INVERSION OF STRONG MOTION ACCELERATION DATA FOR

DETERMINATION OF 3D ATTENUATION STRUCTURES

A DISSERTATION
Submitted in partial fulfilment of the
Requirements for the award of the degree
Of
INTEGRATED MASTER OF TECHNOLOGY
In
GEOPHYSICAL TECHNOLOGY
By
PRANSHU RATRE

DEPARTMENT OF EARTH SCIENCES

INDIAN INSTITUTE OF TECHNOLOGY ROORKEE
ROORKEE- 247667 (INDIA)
JUNE, 2014

CANDIDATES DECLARATION
I, here by, certify that the work which is being presented in this dissertation entitled Inversion
of Strong Motion Acceleration Data for determination of 3D attenuation structures in
partial fulfilment of requirement for the award of INTEGRATEDM.TECH Degree in
GEOPHYSICAL TECHNOLOGY submitted in the Department of Earth Sciences, Indian
Institute of TechnologyRoorkee, is an authentic record of my own work, carried out under the
supervision of Dr.Anand Joshi, Department of Earth Sciences, IIT Roorkee.The matter embodied
in this dissertation has not been submitted by me for the award of any other degree of this or any
other institution.

Date:

PranshuRatre
Geophysical Technology

Place: IIT RoorkeeDepartment of Earth Sciences

Indian Institute of Technology Roorkee

This is to certify that the above statement made by the candidate is true to the best of my
knowledge.

Dr.Anand Joshi
Professor
Department of Earth Sciences
Indian Institute of Technology Roorkee

CERTIFICATE
I, PranshuRatreherebysolemnly declare that the dissertation entitled Inversion of Strong
Motion Acceleration Data for determination of 3D attenuation structures being submitted
by me towards partial fulfilment of the requirement for the award of Integrated Masters of
Technology in Geophysical Technologydegree is a record of my own work and that I have not
copied the work of any other person(s) including published literature and material from any
website.
Wherever the work of other person(s) has been used, it has been duly acknowledged and quoted
with proper reference to the original work.
I fully understand the implications of plagiarism and that if at any stage the above statement
made by me is found to be incorrect, I shall be fully responsible for my act(s).

Place- Roorkee

Name-PranshuRatre

Date-

Geophyiscal Technology

ACKNOWLEDGEMENT

Firstly, I would like to extend my deep gratitude to my supervisor Dr. AnandJoshi ,whose
constant guidance and encouragement throughout my thesis helped me generate a keen
interest and focus in my work. Through my interactions with him for one year, Ihave learnt
extensively from him on how to approach a geophysical problem. I am greatly thankful for
his efforts and patience without which I would not have accomplished my work.

In the past one year, I have learnt a lot not only about my subject but also I will carry many
life lessons from this experience. I am really thankful to the Department of Earth Science
for giving me this opportunity to explore my field of study.

I would also like to thank my friend Arpit who helped me throughout the project giving me
ideas to improve my results. I would also like to thank my friends Pulkit and Arushi who
were my constant moral support and helped me in writing this report.
Finally I would like to thank my family for giving me the extra push whenever I needed it.

Contents
ABSTRACT...................................................................................................................................................... 1
CHAPTER 1: INTRODUCTION ......................................................................................................................... 2
1.1 Background ......................................................................................................................................... 2
CHAPTER 2: METHODOLOGY ........................................................................................................................ 4
CHAPTER 3: DATA DESCRIPTION ................................................................................................................. 12
CHAPTER 4: CASE STUDY-3D TOMOGRAPHY OF CENTRAL HONSHU (Kanto and Chubu regions), JAPAN . 15
4.1 Tectonic Setting of the Central Honshu Region ................................................................................ 15
4.2 Numerical Experiment for the Data Set ............................................................................................ 17
CHAPTER 5: CONCLUSIONS ......................................................................................................................... 19
5.1 Results and Discussions .................................................................................................................... 19
5.2 Conclusions ....................................................................................................................................... 32
REFERENCES ................................................................................................................................................ 33
APPENDIX .................................................................................................................................................... 34

