Bull. Disas. Prev. Res. Inst., Kyoto lJniv., Vol.

33, Part 2, No 298, June, l9B3

Semi-EmpiricalEstimation of Strong Ground Motions

During Large Earthquakes

By K o j i ro InrruR e

( l v l a n u s c r i p tr e c e i v e dM a r c h 2 2 , l 9 B 3 )

Abstract

A synthesisme thod is developed for estimating deterministically strong motions during the
mainshock, using the records o[small eventssuch as foreshocksand aftershockswhich occurred
within the area of the mainshock fault. This synthesisformulation is based on the kinematic
source model o[ Haskell type and the similarity law of earthquakes. The parame ters for this
synthesisare determined to be consistentwith the scaling relations between the moments and
the fault parameters such as fault length, width and dislocation rise time. If the ratio of the
mainshock moment Mo to the small event one Mo, is assumed to be lf 3, then the mainshock
fault can be divided into Nx N elements, each dimension of which is consistent with that of
the small event and N events at each element may be sr-rperposed with a specific time delay
to correct the difference in the rise time between the mainshock and the small event and to
keep a constant slip velocity between them. By means of this method, the mainshock
velocity motions are synthesized using the small event records obtained by velocity-type-
strong-motion-seismographs for l9B0 fzu-Hanto-Toho-Oki Earthquake (lvl:6.7). The
resultant synthesizedmotions show a good agreement with the observed ones in the frequency
range lower than I Hz. Further, the synthesisformr-riationis improved to be applicable to
the higher frequency motions, especiallyacceleration motions. This revised synthesisfor the
higher frequency motions is effective when we use the records lrom the small event having the
lault length L,:Vr-r(27: r u p t u r e v e l o c i t y a n d r : r i s e t i m e o f m a i n s h o c k ). T h e s y n t h e s i z e d
accelerogramsby this revised method are in good agreement with the observed ones in the
t'requency range up to 5 Hz.

l. fntroduction

The investigation of the synthesis of strong ground motions in the near field
has significantlv lagged, compared with that of long period motions in the far field.
This is caused by the difficulties of theoretical treatment for high frequency motions
included in the strong motions. The investigators for earthquake engineering have
been concerned with the strong motions lrom the need of engineering. Therefore,
the input motions usually used for the evaluation of earthquake resistant design
criteria have been synthesized for some time. independently of the physical con-
siderations of the earthquake source. Recently, seismologists have begun to take an
active interest in strong motions to study the details of faulting, as strong motion data
have been accumulating in the near field. On the other hand, many investigators
concentrate their attention of engineering interest on reliable estimates of the strong
6-l K. IIIIKL'II.I

motions for earthquake resistant design of critical structures. The studv of strons
motions currentlv is one of the most remarkabie subiects for seismology ancl
earthquake engineering.
In this study, we have the pr.rrposefor engineering seismology ro develop a relia-
b l e a n d p r a c t i c a l s y n t h e s i sm e t h o d o f s t r o n g m o t i o n s . b a s e d o n c a r e f u l c o n s i c l e r a r i o n s
of the physical properties of the earrhquake fault.
T h e f i r s t s u c c e s s f u la t t e m p t f o r t h e o r e t i c a l c a l c u l a t i o n o f s t r o n g m o r i o n s w a s
made by Aki (1968)l)and Haskell (1969)", using kinematic source model, given
by a propagating dislocation over a fault plane in an infinite homogeneous medium.
Their source models are parameterized by five factors, fault length, fault wiclth,
rupture velocity, final offset of dislocation and rise time, which are essentialfor a cteter-
ministic fauit model. Kawasaki ec al. (1972)3) gave exacr expressions of seismic
m o t i o n s d u e t o a d o u b l e c o u p l e p o i n t s o u r c ei n a s e m i - i n f i n i t e m e d i u m . S a t o ( l 9 7 8 ) * ,
proposed an approach to derive exact expressions of a series of 'rays' for layered
media by applying the Cagniard-deHoop methocl, and Sato and Hirata (lgB0)s)
gave a new approach using integral evaluation'to compute the seismic motions for
layered media including the contributions lrom dispersive surlace ,vvaves. Heaton
and Helmberger (1979)6) succeeded also in synthesizing strong motions on the basis
of a generalized ray theory with the Cagniard-deHoop method for layered models.
B o u c h o n ( 1 9 7 9 ) 7 )d e v e l o p e d a m e t h o d t o c o m p u t e s t r o n g m o t i o n s f o r a p r o p a g a t i n g
fault in layered media, based on a discrete wave number method. These synthesis
methods calculating the strong rnotions based on a deterministic fault model have
been successful in lorv frequency ranges (<l Hz). At high frequencies these
methods underestimate the strong motions. when a coherent rupture propagation is
assumed. Indeterminable lactors in the source and the propagating medium may
strongly influence the high frequency ground motions. Hartzel and Helmberger
( 1 9 8 2 ) 8 )a t t e m p t e d t o d e t e r m i n e a l o c a l i z e d a r e a o f l a r g e r d i s l o c a t i o n , b a s e d o n t h e
analysis of some excellent set of records obtained in the 1979 Imperial Valley,
California earthquake and the highly accurate calculation of the discrete wave
n u m b e r i f i n i t e e l e m e n t m e t h o d ( A l e k s e e va n d M i k k a i l e n k o , 1 9 7 9 ; s , . T h e y p r o p o s e c l
a more complex model r,vith two localized sources better to explain the data.
;\ different approach to estimate strong motions is based on inhomoseneous
lault models such as (barrier model) (.-\ki, l979) rr), and (asperity model)
(Macgarr, l98l)tr'. Barriers cause irregular distribution of slip cluring faulting
and a consequence of this is an irregular distribution of stressdrop. The numerical
experirnents performed by Das and Aki (1977;t) and lvlikumo and Miyatake
(1978)ri' demonstrated clearly that barriers control the complexity of rupture and
thev are responsible for the generation of high frequency radiation. Papageorgiou
a n d A k i ( l 9 8 l ) r ' " c o n s t r u c t e d a n e a r t h q u a k e s o u r c em o d e l w h i c h p r o v i d e a c o m p l e t e
description of acceleration power spectra of direct body waves. They call it a specific
barrier model. The fault surf,ace is visualized as composecl of an ag.sregate of
circttlar cracks, and the strong motions are assumed to be generatecl by the stationary
 I

Seni-Empirical Estimatiort o.f Strons Crotmd fu[otions Durins Large Earthquakes 65

occurrence of these localized ruptures as the rupture front propagates. Boatwright

( 1 9 8 2 ) r s )c o n s t r u c t e d t h e s a m e m o d e l f o r t h e f a r - f i e l d a c c e l e r a t i o n b y c o m b i n i n g t h e
Nladariaga (1977)16) theory for the high-frequency radiation from crack models
of laulting with a simple statistical source model. Based on these stochastic f,ault
models. however, we cannot estimate deterministically the waveforms of the strons
motions.
Another approach was proposed bv Hartzell (1978)i?' to synthesize strong
motions, utilizing observed seismograms from small events as Green functions.
It is a most advantageous method because the Green functions include complex
effects of the dynamical rupture process on the fault as well as heterogeneous
structures around the source and an observation site, which are extremely cumber-
some to evaluate. However, Hartzell's method has some problems which need to
be improved. For example, in his method, the discrepancy between the dislocation
rise time of a large event and that of a small event is not taken into account and the
physical meaning of the scale factor Q is uncertain. Kanamori (1979)til and
Hadley and Helmberger (1980)rs)attempted to predict the strong motions from
large earthquakes applying Hartzeil's method. In Japan, Irikura and Muramatu
(1982;zo', Imaeawa and Mikumo (1982)2t,, and Iida and Hakuno (1982;zz,
attempted to synthesize the mainshock motions using small shock motions such as
foreshocks and aftershocks. Irikura and Muramatu, and Imagawa and Mikumo
improved Hartzell's method by introducing the phase delayed summation with
a specific time function to correct for the difference in the source time function
between the mainshock and the small events. From the comparison with the
observed seismograms of the mainshock, Irikura and Muramatu succeeded in the
synthesis of strong motions lower than I Hz for l9B0 Izu-Hanto-Toho-Oki Earth-
quake (M:6.7). On the other hand, Imagawa and Mikumo indicated that synthe-
sized waveforms provide a satisfactory agreement to long-period components longer
than 5 sec for the 1969 central Gifu Earthquake (M:6.6) and a stochastic fault
model has to be introduced for shorter-period motions (T:l-2 sec), lor example
such as the variation of the rupture velocitv on the fault plane.
In this study, a semi-empirical svnthesis method for estimating the mainshock
motions from records of small events has been formulated. based on the kinematic
source model of Haskell type and the similarity law o{'earthquakes. The parameters
for the synthesis are determined to be consistent with the scaling relations between
moments and fault parameters such as fault length, width and dislocation rise time.
The extent of validity and applicability of this sirnthesismethod is discussedin two
ways, (l) the numerical check of validitv of the synthesis formulation, (2) the
comparison betlveen the synthesized mainshock motions and the observed ones.
As far as numerically calculated, we have found that smaller events are more
preferable for utilization of the synthesis, because their source sizesapproach a point
as events are smaller and their records are suitable as Green functions for mainshock
motions. However, accurate records {rom smaller events may be restricted to
 K. IRIKL:R.)

a narrower and higher frequency range, because amplitudes at low frequencies

become smaller than those of ground noises. In addition another problem has been
reported. The scaling law of seismic spectra for smaller events sho'wssome departure
from the similarity assumption, which was studied to explain fairly weli observations
of earthquakes with magnitude greater than 6 (Aki, 196723), L972zt>, Kanamori
and Anderson, 19752s' ). Thus, in order to synthesize large earthquake motions,
we need to use records from small earthquakes larger than appropriate sizes. Then,
we have to check the valiclity of the similarity relation between large and small
earthquakes.
By means of the synthesis method in this study mainshock velocity motions are
synthesized using the records from small earthquakes such as foreshocks and
aftershocks obtained by the velocity-type-strong-motion-seismographs for the case
of l9B0 Izu-Hanto-Toho-Oki earthquake (the mainshock with fuI:6.7). The
synthesized motions show a good agre,,ementwitlr the observed ones in the period
range longer than L...,,,,!h?rt.is,_the rise time of the,lnpinshock. The amplitudes
of the synthesized motion! ,.;a";;'d..o,ii.l;;jri;. than those of the observed ones.
Further, the synthesisformulation has been improved to be applicable for shorter
period motions, especially for strong acceleration motions, not statistically but
deterministically. This revised method is based on an idea that the slips on the f,ault
plane during the mainshock are able to be^approximately replaced to the sparial
distribution of the slips during small events. When we use the records from the small
events having the fault length L,:V,.r (.V,, rupture velocity; r, rise time of
mainshock), we can synthesize effectively the mainshock motions without de-
creasing the amplitudes at the period range shorter than I sec. The synthesized
accelerograms by means of this revised method sholv a good agreement with the
observed accelerograms in the frequency range up to 5 Hz. This synthesis method
implies the possibility of the prediction of strong ground motions for future large
earthquakes using observed seismograms from small events which have occurrecl
within or near the presumed fault area.

2. Relations between the Ground Motions frorn Large Events and Those
frorn Srnall Events

2.1. Sirnilarity of Earthquakes

The present estimation of strong motions from'large earthquakes using the
records lrom small events is based on the similarity assumption for earthquakes.
We sutnmarize here the similaritv conditions between large and small events to
prepare formulations for the synthesis of strong morions.
The similarity assumption was first introduced by Tsuboi2o) in the idea
'earthquake
volume'. He derived that the total energy of an earthquake, E,

is given by E: where p is an effective elastic constant, le is an

) u@4'V,
average strain drop and V is'earthquake volume'. Assumine L:3VV (L: lault
 Semi-EmpiricalEstimationof StrongGroundMotions During Large Earthquakes 67

length and I,V: fault width) together with the above relation, it results in a constant
strain drop. Thus, earthquakes of different sizes are related by a one-parameter
model. Aki (1967;ztt introduced a scaling law in which seismic spectrum grows
with earthquake magnitude. He showed that the seismic spectra are scaled ac-
cording to the fault length. assuming that the seismic moment is proportional to
Lt (L:lault length).
F u r t h e r m o r e , K a n a m o r i a n d A n d e r s o n ( 1 9 7 5 1 2 0a' n d G e l l e r ( 1 9 7 6 ; z t ' d e r i v e d
the following extended conditions of similarity:

Lf W:const. (l)
Df W:const. (2)
Ll Q ,,' r).:c o n s t. (3)
where L and W are the length and the width of an earthquake fault, respectively,
D, the final offset of the dislocation, r, the rise time and u,, the rupture velocity.
These constants vary with the different nature of source type and different source
region. These similarity conditions are derived as "averages" over a data set of
4l shallow earthquakes collected from all over the world, nevertheless, they fit ob-
served data quite well. It may be very useful to relate the source parameters
between thc large and small events, if the ensemble of the events are classified accord-
ing to the source regions and the source types. When two events with different
size have occurbd within the same region, the following similarity relations are
d e c l u c e df r o m ( l ) t o ( 3 ) .

