You are on page 1of 23

Bulletin of the Seismological Society of America, Vol. 82, No. 2, pp.

580-602, April 1992

THE RELATION B E T W E E N SITE AMPLIFICATION FACTOR AND SURFICIAL GEOLOGY IN CENTRAL CALIFORNIA BY F. Su., K. AKI, T. TENG, Y. ZENG, S. KOYANAGI, AND K. MAYEDA ABSTRACT
Site amplification factors at frequencies of 1.5, 3, 6, and 12 Hz were determined for 132 stations of the USGS seismic network in central California from coda waves of 185 local earthquakes in this area using a recursive stochastic inversion method. We found that the site amplification at a station is systematically related to the geology underlying that station. The site amplification is high for young, Quaternary sediments and decreases with increasing geologic age at all frequencies between 1.5 and 12 Hz. The rate of decrease varies with frequency where site amplification at low frequencies shows a faster rate of decrease with age than at higher frequencies. To quantify the relation between site amplification factor and site geology, the surface geology of station sites was classified into five groups, namely, (1) Quaternary sediments, (2) Pliocene sediments, (3) Miocene through Cretaceous sediments, (4) Franciscan formation and Mesozoic granitic rocks, and (5) Pre-Cretaceous metamorphic rocks. The station site amplification factors for each group were logarithmically averaged and the mean value was assigned to the median geologic age of that group. A smooth power law relation was found between the mean site amplification and the median geologic age. This relation provides a simple way of estimating site effect at a specific site with known surficial geology. Our site amplification factors were compared with magnitude station residuals determined by Eaton (1990) for the same network. A remarkable linear correlation was found between the logarithmic coda amplification factor and the magnitude station residual value, confirming that the coda method provides an effective means of estimating site amplification factor. Comparison of our site amplification factor with strong-motion results obtained from the Loma Prieta earthquake (Boore et al., 1989; Maley et al., 1989; Shakal et al., 1989; Chin and Aki, 1991) suggests that weak and strong-motion site amplification factors correlate well in the region outside the epicentral source region beyond epicentral distances of about 50 km. Within the epicentral source region, however, the weak motion amplification factors estimated from coda waves do not agree with the observed for strong ground motion, suggesting a nonlinear behavior at sediment sites. These results together with studies on the correlation between weak-motion site amplification and earthquake intensity (Borcherdt, 1970; Borcherdt and Gibbs, 1976; Chavez-Garcia et ah, 1990; King et ah, 1990) demonstrate the importance of in situ determination of weak-motion site amplification for seismic zonation.
INTRODUCTION

The importance of recording site effect on seismic ground motion has been well recognized by seismologists and earthquake engineers. Traditionally, direct waves were used for the site effect study (Gutenberg, 1957; Borcherdt, 1970; Rogers et al., 1979, 1984). By comparing the results from direct S waves and coda waves, Tsujiura (1978) and Tucker and King (1984) concluded that using coda waves provides more stable estimates of site amplification. Since coda waves are considered to be composed of back-scattered waves coming from many heterogeneities surrounding the source and receiver, the result is an 580

SITE AMPLIFICATION

FACTOR AND SURFICIAL GEOLOGY

581

average over all directions. Furthermore, the separation of source, path, and site effects is drastically simpler for coda waves than for S waves. Another important issue regarding the study of site effect is the local geology. It has long been known that surface geology exerts significant influence on the site amplification (Borcherdt, 1970; Borcherdt and Gibbs, 1976; Rogers et al., 1979). However, for a quantitative evaluation of this influence, we need to accumulate more observations using data recorded at sites with various types of geology. In this paper, we used a recursive stochastic inverse method to determine the site amplification factors from coda waves using stations in central California located at a variety of surficial geology. The USGS central California seismic network operates hundreds of stations in this area, providing an excellent data set for this study. We found that the site amplification of a station is systematically related to the geology underlying that station. The site amplification is high for young, Quaternary sediments and decreases with increasing geologic age. The rate of decrease varies with frequency. Site amplification at low frequencies show a faster rate of decrease with age than at higher frequencies. Our site amplification factors were compared with Eaton's (1990) magnitude station residuals. A remarkable linear correlation was found between the logarithmic amplification factors from coda waves and the magnitude residual values, confirming that our method provides an effective means of site amplification estimation. The comparison of our weak-motion site amplification results with that of strong-motion results is discussed in a later section.
DATA AND ANALYSIS