LIST OF FIGURES
Figure 1: Initial model and model obtained after first iteration having new quality factor value (Qijk) in
different blocks; i, j,k.
Figure 2: Flow chart of the process of inversion (After Joshi et. al. 2010)
Figure 3 : Topographic Map of Central Honshu Region (Ref. http://www.glgarcs.net/)
Figure 4 : Contour of quality factor values for 1 Hz frequency obtained using the data set at the surface (0
km)
Figure 5 : Contour of quality factor values for 1-Hz frequency obtained using the data set at 5 km beneath
the surface
Figure 6 : Contour of quality factor values for 1-Hz frequency obtained using the data set at 10 km
beneath the surface
Figure 7 : Contour of quality factor values for 5-Hz frequency obtained using the data set at surface
Figure 8: Contour of quality factor values for 5-Hz frequency obtained using the data set at 5 km beneath
the surface
Figure 9 : Contour of quality factor values for 5-Hz frequency obtained using the data set at 10 km
beneath the surface
Figure 10 : Contour of quality factor values for 10-Hz frequency obtained using the data set at the surface
Figure 11 : Contour of quality factor values for 10-Hz frequency obtained using the data set 5 km beneath
the surface
Figure 12 : Contour of quality factor values for 10-Hz frequency obtained using the data set 10 km
beneath the surface
Figure 13 : Resolution matrix for 1Hz frequency
Figure 14 : Resolution matrix for 5Hz frequency
Figure 15 : Resolution matrix for 10 Hz frequency

ABSTRACT
Three island arcs, the Northeast Japan Arc, the Southwest Japan Arc, and the Izu-Bonin Arc
meet in the Central Honshu (the Kanto and Chubu regions) region of Honshu Island in Japan.
This region has very complicated geomorphological and geological structures due to intensive
crustal movement. Three dimensional S-wave quality factor (Q) values are obtained for the
region through the inversion of strong motion acceleration data of twelve earthquakes in the
region (M 4 -5.4). The algorithm used is based on the method of inversion given by HASHIDA and
SHIMAZAKI for determination of three-dimensional attenuation coefficients. It has also been used
and further modified by JOSHI et al. (2010).The earthquake data used was recorded by strong
motion stations of Kik-net network. Borehole data has been used as it has least possible site
effect and a high signal to noise ratio. The entire region has been divided into 25, 3-dimensional
blocks having the same thickness but different frequency dependent S-wave quality factor. A
comparison is made between the obtained attenuation structure and the geological features
present in the region and it shows that the obtained structure is capable of resolving relevant
tectonic features present in the region.

CHAPTER 1: INTRODUCTION
1.1 Background
Seismic waves when passing through a medium decrease in amplitude due to various factors like
spherical spreading, mechanical and other loss mechanism. This loss in energy is measured in
terms of attenuation which can be mathematically defined as the inverse of Quality factor (Q)Quality factor is the ratio of stored energy to dissipated energy in one cycle of the wave (wave
length). The heterogeneity and inelasticity in the earth cause the attenuation of seismic waves as
they travel from the source to a particular site. It is observed that as a wave travels from the
source, its travel path and local site conditions effects the amplitude of the wave. It is the
attenuation property of the medium that affects the ground motion of a particular site during
tectonic movements.
The strong motion data is a record of the ground acceleration during an earthquake event. It is
observed that the peak ground acceleration recorded is connected with the S-wave arrival
(HADLEYand ORCUTT 1982). It is also demonstrated by MIDORIKAWA (1980) that the Quality
factor and shear wave velocity can be linked with each other empirically. Thus, by estimating the
Shear wave (S-wave) Quality factor (Q), we can directly estimate the rock properties and seismic
hazard in the region. In this dissertation, S-wave quality factor is utilized to determine the 3-D
attenuation structures for Chbu region in (Central) Honshu region of Japan.
HASHIDA AND SHIMAZAKI (1984) have developed an inversion algorithm to find the attenuation
structure which they used to determine the 3-D attenuation structure in northeastern part of Japan
beneath the Tohoku and Kanto regions (HASHIDA AND SHIMAZAKI, 1985, 1987). This algorithm
was further modified in JOSHI (2006, 2007) and JOSHI et al. (2010) for determination of 3-D
2

attenuation structure. The algorithm proposed in JOSHI (2006), minimizes the error between
initial and final model to derive the attenuation structure.
Attenuation structures are used in seismological studies to construct seismic zonation map of a
region. High seismic attenuation properties signify low seismic hazard in a region.