LlL,:WlW,:DlD,:rf r,:(Mof Mo,)tis (4)

where the parameters without subscript are for a large event and those with subscript
e, for a small one.
The important parameters neccessaryfor synthesisas well as the source geomerrv
are rupture velocity it, and rise time r. Rupture velocities have been obtained
and noted to have roughlv a constant for earthquakes of different sizes. Geller
(1976;zt' obtained the relation u,:0.720 as an average for reported rupture
velocities.
The rise time r is very difficult to determine since it involves assumptions
on the fault model and is dependent on the rupture velocity. Geller obtained

r: l 6 5 1' 2 1 ( 7 r 3 ,B' 2) , (5)

based on various theoretical assumptions and observational values of l.t earthquakes.
Abe (1975;za' reached a similar conclusion from a data set of five Japanese earth-
quakes. In the next section, using these similar conditions we will approach the
synthesis of strong ground motions lrom small equarthquake motions.

2-2. Relation between the source-tirne-function of a large event and that of

a srnall event
We consider a large event and a small event which occur within the same resion
6B K. IRIKL|R.4

and have the same mechanism.

We call here the large event a mainshock and the small one an elementary earth-
quake. From the similarity conditions mentioned above, a relation is deduced
between the source time functions of the mainshock and the elementary earthquake.
A simple dislocation model of the Haskell type with a smoorh coherenr ruprure
propagation and a constant slip motion over a rectangutar lault (Haskell. 1964;zsr
is useful to expressthis relation.
A schematic model is illustrated in Fig. l. The far-field displacement (J, (x, t) at
any point Q in an infinite homogeneous elastic medium due to dislocatio n / Lr (t, ,1,
I ) over the fault plane J can be written as

(J ,(x ,l ) : (R .(0 ,p ) 4 rp u 3r)'p'

, A D ( ( , 1 1 ,t - t , ) d ( d r l
f f
J O
I
J O
(6)

where t":rla,+J{{t-i lr,,

,rzis the rigidity, r.r,is the wave velocity, r is the distance between the fault plane J
and the point Q, R, is the radiation coefficient, rp is the strike, 0 is the dip angle,
and subscript c indicates an appropriate wave type, P or SV or SH. The source
time functionS(x, l) associatedonly with source parameters is defined by u simple
inteeral of the form,

S(r,t):,u/,t (7)
/,"Aj',(t,\,t-t,)d(h:.
Dividing the fault plane of the large event into tVrx Nw, we take the dimensionol-

a
/

t
I

t - - I
t
J
s
- - .

{/

( 7t , l ^ ) '
Fig. l. Coordinate and fault plane geomerry. F i S . 2. Vlainshock fault plane divided into rVl
The Haskell rrrodel of a rectangular X rVs elements. An element corres-
fault is used, ponds to the latrlt dimension ola small
even t.
 Semi-EmpiricalEstination of StroneGroundfu[otionsDuring Large Earthquakes 69

each element corresponding to that of the small event as shol,vn bv Fig. 2. The
element has L, in length and We in width. Equation (3) is rewritten in the follow-
ing summation,
-! | '\ it'

s(r, r): a L ( { t ,r ^ , t - t , , . ) d { & 1 . (B)

, { ^ \ , , u/ , ' , ' ' " J ' , ' : ' - * '
The dislocation function /L'(t,q,t\ at a point (t,n) on the fault plane J of the
larse event is taken to be a ranlp function with a rise time r and a finai offset D.
That is,

ALl(t,T,/):0 t<0,
: Dtlr 0<t <r
-D t{r

S i m i l a r l y , t h e d i s l o c a t i o n f u n c t i o n , 4 L | , ( { , 1 1 ,t ' S a t t h e p o i n t ( t , l )
on the fault
plane 2, of the small event is also taken to be a ramp function with a rise time r,
and a final offset D,. Now, the similarity condition in Eq. (+) shows that slip velocity
may be assumed as a constant for most earthquakes of different sizesoccurring in the
same area. That is, we can write, sLip ueloictlt:Vo:Dlr:D,lr,, or Df D,:rlr,
:ConSt + /VD.
When the above ratio is approximated by an inteeer, rVr, the following relation
between the dislocation function of the large event and that of the small event is
approximately made up as shown in Fig. 3.
v,
l U ( t ,r , t ) : t-(k- l)",1, (9)
A/u,lt, T,
M^
and / U ( t ,n ,t ) : t-(k- l)',1.
A/U,lt, n,
P L r t t i n g ( 9 ) i n t o ( B ) a n d c h a n g i n g t h e o r d e r o f t h e i n t e g r a l a n d t h e s L r m m a t i o n .w e g e r ,
'\'L vtt, vD

S(x,t):
f it-1, fn--tv "" / U , l € , r ,
I
I = l
I2,"1.'"1
m = I h = t J ; l J n m
t-(k-l)",1. (l0)

aU(t)
aU"(t)
ol. T
I

au(t)-.-- n n
fg=5=Y'-
" 1 { , Ll lJl l l z = T e = ' o
OTe T
_-> t
Fig. 3. The relations between the dislocation function of a large evenr and that ol- a srnall
event and between their derivatives. DlDr:a1rr=iVo and iVp is approxirnated ro
be an integer.
 K. IRIKLRA

The term expressed bv the double integral is made of the source time function
of the small event 5,,^(x, t) having the starting point at a point ({t,r1-). when
the starting point of the large evenr is located at (0,0), we can write
NL Nry ND
s (x , t): S,i . (x, t-taea^), (ll)
,I- .4 , ,!,
where t a p 1 , ni s g i v e n as

t apt ^ : r r ^ f u , * r ' € t 2+ r l ^ t l r , * ( k - l ) r , 1^ .

In the far-field, if each small event taking place on each element is assumed to have
the same source mechanism, the source time function of each event may be regarded
as approximately the same. Then we can rewrite (l l),
Vr ,Yut .V^
,l(x, t)- I I I S, (r, t-tae1^), (l2)
! = l m = L h = L

and
tdh !^ :r t^I a ,+ J F -, rfi l r , * (k- l )r,. (13)
Equation (12) shows that the source time function,S(x, /) for a large event may be
approximated by the phase delaved summation of the source time function
,S,(x, l)
for asmall event.
Then the parameters y'y'r, Ir{w,J{o, u" and re can be estimated by employing
the similarity conditions described in the above section. That is, ^r\y'1,
Nw, No are
determined from the cubic root of the moment ratio between the two earthquakes,
V \W[/6'. For the sake of the simplification for the computation, it is desirable
for the cubic root value to be close to an integer.

2.3. Synthesis Method of strong Ground Motions using

Observed Seisrnograrns of Srnall Events
We will continue to use the Haskell modei with a rectangular fault. The
synthesis method of strong motions from a mainshock is developed as follows, using
observed seismograms of small events such as foreshocks and aftershocks that occurred
within the fault area of the mainshock. In order to simplify the expression, we
assume that all small events used for the synthesishave the same moment. It is easv
to modify the formulation obtained here fior the case of using some events of different
sizes.
Primarily, the ratio of the mainshock moment to the small-evenr moment,
:.YIof
Mo,, is determined. When the ratio NIof Mo, is N3, the mainshock f,ault plane
2(:Lxl/V) isdividedinto.l/xiy'elements. Then, the area of an element,lE,is
taken as the fault size of the small event, t,(:L,xW,). We call the element a
subfault. The displacement (J, in the far-field caused by a subfault Az in an
infinite homogeneous elastic medium is written in the form of

U ,(x , t) :l R ,(0 , < pf4

) rp u , ' rl .r A U , (t, r, t-t,)d(dr1, (14)
/,t' /,* '
 Semi-Empirical Estimation of Strong Ground fu[otionsDuring Large Earthquakes 7l

where t,:rlu,-jF+i'/r, and other notations are the same as in the previous
section.
We consider the surface ground motions amplified by the effect of surface laver-
ing beneath an observation site. This transmission function due to the propagarion
medium is given by T(x, t). Then, if the contribution to the motions from the sub-
lault is virtually equivalent to that from a point source, the surface motions G,(x, t)
are obtained by convolving LI,(x,l) with T(x, t'). That is,

G,(x,,): ( t , t - t ' )- u , ( x , t ' )d t ' (15)

/:T
Accordingly, the ground motions G,1^ which result from the dislocation /Li,,^ of a
small event on an arbitrary element AIr^ located at ((t,t7.), as shorvn in Fig.4,
are written bv

"' ^-* '

G , , ^ ( x ,t ) : T ' . ( x , t)' x ct.1 t,[ -" [' a u,, ^(t, r l , t - t , 1 ^ ) d ( d r 1 , (16)
Jit Jnm

where t,1^:r r-ft),+ J (t1;12'-1(t -rl"f l u,, and

cya:(l l 4 n p u , 3 ) R r , - ( 0 t ^ , g r n )f r t n .

On the other hand, the contribution to the mainshock motions, Gt^, due to the
dislocation A(Jt^ on an element AEt^ during the mainshock are written by

G , ^ ( x ,t ) : T ' ^ ( x , t)*r,*F .','"" t-t,,^) . (l 7)

f /,::n*'aLIt^((, T ,
The relation between the dislocation function of the mainshock, /Ut^, and that of
the small event, /(J,1^, has been given by (9) from the similaritv condition. We
get from (9), (16) and (17),

ND
Gt*(x, t): I G,,^lx,t-(k-l)r,1-). (lB)

Thus, the surflace motions G(x, t) from the mainshock are given by the time-
lagged summation of G;. over the fault plane, that is,

Nt Nw NL NW ND
G(x,t): I 2 G , ^ ( r ,t - t , t . ) : I I T Gr1*(x, t-tanr^) ( te )
I=L m=I I=t m=t h=l

where tapl^ is given as

t ap1^ : r 1 ^| u ,I J # + n ^ , l r , + ( k - l ) t , 1^ . (20)
Then, the ground motions G(x. t) from the mainshock can be calculated from (i9),
if all the records at a given site from every event corresponding to every element
were obtained.
Now, a synthesis method is developed using seismograms from a few small
events which occurred within the fault area of the mainshock. We consider a case
of obtaining only one seismogram from a small event corresponding to a subf,ault
 72 K, IRIKL'R.1

t
-\\- - -
{ c - - -
-----\v
\ \
\ .

t L2,no oLI ^

Fig. 4. Geometrical relation between the fault element /Xto^o of an observed small event
and the observed point Q.

/270^o as shown in Fig.4. The ground motions G,1. from an arbitrary element
may be estimated lrom the observed motions G,1o^oin the following equation, if the
propagation effect T,^ is approximately equal to T1o^o:

G ,1^ ( x , t ) : l R , ( 0t ^ , g t ^ ) f R " ( 0t " ^ o , g , o ^o ) l ( r 1" ^ of r 1^ ) G , ," ^ o ( * , t - t , ^ )

r ,
t,!^:(r,--r,o^o) fu, (21)

Putting(21) into (20), the surface motions G(x,t) from the mainshock are given by
r V , V N
G(x. t):,1 L : [ R , ( 0 / . , 9 t ^ ) fR , , ( 0 t o ^ 0p,t o ^ o ) 1 . 1( r" ^ " f r 1 ^ \
-
()r\
l=L m=t l=l

. G , 1o ^ o ( x , t - t , 1^ - t an t^ ) .

We can easily improve this synthesis method for the case of using seismograms
from several events. When the synthesis is made from the severai events with
different moments, some care must be taken of scaling the observed data for the
moment in regard to the element size and the number of summation.
The starting point in this formulation is taken to be the origin of the coordinate
G,d on the mainshock fault. When the starting point is located ar an arbitrarv
point (€,, r"), it is necessary to change the tapl^ as follows:

tat 1^ :r 1 ^f ,, * ,1 @ '- { )F h ;4 f l r, * (k- l )r, 1, (23)

Now we consider the extent of the validity of this synthesis formulation. The
ground motions from an arbitrarv element shown by (16) and (17) can be usuallv
represented when the observed wavelength is much longer than the source-space
climension-i.e. when the point source approximation is valid. In the case of con-
sidering the ground motions by separating P and S wave parts, even if the source size
is greater than a wavelength, as long as the distance between the source and the re-
ceiver is much greater than the source dimension, the effect of the finite source size
c a n a l s o b e s e p a r a t e d a s d i s c u s s e di n c h a p t e r 1 4 o f A k i a n d R i c h a r c l ( 1 9 8 6 ; t 0 , .
That is, the approximation by (16) and (17) isjustified for

L ,2K Ar1 2 , (24)

 Semi-Empirical Estimation of Strong Ground Motions During Large Earthquakes 73

where Z, is the length of the element source. i.e. the fault length of the small event,
A, the wave length and r,, the distance between source and receiver. This relation
determines the extent of the validity of the distance and the wavelength in the
synthesis formulation obtained here. If the records from the smaller events with
the smaller fault lengths are used, this synthesis method is effective up to the shorter
wavelength and up to the shorter distance from the fault. It must be cautioned that
if the motions of various wave types coexist, this will lower the accuracy of the
approximation of this formulation. We consider this synthesis is valid for the S
wave motions which constitute the main parts of the strong ground motions. The
extension to the case of surface waves is easily made after slight modification.