Data

Records of 185 earthquakes obtained by the USGS central California seismic network during the period from 1984 to 1990 were collected for this study. The magnitude of these earthquakes ranged from 1.8 to 3.5 and their focal depths are less than 20 km. Figure l a shows the distribution of the earthquakes (open circles). All waveforms of these events were digitized at 100 samples per second. The solid stars in Figure l a are the stations used to calculate site amplification factors. These stations are located on a wide variety of geologic settings ranging from alluvium to Mesozoic rocks. A simplified geology map for central California is shown in Figure lb. All the stations used were standard USGS vertical instruments with a natural frequency of 1 Hz. Using calibration information provided by Eaton (1980), all recorded seismograms were corrected for instrument response. Since instrument gain settings changed frequently over the time period spanning our collected events, special care was taken to ensure that proper gain corrections were transcribed from station history files (J. Eaton, personal comm., 1990). The shape of the instrument response curve between 1 and 20 Hz is the same for different stations, while it becomes complicated beyond 20 Hz, depending on the type of discriminator used at the station. To avoid any possible mistake in the instrument correction, our oqtave bands were limited to center frequencies between 1.5 to 12 Hz.
Method

Since the pioneering work of Aki (1969), the fundamental separability of source, site, and path effects in the coda wave power spectrum has been confirmed by many researchers (Aki and Chouet, 1975; Phillips and Aki, 1986;

582

F. SU E T AL.

Su et al., 1991). The power spectrum P(co I t) of coda waves can be expressed as P(co[ t) = source(co) -site(co). path(col t),

(1)

where co is the circular frequency and t is the lapse time measured from the event origin time. The above formula is valid, as a rule of thumb, for t greater than twice the travel time of shear waves from the source to the station (Rautian and Khalturin, 1978). Taking natural log of both sides of (1), we have l l n Pij(~z, tk) = dijkl = ri(wl) + sj(~l) + c(~l, tk),

(2)

where ri(cot) is the site term, sj(col) is the source term, and c(coz, tk) is the coda propagation term. Indices i, j, k, and l represent the station, source, lapse time, and frequency, respectively. A factor of one half is introduced so that later results can be expressed in terms of relative amplitude. Since the coda energy is assumed to be the sum of back-scattered wave energy from heterogeneities distributed more or less uniformly throughout the lithosphere, the path term c(coz, th) depends only on the regional average properties of the medium and is independent of any particular source-receiver location. Taking an average of dijkl in equation (2) over all available station i in the same region with fixed indices j, k, and l and then subtracting it from the original d~jkl values, we arrive at the following equation

d , j k t - dJ~l : r i - r~,

(3)

where Fi is the average of r i over the available stations for fixed indices j, k, and l. We can further write r i = ~mSimrm and ri = (1/Njkl)Z~Imrm, where 5ira = 1 if i = m and 0 otherwise, Njk z is the total number of usable stations for fixed indices j, k, and l and I m = 1 if station m is used, or I m = 0 otherwise. Substituting these expressions into (3), we obtain

dijk

= E ( im -m

mlgjk )r .

(4)

Writing (4) into matrix form, we get G r = z~d, (5)

where h d is the data vector that can be obtained from the coda power spectrum and r is the model vector, which contains the relative site amplification factors to be determined. G is a matrix of real elements. In practice, this matrix could be huge and very sparse. A singular value decomposition and generalized inverse technique used by Phillips and Aki (1986) is too expensive and inefficient when a large number of model parameters and observation data points are involved. In this work, we apply a fast recursive stochastic inversion method (Zeng, 1990) to obtain solutions of our inverse problem. By taking the noise into consideration, we can write equation (5) as
Gr + n = (6)

SITE A M P L I F I C A T I O N F A C T O R A N D S U R F I C I A L G E O L O G Y

583

38
0

.o

~
37

o~ ~.
.

\*

Station

Earthquake

20

KM

122

la)

121 o

120

FIG. 1. (a) Map of the central California showing the distribution of the e a r t h q u a k e s (open circles) and seismic stations (solid stars) used in this study as well as fault traces (solid lines). (b) A simplified geology map for central California. The names of the faults and places are referred in the text.

584

F. SU E T A L .

58

37

56

[~

Quoter Sedim,

Plioce,
Sedim

'~ Miocew
Sedim
Granit

'~ Fronci~

~
35
i

Pre- C Meton"

122 (b}

121 FIG. 1. (Continued),

120

SITE A M P L I F I C A T I O N FACTOR AND SURFICIAL GEOLOGY

585

where n is a noise vector. The I t h component of this equation can be written as G I r + n I = Adi, I = 1 , 2 , . . . , N.

