Open Access Library Journal
How to cite this paper: Sayil, N. (2014) Application of Discrimination Filter Based on the Polarization to the Surface Wave
Records. Open Access Library Journal, 1: e724. http://dx.doi.org/10.4236/oalib.1100724
Application of Discrimination Filter Based
on the Polarization to the Surface Wave
Records
Nilgun Sayil
Department of Geophysics, Engineering Faculty, Karadeniz Technical University, Trabzon, Turkey
Received 2 May 2014; revised 12 June 2014; accepted 24 July 2014
Copyright © 2014 by author and OALib.
This work is licensed under the Creative Commons Attribution International License (CC BY).
http://creativ ecommon s.org/l icens es/by/4.0/
Abstract
As well known, each type of seismic waves has a specific particle motion. The basic surface waves
Love and Rayleigh show the particle motions polarized linearly in the transversal-horizontal plane
and elliptically in the vertical-radial plane, respectively. Like in the body waves, polarization
properties can be used to design the surface wave discrimination filter. The process consists of
weighting the amplitudes of vertical (Z), radial (R) and tangential (T) components of the ground
motion at each frequency according to the particle motion. The weighting process is applied to en-
tire length of each component for selected window length and moving interval, but weights are not
applied to the original phase values. The weighted parts for each window are transformed to the
time domain and filtered signals are obtained as the arithmetic average of values of the overlap-
ping points. The method has been applied to the broad-band digital three-component records at
stations having about 10˚ epicenter distances of Bogazici University Kandilli Observatory and
Earthquake Research Institute (KOERI) of Erzurum earthquakes and noticed that the window
length and moving interval in proportion to epicenter distance a ffec t the results on a large scale.
For the cases in which the best results are obtained, it has been determined that the ratio between
the window length and moving interval for increased epicenter distances are 3.95, 4.5 and 4.8,
respectively.
Keywords
Discrimination Filter, Polarization Properties, Surface Wave
Subject Areas: Applied Physics, Geophysics, Particle Physics
1. Introduction
The parameters such as group arrival times, phase angles and a mplitude values of surface waves are used to re-