2.4. Nurnerical Check of Synthesis Method

The source tirne functions of earthquakes are calculated here as a superposition
of contribution from AU at infi,nitesimal surface elements dz , i.e. (7).
Our formulation for the synthesis of strong ground motions is based on an idea
that the source time function of a large event can be expressed by a superposition of
the source time functions of small events having certain fauit dimensions, without
direct estimation of llJ for the large event, i.e. (12). We need to examine the
frequency range, of seismic waves radiated lrom the fault, in which equation (12)
can be valid as an approximation, relating to the number of small events used and
the element sizes corresponding to the fault dimensions of small events. For this
purpose, the source time function given by direct numerical-integral of lU on the
mainshock fault plane, i.e. (7), is compared with the source time function synthesized
by ( l2) and ( l3) using that of small events. The source time function of each small
events is calculated by the integral of (7), given ALI, for the small event. We call
the former, theoretical seismogram and the latter, synthesized seismogram for
convenience.
Now, we will examine the range of applicability of our formulation for the follow-
ing 3 casesshown in Fig. 5.
l. Model Hl-l: rupture starts at r:0 simultaneouslyover the whole width W of

M O D ELH r-r MODEL H 1 - MODEL Hz-z

Q(x.y,z)
a--r O^, Qr-
,'l

I
z z

L = ' l 5 K m W =7 . 5 K m
Q = ( 4 K m ,2 0 K m . l 8 K m )
Fig. 5. Three fault modelsusedfor numericalcheckof synthesis
method.
74 K. IRIKURA

the fault and extends unilaterally at a constant rupture velocity.

2. IVIodel Hl-2:rupture extends bilaterally. The others are the same as Hl-1.
3. Vtodel H2-2: rupture extends circularly from a point.
The rectangular f,ault, L:2w (L: length and w: width), and a(/(x,7, /) with
a linear ramP function uniformly over the fault plane are assumed as being
the same
in all three cases. The fault dimension, the rise time and the relative location
of the
fault plane and the observational point are taken to coincide with the case
of the
observed seismograms at the JIZ station during the l9B0 Izu-Hanto-Toho-Oki
earthquake, which will be analyzed in the next section. That is, L:15 km,
w:7 .5 km, r: I sec, the coordinate of the observed point : (4 km, 20 km, I B km)

o
l----
ELEMENIARYS€ISMOGRAM S Y N T H E S I SM O D E LH t - r i
e2.5 -r zP'
fE?P
| *lJ-'

S Y N T H E S I Z E DS E I S M O G R A M
MODEL H1_1

Fig.6. C o m p a r i s o n b e t w e c n ' t h e o r e t i c a l ' s e i s m o g r a ma n d ' s y n t h e s i z e d . 'o n e f o r m o d e l H l - l

drawn in the right figure. The 'theoretical'seismogramis calculated as a superposition
o[ contribution from AI] at an infinitesimal elementdl over the whole fault plane
and an elementary seismogram is calculated in the same manner over the hatched
plane- The 'synthesized'seismogramis calculated by the delay and summation
of
the elementary seismogram, following (ll).

E L E M E N T A R YS E I S M O G R A M
E1
n
I

SYNIHESIZED SEISMOGRAM
MODEL Ht-Z

Fig. 7' C o m p a r i s o n b e t w e e n ' t h e o r e t i c a l ' s e i s m o g r a ma n d ' s y n t h e s i z e d ' o n e f o r m o d e l H l - 2

drawn in the right figure. Two events,E I and E2 are used as elementary earthquakes
for synthesis' The calculated method is similar to the method described in Fig. 6.
 I

Semi-EmpiricalEstintationoJ StrongGroundMotions During Large Earthquakes 75

when the starting point is the origin. The moment ratio between the mainshock
and the small events is 63. Since the similarity condition (4) is assumed to be valid,
the mainshock fault plane is divided into 6 x 6 elements and the element size of the
small event is taken as L":2.5 km and LV,:I.25 km, and the rise time, r,:116
:0.166 sec.
The theoretical seismogram and the synthesized seismogram are compared in
Figs. 6, 7 and 8. In each figure, the left upper (one seismogram in Fig. 6 and two
seismograms in Figs. 7 and 8) shows the small event seismograms generated from
subfaults depicted by hatched areas, and the left bottom shows the mainshock seismo-
gram synthesized using the small event ones. The number of the small events for
this synthesis is one for Hl-l and two for Hl-2 and H2-2. The theoretical seismo-
gram for each model is drawn by a dotted curve together with the synthesized one.
In all figures, the seismograms calculated by two different methods almost agree
and the discrepancy of the maximum amplitude between the two waveforms is
within l0%. The Fourier spectra of the theoretical and synthesized ones are
compared in Fig.9. The spectra calculated by the two methods are in good agree-
ment in the frequency range lower than I Hz and rough asreement in the frequencv
range higher than I Hz, although the two spectra have some discrepancies in detail.
The higher limit of the frequency in which this synthesismethod is applicable may be
near 6 Hz, i.e. 11",.
These results show that the synthesis method given by (12) is applicable lor
estimating the source time function of the mainshock in the wide frequency range
even at the short distance comparable to the fault length if the Haskell-type model
is valid for actual earthquakes. We consider the Model H2-2 to be more realistic
for actual earthquakes than the Models Hl-l and Hl-2, because difficulties in

ELEMENTARY SEISMOGRAM
tr'l A
El
/t
I t3 a
/}----

EZ'

S Y NT HE SIZ ESE
D ISMOGR AM
t
MODEL HZ-Z

E2'

Fig. B. Comparison between 'theoretical' seismogram and 'synthesized' one for model H2-2
d r a w n i n t h e r i g h t f i g t r r e . T w o e v e n t sE l , a n d E 2 , are used as elementary earthquakes
for synthesis. The calculated method is similar to the method discribed in Fig. 6.
 76 K. IRIKUR.I

MODELHI-'I M O D E LH 2 - 2

trJ
Q . ^,
F ' -

u.t'"
=
F-

-J
t ! , ^ I
G I U

0. I 0. I l.
nL

- Fig' 9'
Comparison between Fourier spectra of 'theoretical' seismograms
'synthesized' and those of
sesmogramsfor the three models shown in Fig. 5. ,SyN.'and,THEO.,
in figures indicate 'theoretical' and 'synthesized',respectively.

unlimited rupture velocity toward the z-direcrion can be avoided in this

model.
However, the accuracy of the approximation of the synthesis .gE12-z
in the Model
shown in Fig- 8 drops to some extent compared with that of Hl-i and
Hl-2, shown
in Fig' 6 and 7. We have to take care in the case of the synthesis
for higher fre-
quency motions. If a larger number of smaller events appropriately distributed
over the fault are used for the synthesis,the approximation is improvable.
However,
another problem is indicated by Chouet et al. (1978)3r) that the similarity
assump-
tions of earthquakes show some departures for smaller earthquakes.
Thus, for our
synthesis we should use small earthquakes with appropriate size which
can be
related to the mainshock in accordance with the similarity condition.
We have not been checking here the synthesis formulation (19) or (22).
This
examination needs the calcuiation of the theoretical seismograms. taking
into
account on the propagation effects due to the geological structure of
the medium,
for example, those given by Refs. (5)-(7) and (9). This calculation is too com-
plicated and immense, while our synthesis method needs only a simple
and
small quantity of computation. In case of regarding the propagation effects of
seismic waves from each element to the station as being approximately invariant
to
one another, the synthesis formulation (19) become equivalent to (12),
when S, is
considered to be a small event seismogram. From the simple numerical
check
mentioned above, we can estimate a rough extent of the applicability
for our
synthesis method.
 Senti-EtnpiricalEstinmtionoJ'StrongGrowtd Motions During Large Earthquakes

3. Synthesis of the Velocity Motions frorn the Mainshock (.IW:6.7) of the

1980 fzu-Hanto-Toho-Oki earthquake

3.1. Observed Data

A shallow earthquake with a magnitude of 6.7 (the l9B0 Izu-Hanto-Toho-Oki
earrhquake) occurred onJune 29, 1980,offthe eastcoastof the Izu Peninsula. We
obtained the velocity seismograms from the mainshock as well as small events
such as foreshocks and aftershocks at three sites located at short distances lrom about
20 km to 100 km, which were recorded by velocity-type strong-motion-seismographs
designed by Muramatu (Muramatu, 1977)32). The maximum velocity of B kine
was recorded at theJlZ station on hard rock, about 20km away from the epicenter.33)
The locations of the observed stations, JIZ, SMC and OMM are shown by (*)
mark in Fig. 10. The observation system was designed to record exactly ground
velocity motions with the dynamic range from 100 to 0.01 kines over the period range
from 0.05 to 50 second. These seismograms obtained by the velocity-type strong-
motion-seismograph are useful for predicting the ground motions from large earth-
quakes, since they have a wide dynamic range over a wide period range.
An active earthquake swarm occurred in a small area north of the mainshock

SZM
+
t\"a
MAIN
{

\a,

f-.,
l l
L \_l
( ,t43
N
A
1
TI
50 km 0
F i g . 1 0 . T h e l o c a t i o n o f t h e o b s e r v a t i o n sites and the epicenters of the mainshock and the
small events used for synthesis. T h e m a i n s h o c k i s t h e l 9 B 0 I z u - H a n t o - T o h o - O k i
e a r t h q u a k ew i t h M : 6 . 7 .
7B K. IRIKURA
'Sett

hypocenter for
J U N E2 5 J U N E2 9 JUNE29 JULY3 JULY5 - JULY28
nisms of the r
'lsN
1q'A plotted the hy1
ing the events i
within 5 days a
Iy after the larr
about 15 km ir
We presur
distribution of
indicated by a
aftershocks ern r
plane solutions
Imoto et al. are
cl
o o projection. N,l
! o
E
J
J
mechanisms.

3.2. Estirnat
The synth
seismograms ol'
Fig. I l. Hypocentral distribution of earthquakes for the three
period, I ) before the parameters for
mainshock, 2) within 5 days after the mainshock, 3) after
that. (after Imoto et
, al. l98l32') (u) The fault
I,V) and r
(b) The mom
(.) The rise t
(d) The rupr
The pararr
of aftershocks,
conditions of fa,
for a future lar
geology and sei
We summ,
descriptions in
The fault
distribution as ,
The mom
mainshock anr'

Tablt'
o

F i g . 1 2 . Fault plane solutions ol' large earthquake (M >4.0) projected a",

on the lowcr
hemisphere of equal arc projection. Dark and light areas
indicate compressional
and dilatational quadra.ts, respectively. (after Imoto et al.
lgBl32))
 Senti-EnQiricai Eslimation rl'Strong Groutd fu[otions During Large Earthquakes

hvpocenter for about 40 days before and after the mainshock. The focal mecha-
nisms of the earthquake swarm were studied bv Imoto et al. (1980)3{). They
plotted the hypocentral distribution of the earthqtrakes, as shown in Fig. 11, classify-
ing the events into three periods, (l), before the largest shock of the mainshock, (2),
within 5 days after the largest shock, and (3), after that. The aftershocksimmediate-
ly after the largest shock, or in the second period, are distributed within the range of
about l5 km in length along the direction Nl5'w and l2-19 km in depth.
We presumed thefault plane, L:I5 km and lV:7.5 km, from the hypocentral
distribution of aftershocks in the second period. In Fig. 10, the inferred fault is
indicated by a dotted line, and the epicenters of the mainshock, the foreshocks and
aftershocks employed for the present synthesis are shown by ( x ) marks. The fault
plane solutions of the earthquakes rvith the magnitude larger than 4 determined by
Imoto et al. are shorvn in Fig. 12, projected on the lower hemisphere of the equal arc
projection. Most of the earthquakes as well as the mainshock show strike-slip type
mechanisms.