CHAPTER 2: METHODOLOGY
It is generally assumed while interpretation of strong motion record that peak ground
acceleration A is related to arrival of S-waves that are generated from a point source below the
ground. In order to obtain vertical and spatial distribution of shear wave quality factor Qb(f)
along three different perpendicular directions the study area is divided into 3 dimensional blocks
of different Q(f) values. This is the same as representing the entire study area into rocks having
varying attenuation properties. For computing spectral acceleration values at different
frequencies the relation given by HASHIDA and SHIMAZAKI is used:

( )

( )

(1)

Where:
A(f) : Spectral acceleration values
S(f) : Source spectral acceleration at frequency f
Gr: Geometrical factor (assumed as the inverse of hypocentral distance)
Tijk : time spent in the ijkth block of attenuation coefficient Dijk
g

: amplifying effects at the surface of earth ( = 2.0)

The coefficient of attenuation represents the property of the medium and it is the inverse of the
quality factor. Its values determine the attenuation characteristics of the medium; a high value
causing the wave to attenuate faster and low value allowing easier passage of the wave through
the medium. The parameters Qijk(f) and f represent quality factor and frequency respectively.
The subscripts i and j are used for identifying blocks in X and Y directions which are
4

perpendicular to each other, in the horizontal plane, as shown in Fig.1. The subscript k is used
for identifying block in downward direction. In the iterative process of inversion, an initial
distribution of Qijk,o(f) is assumed within each block. The initial distribution is assumed from
available geophysical data of the region. The value of S(f) is required for forward modelling
using Eq.1. The S(f) is mean of several frequency dependent source strength

(f) calculated at

different observation points op and is represented by following equation.

( )

( )

(2)

In the above relation Aop(f) is the obtained spectral acceleration value at frequency f at the
observation point,op in the surface grid.

Initial Model
Observation points for Acal

Figure 1: Initial model and model obtained after first iteration having new quality factor
value (Qijk) in different blocks; i, j,k.

To compute

(f) from Eq. 2, initial subsurface model defined by velocity and Q in every

layer is required. If the shear velocities are varied in different blocks to be used as independent
parameters, an extra set of data is required. So, to minimize the number of independent
parameter layered velocity model is used as an input model as represented by H ASHIDA and
SHIMAZAKI(1984) and JOSHI (2007). The Scal(f) is computed by using mean value of

(f)

obtained from all observation points. Initial model having attenuation coefficient Dijk,o and Scal(f)
is used to compute Acal(f) and is given by following relation:

( )

( )

Assuming the observed spectral value at ij observation point to be

observed spectral acceleration

(3)

(f), the relation between

(f), actual source strength So(f) and actual attenuation

coefficient Dijk is given by the following expression.

( )

( )

, (4)

The parameter Tijk is the time spent by the ray from source to the observation point in the 3-D
block given by subscript ijk. On dividing equation 4 and 3 we get:
( )
( )

( )

( )

(5)

In the above equation the parameter E defines the computational error which can be due to
inadequate model parameterization. Replace Dijk Dijk,o= Dijk in Eq. 5. Hence in Eq. 6 Dijk
represents the difference between initial and final attenuation coefficient in the ijkth block.

( )

( )

( )

(6)

( )

The above equation is a non-linear equation which can be converted into linear equation by
taking its natural logarithm. This gives following form of above equation:

( )
( )

Assuming,

( )
( )

(7)

( )
( )

), as acceleration residual, Eq. 7 can be written as:

( )

( )

(8)

In the above equation the LHS is known and the parameter Tijk is calculated by ray theory. The
analytical approach given by LEE and STEWART(1981) can be used for tracing ray from source to
observation point. Many equations have been obtained for particular earthquakes. The equation
representing data for first earthquake can be written as (JOSHI2007):

(9)
The parameter e in the Eq. 9 denotes error estimate. The parameter

denotes time spent by

the ray from first event in various blocks to reach the station represented by superscript op from

event 1. The data of N earthquakes can be represented in the below matrix given by (JOSHI 2007).

the above matrix equation can be written as:

G m = d.

(10)

Matrices m andd represents model parameters and data respectively. The variation of
subscripts i,j and k ranges from 1 to L, M and K respectively, and the variation of superscripts o
and p ranges from 1 to Ls and Ms respectively. The total number of blocks Nb=L*M*K and total
number of observation points are Ns= Ls*Ms. Hence we have Ns=Ne*N and Np=N+Nb as
equations and parameters respectively. Since matrix G is a rectangular matrix hence by using
Newtons method we get the following model matrix:
m = (GTG)-1 GT d

(11)

In practical case some of the eigenvalues of GTG are very small hence the variance of solution
becomes unacceptably large.Damped least square method of LEVENBERG(1944) is used to avoid
9

the above difficulty because it does not require eigenvalues analysis. The inversion equation
using damped least square method is:
mest = (GTG + I)-1 GT d

(12)