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 2 July 2014 | Volume 1 |
e724
searching elastic properties of the Earth. The environmental noise, the noise related to signal, the equal sharing
of seismic energy between components and the other factors influence discrimination of surface waves forms on
seismograms. The differences between polarization properties of surface waves and microseismic noises enable
to filter a desired type of surface waves on three components records. Discrimination filter based on the polari-
zation properties can be used to substantially improve the discrimination of surface wave s on three components
records. The analysis of polarization depended on rectilinearity and directionality properties. These properties
provide to design weighting functions for seismic wave discrimination. If the signals of the different compo-
nents have same phases, they are called as P- or S-wave s an d have li nea r move ment. If th e s ignals have d iffe rent
phases, they are called a s Rayleig h waves and have el liptical moveme nt. Therefore, Ra yleigh wave with ellipti-
cal particle motion in the vertical-radial plane and Love wave with linear particle motio n in the tangential-hori-
zontal plane can be discriminated on seismograms by we ighting functi ons.
The basis of the surface wave discrimination filter depended on polarization properties was explicitly given
by [1]. Preliminary research about this subject has been applied by [2]. Later, [3]-[5] have been developed dif-
ferent algorithms related to the polarization analysis. Reference [6] noted that rectilinearity and directionality
properties are simply obtained by data covariance matrix. Reference [7] has deter mined t he weig htin g functi ons
as depend on frequency. References [8] [9] have investigated the data belonging to three component stations
network. Reference [10] has applied polarization analysis to multichannel seismic data. Reference [11] applied
to surface wave discrimination filter based on polarization properties on long-period three components records
of the three major earthquakes with different epicenter distances larger than 40˚. They fo u nd tha t i n c ase o f exi s-
tence of the surface waves having substantially great amplitudes recorded on seismograms, it can be filtered
perfectly. Body wave polarization was investigated by [12]-[15]. References [16] [17] have designed to modi-
fied weighting functions for epicenter distances smaller than about 20˚ to corresponding with angular distribu-
tion of polarization parameters obtained from computed synthetic seismograms and applied to three component
earthquake r ecor ds for discrimination o f surface waves. In this stud y, it has been investiga t ed a pplicab ility of the
surface wave discrimination filter to the earthquake records having about 10˚ epicenter distances .
2. Surface Wave Discrimination Filter Based on Polarization Properties
This pro cess is a digita l pr oce dure for extracti ng lo ng-perio d surface wave signals from microseismic noise. The
technique is deter ministic and means basically to frequency filtering using measurements on particle motion to
shape the filter response. Filtering process is performed in the frequency domain-because of dispersive charac-
ters of surface waves. The discrete Fourier transforms of vertical, radial and tangential components of the ground
motion are computed for a selected window length and moving interval. The amplitude coefficients at each fre-
quenc y are weighted accordi ng to how closel y the three-dim ensio nal particle mo tion pattern at t hat frequency cor -
responds to theoretical patterns for Love and Rayleigh waves, arriving from some pre-assigned direction. The
weights or adjustments are not applied to the original phase values. Weighted segments for each window are trans-
formed to the time domain, and filtered signal is obtained as the arithmetic average of the overlapping amplitude s.
In application of this process, a long-period greater than 2.0 sec is considered since it has substantially high
the signal-to-noise ratios of surface waves which were barely discernible on the unprocessed seismograms. The
physical reason for to achieve this filtering lies the inherent spectral difference between the sought-for signals
and the backgr ound noise. Thus, it is possible to obtain favorable signal-to-noise ra tios in t he freq uenc y domain
from data which appears highly contaminated in the time domain. This spectral difference is generally realizable
in the long-period band because of the characteristics of the noise field.
Whereas surface waves from both earthquakes and detonations usually have energy distributed over a broad
frequency range, microseismic noise energy tends to be concentrated in two fairly restricted spectral peaks,
around periods of 6 sec and 15 sec [18] . In the frequency bands around the microseismic peaks, the particle mo-
tion pattern of a low-level signal is confused by the noise (interfering multidirectional Rayleigh, presumably),
and the resultant error causes to attenuate those frequencies. However, the rest of the signal spectrum remains
relatively uncovered and the reconstructed time traces preserve the basic character and envelope of the whole
surface wave train. The spectral concentration of microseismic noise can further be exploited by straightforward
elimination of the certain frequencies. The following basic functions are performed:
1) The magnifications of the three digitized seismograph traces (a vertical and two horizontal components) are
equalized, then the horizontal axes are rotated so that one component (the “radial”) is in li ne with the azi muth a t
which the desired signal is expected to arrive (not necessarily the great circle path). Theoretically, then, one can

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 3 July 2014 | Volume 1 |
e724
expect to seek Rayleigh motion appearing only on the vertical and radial trace Z( τ ) and R( τ ); Love motion
should appear only on the transverse trace T( τ ). No correction is done for the seismographs’ frequency response;
both input and output traces are analogous to the seismogram that would be normally recorded.
2) The ro tated trac es are divided into si multaneou s time se gmen ts with le ngth T, the n for each three direction-
al components the amplitude,
( )
i
Af η
and p hase,
( )
i
f η Φ
information are computed by discrete Fourier se-
ries as in Equations (1) and (2).
( )( )( )
12
22
i ii
Afa fb f η ηη

= +

(1)
( )( )
( )
arctan, 0,1,2,,1
i
ii
bf
fN
af
η
ηη
η

Φ== −




(2)
where i = Z, R, T which represents the vertical, radial and tangential components of the ground motion, respec-
tively. NF, Nyquist Frequency, f = 1/T.
( )
i
af η
and
( )
i
bf η
are discrete Fourier series coefficients.
3) The apparent horizontal azimuth,
β ( η f) of each harmonic is determined by Equation (3), as if both the ho-
rizontal components were in phase, in terms of the angle from the radial direction (Figure 1). A measure of the
eccentricity of the particle motion ellipse,
ψ ( η f) is also calculated by Equation (4), and the phase difference,
α ( η f) between the vertical and radial components is determined by Equation (5).
()()
( )
arctan
T
R
Af
fAf
η
βη η