3-2- Estirnates of the Fault Pararneters for the synthesis

The synthesis of the mainshock motions is made, using (22), from the observecl
seismograms of foreshocks and aftershocks. It is necessary to determine the following
parameters for the synthesis.
(u) The fault dimension of the mainshock synthesized, (the length L and the width
W) and the fault geometry
(b) The moment ratios,,MolMo,, between the mainshock and the small events.
(c) The rise time, r(or r,), of either the mainshock or the small events.
(d) The rupture velocity, t)r.
The parameters in this study were determined from the hypocentral distribution
of aftershocks, the Fourier Spectra of observed seismograms and the similarity
conditions of lault parameteru. For.the purpose of predicting strong ground motions
lor a future large earthquake, these parameters have to be supplied from maps of
geology and seismicity and the similarity conditions for a given seismic region.
We summarize the fault dimension and the geometry in Table l, according to
descriptions in the foregoing section.
The fault dimension and geometry of (a) are estimated lrorn the aftershocks'
distribution as described in the above section.
The moment ratios of (b) are estimated from the spectral ratios between the
mainshock and small events. The observecl seismograms and the Fourier spectra

Table I T h e f a u l t p a r a m e t e r so f t h e l 9 B 0 l z u - H a n t o - T o h o - o k i e a r t h q u a k e

strike N I 5"\\' far.rlt length 1 5k m

dip 90" lault width 7.5 km
rise time 1 . 0s e c
rupture vclocity 3.2 km/sec
 K. IRIKURA

of the mainshock and the aftershocks (Al and A3) at

JIZ are shown in Fig. 13 and
Fig- 14. The spectral ratios between the mainshock and the two aftershocks are
shown in Fig. 15. The observed seismosrams, the Fourier spectra of the main-
shock, foreshock P4 and aftershock Al at SMC are shown in Fig. 16 and Fig. 17
and the spectral ratios are shown in Fig. 18. The seismic moment is estimated from
the low frequency level of the spectra, based on the dislocation theory. Accordingly,
the moment ratio Mof Mo" is given from the flat level, in the low frequency range,
of the spectral ratio shown in Fig. 15 and Fig. 18. The spectral ratio of the main-,
shock to aftershock Al is estimated to be about 200 (=63) at theJIZ site and at the
SMC site as well. Similarly, the moment ratio of the mainshock to aftershock A3
is given to be about 350 ( = 73) and that of the mainshock to foreshock P4. about
200(=6').
Then, the scalingparameterly' (=NM;[MA) correspondingto the ratio of
fault lengths between the two earthquakes (see eq. (a) is estimated to be 6 for the
)
mainshock versus Al, 7 for the mainshock versus A,3 and 6 for the mainshock versus
P+.
The rise time of (c) is estimated as follows. The spectra of the Haskell-type fault
model, based on the dislocation with a ramp function are characterized by a signifi-

'!.o

Fig. 14. The Fourier spectra o[ the mainshock

and two aftershocks, Al and A3, at
:.JIZ,E O F F I Z U P E N .( A F I E A S I o C K T , .
-r ' l{ u' Ar the JIZ starion.
A3
'r$'*r.q,gl*qp.^--- 'il-'rilru$fi\\^/
: I
(MAIN,Hr6.?)lAr,Mi4.9) (MA1N,MrA7L4A;,M.4.6)

!tc
t0@
:
Nr ; l
-" I
= l
r l
r
I I \rt

ry M
i l-rrfih,tfrl'\Atr4|na
ao Y 1!il
:. .I
Id
3tc I
tt11
fl'
l vr ll
tJr
: . 1 |
=iLffi[r*p'1"5a
r€c I
I
2h23n JUN.30'EO t8H06n JUL.27'80 0.1 ! lOHz O.t I tOHz
n.4. 9' fi. | 0Kn n'1. 6. H. t0Kh _ lr5 -____ 6w

Fis' 13. The observed seismograms of the F i g . 1 5 . The spectral ratios between the
mainshock and two aftershocks Al mainshock and aftershock Al and
and A3, at theJIZ station. between the rnainshockand aftershock
A3 at the.f lZ station.
 Seni-Empirical Estinmtion of Strong Ground Motions During Large Earthquakes BI

7 't 1l
l
8t,

lr

-)
t I
Fig. 17. The Fourier spectra of the mainshock
and aftershock Al and foreshock P4
at the SIvIC station.
f
e sl.lC, ( MAIN, M.6.7,/(At. M.4.9)

s M C , ( M A | N ,M . 6 . t ) r ( P a , M r a . 9 )

o.t t, 100.1 t. t00.t t. lo too

Fig. 16. The observed seismograms of the Fig. lB. The spectral ratios between the main-
mainshock, aftershock Al and fore- shock and aftershock Al and between
. shock P4 at the SMC starion. the mainshock and foreshock P4 at
the SIvtC station.

cant trough around llr (r: rise time) as well as by the corner frequency related
to the fault dimension, as clearly shown in Fig. 9. However, the rise time determined
bv picking the trough frequency has larger uncertainties and may often be non-
unique, because the spectra of f,ar-field observation data are srrongly modifiecl by
scattering and absorption in the propagation medium and further by local ground
effects. Therefore, some care has been taken to estimate the rise time.
The JIZ station is in a drift lormed of a hard rock and at a short-distance of
about 20 km away from the epicenter. We consider the seismograms
obtained at
JIZ to be less influenced by the propagation effect. In Fig. 14, the spectra of the
mainshock at JIZ have a common significant trough around I Hz for
the NS, EW
and V components. On the other hand,
the spectra of aftershocks Al ancl A3 have
no significant troush around I Hz.
Thus, the trough around I Hz for the mainshock
 K. IIIIKL'R:I

may be independent of the local ground effects. The troughs of the spectra for
aftershocks Al and A3 appear to be at about 6Hz and 7 Hzas indicated by arrows
although they are not so significant, as compared with the case of the mainshock.
After the above considerations, we determine the rise time of the mainshock to be I
sec. Since the trough frequencies of the aftershock are non-unique, we determine
t h e r i s e t i m e s t o b e c o n s i s t e n t w i t h t h e s i r n i l a r i t y c o n d i t i o n , i . el./ 6 s e c f o r A l
and l/7
sec for A3. These values also correspond to the trough frequencies shown by arron,
marks in Fig. 14.
I n t h e s e c t i o n 2 . 1 , w e n o t e c lt h a t t h e r i s e t i m e i s g i v e n f r o n r t h e s i m i l a r i r y r e l a t i o n
(5). The rise time is obtained to be about 1.2 sec when the inlerred fault plane area
o[the mainshock and S wave velocity of the medium are put into (5). This value
is very close to the rise time estimated here lrom the spectral shapes of the observecl
seismosrams in the short distance. Thus, in the synthesis we may use the value of
the time given by the relation (5) as a first approximation.
We do not have any evidence for a, estimates, but'tentatively
assume to be
3.0 km/sec, based on the empirical data.
The validity of this.assumption is later
examined by a comparison between the synthesized seismograms and the observed
spectra.

3.3. Synthesis Results

The hypocenter of the mainshock is located at the center bottom of the 1ault
plane inferred from the aftershocks' distribution shown in Fig. 11. We consider
that this hypocenter is consistent with the starting point of a rupture and that the
rupture spreads circularly from the starting point, Pot over the fault plane. That is,
this type of rupture propagation corresponds to that for ivlodel H2-2 shown in
Fig. 5. The locations of the hypocenters of the mainshock and the small evenrs
employed for the synthesis are summarized in Table 2.
First, we describe the synthesis of the mainshock motions at SMC. We use
the seismograms from two small earthquakes; one is event P4 with fuI:4.9 in an area
north of P" and the other is event Al with the same magnitude in an area south of
P,. The locations of the epicenters of Al
and P4, and the mainshock lault plane
are shown in Fig. 10. The fault plane solutions of Al and P4 are similar to those o['
the mainshock as shown in Fig. 12. The moment ratios of the mainshock of fuI:6"7
to events P4 and Al are estirnated from the spectral ratios to be MolMo":N3 =200.

Table 2 The origin times and hypocenter of the earthquakes trsedlbr the synthesis.
(afterImoto et al.. l98l31))

M D H M I_,AT. LONG. DEPTH MAc.

(deg.) (d.s.) (k-) (JlvrA)
.
P4 6 28 t2 5 34.934 139.234 I8.9 +.9
IvtAIN 6 29 l{i 20 34.904 139.230 | /.9 (i.7
AI 6 30 2 23 34.8+7 139.245 lB.+ 4.9
A3 tB 6 34.942 r 3 9 . 2 2 0. 17.3 +.6
 Semi-EmpiricalEsthnationof StrongGroundfulotionsDuring L4ge Earthquakes B3

)l' rvhich vields the scaling parameterrV=6.

vtS Then the fault plane of the mainshock is
divided into 6 x 6 elements as shown in
I Fig. 19. The rise time of the mainshock
re is estimated to be I sec as mentioned in
7 section 3.2. The rise time of each small
event, r,. iS estimated to be z/N, from
the similaritv condition. The correction
II lactors for the radiation amplitucles
'a R , ( , 0 r ^ ,g y ^ ) i R , r ( . 0 , 0 ^ 0g,t o ^ o ) a r e t a k e n t o
IC be unity as a first approximation for
cl simplification of the computation. S
rf lVe make deterministically a synthesis Fig. 19. f'he schematic model of the lault
fbr the mainshock motions in the follo- plane for synthesis. Rtrpture spreads
re rving two stages. In the first stage, the cocircularly from Po.

lr
d 5 1 1 C . EO F F I Z U P E N . , J U N . 3 O ' E O 2, H 2 3 N
= 4. 9, rfl= I 8K11

It

P E N . ,J U N .2 8 ' 8 0 , t 2 H 0 5 f 1
NS-C0f1P.
e
S,

n
IS

,e
' i?EEb.ie3,FISiJliliAel
; II S vYNNr T
HHt rE
< rS IZ
7 tEr nO
5 rf0^rp
i Y H ! ! r i n i y F I i l A : ' i i A=e*l' S
q
r .= e o
L ,=r5( 3 / 63, / 6 )
I a ^

a
)f 2 9' 8 0

e 2.. 00 2!. 00 12.00 l!.00 t 0 .0 o

5EC
rf'
P E N ., J U N .2 - a' 8 0 . I 6 H 2 o t 1
7 Ns-c0tlP.

),

Fig' 20' Comparison ol thc synthesizedvelocity seismogram with the observed

one for the
mainshock at the SMC station. The left upper two traces are the observed
seismograms of aftershock Al and foreshock P4 used
as elementary earthquakes,
the left third trace, the synthesized seismogram for
the mainshock, and the le{t
bottom trace, the observed seismogram of the mainshock.
The right Llpper figure
is the Fourier spectrum of the synthesizedseismogram
and the right lower figure is
that o[ the observed mainshock seismogram.
 K. IRIKURA

mainshock fault plane is divided into two areas, one is S, for the northern half area,
the other, ^92for the southern half area, each of which has 3 x 6 elements, respectively.
The synthesized seismogram/ is obtained, using P4 for Sr and 1f2, using Al for Sz,
respectively. In the second stage, after summing up .fi and fz, we obtain the
synthesized mainshock motions.
The synthesized velocity seismogram of the NS component for the SMC station
is shown in Fig. 20, together with the observed seismogram of the mainshock. The
upper two are the P4 and Al seismograms used as elementary earthquakes, the third
is the synthesized and the fourth is the observed seismograms. These are outputs
from a 4 Hz low-pass-fi.iter, by which an apparent predominant frequency of I lr,
(:6 Hz) involved in the synthesized seismogram has been removed. This periodic
motion is discussed in later section. It is found that the synthesized seismogram is
in good agreement with the observed seismogram, except for a few portions with
periods around I sec.
Similarly, the synthesis of the mainshock velocity motions is made for OMM
using P4 and Al and for JIZ using A3 and Al (because event P4 faiied to be observed

0 t 1 f 1E, 0 F F I Z u p E N ., J U N .3 o . 8 0 . 2H23fl
= 4 . 9 . H =l 8 K l 1 NS

P E N .. J U N .2 8 ' 8 0 . r 2 H O s f j
NS- C O N P .

n 0 0 E L2 , , s t AI , P 4 )
C t flCULAR. , n A | 1 P5,I = NI S t l . ( J / O , J / b J
S Y N TH

x I N E1 5 E F s x c
16'--1lLll-
I Z U P E N ,. J U N .2 9 ' 8 0 ,

20.00 1t.00 2r.oo :**'

_ 0r1f1 E, o F F i z u P E N,. N.29 '80, t 6H20t1

t,\'h,.,,,
I 1 Y = 67.. H =[ 8 K 1 1 , - , .

r oo
IIUI
t,,
lll
to - '
I
0.01 o.t t.
,, lo

Fig.2l. Comparison of the synthesized velocity seismogram with the observed one for the
mainshock at the OMM station. The arrangement of the figures is the same as
that of Fig. 20. The observedseismogram at the OlvtM station from the mainshock
is recorded in a saturated form over I kine becarlseof poor conditions of the auto-
gain-controlsystem.
 Semi-Empirical Estimation of StrongGround Motions Durine Large Earthquakes

U I T I ' U P E N ,. J U N .3 0 ' 8 0 .
H- I Q'(M

^'t^,','8?REb.i;
l n
5ill Jliri 3i :
3'FI f AnP, 5T=N l5l.l,

:t J I Z , E O F F I Z U P E N ., J U L . 2 1 ' 8 0 . t 8 H 0 6 N = SYNTHE
- , A H NS-C011P. F
e
I
s

a f 1 00 E L
C I R CUUL , ii 3rEI5i[]ililelE't:* Drp=e0.
r5r"r. L=15(4/1'3/6)
S Y NT H E SI Z E O
2 l t t
I

0.00 2.00 {.00 !. 00 !. 00 10.00 l:. 00 11.00 lt. 00 lt' 00

)tL

e J t L , E o F F t Z U P E N ,. J U N .2 9 ' 8 0 . l 6 H 2 0 N
i fi=0. 7. H=l0Kf1, NS-CoNP.

