You are on page 1of 15

Bulletin of the Seismological Society of America. VoL 61, No. 6, pp. 1739-1753.

December, 1971

A M E T H O D FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION


BY ~/[IHAILO D. TRIFUNAC
ABSTRACT

A new method of synthesizing artificial accelerograms, based on knowledge of the temporal and spatial properties of the energy-release mechanism, incorporates the effects of local geology through theoretical group-velocity dispersion data for a given site. It is assumed that most of the near-field strong ground motion caused by shallow or surface faulting or explosions is represented by energy propagating through the low-velocity surface-wave guide. The new method of generating artificial accelerograms realistically models strong ground motion for applied structural analysis and theoretical investigations of the response of hysteretic-type structures, soil-structure interaction, and statistics of failure.
INTRODUCTION

The most realistic approach to the design of earthquake-resistant structures is through the dynamic analysis of structural models excited by realistic ground motion. Except for a few cases, no typical recording is available for past strong ground motion at a given site: Even if such a recording were available, there is at present no basis to expect that a future earthquake might generate the same or similar ground motions. Thus, it becomes necessary to develop a method of synthesizing strong ground-motion records. This synthesis must be based on all of the physical properties of strong motion that are presently known and must incorporate the probabilistic nature of earthquake occurrence. This paper outlines a method of accelerogram synthesis. This method is based on a detailed interpretation of several recorded strong-motion accelerograms (Trifunae, 1971) and incorporates the properties of the local geological site and source mechanism whenever these are known. To demonstrate the method, it is necessary to assume some reasonable spectral properties of the modeled ground motion. Since many actual properties of the near-field strong earthquake ground motion are unknown, the assumed functional forms for the spectra used in the synthesis may, at best, only approximate the actual spectra.
HISTORICAL REVIEW OF THE ACCELEROGRAM PREDICTION AND INTERPRETATION

The first strong-motion recordings were obtained during the Long Beach, California, earthquake March 10, 1933, and seven years later during the Imperial Valley, California, earthquake of May 18, 1940. Studies dealing with interpretation and use of strong-motion accelerograms were then initiated. Housner (1947) gave the first engineering interpretation of strong-motion accelerograms. He concluded that the extreme irregularity of recorded ground acceleration could be explained by large numbers of waves whose arrivals are determined by factors other than the origin of the disturbance or the properties of the medium through which the waves propagate. In a later paper, Housner (1955) assumed that an accelerogram could be formed by adding a large number of one-cycle sine-wave acceleration waves occurring randomly in time. He showed that this representation leads to a response spectrum that favorably agrees with spectra derived from recorded accelerograms.
1739

1740

BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA

Housner's interpretation was used by many to investigate various aspects of structural response to the random earthquake-like excitation. Goodman et al. (1955), Hudson (1956), Rosenblueth (1956), Bycroft (1960) and Rosenblueth and Bustamante (1962) modeled strong motion with series of pulses distributed randomly in time. To study the response of hysteretic structures, Housner and Jennings (1964) developed Gaussian random accelerograms. The power spectral density for this process was determined from the average undamped velocity spectra of recorded ground acceleration. Tajimi (1960), Barstein (1960), Goto and Kameda (1969), and Penzien and Liu (1969) used probability methods to design for the effects of seismic forces and to determine the maximum response of a building structure during an earthquake represented by such stationary Gaussian processes. As the number of recorded accelerograms increased, it became evident that many accelerograms do not resemble stationary random processes. Although stationary Gaussian processes were designed to model only the "primary phase" or "strong motion" part of an accelerogram (Housner and Jennings, 1964), the recorded accelerograms demonstrated that even this strong-motion part is not stationary. Bolotin (1960) made some basic observations dealing with this nonstationarity. Bogdanoff et al. (1961) modeled strong motion by a nonstationary, random process represented by a sum of decaying sinusoids. Other nonstationary random processes were generated by Cornell (1964), Amin and Ang (1966), Goto et al. (1966), Shinozuka and Sato (1967), Jennings et al. (1968), and Goto and Toki (1969). A nonstationary process can be derived from a stationary process by multiplying the latter by a timedependent envelope. Amin and Ang (1966) used such an envelope to modify the stationary Gaussian process. A choice of envelope is made to agree with different types (Jennings et al., 1968) of recorded accelerograms. On the basis of reasonable physical assumptions, Rascon and Cornell (1969) synthesized a strong-motion accelerogram by a sequence of elementary double-couple sources located along a line. Their model incorporates the effects of spherical spreading and attenuation (Honda, 1957) but is applied to P and S waves only. From the brief outline of some major contributions to the problem of constructing artificial accelerograms, it may be seen that the early interpretations of strong earthquake ground motion in the near epicentral region (Housner, 1947; 1955) played an important role by defining the general gross features of the first recorded accelerograms as similar to a sequence of random pulses. Although the source mechanism differs, the strong motion caused by an underground nuclear explosion may be predicted by the same methods developed in this paper. One approach to the synthesis of nonstationary random functions to simulate explosion-generated ground motionwas recently outlined by Beaudet (1970). Although his model introduces frequency dependence in time, it fails to relate this nonstationary frequency variation to the actual in situ properties of the ground. A simple approach to modeling artificial accelerograms by comparing their relative response spectra, or the Fourier amplitude spectra, with those from actually recorded earthquake aecelerograms has serious limitations. When the phase spectrum is neglected, as is usually done in such comparisons, there are infinitely many functions for which response spectra may be identical. The June 27, 1966 Parkfield, California earthquake, for example, generated relative velocity spectra with essentially the same spectral amplitudes as the May 18, 1940 Imperial Valley, California earthquake (Housner and Trifunac, 1967). The ground motion during the Parkfield earthquake

A METHOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1741

caused little damage, presumably because of its short duration, whereas the Imperial Valley earthquake of 1940 caused significant damage (Richter, 1958). Hence, to generate physically realistic accelerograms, a more detailed time-dependent interpretation of strong motion is needed. Amin et al. (1969) considered the significance of the acceleration nonstationarity in time. They showed that for lightly-damped linear systems with a single degree of freedom, nonstationarity does not appear to be too important. For more realistic nonlinear inelastic systems, however, the effect of nonstationarity cannot be neglected. In a simple Fourier analysis of a complex record, information on time dependence of incoming wave forms, i.e., the time of their arrivals, duration, amplitudes and dispersion properties, are not directly displayed. To retain this information, a form of the multiple filtering technique (Trifunac, 1971) can be used. Some results of applying such a technique on several typical aeeelerograms and an interpretation of strong ground motion are given below.
NATURE OF STRONG ~OTIONS NEAR SHALLOW SOURCES

Until about 10 years ago, seismic-source studies were based on far-field measuremeats of seismic waves and were concerned mainly with a source of energy release represented by a point source. More recently, studies dealing with large earthquakes motivated the development of source representation in terms of simple dislocation theory. Although it may represent a etude and grossly average way of describing actual source mechanism when compared to near-field measurements, this theory is sufficiently accurate for the far-field observations. Since high-frequency seismic waves are attenuated and scattered most rapidly with distance, nature limits the possible accuracy of resolving details of source mechanisms from distant records. The degree to which a pattern of seismic-energy release can be determined by recording within 30 km of the source with a high-frequency (about 15 eps) aecelerometer is indicated by a study of the May 18, 1940 Imperial Valley, California earthquake (Trifunac and Brune, 1970). The "pattern of seismic energy release" is meant here to represent a sequence of one or more seismic events in time. These events superimposed represent an average dislocation model that characterizes distant long-period waves. The meaning of an "event" is here associated with any occurrence of energy that generates P, S and surface waves that can be resolved and identified and that might be associated with a relatively discrete single dislocation. Such a definition of an "event" clearly depends on the frequency resolution of the recording instrument and its distance from the source. A typical simple event as recorded in E1 Centro during the Imperial Valley earthquake is shown in Figure 2. The idea of the "immediate source region" and the "extended source region" is introduced to systematically separate the first stage of wave radiation from the later effects of wave reflections and refractions caused by the material discontinuities in the surrounding medium. Together they form the near-source region. Immediate source region. Tile lack of high-frequency recordings of waves generated at the causative fault precludes precise interpretation of the actual mechanism of the fault slippage. Therefore, although the following description of the immediate source region possibly may be realistic, it represents only a speculation. Following Housner (1955), we assume that a fault fracture may be described by a swarm of shear dislocations. This swarm dusters along a surface of weakness in the extended medium and releases shear stress that existed on the fault prior to the earthquake. The size of each dislocation, the frequency of the radiated seismic pulses, and

1742

BULLETIN

OF T H E

S E I S M O L O G I C A L S O C I E T Y OF A M E R I C A

the occurrence of dislocations in time and space depend on the initial stress distribution and material properties along the surface of weakness. To simplify the analysis, we imagine that one can surround the final fractured fault plane b y a closed surface only a short distance from the final fracture (Figure 1). The region inside this closed surface is then called the "immediate source region." Its shape will be such that it incorporates only the fractured fault plane. Also, we neglect any energy radiated into this region and consider only outward radiations from the fractured fault plane. Extended source region. Without loss of generality, we can consider a two-dimensional model (Figure 1) consisting of several horizontal surface layers and intersected by a vertical right-lateral strike-slip fault. Of course, this simple geometry may not be found in many real fault areas, but it may approximate the average configuration of a given strong-motion site.