=


(3)
( )( )
( )( )( )( )
()
12
22
arctan,
RT
Z
Af
fAfAf Af
Af
η
ψηηη η
η
 
== +
 


(4)
( )( )( )
RZ
fff αηφ ηφ η = −
(5)
4) T he Fourier amplitude coefficients of each direction component, AR(
η f), AT( η f) a nd AZ( η f) ar e t he n wei g ht -
ed according to the formulas in Equation (6).
( )( )( )( )()
( )()( )( )( )
( )()( )( )
cos cossin
cos cossin
sin sin
MK N
ZZ
MK N
RR
MK
TT
AfAf fff
AfAf fff
AfAf ff
ηηβηψη θαη
ηηβηψη θαη
η ηβηψη
′= ⋅⋅−⋅
 
 
′= ⋅⋅−⋅
 
 
′=⋅⋅
  
  
(6)
where
( )()
sin,π2π
N
nf f α αη ≤≤ 

.
( )
T
Af η
′
,
( )
R
Af η
′
and
()
Z
Af η
′
are the weighted vertical, radial
and tangential components of the ground motion. No weights or adjustments are applied to the phase angles.
Note that the Z and R components have identical treatment, and that all weighting factors vary from 0 to 1 as
powers of sinus or cosinus depending upon the degree to which the particle motion corresponds to pure Love or
Rayle igh wa ve be havio r. To d ate, the most satisfactory results generally have been obtained with the weighting
exponents M, K, and N set to are empirically determined as 8, 8 and 4, respectively, operating on segments 128
sec long with a 1/16 (8 sec) time increment [1].
Figure 1 . The relatio n b etween th e ap par ent hori zont al azimuth β , the eccentr icity ψ and
three orthogonal components of ground motion
()
R
Af η
,
( )
T
Af η
and
()
Z
Af η
.

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 4 July 2014 | Volume 1 |
e724
The effects of the first weighting factors (functions of β ) are to attenuate transverse-tending energy on the Z
and R components and radial-tending energy on the T component. In other words, in case the motion in the
horizontal plane is radial perfectly ( β ( η f) = 0), it is seen that the amplitude coefficients of the Z and R compo-
nents are stable and that of the T component are decreased. This situation is also derived from the Rayleigh
wave particle motion. On the other hand, in case some of dominated periods are on the T component (
β ( η f) =
π/2) and the amplitude s on the Z component of the ground motion are very small, this situation is corresponding
to the Love wave particle motion.
The second set of weighting factors depends upon the angle Ψ as a measure of the eccentricity of the Ray-
leigh orbit. On the Z and R components, the angle desired (
θ = 0.21π) is the one corresponding to a theoretical
horizontal/vertical displacement ratio (~0.8) for fundamental long-period Rayleigh waves assuming the Guten-
berg earth model [19]. This value is fairly close to what is actually found at ins tallations o n competent, massive
rock. However, this weighting function will attenuate higher mode Rayleigh waves at the shorter periods (less
than 10 sec), as well as short-period fundamental Rayleigh on incompetent surface layers. T he a mpl it ud es of the
Z and R components are including unit weighting factor according as a function of
( )
cos
K
f ψη θ − 

in the
case of
ψ ( η f) = θ . The amplitudes at the T component as to function of
( )
sin
K
f ψη 

are applied unit
weighting for the case of perfectly horizontal motion (
ψ ( η f) = π/2). The amplitude values of the R and Z
components are decreased by the function of
( )
sin
N
f αη 

