surface-wave observations also indicate anoma-
presented in Table 2. Frequencies within which values lous arrivals in recorded seismograms. Earlier studies
of back azimuth and angle of incidence are in good indicate that the presence of anisotropy and lateral
agreement with expected values indicate that arrivals heterogeneity in the subsurface could cause refrac-
may only be considered plane polarized below *2 Hz. tion, scattering, focusing, defocussing, and mode
Author's personal copy
Vol. 174, (2017) Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India 903
Author's personal copy
904 A. K. Rai Pure Appl. Geophys.

coupling resulting characteristic anomalies in phase, which the earth structure beneath the seismic
amplitude, and polarization of surface waves stations may be considered relatively
(Woodhouse and Wong 1986; Park 1987; Park and homogeneous.
Yu 1992, 1993; Laske and Masters 1996). Some of Surface wave also shows indications of anomalous
the seismograms recorded in the study area show the phases, such as quasi-love waves and polarization
presence of quasi-Love waves on long-period sur- anomalies.
face-wave data indicating that surface waves have
The contaminations of direct arrivals from out-of-
been affected by the presence of anisotropy or lateral
plane energy are not an exception but are likely to be
heterogeneity in the subsurface (Fig. 5). The multi-
present in most of the cases in almost all tectonic
taper method indicates at least 1015 deviation of
settings. Therefore, analysis of seismological data
surface-wave polarization from their expected back
must consider these effects.
azimuths (Fig. 6). Although the source of such
anomalies and perturbations cannot be defined with
greater accuracy only by polarization measurements, Acknowledgements
these observations indicate the extent of anomalies
present in the data. Therefore, surface-wave studies The author thanks to the various academic and
should carefully estimate effects of lateral hetero- research institutes for providing seismological data
geneity and azimuthal anisotropy that may for completing this work. Special thanks to Prof.
contaminate the seismic records in the southern India. K. Priestley and Dr. S.S. Rai for providing invaluable
suggestions and guidance during this work. My
sincere thanks are also due to the two anonymous
4. Conclusions reviewers whose constructive suggestions immensely
improved the original manuscript. Figures were cre-
The Indian lithosphere, which is an amalgamation ated using GMT (Wessel and Smith 1991).
of a variety of low-grade and high-grade terrains of
ages ranging from Archaean to neo-Proterozoic, has REFERENCES
experienced episodic deformation during the geologic
history. Therefore, crust and mantle are likely to be Anderson, D. L. (1961). Elastic wave propagation in layered ani-
highly heterogeneous dominated by structural fea- sotropic media. Journal Geophysical Research, 66, 29532963.
Backus, G. E. (1962). Long-wave elastic anisotropy produced by
tures of various dimensions. These features
horizontal layering. Journal Geophysical Research, 67,
distributed at various depths and dimensions are the 44274440.
most likely source of scattering observed in the Bokelmann, G. (1995). P-wave array polarization analysis and
seismic data. Following are the main conclusions of effective anisotropy of the brittle crust. Geophysical Journal
International, 120, 145162.
this study: Crampin, S., Stephen, R., & McGonigle, R. (1982). The polariza-
tion of P-waves in anisotropic media. Geophysical Journal
Teleseismic waveforms recorded in the Indian International, 68, 477485.
shield indicate polarization anomalies and contam- Deshpande, A. A., & Mohan, G. (2016). Seismic evidence of
ination of direct arrivals. crustal heterogeneity beneath the northwestern Deccan volcanic
province of India from joint inversion of Rayleigh wave dis-
Although deviations in arrival azimuths for most of persion measurements and P receiver functions. Journal of Asian
the station-event pairs are less than 5, this Earth Sciences, 128, 5463.
indicates the presence of seismic heterogeneities Farra, V., & Begat, S. (1995). Sensitivity of qP-wave traveltimes
and polarization vectors to heterogeneity, anisotropy and inter-
in the subsurface that appear to scatter seismic
faces. Geophysical Journal International, 121, 371384.
energy in out-of-plane directions and contaminat- Hu, G., & Menke, W. (1994). Polarization tomography for P wave
ing direct arrivals. velocity structure in southern Calfornia. Journal Geophysical
The multi-taper spectral estimates of back azimuth Research, 99, 1524515256.
Jurkevics, A. (1988). Polarization analysis of three-component
and incidence angle as a function of frequency array data. Bulletin of the Seismological Society of America, 78,
identifies the frequency bands (\2.0 Hz) within 17251743.
Author's personal copy
Vol. 174, (2017) Frequency-Dependent Scattering Observed in P- and Surface-Wave Arrivals From South India 905

Kanasewich, E. R. (1975). Time sequence analysis in geophysics Rai, A., Rai, S., Gaur, V., & Priestley, K. (2009). Seismic char-
(pp. 292296). Edmonton: The University of Alberta Press. acteristics of the south Indian granulite crust. Journal of Virtual
Kawahara, J. (2011). Scattering attenuation of elastic waves due to Explorer, Geological Anatomy of Indian and the Middle East,.
low-contrast inclusions. Wave Motion, 48, 290300. doi:10.3809/jvirtex.2009.00252.
Laske, G., & Masters, G. (1996). Constraints on global phase Ramakrishnan, M., & Vaidyanathan, R. (2008). Geology of India
velocity maps from long period polarization data. Journal Geo- (Vol. 1). Bangalore: Geological Society of India.
physical Research, 101, 1605916075. Samson, J. (1983a). Pure states, polarized waves and principal
Morozov, I., & Din, M. (2008). Use of receiver functions in wide- components in the spectra of multiple, geophysical time-series.
angle controlled-source crustal data sets. Geophysical Journal Geophysical Journal International, 72, 647664.
International, 173, 299308. Samson, J. (1983b). The spectral matrix, eigenvalues and principal
Park, J. (1987). Multitaper spectral analysis of high-frequency components in the analysis of multichannel geophysical data.
seismogram. Journal Geophysical Research, 92, 1267512684. Annales Geophysicae, 1, 115119.
Park, J., Vernon, F., & Lindberg, C. (1987). Frequency dependent Slepian, D. (1978). Prolate spheroidal wave functions, Fourier
polarization analysis of high-frequency seismogram. Journal analysis and uncertainty. Bell System Technical Journal, 57,
Geophysical Research, 92, 1266412674. 13711429.
Park, J., & Yu, Y. (1992). Anisotropy and coupled free oscillations: Thomson, D. (1982). Spectrum estimation and harmonic analysis.
simplified models and surface wave observations. Geophysical Annales Geophysicae, 70, 10551096.
Journal International, 110, 401420. Wessel, P., & Smith, W. H. F. (1991). Free software helps map and
Park, J., & Yu, Y. (1993). Seismic determination of elastic aniso- display data. EOS Transactions of the American Geophysical
tropy and mantle flow. Science, 261, 11591162. Union, 72, 441.
Rai, A., Rai, S., & Gaur, V. (2008). Crustal stress patterns: analysis Woodhouse, J., & Wong, Y. (1986). Amplitude, phase and path
of Ps splitting for seismic anisotropy. Current Science, 94(7), anomalies of mantle waves. Geophysical Journal International,
911915. 87, 753773.

(Received May 26, 2016, revised October 31, 2016, accepted November 3, 2016, Published online December 20, 2016)

You might also like