In the preceding equation is the damping factor and Iis a unit matrix. The above equation can
be solved by minimizing |d-Gm|+ mt(I)m instead of |d-Gm|2, where (I) is a diagonal matrix
with damping factors. The is chosen based on method given by DIMRI(1992) and used by
JOSHI(2007).
The estimation of unknown quantities like

( ( )

( ))and Dijk =(Dijk Dijk,o) can be

done by inversion algorithm. This is an iterative procedure in which after each inversion Qijk,o is
replaced by Qijk which can be represented by following expression:
Qijk(f)=Qijk(f)+Qijk,o(f)

(13)

In above equation Qijk(f) represents small change in quality factor. In each consecutive iteration
the value of Qijk(f) is taken as initial model, replacing Qijk,o(f). Now themodified initial model
So(f) is again calculated at every observation point and its mean value is used to calculate Acal(f).
This process againgenerates a new set of equations for inversion. The inversion is repeatedly
performed until solution with minimum error is reached. The entire inversion can be summarized
in the form of a flow graph in the Figure 2.

10

Figure 2: Flow chart of the process of inversion (After Joshi et. al. 2010)

11

CHAPTER 3: DATA DESCRIPTION

The nationwide seismic observation networks have been constructed by National Institute for
earth science and disaster prevention. The sub set of HI- net (high sensitivity seismograph
network) is kik net (KIBAN kyoshin network) and it consists of 660 strong motion observation
stations. This network also has surface as well as borehole data. The KiK-net stations have
borehole of 100-200m in depth. Records from borehole sensor at different KiK stations have
been used to use records which are free from site amplifications. Each KiK station is having a
digital strong motion seismograph having a wide frequency band and wide dynamic range of
measurable acceleration which is 2000 gal. The frequency of sampling the data is 200 Hz. The
Kik-net stations are equipped with double three component accelerometers of which one of them
is placed at the surface and other is placed in the borehole at every station. Hence strong motion
seismographs of Kik-net have 6 acceleration sensors. The acceleration data has been acquired
from official website of Kik-net. In the present work 12 earthquakes readings from 30 stations
across Honshu region have been used. The details of the 30 stations that have recorded these
earthquakes have been given in Table 1. Input velocity model has been taken from HONDA et. al.
(2005). The locations of earthquakes and recording stations are shown in Fig 3. The records have
been processed by procedure suggested by BOORE and BOMMER 2005. The processing involves
frequency filtering, baseline correction and instrument scaling. A MATLAB script is developed
for the pre-processing of data. (Appendix 1). The size of study area is 110*110 km in central
Honshu region of Japan. For attenuation tomography the whole region is divided into 5, 5 and 3
rectangular blocks aligned along X, Y and Z axis. This provides with 36 points of observation. In
present scenario strong motion data of 12 earthquakes gives 432 equations and 87 parameters.

12

Table 1 Detail of the recording stations used in present work (http://www.kik.bosai.go.jp).

STATION CODE

LATITUDE (N)

LONGITITUDE (E)

GIFH11
GIFH14
GIFH15
GIFH16
GIFH18
GIFH19
GIFH20
GIFH22
GIFH24
GIFH27
GIFH28
NGNH03
NGNH07
NGNH08
NGNH09
NGNH10
NGNH11
NGNH12
NGNH13
NGNH14
NGNH15
NGNH16
NGNH17
NGNH18

35.4865
36.2493
36.1338
36.094
35.8991
36.0216
35.7991
35.6682
35.6401
35.4527
35.4571
35.4786
36.7434
36.2541
36.2859
35.9632
35.9157
35.9696
35.5143
35.3095
36.0088
35.9465
36.1425
35.9324

137.2464
137.5174
137.2208
137.3438
137.1495
137.3906
137.2531
137.1054
137.3187
137.004
137.4706
137.7346
138.376
137.8591
138.2491
137.7669
138.3052
138.4797
137.8767
137.6261
137.9305
138.1848
138.5504
137.595

13

Table 2 :Hypocentral parameters of the events used in present work

Events
2003/05/18
2004/07/27
2012/01/28
2002/12/04
2005/01/02
2008/06/13
2004/01/11
2008/05/12
2012/01/28
2013/02/15
2003/04/01
2004/01/11

Origin Time
(hh:mm:sec)

Epicenter
Depth(km)