in interval varied from 0 to 1 for fundamental
mode Rayleigh waves.
The third weighting factor as a function of
α attenuates the Z and R components by an amount which de-
creases from 1 to 0 as the phase departs from the theoretical 90˚ retrograde r e la tionship for Rayleigh motion o n a
laterally homogeneous half space. No corresponding weight is possible for the T component. The process clearly
will not wor k at all if Love and Rayleigh wa ves o f similar frequency content (and comparable amplitudes) arrive
in the same time segment. In such a case, the time and space p atterns of both wave groups will be mutually con-
fused, resulting in little or no output.
5) All three seismograms are reconstructed in the time domain using the weighted Fourier amplitude coeffi-
cients,
()
Z
Af η
′
,
( )
R
Af η
′
and
()
T
Af η
′
and the phase angles determined initially
( )
Z
f
θη
′
,
( )
R
f θη
′
and
()
T
f θη
′
.
6) The filtered signal is obtained by using the arithmetic mean of overlapped amplitude values.
3. Application of Method
In thi s st udy, the surface wave discrimination filter based on p olarization properties was applied to the broad-band
digital three component seismograms recorded at eight stations (Table 1, Figure 2) having about 10˚ epicenter dis-
tances of Bogazici University Kandilli Observatory and Earthquake Research Institute (KOERI) for Erzurum earth-
quake occurred in Turkey. Focal parameters of Erzu ru m earthquake are g iv en in Table 2. The ti me sampling is 1 sec.
The magnifications of the three component seismograms are firstly equalized, then the horizontal axes are ro-
tated as shown from Figure 3(a) to Figure 10(a). The rotated seismograms are divided into simultaneous time
segments of length T, then for each of the three directional components the amplitude and phase terms are com-
puted by discrete Fourier series.
The apparent horizontal azimuth, β ( η f) of each harmonic is determined by Equation (3). A measure of the ec-
centricity of t he particle motion ellipse, ψ ( η f) is also calculated by Equation (4), and the phase difference, α ( η f)
between the vertical and radial components is determined by Equation (5). The Fourier amplitude coefficients of
each directional component, AR(
η f), AT( η f) and AZ( η f) are then weighted according to the Equation (6). In thes e
equations, M, K, N constants and the angle
θ corresponding to horizontal/vertical displacement ratio have been
valued as 8, 8, 4 and 0.8, respectively (as suggested by [1]). After, it is back-transformed to the time domain
with original phase value and scaled amplitude values,
( )
Z
Af η
′
,
( )
R
Af η
′
and
( )
T
Af η
′
. T he same process i s
rep eated for the o ther windo wing by selec ted moving i nterva l and this pr ocess is co ntinued b y scanning all of
the signals. Finally, filtered signals are obtained as the arithmetic average of the overlapping amplitudes
(Figures 3(b)-10(b)). It has been examined on records having to difference epicentre distance for several win-
dow length and moving interval (Table 3). The original and filtered three component seis mograms of recorded
at ISP (Figure 3), YL VX (Figure 4), ANTB (F igure 5), ISKB (Figure 6), BALB (Fi gure 7), MR MX (Figure 8),
MLSB ( Figure 9) and EDRB (Fig ure 10) stations are compared. The start of time axis is 19:33:00 at all records
in figures.

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 5 July 2014 | Volume 1 |
e724
Table 1. Information of stations used in the application.
Station
Code C oordinates (˚N) (˚E) Epicenter ∆(˚) Azim uth Az(˚)
YLVX 40.34 29.22 8.16 90
MLSB 37.18 27.47 10.07 71
MRMX 40.36 27.35 9.59 89
ISP 37.49 30.31 7.82 70
IS K 41.04 29.04 8.33 95
EDRB 41.51 26.45 10.32 95
BALB 39.38 27.53 9.52 83
ANT B 36.54 30.39 8.15 63
Table 2. The focal parameters of Erzu rum earthquakes used in the application.
Date
(d m y) Origin Time (h min sec) Coordinates (˚N) (˚E) Focal Depth (km) Magnitude Ms
25.03.2004 19:30:46.3 39.92 40.82 10 6.0
Table 3. Parameters used in the analysis.
Station
Code E picent er ∆(km) Data Length (sec ) Window Length (sec ) Window Length/ Moving Interval S moothing Operator
ISP 927 720 75 3.95 9
YLVX 928 950 75 3.95 9
ANT B 950 720 75 3.95 9
IS K 1011 720 90 4.10 7
BALB 1113 720 90 4.50 9
MRMX 1134 720 90 4.50 9
MLSB 1176 720 90 4.50 9
EDRB 1211 720 120 4.80 9
Figure 2 . Location of the event (•) and stations () are shown at tecton ic map of Anatolia [20].