0.01 0,I l. r0
rz
Fig. 22. Comparison of the synthesizedvelocity seismogram with the observed one lbr the
m a i n s h o c ka t t h e . J I Z s t a t i o n . T h e a r r a n g e m e n t o f t h e figures is the same as that
of Fig. 20.

at theJIZ station). These results are shown in Fig. 2l and22. The

moment ratio o[
the mainshock to 43 is almost 350 and then the scaling factor is estimated
to be 7.
For this reason, the northern half area s, is divided into 4 x7 (+ l{12xi/
for N:7).
The svnthesized seismogram at OMM is also in good agreement with
the observed
one, similar to the case of SMC. The synthesized seismogram at JIZ agreeswell
lvith the observed one, inclusive of portions with periods
around I sec, although the
spectral amplitudes of the synthesized one higher than
I Hz are underestimated as
compared with those of the observed one.
To examine the validity of the assumed parameters, we calculate
the synthesized
seismograms for a function of each parameter,
and compare them with the observed
one in the time domain' Three kinds of measure to evaluate coincidence between
the synthesized seismogram (t) and the
-f observed one g(t) are used: correlation
function d, amplitude ratio a, and residual
function r, where these parameters are
defined 3S,

o:lf ,'rrt')g(t-,t')0,'
/(f ,'f,(,)0,
f ,'s,(t)dt)"'l
o:l/,'f,dt
f f ,'s,dt]"'
': f ,'t-f-s)'dt(1,'-r,0,
/ f,'s,at)"'
 K. IRIKL|R.4

We determine an optimum model by changing three parameters in

the fbllowing
order: (l) the relative position of the starting point on the mainshock fault
plane,
(2) the rupture velocity and (3) the rise time. The fault geomerry,
and the number
of the elements assigned are fixed in all cases.
Examinations with regard to the location of the starting point po ara-shown
in
Fig. 23 for the casesof the SIWC and
JIZ station. Let the coordinate of p, be (x,1),
located at a grid point on the lault plane as shown in the right upper
figure. 6, o
and r are computed for the time length of 40sec for SN4C and 12 sec
for JIZ after
low-pass-filtering with a cut-off-frequency of 4 Hz. Fig. 23 shows that,
whe \ po
is located at (3,0), the correlation is clearly highest, the amplitude ratio
is closest to
unity and the residual is least, although those values are different for SMC
and JIZ.
The difference of $, a and r for the two stations is mainly due to the difference
in
the data length used. This optimum location of the starting point is consistent with
the relative position of the mainshock hypocenter within the distribution
area of
aftershocksin Fig. ll.
Similar examinations with resard fb the rupture velocity V, are shown in Fig.
24, when the starting point is locatedat (3, 0). The influenceon the waveform
due

0.6 SEt
-+
r.0 (3.0) (6.0)
porometers

v.o
e = N15"W
6 = 90o
02
L = 15 km
1.0
w = 7.5 km
T :l sec
V. = 3.0 km/sec

( 1,0) (3.0) (s,o) (1.0) (30)(s.0)

START POINT START
POINT
Fig' 23. Tire examination of the variation of the synthesizedseismogramsdue
to the location
of the starting point within the far-rlt plane. {, a and r are correlation
function,
amplitude ratio and residual function between the synthesized seismogram
flor the
starting point varying from (0, 0) to(6, 0) at the lault plane, as shown in
the rig5t
upper figr.rre,and the observed seismogram.
 semi-Empirical Estimation of Strong Ground Motions During Large Earthquakes

[ '
u
) n q _ sMc nq - Jlz
')
r
S MC
I
JIZ
\
""fn'','
la- f'
0.8

l4
ll
o.7

) I .,t__
r
0

)
t__ u.o

l-:
l a
i
l 0,9 0.8
I

T f , ,, , [,
r
tt1
r.0 t \--,
F I

t-
0.6 06 l0 lt, I I sec
u-o
T
Fig. 25. The examination of the variation of
0.2 0.2 the synthesized seismograms due to
)A 28 3.0 3.2 2S 3.0 32
the rise time o[ the mainshock. In
Vr kmrbec Vr km/scc
t h e s y n t h e s i s ,t h e r e l a t i o n , . r e : r l N ,
Fig. 24. The examination of the variation of is kept between the rise time of
the synthesized seismograms due to elementary earthquakes and that of
the rupture velocity. the mainshock,.

to the change in the rupture velocity, ranging between 2.8 and 3.3 km/sec is found
to be very little. We may not have the resolution enough to determine the rupture
velocity from this analysis. We have chosen V,:3.2 km/sec as the optimum
ivhich gives the minimum residual value.
In order to examine the validity of the rise-time value of the mainshock, the
influence on the synthesizeclwaveform due to the change in the rise time ranging 0.6
to 2'0 km/sec is shown in Fig. 25. The rise time of elementary earthquakes is
estimated from the similarty condition r, :rll{. For the case of the JIZ station, the
oPtimum value from the correlation and the residual function is given to be 1.0 sec,
rvhich is consistent with the estimated value from the spectral shape at
JIZ, as men-
tioned in section 3.1. On the other hand, the rise time for the case of SMC
is given
to have best fit for about 1.5 sec, which is somewhat larger than that for
JIZ. For
the case of SMC, the synthesized seismogram having z:1.5 sec is compared
with
the observed one in Fig. 26. It seems that the waveform and spectra of the
s v b t h e s i z e dw i t h r : 1 . 5 s e c a r e i n b e t t e r a s r e e m e n t w i t h
those of the observed, in-
clusive of the period range from I to 1.5 sec, while its range is in
disagreement between
the svnthesized and the observed in Fig.
20. For the case of OIvIIvI we obtain better
fit lorr:1.5 sec rather than r:1.0 from the comparison between the synthesized
and the observed (we can not
accuratelv compute the correlation and residual
 K. IRIKLRA

20.00 2t. 00 2 1 .0 0

r <t N i r S t i r n r
ru
=
=-

S N C , EO F F I Z U P E N , J U N .2 9 ' B O , l 6 H 2 0 1 1
-
I
= 6 . 7 , H =I B K f l
OBSERVED NS-C0f1P. "a
r

Fig. 26. Comparisonof the synthesizedvelocity seismogram'r"ur* l:lt i.. *,,it.n"

observedone for the mainshockat SMC.

function between the two seismogram because of the saturated form of the observed
seismogram).
The results, showing a discrepancy between the optimum values of the rise time
for the three stations, correspond to difficulties of unique determination for the
rise time lrom seismic observations as mentioned above. For far-field observations,
especially, the information on the rise time which control the short period
motions seems to become more ambiguous with distance for attenuation and scatter-
ing in propagation medium. We can not discuss further which value is more
probable only from this information. The rise time of I sec here is adopted as the
optimum, determined from the data for JTZ at relatively short distance, since we have
one purpose of predicting strong earthquake motions in near field.
In the synthesis done so far, we have given constraints on the mainshock rise-
time and the elementary earthquake rise-time from the similarity condition (4). It
is reported in some papers that the stressdrops during aftershocks are not always con-
sistent with the stress drop during the mainshock and then the similarity condition
based on a constant slip velocity is not always valid, as indicated by e.g. Imagawa and
Mikumo (1982). The following examination is made, to check the validity of the
similarity condition (a) for the l9B0 Izu-Hanto-Toho-Oki earthquakes. We synthe-
size the mainshock motions, assuming that the similarity conditions Lf L,:rVlW,:
DlD, are valid but the slip velocity is nor constant, i.e., Df D,*rfr,. Then the
mainshock dislocation function is related to an elementary earthquake one as follows:

A / n
A L t ( { , r ? ,t ) : ; AU,l€,I, t-(k-l)rln)
_I=,
 Semi-Empirical Estimation of Strong Ground Motions During Large Earthquakes B9

where i/ is scaling parameter. re is

equal to r for n:\, r12 for n:2,..., and
rf l,{ for n: N. The apparent oscillatory
motions in this synthesis as described in
the next section are reduced by the 0.7 0.7

smoothing operation and filtering for the

06 u.o
period range lower than 4 Hz. When
t.o t.o
r a v ? r i e Sr t o r l I { , i . e . n : \ , 2 , 3 , 4 , and
l/ for a fixed value of r: I sec, the t 2 t . t

coincidence between the synthesized and

the observed is examined in Fig. 27. 08
The best fit is obtained for r,:rlI,{.
This relation is consistent with the simi-
larity condition, DlD,:rlr,. We com-
Pare the two synthesized seismograms,
one is based on za:7 and the other, based
oo re :rlN, with the observed seismo- N:5forP4,Al ; N=7forA3
grams in Fig.28. The synthesized seis- Fig. 27. The examination of the variation of
mogram for z":z has larger amplitudes the synthesized seismogram due to
at high frequencies than that for z,: the relations between the rise time of
,lN. The overall waveforms clearly show the elementary earthquake and that
of the mainshock, aa:rt rl2, tl3, tl4,
better fit for ,,:rl]{.
r/l/ (i/:6, for Al and P4,.lf:7, for
As the result mentioned above, we A3).
show the synthesized seismograms of NS,
EW and V componentsfor the optimum parametersat the SMC and JIZ station in
Fig. 29 and 30. It is to be emphasizedhere that the synthesizedseismogramsare
in very good agreementwith the observedrecords for all the three componenrsar

I5IIlJlAnl?Ae=
L r.rBs.t i / 7 .3 / 6 1

ttc

' i?9Eb.ii3'FI5ijlillAeltAe,tnl6, L.rs(3/s. 3/61 : i ? 8 8 8 8 8 E b . i a 3 ' f i l 5S ly 5x rlHR / 6r. -i ?r A

t€i SA A/ei .t.Lt = t st l / 7 . 3 / 6 )
S Y N T r r s t z E D t2. ., r r 5 -l
: I tZEOt,f..f/6
'r||*
1;-\/ A/ 'U^/
\"
v /\ -^
f,"\nf\"/\-^\ ..A ^-/\

a.{ ra.{ !t.r t.n r{

tcc

. 9'80. t 6H2ort
P€N. JUr{2 J l z , E o F F t z u P E N ,. J U N2. 9 ' 6 0 . l 6 H 2 0 n
t Hs-collP' . 6 . 7 . i ' 1 I. O K n N
, 5 NS-COiP.
or*r"ao

Fig. 28. Comparisons between the velocity seismograms synthesized for rr:r drd for t,
:r/6 and the observed seismogram. (a) The case for the SMC station. (b) The
case for the JIZ station.

g
 90 K. IRIKL'RA

r 1 c 0 i L2 . 2 E v E N r S l lf . ? { l
:: i !sE [ ! ! : i l i : 1 1 F l 5 ] i i i i i r lri .i?==* ,e,cr . , - . .: ,.)s,,.. 4 ,
s trtBCr_JLA vB a :. 3 . 2 . F { f i ) . 5 r = N l S r . l : 9 . 3 0 . ! = l S { 3 / S . 3 , / 5 1
_- s Y N TH e S t Z E o f f - c c i t P / N r H E Sl Z q 0 v _ c o r , P
]
:r. '-,-
'\*-'*f"/ r r 1
.. ,L\/ /'- .t'---..^==...=--
noorr,,\j t
*1, "f
\-/'\ 1----*-, v
stc

5EC
: -lZ.E JFF IZj PE\..-J\.29'tC
"- r = 6 . 7 , H : l 0 K r . ? - C t r P .
085ERvE0

:,. 0rp=e0. .-=i5 i 4,1.3,s)

:: i!Ei!:il||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
] 5 Y N T H E S I Z E DE,w . C O M P

i ;--;-u^ftur/t ^r--v-!r.*------------

stc
g.-.2:E ir. IZU 9E\..Jf\,.'3'3C. l6-21:
-; ; ' : 5 . ? . H : | C K H . Ev.Cl-'
- - Ii 0 oESERV EE o0
;
1-s*-,,.rilJ" n .^r*-.-rr\J/.\=,\-^.*-

0ip=e0
5:r.
: i!fltIigi:F15]llJlieiSi.*, s(3/6
L=, 3 hp
! ? B E b . ! i , 3 1 E I 5 i 5 l J iD
: | 5YNtHESIZm NS-COMp
l,B=ig. -e, r .

..0. r..oo .!f: r..oo r..o !r.d ,..6

,lc

" s n c . 5 o F F I Z u . e N . , J U N .2 9 . 8 0 . I 6 H 2 O r P E r , J, ,J U N .2 9 . 8 0 , t S H 2 o n
I i=6. r, H:l8KH.v-CoilP. \S-C0irp
:' I OBSERv€0
: 1
'r
n*r*"rfr

r'oo ": '6

l;t
Fig. 29. Comparison of the three-components Fig. 30. Comparison o[ the three-componenrs
velocity seismograms(NS,EW and V) velocity seismograms(N.S,EW ancl V)
synthesized using the optimum para- synthesizedusing the optimum para-
meters with the observed seismogram meters witlr the observed seismogram
at the SIvIC starion. (a) The case a t t h e J I Z s t a t i o n . ( a ) T h e c a s eo f
of NS-component. (b) The case of NS-component. (b) The case of
ElV-component. (c) The case of V- ElV-component. ( c ) T h e c a s eo f V -
component comDonent,

the both stations.