FIG. 1. Idealized model of the source region. During an earthquake, a swarm of dislocations along the fault plane generates waves that are radiated from the immediate source region and are represented by several characteristic ray paths as in Figure 1. Depending on the initial direction , and assuming that the shear-wave velocities satisfy the following relation, fil < fi2 < t~3 < ~4, a typical ray may be trapped in one layer (ray with 1), two layers (ray with 2), or three layers (ray with 3). After constructive interference, such energy travels away from the source in the form of a surface wave. Another possibility is that a ray emerges from the immediate source region so that the angle ~ at the interface between layer 3 and the half-space is less than the critical angle c ; thus, a fraction of energy is radiated to the half-space. To simplify the analysis and interpretation of the strong-motion aceelerograms, we will use the mode representation for all horizontally-guided energy. Although from the ray theory point of view, P and S waves could be considered separately, we will represent them by the higher-mode surface-wave arrivals.

A METHOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1743

By successive applications of Snell's law, it can be shown that for ~1 < 82 < ~ and for the realistic values of fh, well over two-thirds of the energy generated in the shallow or surface source region may remain trapped in the surface layers to travel horizontally through the surface-wave guide. This is correct only when the immediate
FREQUENCY 5 - RADIANS I0 / SECOND 15 ' 0.1 g '

Dmm
0 I
2

.m
3

m
4 5

FIG. 2. The ~o~A(t, co) l~esponse E n v e l o p e S p e c t r u m (RES) for a single e v e n t as recorded in E1 C e n t r o ( E v e n t 9 NS) during the Imperial Valley, California, E a r t h q u a k e 1940 (Trifunae and B r u n e 1970).

source region is essentially in the surface-wave guide, which is of special interest for engineering seismology. The analysis of several typical strong-motion aecelerograms recorded on deep alluvium supports this interpretation (Trifunae, 1971). Nature of strong-motion aceelerograms. To determine time-dependent variations in the frequency content of the recorded strong ground motion, a form of the multiple filter analysis called Response Envelope Spectrum (RES) (Trifunae, 1971) can be used. This technique involves determination of the response of an ensemble of vis-

1744

BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA

cously-damped single-degree-of-freedom oscillators and mapping of the relative displacement response envelope A (t, co) as a function of oscillator frequency co and time t. The particular choice of a single-degree-of-freedom lightly-damped oscillator, which is essentially a narrow band-pass filter, was motivated by the potential use of RES in structural response studies. If A(t, co) is multiplied by co2,the result is the envelope of oscillator acceleration. In Figure 2, co2.A (t, co) is plotted for the time interval of 30 see and for frequencies ranging from 0 to 17 radians/sec. Amplitudes are arbitrarily normalized so that the peak value equals five. An accelerogram, plotted on the side in Figure 2 represents the gTound acceleration recorded in E1 Centro during the Imperial Valley, California 1940 earthquake (Trifunac and Brune, 1970). The solid (Love waves) and dashed (Rayleigh waves) lines superimposed on RES are calculated theoretical arrivals of the surfacewave modes and are based on the group-wave velocities for that site. On the average, these arrivals agree with the high amplitudes of RES. This suggests that a significant part of the strong motion may be represented by horizontally guided energy through the soft surface layers (Trifunac, 1971). The above interpretations give an adequate picture of strong motion near the source if there is no major discrepancy between the model of Figure i and the actual geological configuration. The effect of the geometrical and material discontinuities and irregularities likely to be encountered as the station is moved away from the immediate source region may require further refinement of the simple model proposed. In general, these additional features may be represented by a variety of filtering effects that will diminish some frequency components and amplify others.
A METHOD FOR SYNTHESIZING ARTIFICIAL ACCELEROGRAMS