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 6 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 3. Z, R and T componen ts recorded at ISP (Isparta) station. (a) Original r ecords; (b) Filtered
cases.
(a)
0200 400 600 800
-80000
-40000
0
40000
80000
0200 400 600 800
-80000
-40000
0
40000
80000
Z-(ISP)
R-(ISP)
0200 400 600 800
-80000
-40000
0
40000
80000
T-(ISP)
?
0200 400 600 800
-10000
-5000
0
5000
10000
0200 400 600 800
-10000
-5000
0
5000
10000
Z-(ISP)
R-(ISP)
0200 400 600 800
-10000
-5000
0
5000
10000
T-(ISP)
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 7 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 4 . Z, R and T components reco rded at YLVX (Yalova) station. (a) Origi nal records, (b ) Filtered cases.
0200400 600 8001000
-150000
-100000
-50000
0
50000
100000
150000
0200400 600 8001000
-150000
-100000
-50000
0
50000
100000
150000
Z-(YLVX)
R-(YLVX)
0200400 600 8001000
-150000
-100000
-50000
0
50000
100000
150000
T-(YLVX)
(a)
?
0200 400 600800 1000
-20000
-10000
0
10000
20000
0200 400 600800 1000
-20000
-10000
0
10000
20000
Z-(YLVX)
R-(YLVX)
0200 400 600800 1000
-20000
-10000
0
10000
20000
T-(YLVX)
POLAR0ZASYON SÜZGEC0 UYGULANMI VER0LER
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 8 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 5. Z, R and T components recorded at ANTB (Antalya) station. (a) Original records, (b)
Filter ed cases.
0200 400 600 800
-600000
-300000
0
300000
600000
0200 400 600 800
-600000
-300000
0
300000
600000
Z-(ANTB)
R-(ANTB)
0200 400 600 800
-600000
-300000
0
300000
600000
T-(ANTB)
OR0J0NAL VER0LER
(a)
?
0200 400600 800
-30000
-20000
-10000
0
10000
20000
30000
0200 400600 800
-30000
-20000
-10000
0
10000
20000
30000
Z-(ANTB)
R-(ANTB)
0200 400600 800
-30000
-20000
-10000
0
10000
20000
30000
T-(ANTB)
POLAR0ZASYON SÜZGEC0 UYGULANMI^ VER0LER
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 9 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 6 . Z, R and T components recorded at ISKB (Istanbul) station. (a) Original records, (b) Filtered cases.
0200 400600 800
-60000
-40000
-20000
0
20000
40000
60000
0200 400600 800
-60000
-40000
-20000
0
20000
40000
60000
Z-(ISKB)
R-(ISKB)
0200 400600 800
-60000
-40000
-20000
0
20000
40000
60000
T-(ISKB)
OR0J0NAL VER0LER
(a)
?
?
?
0200 400600 800
-10000
-5000
0
5000
10000
0200 400600 800
-10000
-5000
0
5000
10000
Z-(ISKB)
R-(ISKB)
0200 400600 800
-10000
-5000
0
5000
10000
T-(ISKB)
(b)
LQ
LR
LR

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 10 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 7. Z, R and T components recorded at BALB (Balıkesir) station. (a) Original records, (b) Fil-
tered cases.
(a)
0200 400600 800
-90000
-60000
-30000
0
30000
60000
90000
0200 400600 800
-90000
-60000
-30000
0
30000
60000
90000
Z-(BALB)
R-(BALB)
0200 400600 800
-90000
-60000
-30000
0
30000
60000
90000
T-(BALB)
?