4. Synthesis of Strong Acceleration Motion

4.1. rrnprovernent (1): Rernoval of Ghostly oscillatory Motion

We have so far described the svnthesisof velocity seismograms ancl our attention
was focused on frequency components lower than I Hz. However, ground motions
with frequencies higher than I Hz play an important role in acceleration
seismo-
grams. We need to estimate the hieh frequency contents of strong motions, especially
for engineerins interest. The synthesized accelerogram for the mainshock can
be
obtained, if the accelerogrami for the elementary earthquakes are put into G,(x, t)
in (22). However, we have some problems in applving (22) directly to the svnthesis
 \cttti-linliricul li.rtirtutt iott ttl ,(itrottg Grountl I[oliotts Drring l]trge liartlqtutke.t 9t

J I , Z ,E O F F I Z U P E N .. J U N .3 0 ' 8 0 , 2H2311
= 4 . 9 , H =l 8 K t 1 A l NS
t

'4
J /-z
f '/z
A3
^" ^. :, : ' -, "Jt11 0R0c u
e LL2A.R2. E V E N I S ( A 3 . A l )
v B = 3 . 2 , f l A t 1 pI,R = 0 .
J I Z , E O F F I Z U P E N . ,J U L .2 7
'80,
I 8H0611
-
: SYNTF€SI EO
o
o = 4 . 6 , H = l7 K H A 3 NS
J

[lryfI
lo
t I

o
lo
I

OEL2,2 EVENTS(A3,AI)
R C U L A RV, B = 3 . 2 ,R A t 1 PT, B = 0 . | | t t l l | | | t t l I | | l l l
S YN T HE SI Z E D l0
0 .0 l l'
Hz
c

5 ' t t r l E c l r zE. o F F I z u P E N .J, U N . 2 9

-: oESERVE
4.00 !.00 1 0 .o 0 11.00 ta.0c
s€c
o
J I Z , E O F F I Z U P E N . J U N2
. 9 ' 8 0 ,t 6 H 2 0 1 1 l o '
^ , n , lA
\ iil,,
;
VIrITld
a

= 6 . 7 , H =l 0 K l 1 , N5
O B 5E R V E O
o o
o lo
qi
o0 I
I
| | t t l l I I I t t t l
I
ro - '
0.0t 0.r l. l0
HZ

I;ig. 31. C o m p a r i s o n o l ' t h r : a c c e l c r a t i o n s e i s m o g r a m s v n t l - r e s i z e db y ( 2 2 ) l v i t h the observed

seismogram the JIZat station. The leli upper two traces are the observed
accelerograms ol'altershock Al and A3 used as elementary shocks. and the leli
third trace, the synthesized accelerogram for the mainshock, and tl-re leli botrom
trace, the observed accelerogram of the mainshock. The right Lrpper: the schematic
m a i n s h o c k l a u l t a n d t h e e l e m e n t a r y e a r t h q u r a k e s r " r b l a u l t s r . r s e d{ b r s y n t h c s i s . The
right middle and bottom: the Fourier sperctrum c'l'the synthesizcd accelerograrns
and that of the observed accclerosram for the mainshock.

ol- the accelerogram.

One is a problem generating an apparent predominant-frequency in the
svnthesis following (22). The waveform and spectrum of the svnthesized accelero-
gram for the
JIZ station are compared with those of the observed one in Fig. 31.
The waveform of the synthesized one is similar in its envelope with that of the
observed one, but the two spectra are significantly different lrorn each other in
the high frequency contents. In particular, the synthesized accelerogram involves
predominant frequencies around 6-7 Hz, while the observed one does not. These
ghostly oscillatorv motions appearine in the synthesized one are owing to the
 K. IRIKURA

following reasons.
The synthesizedseismogramexpressedby (22) is rewritten bv the convolution
of an elementaryseismogramwith a discretetime series-f(t),

-f(t): A F, p,(*)ttr- trtm-(,r-r)z,l (24)

where E is the Dirac delta function. The above parameters ca, c,^ and lr1^ vte
employed here to simplify the expression. That is, the mainshock
motion G (t ) is
given as

G( t ) : f Q ) * G , ( t ) .
The discrete function-f(l) has an apparent periodicity ra, ?s is obvious lrom an
inspection. This periodicity is owing to (9), in which the dislocation time function
of the mainshock, AU (t), is given by the phase delayed summation,
with a constant
time shift r,, of thar of an elementary earthquake. a(J,(t). If both /u(l) and
AU'(t) are exact ramp functions and z is equal to .ly'.z,,.the relation
(9) is exact
and then the periodicity of zr would not appear in the synthesized
motions" This
is illustrated in Figs. 32a and 32b. However, the above-mentioned
conditions are

(a) Smoll Event Lorge Event

a u " ( t) a u (t ) = 9 a U " (-t( i - t ) r . )

,f_-
, / '
- /
-__,/.
q
,/ ..

^u.(t) . : U ( )t = l a U " [-t ( i- l ) 1 . ]

rl
-r [_-__
--Jri .,' "" fli.-
-

Fig. 32. (a) The dislocation time function ol'

the small event and that of the
. WhZn j
f.!c> {c-:,mc

;
t,lilflil
t:-.:i:-::-
large event, assumed to be a
-;-t.L ..-, . : l l : ; l : : l : r ; l ramp function and re:rlN.
!i.r!:-.r'i'c rr
(b) The time derivative of the dislo-
cation function o[ the small event
aU"(t)
and that of the large event.
_l-L_ (c) lVhen an assumed ra is different
from a true rr, the synthe;ized.
dislocation of the large event
_ f-l_ ___
results to have an apparent
/'.-:---1
a
f t
, r r - - , . r - . , , . ,
L
L
periodicity at r,.
(d) When the operation by (22,) is
WhCn lkrjc> {0.:umr:t
made, the synthesizeddislocation
--f
function is smoothed as illustra-
L
ted.
 Semi-Empirical Eslimation of Strong Ground Motions During Large Earthquakes 93

unrealistic for actual svnthetic problems. Even if the relation (9) is exact, it is very
difficult to estimate the true value of r,.
When the estimated value of r" is different
from the true value, the synthesized dislocation function has an apparent predomi-
nant frequency as shown in Fig. 32c. When the relation (9) is not exact but holds
approximately, similar oscillations are generated in the synthesis. This is the
reason why the synthesizcd accelerogram in Fig. 3l has ghostlv predominant
fre quencies. Since r, of A I and that of A3 has been taken to be I /6 sec and
l/7 sec, respectively, the synthesized one has apparent predominant lrequencies
around 6-7 H.z.
we can avoid this difficulty by changing (22) to the lollowing form:

rV tr' Nxn
G \( r/ ) : I
t7t
)t
.7t
t
lTr ffi* ,,lt-t.m-&-\!,,) (19'\

The meaning of this operation is a kind of smoothing as illustrated in Fig. 32d.

2H23N
N5
J

t.

2L, 2 E v E N
u ^ l j ^T-ur '1IB0c0UEL I S ( A 3 ,A I )
A Rv ,B =3 . 2 , R A m P ,
J I Z , E O F F I Z U P E N ,. J U L 2
. 7'BO, - SYNTHE:C C U
I
= 4 . 6 , H = l 7 K NA 3 5,10

n A
t

vh
lo
00
o
5,10

o
ro
f,rlu

I l O D E L2 , 2 E V E N T S I A 3 ,A I )
C I R C U L A RV, R= 3 . 2 . B A N P T , R = 0 .I 6 6 x 6 , L = t 5
S Y N T H E S I Z E DS M O O T H I N(GN ' =5 ) - l

l0
0.0l
N L

00
GALxS_ECrrz tr nqtr t-
-
: OBSERVE
) r t u
6.00 6.00 10.00 12,00 It.00

'lAilil, ,
Tt
'80, t6H20f1
J I Z , E O F F i Z U P E N ,. J U N 2
. 9 r o t

I il$lilr
= 6 . 7 , H =l 0 K l 1 N
, S N rq
5.10
o
OBSERVEO

lo
o llr
5xl0
- l
II
I l l l t l | | l l t l
I
t0
0 .0 l l. l0
HZ

Fig. 33. Comparison o[ the synthesized acceleration s e i s m o g r a ms m o o t h e d b y ( 2 2 ' ) w i t h

the observed seismogram at the JIZ station.

;
c
i
,
I
..
f,
 K. IRIKURA

Applying (22') to the synthesis, the apparent periodicity is shiftecl to the shorter
period, r,f n''Then we can obtain the synthesized motions up to the frecuencv
range of engineering interest.
Applying the revised formulation (22) with //,:5, u,e make a svnrhesisfor the
mainshock accelerogram atJlZ. The waveform and the specrrum of the svnthesized
seismogram are shown in Fig. 33, after low-pass-filtering with a l0 Hz cut-off
lrequency. The ghostly oscillatory motions owing to r, clisappear in the synthesiz-
ed seismosram. However, the synthesized one has significantly small amplitudes
for the frequency range higher than I Hz, comparecl with the observed one" This
is discussedin the next section.

4.2. Irnprovernent (,2); Revised Synthesis Method for High-Frequency

Motion.
Our synthesisformulation (19) is based on the representation of the source time
function in the far-field due to a constant dislocation over a fault plane in an infinite
homogeneous elastic medium, i.e. (6). When a coherenr rupture propagation and a
constant dislocation represented by a linear ramp time function are assumed over
a rectanguiar fault plane, the source time function expressed by (6) has flat spectra
at low frequencies and <o-3 high-frequency asymptotes. For the case of lvlodel
Hl- I in section 2.4, the spectral amplitude of the source time function can be
o b t a i n e d i n t h e s i m p l e e x p r e s s i o n( M i k u m o , l g 7 l 3 s ) a n d G e l l e r , t n t 6 z z > ) :

r^s(-)r:ri"*3],t*f.L.1 'r!ji:ll (2s)

where xt:lL(l lun-cos plu,) l2l, xr:t,lv c o sg s i n 9 l ( 2 u , ) l a n d x , : r 1 2 . It is clear
from (25) that the source time function of Nlodels Hl-1 and Hl-2 have cu-3 decav
at the high frequencies. Similarly, for the case of lvlodel H2-2 the source time
function has a flat spectrum at low frequencies and u)-3 decay usually at the
frequencies higher than I fr, as shown in Fig. 9, although the Fourier spectrum
can not be obtained in an analytical expression.
Thus, in our synthesis formultion, the spectrum of G(x, t) expressed by a triple
integral such as (19) yields c.r.r-3
decay at frequencies higher than lfr, if G,,_ (x, t)
has a flat spectrum. The synthesized seismogram shown in Fig. 3l has smaller
amplitudes than the observed one at frequencies higher than about I Hz (consistent
with I /r). This means that there is large discrepancy between the characteristic
leatures involved in the observed seismogram ancl the assumption in our synthesis
in the frequency range higher than llr. This problem is related to the basic
assumption in our formulation, that is, a smooth rupture propagation over a rectan-
gular fault plane. To recover the fall-off of the spectral amplitudes ar frequencies
higher than I Hz to the observed level. it would be necessary to introduce inhomo-
g e n e o u sf , a u l t m o d e l s s u c h a s t h o s e i n c l u d i n s ' b a r r i e r s ' ( e . g . D a s a n d A k i , l g 7 7 t z , ) .
However, from another point of view we can apply the present synthesis methocl t6

I
 Seni-Empirical Estimation of Strong Ground Motiotts During Large Earthqtnkes 95

I
I
f--f--r-r-r-r- v
---f---#--.1 -z
i_-i--+-i-+-;{
|---f----t--f-?F-i
e --r-r-i--#
I
f-#
____fti*___,
d

tr u , /
./,/
i /
I
I
)S I
J
ffi,
(a) (b)

Fig. 34. lal Thc distribution of strblaults during a large event. Each sub{ar,rlt corresponds
to a small event.
( b ) The relation between the dislocation time lirnction of the small event and that
of the large event in
the space and time domain when is considered the
distribution of srrbfar.rlts as shown in (a).

estimate the high lrequency motions by modifying our formulation without directly
presuming 'barriers', if we use elementary earthquakes having a characteristic scale
involved in the mainshock phenomena.
Now, we return this discussion to the relation (9),

l v n

A {l (t, \, t) : Z / ti,l(, r, t- (k- l)r,l again (9)

h=L

and start to revise the formulation. In section 2.2,we introduced (9), onlv in the
time domain, to satisfy the similarity condition,

D l D ,:tl r,- No (26)

The relation (9) is here reconsidered to have a physical meaning in the space and
time domain during the mainshock. Suppose that i/o subfaults are aligned in
a certain space interval lx contacting one plane after another as shown in Fig. 34a.
The length of a subflault corresponds to that of an elementary earthquake, L,.
Accordingly, a uniform dislocation on a subfault is taken to be consistenr with the
same distribution as the dislocation during an elementary earrhquake. Then
the dislocation during the mainshock is expressed by the delavecl surnmation of the
dislocation during the elementary earthquake along the x-axis,

.Vn

/ U (.r): I ALr,l.u-(fr- I )Arl. (27)

h = I

When the rupture of the subfaults propagates along the positive direction of the
x-axis with the velocity u,, (27) can be rewritten in time domain,

ArJ(i): : tr,r,lt-$-D i:) (28)

:
tI

,
!
 K. IRIKURA

If Ax is replaced by ur're;equation (28) becomes equivalent to (9). Thus, we can

rewrite the relation (9) by presuming the space and time distribution of the sub-
f,aults as shown in Fig. 34b as follows:

t"D

/ U ( t , , t ,t ) : 1 , .A U , l {- ( f - I ) n , r , , , 1t ), , (2ea)
h=L

ND
and / tt ((, r, t) : Z. A U,l(, \ - (k- l)u,r,, tl. (2eb)
A=l

Since f and rl are taken along the strike and the dip direcrions, respectively, (2ga)
is an intuitively clear relation to express the case of strike-slip type fault, while (2gb)
is suitable for the case of the dip-slip type fault.
We have adopted (29a) in this study, since the l9B0 Izu-Hanto-Toho-Oki
earthquake we have analvzed here has the strike-slip type mechanism. Conseque-
ntlv, equations (19) and (20) are rewritten as follows:

ND NL Nly
G ( x ,t ) : I I X Gr1^(x, t-t'ap1-) (30)
h=L l=l m=l

,:!: ) qS+-w--Md:tft
uc A,
(31)
Axr:p ra" (32)
If we use elementary earthquakes with the fault length L,:Ir{oax" fior the
svnthesis, (30) and (31) are further rewritten.