03:23:00
00:55:00
07:43:00
08:09:00

7.000000
11.000000
18.000000

01:30:00
11:21:00
16:57:00
11:47:00
07:39:00
18:18:00
09:25:00
16:57:00

6.000000
13.000000
8.000000
10.000000
18.000000
11.000000
8.000000
8.000000

8.000000

14

Magnitude
Latitude(N)

Longitude(E)

4.5
4.5
5.4
4.2

35.866000
35.758000
35.488000

137.595000
137.107000
138.977000

35.870000

137.594000

4.2
4.7
4.0
4.0
4.9
4.2
4.1
4.1

35.868000
35.910000
36.398000
35.792000
35.492000
35.777000
35.941000
35.940000

137.574000
137.703000
137.980000
137.973000
138.978000
138.038000
137.520000
137.529000

CHAPTER 4: CASE STUDY-3D TOMOGRAPHY OF CENTRAL HONSHU

(Kanto and Chubu regions), JAPAN
4.1 Tectonic Setting of the Central Honshu Region
Honshu Island is a mountainous and volcanic region, prone to frequent tectonic activities. The
Great Kantearthquake caused heavy damage in Tokyo in September 1923, and the Tohuku
earthquake of March 2011 moved the north-eastern part of the island as much as 5.3 m.
The broadest region of the Honshu Island is qualified as Central Honshu having the Kanto and
Chubu regions in it. It is characterized by the meeting of three island arcs namely, the North-east
Japan Arc, The South-west Japan Arc, and the Izu-Bonin Arc. The region is divided by the
Itoigawa-Shizuoka Tectonic Line (ISTL), the Median Tectonic Line (MTL), and the volcanic
front. The South-west and the North-east Japan Arc demarcatedby the ISTL, which is a faultzone running north-south in the Chubu region of Honshu. The MTL consists of right lateral
strike slip faults, running for over a 1000 km. The MTL divides the South-west Japan Arc into
two zones: Outer zone (lying on the Pacific side) and Inner zone (the continental side). The
volcanic front runs along the extension of the North-east Japan Arc turning towards the IzuBonin Arc in front of the ISTL. Similar to ISTL, the volcanic front divides the North-east Japan
Arc into Inner (Northern side) and Outer (Southern side) Arcs.
On the eastern side of ISTL lies the Kanto Plain, which is the largest plain in Japan. A thick layer
of Quaternary sediments including huge amount of volcanic ash cover the plain. The volcanic
ash comes from the volcanoes like Hakone and the Fuji which lie on the western side of the
plain. ISTL serves as the western boundary of Fossa Magna, a tectonic depression zone, which
subsided when the Japanese islands displaced from the continent margin to the present
15

locationand then uplifted. The Fossa Magna depression is very deep; the Neogene sediments
accumulated in the region are more than 5 km thick y over 10 km thick in some regions in the
southern part of the zone. Large uplift rate have generated lots of faults and folds in the
sedimentary rocks. During the expansion of Sea of Japan, intensive volcanism led to the
deposition of volcanic rocks in the Fossa Magna.

Figure 3 : Topographic Map of Central Honshu Region (Ref. http://www.glgarcs.net/)

16

4.2 Numerical Experiment for the Data Set

The computer software named 3-D frequency dependent Q structure (JOSHIet. Al. 2010) has been
used in this thesis. This algorithm takes the following inputs: Initial velocity model, initial Q (f)
model, coordinated of the hypocentres of earthquakes, spectral acceleration values at observation
points and the reference frequency at which attenuation structure is estimated. The spectral
acceleration values at the grid points are obtained through the interpolation of observed values at
various recording stations. From the data of 12 earthquakes, 432 spectral acceleration values at
36 observation points for a particular frequency are obtained.
Ratio of the observed and calculated spectral acceleration at source, and final value of Q (f) at
different blocks is given as the output of the program. Root mean square error for data (rms)dat
and the model matrix (rms)mod is also computed for each iteration. The resolution matrix and the
covariance matrix check the fitness of the model matrix which is found out using the damped
least square inversion scheme. The resolution matrix R and the covariance matrix C can be
represented by the expression below: (HAYDAR and MITCHELL 1990).