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 11 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 8 . Z, R and T components recorded at MRMX ( Marmara) station. (a) Original records, (b) Filtered
cases.
(a)
0200 400 600 800
-80000
-40000
0
40000
80000
0200 400 600 800
-80000
-40000
0
40000
80000
Z-(MRMX)
R-(MRMX)
0200 400 600 800
-80000
-40000
0
40000
80000
T-(MRMX)
OR0J0NAL VER0LER
?
0200 400 600800
-8000
-4000
0
4000
8000
0200 400 600800
-8000
-4000
0
4000
8000
Z-(MRMX)
R-(MRMX)
0200 400 600800
-8000
-4000
0
4000
8000
T-(MRMX)
POLAR0ZASYON SÜZGEC0 UYGULANMI^ VER0LER
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 12 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 9. Z, R and T components recorded at MLSB (Bodrum) station. (a) Original records, (b) Filtered
cases.
0200 400600 800
-80000
-40000
0
40000
80000
0200 400600 800
-80000
-40000
0
40000
80000
Z-(MLSB)
R-(MLSB)
0200 400600 800
-80000
-40000
0
40000
80000
T-(MLSB)
OR0J0NAL VER0LER
(a)
?
0200 400 600 800
-15000
-10000
-5000
0
5000
10000
15000
0200 400 600 800
-15000
-10000
-5000
0
5000
10000
15000
Z-(MLSB)
R-(MLSB)
0200 400 600 800
-15000
-10000
-5000
0
5000
10000
15000
T-(MLSB)
POLAR0ZASYON SÜZGEC0 UYGULANMI^ VER0LER
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 13 July 2014 | Volume 1 |
e724
(a)
(b)
Figure 10. Z, R and T components recorded at EDRB (Edirne) station. (a) Original records, (b) Fil-
tered cases.
0200 400 600 800
-30000
-20000
-10000
0
10000
20000
30000
0200 400 600 800
-30000
-20000
-10000
0
10000
20000
30000
Z-(EDRB)
R-(EDRB)
0200 400 600 800
-30000
-20000
-10000
0
10000
20000
30000
T-(EDRB)
(a)
?
0200 400 600 800
-3000
-2000
-1000
0
1000
2000
3000
0200 400 600800
-3000
-2000
-1000
0
1000
2000
3000
Z-(EDRB)
R-(EDRB)
0200 400 600800
-3000
-2000
-1000
0
1000
2000
3000
T-(EDRB)
POLAR0ZASYON SÜZGEC0 UYGULANMI^ VER0LER
(b)
LQ

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 14 July 2014 | Volume 1 |
e724
4. Discussion and Conclusions
In this study, discrimination filter based on polarization properties has been applied to discriminate a desired
surface wave phase on seismograms rec or de d at sta tio ns ha vi ng ab o ut 1 0 ˚ of epicentre distances. For this purpose,
three-component broadband digital seismograms recorded at eight stations of Bogazici University Kandilli Ob-
servatory and Earthquake Research Institute (KOERI) of Erzurum earthquakes were used and filtered seismograms
were compared by original traces.
It has been found that the window length for the minimal epicentre distance (927 km) is 75 sec (for ISP) and,
the window length for the maximal epicentre distance (1211 km) is 120 sec (for EDRB). As can be seen from
analysed records, window length must be increased as related to the ascending epicentre distance (Table 3 ). Tri-
als related to the surface wave discr imination filter technique have been denoted that window length and moving
interval are significantly effect to the results. In this study, it has been determined that the ratio between the
windo w len gth and movi ng inte rval is i n the inter val o f 3.95 - 4.80 from the analysis o f the r ecord s. References
[6] and [11] have found the ratios of 4.4 and 3.0, respectively. Namely, the conclusions of present study agree
with the results of the preview studies for different epicentre distances.
Love waves at records app lied polarization filter have been ob tained perfectly because the amplitudes on the
tangential co mponent (T) are larger than the amplitudes on the vertical (Z) and radial (R) components in all re-
cords (Figures 3-10). Dominate arrivals in some periods are on the T component and the amplitudes on the Z
component of the ground motion are very small. Namely, total effect of weighting factors has been strengthened
to Love waves at some periods arrive at the station. T his re sult i mplie s t ha t t he e ffects of the first weighting fac-
tors (functions of β ) in Equation (6) are to attenuate transverse-tending energy on the Z and R components and
radial-tendi ng ener gy on t he T-component. The d irecti ons of travel o f obvio us but unide ntifi ed Ra yleigh group s
at records have been determined by simply aiming the process for an azimuth angle about 90˚ and consider ing of
the horizontal a ngles,
β . Therefore, filter performance is low on Z and R component s and in extra cting t he Ra y-
leigh wave.
References
[1] Simons, R.S. (1968) A Surface Wave Particle Motion Discrimination Process. Bulletin of Seismological Society of
America, 58, 629-637.
[2] Gal’perin, E.I. and Frolova, A.V. (1960) Azimuth-Phase Co rrelation for Elliptically Polarized Waves, Izvestiya Soviet
Academy of Sciences. G eophysics Seri es , 2, 195-208. (in Russian)
[3] Shimsoni, M. and Smith, S.W. (1964) Seismic Signal Enhancement with Three Component Detectors. Geophysics, 24,
664-671. http://dx.doi.org/10.1190/1.1439402
[4] Flinn, E.A. (1965) Signal Analysis Using Rectilinearity and Direction of Particle Motion. Proceedings of IEEE , 12,
1874-1876. http://dx.doi.org/10.1109/PROC.1965.4462
[5] Mercado, B.J. (1968) Linear Filtering of Multicomponent Seismic Data. Geophysics, 33, 926-935.
http://dx.doi.org/10.1190/1.1439986
[6] Basa, S.H., Ozer, M.F., Osmansahin, I. and Kenar, O. (1994) Polarization Analysis of Three Component Data. Geo-
physi cs, 8, 77-89.
[7] Alkaz, V.G., Onofrash, N.I. and Perelberg, A.I. (1977) Polarization Analysis of Seismic Waves. Shtiintca Press, Ki-
shinev. (in Russian)
[8] Esmersoy, C. (1984) Polarization Analysis, Rotation and Velocity Estimation in Three Component VSP. In: Toksoz,
M.N. and Stewart, R.R., Eds., Vertical Seismic Profiling—Part B: Advanced Concepts, Geophysical Press, Houston,
236-255.
[9] Jurkevics, A. (1988) Polarization Analysis of Three Component Array Data. Bulletin of Seismological Society of
America, 78, 1725-1743.
[10] Perelberg, A.I. and Hornbostel, S.C. (1994) Applications of Seismic Polarization Analysis. Geophysics, 59, 119-130.
http://dx.doi.org/10.1190/1.1443522
[11] Osmansahin, İ., Özer, M.F. and Sayil, N. (1994) Surface Wave Discrimination Filter Based on Polarization Properties.
Geophysics, 8, 99-104.
[12] Zheng, Y. (1995) Seismic Polarization Filtering: Noise Reduction and Off-Line Imaging. M SC, Universit y of Galgar y,
Galgary.
[13] Patane, D. and Ferrari, F. (1999) ASDP: A PC-Based Program Using a Multi-algorithm Approach for Automatic De-