Np*N2 Nry
G(x,t,\:,1, prG,t^(.x,t-t'a1'^) (30)'

t',., -r!:v- -l- Jt,2l'l^[

L dl'm
(31)'
Uc A,

where

q^: (rn- l) rl/,, m:1, 2, ..., I{r,

( t ,: ( l ' - l ) L , I N D , I ' : 1 , 2 , . . , , N t X I \ [ o ,

and G,1^ is a seismogramfrom an element /Ir^(/:l-i/r and m:l-y'/,r)"

The accelerogram synthesized by (30') has a-2 decay when each G,1- has a1
flat spectrlrm so that it has more rich high-frequency motions than that synthesized
by (19) This tends to reduce the above-described discrepancy between the
synthesized and the observed accelerograms. The synthesis by (30') neecl not
apparently eive both the rise time r for the mainshock and r, for elementary
earthquakes. Actually, we have to use elementary earthquakes with a definite f,ault
length, L,:lr{o/x":p,.[{p7e:t)r.r. Since r=l sec and u,=3 kmisec for l9B0 Izu-
Hanto-Toho-Oki, we need to use foreshocks or aftershocks with the length of about
3 km.
 Seni-EmpiricalEstimationof StrongGroundMotionsDuring LargeEarthqtnkes 97

Bouchon (1978;'u' showed that the multiple cracks with barriers are roughl,v
equivalent to dislocation model with a uniform slip of the Haskell-rype source.
Based on this evidence, Aki et al. (1977])tt>showed that one.canmake a roug6 esti-
mate of a barrier interval from the rise time in such a way that

(barrier interval) - (rupture velocitv) . (rise time) (33)

On the other hand, Aki et al. summarized probable barrier intervals inferred from the
observed fault slip for several earthquakes studied by Matsuda ( lg723i:, and others).
They irrferred an average barrier interval of 3 km for the lg74 lzu-Hanto-Oki
earthquake based on Matsuda and Yamashina (1974)3e). It is of interest that
'barrier
interval' from (33) is about 3 km lor the l9B0 Izu-Hanto-Toho-Oki earth-
quake and almost the same as that for the 1974 Izu-Hanto-Oki Earthquake which
occurred in the near region southwest of the l9B0 Izu-Hanto-Toho-Oki earthquake.
It is concluded from the revised formulation (30') that for the synthesis of high
lrequency motions it would be better to use elementary earthquakes with the char-
acteristic scale of the f,ault length,
(fault length of elementary earthquake)-(rise time of the mainshock) . (rupture
velocity).
Under this condicion, the synthesized motions do not miss the high frequency con-
stituents including the source characteristics of the elementary earthquakes. Accord-
ing to the 'barrier model', this means that the optimum elementary earthquakes for
the synthesis should have a fault length consistent with a 'barrier interval' of the
mainshock.

4.3- Synthesized Results of Strong Accelerograrns Using the Revised

Method
The lault lengths of foreshock P4 and aftershock Al with LI:4.9 are estimared
to be both about 2.5 km lrom the similarity condition L,lL:NM;[M;,:iy', where
-l/ is estimated to be about 6 from the spectral ratios between the mainshock and the
small events as discussed in section 3.2. and Z is estimated to be 15 km from the
epicentral distribution of the aftershocks. Similarly, the f,ault length of aftershock
43 with M:4.6 is estimated as about 2.1 km. In order to make a synthesis by the
revised formulation (30'), it is necessary to use the records from elementary earth-
quakes having the f,ault length of about 3 km from the condition Z,:ur.r. The
three events P4, Al and A3 are regarded as having the fault length of the same
order. Therefore, using the observed seismograms from these events, we make a
synthesis of the mainshock motions by means of (30'). An operation of smoothing
has been made on the basis of (22'\, to reduce the ghostly oscillatory motions clue to
the apparent periodicity of lxfu, corresponding to 7, mentioned in section 4.1.
The waveform and the spectrum of NS-component of the synthesized seismo-
gram for the.|IZ station are compared with those o[ the observed one in Fig. 35.
The upper two traces are the observed seismograms from events Al and A3 used as
 98 K. IRIKURA

P E N ,. J U N 3
. 0'80, 2H231

" " f j ". a. n' *0E0tvsEEL2Dt v,. 2 E V E N T( ^S3 .A t )

'80. l8H06f1 R = 3 . 2 v, s = 3 .s , L '
J I Z . E O F F I Z U P E N ., J U L . 2 7 -
= 4 . 6 , H = l 7 K t 1A 3 NS

O E i -2 1 , 2 E V E N T(SA 3 ,A I )
v I s E 0 ,V R = 32. , V 5 = 3 5
. , L = t 5 , t ^ l = 75.
S Y NT H E S I Z E D lo-'
0 .0 l

G A L ' S , tE7C. F n F F r 7 u p E N . .J U N . 2 s
' n 1
-
0BSER'i'-u
t0.00

J I Z , E O F F I Z U P E N. J U N 2
= 6 . J , H = l 0 K MN
, S OBSERVED
. 9 '80, r6H20r1
N5
ro
t n
'I'llTv
VI
Ailil
,
n lll
/\

o
lo

ill
l0
tJ,
Fig. 35. comparison of the synthesizedacceleration,.tr*o*.u,] *rt...riio.,l.,, b, ,ht
revisedfiormulation(30') with the observedseismogram at the JIZ station. The
leli upper two tracesare the observedaccelerograms of aftershockAl and A3 used
as elementaryearthquakes,the left third trace,the synthesized accelerogram and
the left bottom trace, the observedaccelerogramof the mainshock. The right
upper figureis the Fourierspectrumof the synthesized accelerogram and the lower
figure,that of the observedmainshockone.

the elementary earthquakes. The parameters, except for r,, are

given to be the
same values as the case of the synthesized velocity motions in section
3.3. It is not
neccessary to give r, but to give Ax:L,lN. The third and the fourth trace in
Fig. 35 show the synthesized and the observed accelerogram. The
envelope of the
synthesized waveform agree well with that of the observed one, although
the two
waveforms do not always correspond to each other in individual phases.
The spect-
ral amplitudes of the synthesized accelerogram (the right upper
figure in Fig. 35)
agree well with those of the observed one (the right lower figure)
at the frequencies
up to 5 Hz.
The waveform and the spectrum of NS-component of the synthesized
accelero-
,sram flor the SN4C station are compared with those of the observed one in Fig. 36.
 Semi'Empirical Estimation of Strong Ground Motions During Large Earthquakes

S i 1 Ci, O F F I Z U P E N ., J U N .3 O ' 8 0 . 2 H 2 3 N
=4.9,H=t8KfA 1l N5

S I l C ,E O F F t Z U P t N . . J U N .2 8 ' 8 0 . l 2 l l 0 ' 5 N ^U ^.L.x -^l lm 0 0 E2Lt , 2 E v E N T S { P 1 , A l )

i t v t q t r n
\ r a - ? n r - r q r r - t c
= 4 . 9, H = l 9 K NP 4 N5-CONP. -

R.
sEc

f r 0 0 E L 2 , 2 E V E N T (SP 4 ,A I )
F E VI S E Ov B = 3 . 2 , L = t 5 , r ^ l = 75.
S / NT H E S t Z ED

r.00 8 .0 0 l:,00 16.00 20.00 2r.00 24,00 t2.00 t8.00

5EC

5 1 1 CE, O F F I Z U P E N . , J U N . 2 9 ' 8 0 , I 6 H 2 O I 1
N5-C0flP.

n.

0.01 0,I !. l0
Hz

Fig. 36. Comparison of the synthesized acceleration seismogram of NS-component by the

revised formulation (30') with the observed seismogram at the SMC station. The
arrangement of the figures is the same as Fig. 33.

The two events P4 and A I are used as the elementarv earthquakes for the present
synthesis. Similar to the case of the JIZ station, the paramerers are given to be the
same as the case of the svnthesized velocity motions. The envelope of the synthesized
waveform also agrees well with that of the observed one. As compared in the
right fi.gures, the spectral amplitudes of the synthesized one agree well with those of
the observed one at the frequencies up to 5 Hz, similar to the case of the
JIZ station.
Fig. 37 shows the V-component seismograms of the synthesized acceleration
motions and the observed ones for JIZ station and SMC sration. We can see the
vertical component synthesized-accelerograms are also in a good agreement in its
envelope with the observed ones for the both stations.
These results show that the revised method (30') is extremely useful for the
synthesis of high frequency motions up to 5 Hz. We consider that it is difficult to
synthesize deterministically higher frequency motions beyond 5 Hz, because these
high-frequency motions may be represented as having statistical natures in source
effect and path effect as discussedby Andrews (lgBl)a0) and others.
The frequency range effective for the synthesis by the revised method (30')
100 K. IRIKURA

JI Z , N S
. , 1 1 ! r 1 1 . I ' . Y [ : , ; l A ] , \ 1 ,
' . 1 . . . i : , . ] . S .
1 , '/ , , : t | , . ' = : r . L r . )

syNrhEStzED
: I -\,^
'.! :
6 i _ ' : - r
\
, ' . - ., J1 '
lr/.\-/
s r c \ ; " v
JIZ,V
, , - J i 1 . E0 F ' : I I U p t N . , . J ! f t . 2 . j . 8 t03. i i 2 L r r j
=5Y5:';ari:sl,
i,fr
3,{89ibr6,' ! rs.,r=, _11r.6-
r . r i . l r l ( i l !. 5
OASERVED
SVNTHESIZEO
"
:'] | t*
L- --a,, M '-u, ,. 7
,'"/
1-'
"'\V---"\--r''\- ^=^--'
r,,,t.",t-''."'
; l"rg*uh,,J^ri{1'l,l/Xip^f/^*
r;.-
5.:.
,*!-',,/
,,jV

3.0l I0.c0 !). ta

sEc

2s'
; 1 : A l ; 1 " 9 i [ ^ i 1 ; , ! ; ; ; ; o 'B" \C' . ' i' !, , , - n
= l r

"'"^,'q';',''*,s"*,,'-".
i h;,',1^jh:,t'41","""11y\rti

,r^\I^.V4.-^-,."."./-\
SMC,V
s.r99!b:3,
?;.5Y5)::131:l
i: SYNT|€S|ZED
-- -
-iO1;ll,,l,ifJ,i[:;i,li*,'""';;,1i'4,^+r1r.r.,'.,1,r'\.rnt"'*i',,,,r,
5 . , 1 , t .I r , l

Fig. 37. Comparison of the synthesizedaccele- Fig. 38. Comparison of the synthesized velo-
ration seismograms of V-component city seismograms of NS-component
by the revised formulation (30') with by the revised formulation (30') with
the observed seismogramsat the JIZ the observed seismograms at the JIZ,
and SlvtC station. SMC and OMM station.

is not restricted to high frequencies. By means of this revised method the synthesized
velocity motions for theJIZ, SMC and OMM station are compared with the observ-
ed ones in Fig. 38. We can see an extremely good agreement between the synthesiz-
ed and the observed seismograms for the three stations.
This revised formulation is also based on a smooth rupture propagation over a
fault plane. However, seismic effects due to the complex nature of the rupture process
inside the blocks with the length I,, corresponding to the source sizes of small events,
are kept in the synthesized results rvithout being filtered in time domain through the
synthesis procedure. This formulation is based on an idea that the similarity con-
dition between the dislocation of the mainshock and that of small events is satisfied I

in each block with the length u,'r. If the length of each block is regarded as a r 1

kind of barrier interval, this synthesis method is interpreted to be close to computa- I

tional models of multiple cracks with barriers shown by Bouchon (1978)36) and o
Madariaga (1974)a". fl
cl
 Semi'Empirical Estimation of StrongGround Motions During Large Earthquakes l0l

5. Conclusion

l. The synthesis method described in this paper is based on the representation,

including the source effects, relating to the dislocation at every point on the fault
plane and to the rupture propagation over the fault plane, and the path effects,
relating to the wave propagation from the source to the site, although some approx-
imations are made. If there is a certain similaritv relation between large earthquakes
and small ones within the same source area, the equation for the synthesis is an accu-
rate approximation for the wave field from the source to the site.
2. The synthesis method is checked by synthesizing the velocity motions of the
mainshock for 3 stations at the short distances (/:20-100 km) from the epicenter
in the case of the l9B0 Izu-Hanto-Toho-Oki earthquake, using the records of two
small events, whose hypocenters are located at the northern half area and the
southern half area of the mainshock fault plane, respectively. The synthesized
seismograms show a good agreement with the observed seismograms in the frequency
range less than I Hz.
3. The validity of the assumed source parameters, such as the starting point of rup-
ture, the rupture velocity, and the rise time is examined by means of three kinds of
measure, a correlation function, an amplitude ratio and a residual function between
the synthesized seismograms and the observed seismograms of the mainshock. We
find that the best agreement between the two seismograms is obtained for the model
expected from the epicentral and depth distribution of the aftershocks and the similar-
ity condition between the mainshock and the small events. These results show that
the present synthesis is physically meaningful.
+- A synthesis method for higher frequency motions is further revised by changing
the relation in the time domain between the dislocation function of the mainshock
and that of small events to the relation in time-space domain between them. When
we use the records from the small events having the f,ault length L":u,.r (u,:
the rupture velocity, z: the rise time of the mainshock), this revised synthesis is
effective for higher frequency motions. The synthesized accelerograms by this
revised method show a good agreement with the observed accelerograms of the main-
shock in the frequency range up to 5 IHz.