)
(

(14)

(15)

Where, d 2is the variance of error in data given by :

..(16) & (17)

17

Resolution matrix provides the association between estimated and actual model. A Unique
solution will be obtained when R is a unit matrix. Deviation of R from unit matrix shows poor
resolution. In this thesis, the total number of parameters (Np) is 87 and amongst these, 12
comprises of the source acceleration spectra of the events in consideration and rest 75 signify the
attenuation coefficient of various blocks. When all errors are minimized simultaneously in a
model, a reliable solution can be obtained. To arrive at a model that simultaneously minimizes
rmsdat and rmsmod, normalized value of the two variables is used :

(18)

18

CHAPTER 5: CONCLUSIONS

5.1 Results and Discussions

Three dimensional attenuation structures have been found for frequencies 1, 5 and 10 Hz. The
subject are consists of the ISTL and MTL. On the eastern side of ISTL, Quaternary sediments
along with huge amount of volcanic ash cover the region. High quality factor values were
determined on the eastern side as compared to the western side of ISTL, which is because of the
existence of sediments in the depressed area on the eastern side. Also, it is observed that the
surface layer has low quality factor values as compared to the subsurface layers. This is due to
the presence of lose sediments in the upper layer. As we go deeper, the trend of contour lines
become parallel to the ISTL, revealing that the Q structure obtained matches with the geological
structure at different depths from 0-10 km.

19

Figure 4 : Contour of quality factor values for 1 Hz frequency obtained using the data set
at the surface (0 km)

20

Figure 5 : Contour of quality factor values for 1-Hz frequency obtained using the data set
at 5 km beneath the surface

21

Figure 6 : Contour of quality factor values for 1-Hz frequency obtained using the data set
at 10 km beneath the surface

22

Figure 7 : Contour of quality factor values for 5-Hz frequency obtained using the data set
at surface

23

Figure 8: Contour of quality factor values for 5-Hz frequency obtained using the data set at
5 km beneath the surface

24

Figure 9 : Contour of quality factor values for 5-Hz frequency obtained using the data set
at 10 km beneath the surface

25

Figure 10 : Contour of quality factor values for 10-Hz frequency obtained using the data
set at the surface

26

Figure 11 : Contour of quality factor values for 10-Hz frequency obtained using the data
set 5 km beneath the surface

27

Figure 12 : Contour of quality factor values for 10-Hz frequency obtained using the data
set 10 km beneath the surface

28

Figure 13 : Resolution matrix for 1Hz frequency

29

Figure 14 : Resolution matrix for 5Hz frequency

30

Figure 15 : Resolution matrix for 10 Hz frequency

31

5.2 Conclusions
The 3 Dimensional attenuation structure is obtained for the Central Honshu region on the basis
of Q(f) at varying frequencies by using strong motion data recorded by KiK-net, Japan network
for 12 earthquakes.
The attenuation structure thus obtained gives comparable trends with the geological structure of
the study area. The results, obtained prove the robustness of the inversion algorithm developed
for the determination of attenuation structure in a region.

32

REFERENCES
BOORE, D. M., and BOMMER, J. J. (2005), Processing of strongmotion accelerograms: needs, options and
consequences, SoilDynEarthq Eng. 25, 93115.
DIMRI, V. P. (1992), Deconvolution and inverse theory: applicationto geophysical problems. Elsevier
Science, Amsterdam. 230.
HASHIDA, T., and SHIMAZAKI, K. (1984), Determination of seismicattenuation structure and source
strength by inversion of seismicintensity data: method and numerical experiment, J Phys Earth.
32, 299316.
HASHIDA, T., and SHIMAZAKI, K. (1985), Seismic tomography: 3-Dimage of upper mantle attenuation
beneath the Kanto district,Japan, Earth and Planet. Science letters, 75(4), 403409.
HASHIDA, T., and SHIMAZAKI, K. (1987), Determination of seismicattenuation structure and source
strength by inversion of seismicintensity data: Tohoku district, the northeastern Honshu, J. Phys.Earth.
35, 5792.
HAYDAR, J., SHUKRI, A. l., and MITCHELL, B. J. (1990), Threedimensional attenuation structure in and
around the new Madridseismic zone, Bull Seismol Soc. Am. 80, 615632.
HONDA, R., AOI, S., SEKIGUCHI, H., MORIKAWA, N., KUNUGI, T., andFUJIWARA, H. (2005), Ground
motion and rupture process of the2004 mid Niigata Prefecture earthquake obtained from strongmotion
data of K-NET and KiK-net.
JOSHI, A. (2006), Three dimensional attenuation structure of thecentral seismic gap region of Himalaya
obtained from inversionof seismic intensity data, Curr Sci. 90, 581585.
JOSHI, A. (2007), Inversion of seismic intensity data for the determinationof three-dimensional
attenuation structures in thecentral gap region of Himalayas, Nat Hazards. 43, 129146.
JOSHI, A., MOHANTY, M., BANSAL, A. R., DIMIRI, V.P., and CHADHA,R.K. (2010), Use of spectral
acceleration data for determinationof three-dimensional attenuation structure in Pithoragarh regionof
Kumaon Himalaya, J. Seismol. 14, 247272.
LEE, W. H. K., and STEWART, S. W. (1981), Principles and applicationsof microearthquake
networks.Academic, New York.293.