N. Sayil
OALibJ | DOI:10.4236/oalib.1100724 15 July 2014 | Volume 1 |
e724
tectio n and Location of Local E arthquakes. Physics of the Earth Planetary Interiors, 133, 57-74.
http://dx.doi.org/10.1016/S0031-9201(99)00030-8
[14] Du, Z., Foulger, G.R. and Weijian, M. (2000) Noise Reduction for Broad-Band, Three Component Seismograms Using
Data-Adaptive Polarization Filters. Geophysical Journal International, 141, 8 20-828.
http://dx.doi.org/10.1046/j.1365-246x.2000.00156.x
[15] Franco, R. and Musacchio, G. (2001) Polarization Filter with Singular Value Decompozition. Geophysics, 66, 932-938.
http://dx.doi.org/10.1190/1.1444983
[16] Kutlu, Y.A. (2006) Surface Wave Discrimination Filter Based on Polarization Properties and Its Applications. Ph.D.
Thesis, Karadeniz Technical U niversit y, Trabzon.
[17] Kutlu, Y.A. and Sayil, N. (2013) The Modified Surface Wave Particle Motion Discrimination Process. International
Journal of the Physical Sciences, 8, 395-405.
[18] Haubrich, R.A. and McKen zie, G.S. (1965) E arth Noise 5 to 500 Millicycles per Second. Part 2. Journal of Geophysi-
cal Research, 70, 1429-1440. http://dx.doi.org/10.1029/JZ070i006p01429
[19] Dorman, J. and Prentiss, D. (1960) Particle Amplitude Profiles for Rayleigh Waves on a Heterogeneous Earth. Jour-
nal of Geophysical Research, 65, 380 5-3816. http://dx.doi.org/10.1029/JZ065i011p03805
[20] Saroglu, F., Emre, O. and Kuscu, I. (1992) Active Fault Map of Turkey. General Dir ectorat e of Mineral Res earch and
Exploration, Ankara.