The method. Once the statistical properties of the recorded strong-motion accelero grams become available and the earthquake source processes are known, it will be simple matter to predict possible future motions. We can expect to arrive at empirical and/or theoretical relationships between the time-dependent Fourier amplitudes displayed by the RES and the parameters that describe the source mechanism in time and space. We shall now assume that all of the relationships are known and construct artificial accelerograms for a given geological site. We assume that a local site may be approximated by a model with several layers as in Figure 1. As a first approximation, the spectrum of radiation from the immediate source region can be modeled (Housner, 1955) by assuming a random sequence of elementary dislocations. Knowing source size, depth, and orientation in space, we assume that it is possible to calculate the amplitudes of the resulting surface-wave modes for a given direction. We propose to model ground acceleration in the following way
o0

a(t; A) = Aco ~ f ~ ( t ; A) sin (co~t + e~)


where

(1)

a(t; A) = acceleration, f~(t; A) = amplitude envelope of a sine wave with frequency co~ at time t and for
= = = = distance A, initial phase e(co) at the frequency co,~, time, circular frequency (co = 2~rf) ; co~ = nAco, n = 0, 1, 2, , distance.

t
COn

A METHOD

FOR SYNTHESIZING

R E A L I S T I C STRONG

GROUND MOTION

1745

Function f~(t; A) which describes the filtering effects of the extended source region might be given b y
M

f . ( t ; A) = ~

m=l

A~,m(A)4, .... (t -- t.,m)

(2)

where A ..... (A) = m TM mode participation factor, as a function of distance, )n,~(t - tin) = amplitude envelope of the sine wave with frequency co~ belonging to the m *h mode, t~,m = time of the arrival of energy propagating with a group velocity U~,~ for the mth mode at frequency ~0n. The mode participation factor A~,~(A) might be expressed as

A~,~(~) = C~,~.~,~(~)
where C~,m = mode participation factor for A = 0, -yn,~(A) = attenuation function, such t h a t ~,~,~(0) = 1. TABLE
1

(3)

WESTMORELAND SEISMIC PROFILE* Layer Thickness (kin) Layer Properties a(km/sec) ~(km/sec) p (gr/cm3) t

1 2 3 4 5 6

0.18 0.55 0.98 1.19 2.68 oo

1.70 1.96 2.71 3.76 4.69 6.40

0.98 1.13 1.57 2.17 2.71 3.70

1.28 1.36 1.59 1.91 2.19 2.71

* After Biehler (1964). The densities in the last column, calculated using the empirical relationship p = 0.770 + 0.302 a, are adequate for the dispersion analysis, but are not realistic for actual surface layers. The initial phase function E~ and the mode participation factor C ..... might be determined from the assumed source mechanism. The amplitude envelope function 4~.m(t - A/U~,,~) is t a k e n to be of the form sin x / x where x = Ao~/2(t - A / U + de(co)/do), but for a swarm of finite dislocations in the immediate source region represents a complicated sum of such functions over the time interval during which the swarm takes place. Hence, the resulting 4~n,,~(t - A / U , , m ) could be approximated b y a bell-shaped function in the time domain. Direct body-wave phases P and S can be modeled in two different ways. One way consists of adding two bursts of energy arrivals belonging to P and S waves to the surface-wave motions. These bursts would be sealed in frequency and amplitude according to the source mechanism. The arrival and duration of each burst would be determined b y the body-wave travel times and the expected duration of the source. This approach appears to be suited for more distant recording stations when greater distance leads to better separation of various phases in time. Another approach to modeling the body-wave arrivals is through t h e higher-order modal contributions. As the extended source size generates a great n u m b e r of r a y directions, this method m a y be better for the near-field strong motion.

1746

BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA_

Sample accelerograms. Four sample aceelerograms were synthesized for the site where the Imperial Valley, California 1940 earthquake was recorded. For a number of reasons, this area appears to be ideal for modeling strong-motion accelerograms. Geological configuration is well known (Biehler, 1964) and can be approximated by a set of horizontal layers (Table 1). A number of shocks originating along the Imperial fault have been studied and interpreted in detail (Trifunae and Brune, 1970; Trifunac, 1971). Most of the relationships that lead to specification of the Amplitude Envelope Spectrum (AES) given by fn(t; A) for a given site and focal mechanism have yet to be developed. To demonstrate the method, it is assumed here that the RES given in

41 |

'

I//,,'IJ l/ // 4;
" ;kit
I ~
'

i I I /1~1 ( /I f / " II I

I / ~/" ~I
/'

'

'

f
.

//17 I]
I

/ ,,"
I

"

/
. . . . . --"

/I

o kill X D k , ~

.-I---::__-

,-[- - - - _ _ _

............