33

APPENDIX
MATLAB PROGRAM FOR PRE-PROCESSING OF DATA
%%%%%%%%%%%%%%------%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%SPEC() function used in this program to find the elastic response
spectra was developed by @ MostafaTazarv, Carleton University, May
2011
clear
clc
ext=[{'*.NS1'} {'*.EW1'}];
%%directory containing the strong motion data files for all
earthquakes and different stations
eq_dir='D:\Dissertation\eqdata\';
%% directory containing the processed values
creat_dir='D:\Dissertation\eqdata\data\processedValues\';
if(exist(creat_dir,'dir')==0)
mkdir(creat_dir);
end
k=0;
l=1;
n=1;
%%reading coordinates from file
eq_file=dir([eq_dir,ext{:,1}]);
file{:,1}=eq_file(1).name(7:16); %% for checking which files we have
read already so as to avoid writing multiple values of eqcoord
eq_filename=[eq_dir,eq_file(1).name];
socoord(1,1:2)=textread(eq_filename,'%*s %f',2,'headerlines',1);
socoord(1,3)=textread(eq_filename,'%*s %*s %f',1,'headerlines',3);
for i=1:length(eq_file)
eq_filename=[eq_dir,eq_file(i).name];
chk=eq_file(i).name(7:16);
mat=strcmp(chk,file);
34

if max(mat)==0
l=l+1;
file{:,l}=eq_file(i).name(7:16);
n=n+1;
socoord(n,1:2)=textread(eq_filename,'%*s
%f',2,'headerlines',1);
socoord(n,3)=textread(eq_filename,'%*s %*s
%f',1,'headerlines',3);
end
c=textread(eq_filename,'%*s %*s %f',2,'headerlines',6);
%% Considering stations only in the study area : Lat-35N-37N and
%% Long-137E-139E
if 35<=c(1) && c(1)<=37&&137<=c(2)&& c(2)<=139
k=k+1;
station{:,k}=eq_file(i).name(1:16);
stcoord(k,:)=textread(eq_filename,'%*s %*s
%f',2,'headerlines',6);
end
end
%%writing coordinates of station and source in a separate file
coord_filename='D:\Dissertation\eqdata\coordCoordinates.doc';
coorfid=fopen(coord_filename,'w');
fprintf(coorfid,'Station Coordinates\r\n');
fprintf(coorfid,'SATION
LAT
LONG\r\n');
for i=1:length(stcoord)
fprintf(coorfid,'%s \t %f \t %f \r\n',station{:,i},stcoord(i,:));
end
fprintf(coorfid,'Source Coordinates\r\n');
fprintf(coorfid,'LAT
LONG
Depth\r\n');
for i=1:length(socoord)
fprintf(coorfid,'%f \t %f \t %f\r\n',socoord(i,:));
end
fclose(coorfid);
%%reading data, applying baseline corrections, writing the corrected
%%data
%%in a separate file of the same name.

%%considering only stations whose

35<=lat<=37 and 137<=long<=139

%%Reading first NS file and then the corresponding EW file and making
35

%%corrections simultaneously--taking geometric mean and thus final

%%acceleration values!!
for i=1:length(eq_file)
eq_NS=[eq_dir,eq_file(i).name];
eq_EW=[eq_dir,eq_file(i).name(1:16),'.EW1'];
c=textread(eq_NS,'%*s %*s %f',2,'headerlines',6);
disp(['Scanning ',num2str(i), ' out of
',num2str(length(eq_file))]);
if 35<=c(1) && c(1)<=37&&137<=c(2)&& c(2)<=139
dataNS=textread(eq_NS,'%f','headerlines',17);
dataEW=textread(eq_EW,'%f','headerlines',17);
[sfNS(1), sfNS(2)]=textread(eq_NS,'%*s %*s
%f(gal)/%f',1,'headerlines',13);
[sfEW(1), sfEW(2)]=textread(eq_EW,'%*s %*s
%f(gal)/%f',1,'headerlines',13);
dataNS=dataNS*(sfNS(1)/sfNS(2));
dataEW=dataEW*(sfEW(1)/sfEW(2));