Acknowledgernents

The author is grateful to Prof. S. Yoshikawa for critically reading to improve

this paper and for his encouragement in carrying out this work. The author thanks
Prof. I. Muramatu for his valuable suggestions and comments. The observation
for this study was performed in collaboration with him. The author has benefitted
greatly from discussion of Prof. T. Mikumo and the motivation of this study arose
from his suggestions. The authour thanks Prof. Y. Kobayashi for many stimulating
discussions and comments. Dr. Kinoshita kindly provided the digital data obtained
 t02 K. IRIKURA

at the Nakaizu station (.\ru) of National Research Center for Disaster Prevention,
the Science and Technology Agency. Helpful discussionsduring the course of this
work with Mr. M. Horike, F. Amaike, S. Kasuga and many other colleagues are
aiso gratefully acknowledged. The author thanks aiso to IVIr. M. Nishi for his
co-operation in the observation and Mrs. K. Kutsuki for arranging this paper. The
author would like to acknowledge an anonymous reviewer for very careful help to
improve the manuscript. This research was supported in part by a Grant-in-Aid
for Natural Disaster Scientific Research from the Ministry of Education, Science and
Culture under Grant No. 57025025. The data processing was run on a Facom
N4-140 at the Information Data Processing Center for Disaster Prevention Research,
of Disaster Prevention Research Institute of Kyoto Universitv.

References

l) Aki, K.: S e i s m i cD i s p l a c e m e n t sn e a r a F a u l t , J . G e o p h y s . R e r . , V o l . 7 3 , 1 9 6 8 ,p p . 5 3 5 9 - 5 3 7 6 .
2) Haskell, N. A.: Elastic Displacements in the Near-Field of a Propagating Fault, Bull. Seism.
Soc. Am., Vol. 59, 1969, pp. 865-908.
3) Kawasaki, f., Y. Suzuki, and R. Sato: Seismic Waves Due to a Shear Fault in a Semi-infinite
M e d i u m , J . P h y s . E a r t h , V o l . 2 1 , 1 9 7 3 ,p p . 2 5 1 - 2 5 4 .
4) Sato, R.: Long-Period Surface Velocities and Accelerations Due to a Dislocation Source Model
i n a M e d i u m w i t h S p h e r i c a l M u l t i - L a y e r s , P a r t I I , J . P h y s . E a r t h , V o l . 2 6 , 1 9 7 8 ,p p . 1 7 - 3 7 .
5) Sato, R. and N. Hirata: One Method to Compute Theoretical Seismogram in a Layered
M e d i u m , J . P h y s . E a r t h , v o l . 2 8 , 1 9 8 0 ,p p . 1 4 5 - 1 6 8 .
6) Heaton, T. H. and D. V. Helmberger: Generalized Ray Models of the San Fernando Earth-
q u a k e s ,V o l . 6 7 , 1 9 7 9 , p p . l 3 l l - 1 3 4 1 .
7) Bouchon, M.: Discrete Wave Number Representation of Elastic lVave Fields in Three-Space
Dimensions, J. Geophys. Res., Vol. 84, 1979, pp. 3609-3614.
B) Hartzell, S. and D. V. Helmberger: Strong Motion Modeling of the Imperial Valley Earth-
q u a k e o f 1 9 7 9 ,B u l l . S e i s m .S o c . A m . , Y o L . 7 2 , 1 9 8 2 ,p p . 5 7 1 - 5 9 6 .
9) Alekseev, A. S' and B. G. Mikhailenko: Numerical Modelling of Transient lVave Fields in
Seismology and Seismic Prospecting, Siberian Branch of Academy of Sciences of the USSR,
Novosibirsk. 1979.
l0) Aki, K.: Characterization of Barriers on an Earthquake Fault,J. Geophys.Res., Vol.84, 1979,
pp. 6140-6148.
I I ) Mcgarr, A. : Analysis of Peak Ground Motion in Terms of a Mode I of Inhomogeneous Faulting,
J . G e o p h y s .R e s . , V o l . 8 6 , l 9 8 l , p p . 3 g 0 l - 3 9 0 2 .
12) Das, S. and K. Aki: Fault Plane with Barriers: A Versatile Earthquake Model,J. Geophys.
Res. 82, 1977, pp. 5648-5670.
13) Mikumo, T. and T. Miyatake: Dynamical Rupture Processon a Three-Dimensional Fauit
with Non-Uniform Frictions and Near-Field Seismic lVaves, Geophys.
J. Roy. Astron. Soc.,
Vol. 54, 1978, pp. +17-438.
l4) Papageorgiou, A. S. and K. Aki: A Specific Barrier Model for the
Quantitative Description of
Inhomogeneous Faulting and the Prediction of Strong Ground Motion, I and II, in Proceedings
of Workshop XVI-The Dynamic Characteristics of Faulting fnferred from Recordings of
S t r o n g G r o u n d M o t i o n , V o l . l , J . B o a t w r i g h t , E d i t o r , U . S . G e o l . , S u r v . ,O p e n - F i l e R e p t . 8 2 - 5 9 1 ,
1982, pp. 3l l-504.
l 5 ) B o a t w r i g h t , J . : A D y n a m i c M o d e l f o r F a r - F i e l d A c c e l e r a t i o n ,B u l l . S e i s m .S o c . A m . , Y o l . 7 2 ,
1 9 8 2 ,p p . 1 0 4 9 - 1 0 6 8 .
l6) Madariaga, R.: High-Frequency Radiation from Crack (Stress-Drop) Models of Earthquake
 Serni-EmpiricalEstimationof StrongGroundMotioru During Large Earthquakes 103

F a u l t i n g , G e o p h y s . . J .R o y . A s t r o n . S o c . , V o l . 5 1 , 1 9 7 7 ,p p . 6 2 5 - 6 5 2 .
17) H a r t z e l l , S . : E a r t h q u a k e A f t e r s h o c k sa s G r e e n ' s F r - r n c t i o n sG , eophys. Res. Letters, Vol.5,
1 9 7 8 ,p p . l - 4 .
lB) Kanamori, H': A Semi-Empirical Approach to Prediction of Long-Period Ground lvfotions
f r o m G r e a t E a r t h q u a k e s ,B u l l . S e i s m . S o c . A m . , V o l . 6 9 , 1 9 7 9 , p p . 1 6 5 + - 1 6 7 0 .
l9) Hadley, D. lvl. and D. V. Helmberger: Simulation of Strong Ground NIotions,Bull. Seisrn.
S o c . A m . , V o l . 7 0 , 1 9 8 0 ,p p . 6 1 7 - 6 3 0 .
20) I r i k u r a , K . a n d I . N { u r a m a t u : S y n t h e s i so [ s t r o n g G r o u n d l v l o t i o n s f r o m l - a r g e E a r t h q u a k e s
U s i n g O b s e r v e d S e i s m o g r a m s o f S m a l l E v e n t s , P r o c e e d i n g sT h i r d I n t c r n a t i o n a l E a r t h q .
M i c r o z o n a t i o n C o n f . , S e a t t l e ,V o l . l , 1 9 8 2 , p p . 4 4 7 - 4 5 8 .
2l) Imagawa, K' and T. Mikumo: Near-Field Seismic iVaveforms from lVlajor Earthquakes and
C o n s i d e r a t i o n o n t h e R u p t u r e P r o c e s so n t h e F a u l t , J i s h i n , V o l . 3 5 . 1 9 8 2 , p p . 5 7 5 - 5 9 0 ( i n
Japanese).
22) Iida, lVI. and M. Hakuno: Estimation of Building ResponseDue to Large-scale Earthquake
from Observed Response Records to Small Earthquakes, Proceedings Third International
Earthq. Microzonation Con[., Seattle, Vol. I l, lgB2, pp. 895-908.
23) A k i , K . : S c a l i n g L a w o f S e i s m i cS p e c t r u m ,J . G e o p h y s .R e s . , Y o l . 7 2 , 1 9 6 7 ,p p . l 2 l 7 - l 2 3 l .
24) Aki, K.: Scaling Law of Earthquake Source Time-Function. Geophys..l.R. Astron. Soc.,
3l:3-25,1972.
25) Kanamori, H. and D. L. Anderson: Theoretical Basis of Some Empirical Relations in Seis-
m o l o g y , B u l l . S e i s m . S o c . A m . , V o l . 6 5 , 1 9 7 5 ,p p . 1 0 7 3 - 1 0 9 5 .
26) Tsuboi, C.: Earthquake Energy. Earthquake Volume, Aftirshock Area and Strength o[ the
E a r t h ' s C r u s t , J . P h y s . E a r t h , V o l . 4 , 1 9 5 6 ,p p . G 3 - G 6 .
27) Geller, R.J., Scaling Relations for Earthquake Source Parameters and lvlagnitudes, Bqll.
S e i s m . S o c . A m . V o l . 6 6 , 1 9 7 6 ,p p . l 5 0 l - 1 5 2 3 .
28) Abe, K.: Reliable Estimation of the Seismic lvloment of Large Earthquakes, J. Phys. Earth,
V o l . 2 3 , 1 9 7 5 ,p p . 3 8 l - 3 9 0 .
29) Haskell, N. A.: Radiation Pattern of Sr.rrfacelVaves from Point Source in a lvltrlti-Layered
M e d i u m , B u l l . S e i s m .S o c . A m . , V o l . 5 4 , 1 9 6 4 ,p p . 3 7 7 - 3 9 9 .
30) A k i , K . , a n d P . G . R i c h a r d : Q u a n t i t a t i v e S e i s m o l o g y ,T h e o r y a n d l v l e t h o d , V o l . 2 , C h a p t e r
14, pp. 804-805, W. H. Freman and Company, 1980.
3l) Chouet. B.' K.Aki and M. Tsujiura: Regional Variation o[the Scalling Law cl'Earthquakes
S o u r c e S p e c t r a , B u l l . S e i s m .S o c . A m . , V o l . 6 8 , 1 9 7 8 , p p . 4 9 - 7 9 .
32) Muramatu, I.: A Velocity Type Strong Motion Seismograph with !\'ide Frequency Range,
J . S e i s m . S o c .J a p a n , J I S H I N , V o l . 3 0 , 1 9 7 7, p p . 3 l 7 - 3 3 8 ( i n J a p a n e s e ).
3 3 ) K i n o s h i t a , S . : P a r a l l e l O b s e r v a t i o no f S t r o n g G r o u n d j V f o t i o nb y U s i n g B o t h A c c e l e r a t i o na n d
Velocitv Type Seismometersat Nakaizu Crustal Active Observatory. Research Notes National
R e s . C e n t e r D i s a s t .P r e v . , V o l . 4 4 , l 9 8 l , p p . l - 3 l .
3+) Imoto, M., I.Karakama, R. S. Matsuura, F. Yamazaki, A. Yoshida and K. Ishibashi: Focal
M e s h a n i s m s o [ t h e l 9 B 0 E a r t h q u a k e S w a r m o f f t h e E a s t C o a s t o f t h e I z r r P e n i n s u i a ,J a p a n ,
JISHIN, Vol. 34, l98l, pp. 48l-493.
35) Mikr-rmo, T.: Source Processof Deep and Intermediate Earthquakes as Inl'errcd from Long-
P e r i o d P a n d S l V a v e f o r m s , J . P h y s . E a r t h . V o l . 1 9 . 1 9 7 1 ,p p . l - 2 0 .
36) Bouchon, lv[.: A Dynamic Source IVIodelfor the San Fernando EarthqLrake,Bull. seism. Soc.
A m . , V o l . 6 8 , 1 9 7 8 ,p p . 1 5 5 5 - 1 5 7 6 .
37) Aki, K., ivl. Bouchon' B. Chouet, and S. Das: Quantitative Prediction of Strong N{otion lbr
a P o t e n t i a l E a r t h q u a k e F a u l t , A n n a l i d i G e o g i s i c a ,V o l . 3 0 , 1 9 7 7 ,p p . 3 4 l - 3 6 8 .
3 8 ) N { a t s u d a , T . : S u r f a c e F a t r l t s A s s o c i a t e dw i t h K i t a - I z u E a r t h q u a k e o [ 1 9 3 0 i n I z u P e n i n s u l a ,
'Ihe
Japan, I z u P e n i n s u l a , d e i t e d b y M . H o s h i n o a n d H . A o k i , T o k a i U n i v . P r e s s ,T o k y o . ,
t972.
39) M a t s u d a , T . a n d K . Y a m a s h i n a : S u r f a c e F a u l t s A s s o c i a t e dw i t h t h e I z u - H a n t o - O k i E a r t h -
quake of l974,Japan, Sokuho (inJapanese). Earthq. Res. Inst. Univ. Tokyo, Vol. l+, lg7+
104 T(. TRIKUR.I

pp.135-158.
40) Andrews,D.J., A Stochastic Fault Model,2. Time-Dependent Case,J.Geophys.Res.,Vol.
8 6 , 1 9 8 1p, p . 1 0 8 2 1 - 1 0 8 3 4 .
4l) Madariaga,R.: on the Relation betweenSeismiclvlomentand Stress
Drop.in the presence
of Stressand StrengthHeterogeneity, J. Geophys.Ro., Vol. 84, 1979,pp. 2z4g_22s0.