._.

2 3 PERIOD , seconds

FIG. 3. Love- and R~y]eigh-wave group-velocity dispersion curves. The curves I , 2, 3 and 4 for the periods between 1 and 5 sec were computed for the model in Table 1 (Trifun~c and Brune, 1970). For periods shorter t h a n about 1 sec, all of the curves were extended in agreement w i t h the RES in Figure 2.

Figure 2 is a typical member from the family of such spectra, and the AES for the artificial aeeelerograms in this paper are modeled after the properties of this RES. Group-velocity dispersion curves for the first four Love- and Rayleigh-wave modes were available from our previous studies (Trifunac and Brune, 1970; Trifunae, 1971) for periods approximately longer than 1 see and for the layer system given in Table 1. Higher-mode group dispersion curves become very oscillatory for waves with periods shorter than about i see and overlap in a complicated manner. It was decided to extend the existing dispersion curves (Trifunac, 1971) into the short period range so that they correspond to the arrivals of the high amplitudes in the RES in Figure 2 and represent the predominant average effects of many calculated dispersion curves. The adopted, extended group-velocity curves are given in Figure 3. The S-wave burst of energy was modeled by dispersion curves 6 and 7 (Figure 3) drawn to represent the average expeered arrivals in the model.

A METHOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1747

Once the arrivals t .... = t/U~,,~ are determined, one has to model the duration of each energy burst for a given m o d e number. This was done b y assuming envelope amplitude ep~.,~(t - t~.,~) to h a v e the form of the exponential curve
t.,~)/.o ..... }

(4)

where O-n,m is chosen to decrease for the increasing n, i.e., increasing frequency co. I n the eases presented in Figures 4, 5, 6, and 7, ~ . ~ for m = 1, 2, 3, 4, 5, 6, and 7 was

I0

15

20 - RADIANS

25

50

35

FREQUENCY

/ SECOND

4 5 0 I 2 3 FIG. 4. Synthesized accelerogram and the Amplitude Envelope S p e c t r u m (AES) for A = 5 km. Full lines indicate the arrivals of the Love and Rayleigh modes as derived from Figure 3.

t a k e n to be of the form ~,m ~ a0 exp ( -co~/2co, 2 2)

(~)

w i t h o 0 = l f o r m = 1 , 2 , 3 , 4 , 5 , a n d 6 and o0 -- 3 f o r m = 7 a n d c 0 . = 20 for a l l m . W i t h this choice of ~ . . . . actual d u r a t i o n of each burst of e n e r g y was a b o u t 2 see (for m = 1, 2, 3, 4, 5, and 6) at zero f r e q u e n c y and decreased to a b o u t 1 see for the frequency of 20 rad/sec. Similarly for m = 7, it decreased from 6 at co = 0 to a b o u t 3 see at e = 20 t a d / s e e . T h e mode participation factor A~,m(A) = C~,m7 ..... (A) was t a k e n to be C . . . . assuming for simplicity t h a t % , ~ ( A ) = 1; i.e., a t t e n u a t i o n with distance was he-

1748

BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA

glected. The functional dependence of A .... (A) on m, ~, and A is not yet known but may be determined from the analysis of the existing accelerogram records. We assumed t h a t it might be approximated by the bell-shaped function of the form

A.,m(A) ~lexp {- (m -

mo)2/2Co 2} -1-

CRXR., ) "G,,,,.

(6)

where Co was taken to be 3 for m = 1, 2, 3, 4, and 5 and 2 for m = 6 and 7, and m0 was taken to be 5 for m = 1, 2, 3, 4, and 5, 6 for m = 6 and 7 for m = 7. The second

if)

Z 0

t[5

o
0 5 I0 15 20 - RADIANS 25 / 50 SECOND 55 FREQUENCY

D
0

)
I

m
2

m
5

m
4

m
5

FIG. 5. Synthesized aceelerogram a n d the A m p l i t u d e E n v e l o p e S p e c t r u m (AES) for A = 10 kin. Full lines i n d i c a t e t h e arrivMs of the Love and Rayleigh modes as derived from Figure 3.

term C~XR~ consists of the scale factor C~, taken to be 0.2 for all m, and X ~ , the pseudo random number with values uniformly distributed between - 1 and 1. Finally G.... was approximated also by a bell-shaped normal distribution function G,~,m ~ I Bo exp { - (~o~ - cop)2/2o~8 2} q-

BR'XR. [

(7)

with all parameters given in Table 2. Again XR. was a random sequence with amplitudes uniformly distributed between - 1 and 1. The initial phases e(co) were generated by the pseudo random numbers distributed uniformly between -~r and 7r.

A METFIOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1749

With the above parameters, equation (1) was used to synthesize aecelerograms. The first 25 terms in the summation were used for o~ = nAco and with AoJ = 1.5 rad/sec. The only parameter that varied for the test cases in Figures 4, 5, 6, and 7 was the distance A taken to be 5, 10, 15, and 20 kin. The synthesized accelerograms in Figures 4 to 7 are presented together with the Amplitude Envelope Spectrum (AES) given by f,(t; A). The lines superimposed on the AES in these figures correspond to the surface-wave arrivals determined from the group dispersion curves in Figure 3. From the definition for AES and RES, it is clear

,,c

121 Z 0 W

I0

15

20

25

50

55

F R E Q U E N C Y - R A D I A N S / SECOND

D
0 I

)
2 5

u
4

m
5

FIG. 6. Synthesized accelerogram and t h e A m p l i t u d e E n v e l o p e S p e c t r u m (AES) for A = 15 kin. Full lines indicate the arrivals of the Love a n d Rayleigh modes as derived from Figure 3.

that these two are in many respects very similar. If the amplitudes in AES change slowly for the frequency in question, RES will be a good approximation of AES. Despite great simplification in choosing parameters which describe AES for the artificial accelerogram in Figure 7, the accelerograms in Figures 2 and 7 have many similar properties. The most distinct common feature is represented by the arrivals of various surface modes. The differences between the accelerograms in Figures 2 and 7 may be representative of the differences to be expected among the ensemble of predicted accelerograms for variations in the input parameters. Some later phases observed on the recorded accelerogram in Figure 2 that occurred after 23 sec were not modeled in the four artificial aceelerograms (Figures 4 to 7).

1750

BULLETIN OF THE SEISiV[OLOGICAL SOCIETY OF AMERICA

These later phases cannot be explained by the simple surface-wave arrivals. They might be associated with laterally reflected waves which traveled along indirect paths and then arrived later at the recording station. Another possible explanation is that these phases represent slightly attenuated leaking modes because of the high-velocity gradients in the Imperial Valley (Table 1). Scaling of the artificial accelerograms. All of the AES in Figures 4 to 7 were normalized such that the peak value equals 5. The synthesized accelerograms were also independently normalized to have the same peak amplitudes. These normalizations

IO

15

20

25

30

55

FREQUENCY-RADIANS / SECOND

[]
0

m
I

[]
2

[]
~

[]
4

m
5

FIG. 7. S y n t h e s i z e d acce]erogram and t h e A m p l i t u d e E n v e l o p e S p e c t r u m (AES) for A = 20 km. Full lines indicate the arrivals of the Love and Rayleigh modes as derived from Figure 3.

were performed to improve the display of ground filtering effects through the use of dispersion curves. The desired amplitudes for the synthesized accelerograms can be derived from several methods. We will mention here only the simplest scaling method based On the local Richter magnitude (Richter, 1958). This magnitude scale is defined by ML = logA - - l o g A 0 (8)

where A is the peak amplitude in millimeters of the trace recorded by a standard Wood-Anderson torsion seismometer ( V = 2800, T = 0.8 sec, ~ = 0.8) at the distance of 100 km from the epicenter. A0 is the amplitude with which the same instrument would record an earthquake of magnitude zero at this distance.

& METHOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1751

To scale an artificial accelerogram for the ground motion that would correspond to an earthquake of magnitude ML, it suffices to calculate the response of a single-degreeof-freedom system with natural period T = 0.8 sec and the fraction of critical damping = 0.8 for the input acceleration 2800 a(t). When this response is determined, the peak amplitude is read in millimeters and M~ is obtained. If the desired local magnirude is ML, then the accelerogram must be multiplied by V0 = 10 ~ - ~ L (9)

Multiple events. A number of investigators have suggested (e.g., Florensov and Solonenko, 1963; Haskell, 1964, 1966; Aki, 1967; Wyss and Brune, 1967; Trifunac and Brune, 1970) that the source behavior in time may be quite complex in detail. In their investigation of the mechanism of the Imperial Valley, California 1940 earthquake, Trifunac and Brune (1970) showed that as recorded at great distances, a propagating rupture actually consisted of a sequence of distinct events originating near the instrumentally determined epicenter and propagating to the southeast. The degree of the source complexity and how it should be determined is still an open question. For faulting longer than several tens of kilometers, however, it might be expected that the
TABLE 2
PARAMETERS BJ 1.5 2.0 3.0
Bo , BR , wp ,