%%multiplying data to
%%scale factor

spa_filename=[creat_dir,eq_file(i).name(1:16),'.dat'];
spa_fid=fopen(spa_filename,'w');
meanNS=sum(dataNS)/length(dataNS);
meanEW=sum(dataEW)/length(dataEW);
ftNS=textread(eq_NS,'%*s %*s %3f %*[^\n]',2,'headerlines',10);
ftEW=textread(eq_EW,'%*s %*s %3f %*[^\n]',2,'headerlines',10);
ftNS(1)=1/ftNS(1);
ftEW(1)=1/ftEW(1);
timeNS=0:ftNS(1):(ftNS(2)-ftNS(1));
timeEW=0:ftEW(1):(ftEW(2)-ftEW(1));
%%Applying corrections to the acceleration data
newAccNS=dataNS-meanNS;
newAccEW=dataEW-meanEW;
[x,y] = butter(3,[.04,.99]);
acc_filteredNS = filtfilt(x,y,newAccNS);
acc_filteredEW = filtfilt(x,y,newAccEW);
clear Spa;
36

%%Finding Spectral Acceleration using SPEC()

[TNS,SpaNS]=SPEC(ftNS(1),acc_filteredNS,5,9.81,2);
[TEW,SpaEW]=SPEC(ftEW(1),acc_filteredEW,5,9.81,2);
Spa(:,1)=TNS;
Spa(:,2)=sqrt(SpaNS.^2 +SpaEW.^2);
for n=1:length(Spa)
fprintf(spa_fid,' %g \t %g \t \r\n',Spa(n,:));
end
end
end
spa_file=dir([creat_dir,'*.dat']);
%%Directory that will contain the spectral acceleration files for
%%different frequencies
Hzdir='D:\Dissertation\eqdata\data\processedValues\final\';
if(exist(Hzdir,'dir')==0)
mkdir(Hzdir);
end
a1fid=fopen([Hzdir,'acc1HZ.dat'],'w');
a5fid=fopen([Hzdir,'acc5HZ.dat'],'w');
a10fid=fopen([Hzdir,'acc10HZ.dat'],'w');
%%Spectral acceleration files have been prepared for 1Hz, 5Hz, 10Hz
%%frequencies
for i=1:length(spa_file)
spafilename=[creat_dir,spa_file(i).name];
original=[eq_dir,spa_file(i).name(1:16),'.NS1'];
coord(1,1:2)=textread(original,'%*s %f',2,'headerlines',1);
coord(1,3)=textread(original,'%*s %*s %f',1,'headerlines',3);
coord(1,1)=coord(1)-35;
coord(1,2)=coord(2)-137;
stc(1,1)=stcoord(i,1)-35;
stc(1,2)=stcoord(i,2)-137;
t=textread(spafilename,' %f %*[^\n]');
s=find(t==1);
a1=textread(spafilename,'%*f %f ',1,'headerlines',(s-1));
fprintf(a1fid,' %g \t %g \t %g\t%g \t %g \t %g
\r\n',coord(1,:),stc(1,:),a1);
s=find(t==0.2);
a5=textread(spafilename,' %*f %f ',1,'headerlines',(s-1));
fprintf(a5fid,' %g \t %g \t %g\t%g \t %g \t %g
\r\n',coord(1,:),stc(1,:),a5);
37

s=find(t==0.1);
a10=textread(spafilename,'%*f %f ',1,'headerlines',(s-1));
fprintf(a10fid,' %g \t %g \t %g\t%g \t %g \t %g
\r\n',coord(1,:),stc(1,:),a10);
end
fclose(a1fid);
fclose(a5fid);
fclose(a10fid);
freq=[1 5 10];

%%Frequencies considered

for i=1:3
acc=load(['D:\Dissertation\eqdata\data\processedValues\final\acc',
num2str(freq(i)),'HZ.dat']);
[A,B]=sort(acc(:,1));
accnew=acc(B,:);
fid=fopen([Hzdir,'a',num2str(freq(i)),'final.dat'],'w');
eqs = unique(accnew(:,1));
for j = 1:length(eqs)
eq = eqs(j)
data_for_eq = accnew(accnew(:,1) == eq,:);
[xq,yq] = meshgrid(0:.4:2, 0:.4:2);
aq = griddata(data_for_eq(:,4),data_for_eq(:,5),
data_for_eq(:,6),xq,yq,'nearest');
size=numel(aq);
aq=reshape(aq,[size 1]);
for n=1:length(aq)
fprintf(fid,' %g \r\n',aq(n));
end
end
end

38