2~ND COB USED IN EQUATION (7) FOR I)IFFERENT m COp(rad/sec) 10.0 25.0 30.0 WB(rad/sec) 5.0 15.0 10.0 m 1, 2, 3 4, 5 6

BR 0.10 0.10 0.30

1.5

0.25

30.0

5.0

close-field strong motion might be represented by several simple events of varying magnitude and with progressively displaced hypocenters along the fault plane. Artificial accelerograms might then be constructed by successively adding simple events similar to those in Figures 4 to 7 but of different magnitude and displaced in time according to their progressive occurrence along the fault.
CONCLUSIONS

First interpretations of strong earthquake ground motion were carried out from the point of view of replacing the source properties and propagational effects by stochastic processes. Such processes were used to study failure of hysteretic-type structures and, in general, for applied structural analysis. The same approach might be used to describe the response of the immediate source region, but the filtering effects of the medium along the wave-propagation path cannot be neglected. It is proposed to continue to simulate seismic radiation from the immediate source region using random processes, not because this is conclusively indicated by the physical nature of the energy release, but because a detailed source mechanism is not yet understood. Analysis of several typical strong-motion accelerograms recorded on deep alluvium has indicated that most of the waves radiated from the shallow immediate-source region propagate in the form of surface waves guided through the shallow low-velocity surface layers. Understanding the nature of such near-field strong ground motion is important because the distribution of arrivals of the main bursts of energy in time

1752

BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA

determines the nature of structural response. The arrivals of these bursts of energy are governed b y the surface-wave group velocity. A simple method developed for generating artificial accelerograms can be used for predicting strong-motion effects produced b y earthquakes or underground nuclear explosions. The synthetic accelerograms incorporate the properties of wave propagation through the local geology obtained from theoretical group-dispersion data for a given site. The input data to the layer model of the local site are based on knowledge of the temporal and spatial properties of the energy release mechanism. The functional dependence of these input data on variables such as local earthquake magnitude, length of faulting, and multiplicity of the source has yet to be developed. The new method of synthesizing artificial accelerograms realistically models strong ground motion because it is based on the properties determined from the actually recorded accelerograph records. I t can be used to generate artificial accelerograms for designing nuclear power plants, dams, bridges a n d other i m p o r t a n t structures t h a t will resist severe ground motion. This method does not exclude the simplified models based on the vertically-propagating waves often used to model local site conditions but complements them. ACKNOWLEDGMENTS
I am indebted to T. Matumoto, L. Sykes, X. Jacob, and J. Healy of the Lamont-Doherty Geological Observatory for critically reading the manuscript and offering many valuable suggestions. Mr. F. E. Udwadia, graduate student at the California Institute of Technology, participated in preliminary computer programming for the accelerogram synthesis. His help is greatly appreciated.

This research was supported in part by the National Science Foundation Grant GA 22 709.
REFERENCES

Aki, K. (1967). Scaling law of seismic spectrum, J. Geophys. Res. 72, 1217-1232. Amin, M. and A. H. Ang (1966). A Nonstationary Stochastic Model for Strong-Motion Earthquakes, Structural Research Series No. 806, Department of Civil Engineering, University of Illinois. Amin, M., It. S. Ts'Ao, and A. H. S. Ang (1969). Significance of nonstationarity of earthquake motions, Proc. World Conf. Earthquake Eng., 4th, Santiago, Chile. Barstein, M. F. (1960). Application of probability methods for designing the effects of seismic forces on engineering structures, Proc. World Conf. Earthquake Engineering, 2nd, Japan, pp. 1467-1482. Beaudet, P. R. (1970). Synthesis of Nonstationary Seismic Signals, Bull. Seism. Soc. Am. 60, 1615-1624. Biehler, S. (1964). Geophysical Study of Salton Trough of Southern California, Ph.D. Thesis, California Institute of Technology. Bogdanoff, J. L., J. E. Goldberg, and M. C. Bernard (1961). Response of a simple structure to a random earthquake-like disturbance, Bull. Seism. Soc. Am. 51,293-310 Bolotin, V. V. (1960). Statistical Theory of Aseismie Design of Structures, Proc. World Conference on Earthquake Engineering, 2nd, Japan, pp. 1365-1374. Bycroft, G. N. (1960). White-noise representation of earthquakes, J. Eng. Mech. Dis., Am. Soc. Civil Engrs. 86, 1-16. Cornell, C. A. (1964). Stochastic Process Models in Structural Engineering, Technical Report No. 84, Department of Civil Engineering, Stanford University. Florensov, N. A. and V. P. Solonenko (1963). The Gobi-Altai Earthquake, Isdatel'stro Akademii Nauk SSR, Moscow. Goodman, L. E., E. Rosenblueth, and N. M. Newmark (1955). Aseismic design of firmly founded elastic structures. Transactions Am. Soc. Civil Engrs. 120,782-802. Goto, I-I., K. Toki, and T. Aiyoshi (1966). Generation of artificial earthquakes on digital computer for aseismic design of structures, Proc. Japan Earthquake Eng. Syrup., pp. 25-30.

A METHOD FOR SYNTHESIZING REALISTIC STRONG GROUND MOTION

1753

Goto, H. and H. Kameda (1969). Statistical Inference of Future Earthquake Ground Motion, Proc. World Conf. Earthquake Eng., 4th, Santiago, Chile. Goto, H. and K. Toki (1969). Structural response to nonstationary random excitation, Proc. World Conf. Earthquake Eng., 4th, Santiago, Chile. Haskell, N. A. (1964). Total energy and energy spectral density of elastic wave radiation from propagating faults, Bull. Seism. Soc. Am. 54, 1811-1841. Haskell, N. A. (1966). Part II, A statistical source model, Bull. Seism. Soc. Am. 55,125-140. Honda, It. (1957). The Mechanism of Earthquakes, Geophysics Institute, Tohoku University, 9, 5, 1-46. Housner, G. W. (1947). Characteristics of strong-motion earthquakes, Bull. Seism. Soe. Am. 37, 19-31. Housner, G. W. (1955). Properties of strong ground motion earthquakes, Bull. Seism. Soc. Am. 45,187-218. Housner, G. W. and P. C. Jennings (1964). Generation of artificial earthquakes, J. Eng. Mech. Div., Am. Soc. Civil Engrs. 90, pp. 113-150. Housner, G. W. and M. D. Trifunae (1967). Analysis of accelerograms-Parkfield earthquake, Bull. Sei.sm. Soc. Am., 57, 1193-1220. Hudson, D. E. (1956). Response spectrum techniques in engineering seismology, World Conference on Earthquake Engineering, Berkeley, California. Jennings, P. C., G. W. Housner, and N. C. Tsai (1968). Simulated Earthquake Motions, Earthquake Engineering Research Laboratory, California Institute of Technology, Pasadena, California. Penzien, J. and S. C. Liu (1969). Nondeterministic analysis of nonlinear structures subjected to earthquake excitations, Proc. World Conf. Earthquake Eng., 4th, Santiago, Chile. Rascon, O. A. and A. C. Cornell (1969). A physically based model to simulate strong earthquake records on firm ground, Proe. World Conf. Earthquake Eng., 4th, Santiago, Chile. Richter, C. F. (1958). Elementary Seismology, W. H. Freeman and Co., San Francisco. Rosenblueth, E. (1956). Some applications of probability theory on aseismic design, World Conference on Earthquake Engineering, Berkeley, California. Rosenblueth, E. and J. E. Bustamante (1962) Distribution of structural response to earthquakes, J. Eng. Mech. Div., Am. Soc. Civil Engrs. 88, 75-106. Shinozuka, M. and Y. Sato (1967). Simulation of nonstationary random processes, J. Eng. Mech. Div., Am. Soe. Civil Engrs. 93, 11-40. Tajimi, H. (1960). A statistical method of determining the maximum response of building structure during an earthquake, Proc. World Conf. Earthquake Eng., 2nd, Japan. Trifunac, M. D. and J. N. Brune (1970). Complexity of energy release during the Imperial Valley California, earthquake of 1940, Bull. Seism. Soe. Am. 60,137-160. Trifunac, M. D. (1971). Response envelope spectrum and interpretation of strong earthquake ground motion, Bull. Seism. Soc. Am. 61,343-356. Wyss, M. and J. N. Brune (1967). The Alaska earthquake of March 28, 1964. A complex multiple structure, Bull. Seism. Soc. Am. 57, 1017-1023.
L,C.MONT-DOHERTY GEOLOGICAL OBSERVATORY OF COLUMBIA UNIVERSITY PALISADES, N E W YORK 10964 CONTRIBUTION NO. 1723

Manuscript received May 3, 1971

You might also like