(7)

where G1 is the I t h row of matrix G, and n I and Adl are the I t h element of vectors n and Ad, respectively. By applying the recursive stochastic inverse formula (Zeng, 1990) to our case, we get [2i+ 1 = 21 + KI+I(AdI+I - GI+12t)
T ~K1+1 : R I G x + I / ( G z + I R I G xT + I + 02 1+1)

(s)

|( R z + 1 = R z + K z + t G I + I R I where 21 is the stochastic inverse solution of r given A d l , . . . , Adl, R I is its corresponding error covariance matrix and z2 is the noise variance for n I. Superscript " T " means matrix transpose. We define the initial estimate 2 o = 0, and give R o a priori. Then we compute the updated r according to the recursive procedure given by (8). From (8), we can see t h a t this recursive inverse process is a sequence of scalar inverses t h a t require only matrix and vector additions and multiplications. For the inversion of a large sparse matrix, we just compute the term RIGT+I by tracing only the nonzero elements in G1+1, which can avoid a large number of unnecessary calculations and greatly reduce the computation time. It is particularly effective for inverse problems with a large number of model parameters and observations, such as our site amplification inversion. In addition, this recursive scheme provides some insight to the inversion process. From the recursive process given in (8), we see t h a t our process can stop at any step I and the result is stored. These intermediate results give us a good perspective of how each observation improves our inverse solution. Because results of our inversion can be incrementally stored, we can always update our solution with new data points without computation over the whole data set. Another advantage of the inversion method, as compared to the spectral ratio method (Mayeda et al., 1991), is t h a t it can be applied to a seismic network over a large region, which enables us to compare the site amplification factors over the large region with a common reference. The details of coda power spectrum calculation, noise elimination and data processing technique can be found in Su and Aki (1990).
RESULTS

Coda Site Amplification in Central California


The final inversion results of site amplification factors are listed in Table 1. The standard errors, in n a t u r a l log, are about 0.065 for frequency 1.5 Hz, 0.056 for 3 Hz, 0.051 for 6 Hz, and 0.050 for 12 Hz. To show the spatial distribution of site amplification in central California we plot these site amplification factors on maps (Fig. 2a to d) for the frequency bands centered at 1.5, 3.0, 6.0, and 12.0 Hz, respectively. In these figures, we used six symbols to indicate six different site amplification ranges. For 1.5 Hz, we can see t h a t the highest site amplifications are in the vicinity of Watsonville and fault zone near Hollister. The surface geology of these areas consists mainly of Quaternary sediments. The

586

F. S U E T A L .

TABLE 1
STATION LOCATION, SITE SURFICIAL GEOLOGY, AND SITE AMPLIFICATION FACTOR DETERMINED IN THIS STUDY Site AmplificationLn ( A / A) Sta~on Location 1.5 Hz 3 Hz 6 Hz 12 Hz Geological Symbol Qs

JPRV PPTV BHRV CDUV PSAV HCOV JPLV HPHV JLTV MYLV BEHV HFHV JSGV JSJV NHMV HKRV BSLV HORV NOLV NCFV CMCV CDOV CRAV BBNV JTGV JBZV JPSV JRGV HPRV PBWV PLOV JBGV CBRV CBWV CMJV CACV HQRV HSFV JSMV BPIV BRVV HCBV CSPV PANV BVLV PJLV CMHV JECV JSFV JSCV

37 36 36 38 36 36 36 36 37 37 36 36 37 37 38 36 36 36 38 38 37 37 37 36 37 37 37 37 36 36 36 37 37 37 37 37 36 36 37 36 36 36 37 35 36 36 37 37 37 37

47.70 6.50 43.67 1.78 1.52 53.31 58.62 51.38 21.22 23.02 39.88 53.29 16.96 20.03 9.28 54.10 46.53 55.03 2.50 19.28 46.88 43.80 46.03 30.60 1.71 1.07 11.94 2.22 57.19 18.90 14.79 20.52 48.97 55.45 31.25 58.57 50.02 48.72 12.74 29.40 25.49 55.88 57.45 46.78 34.51 5.39 21.57 3.04 24.31 17.07

122 120 121 122 120 121 121 121 122 120 121 121 122 122 121 121 121 121 122 122 122 121 121 121 121 121 122 121 121 120 121 122 122 122 121 121 121 121 122 121 121 121 122 120 121 121 121 121 122 122

28.43 43.27 15.83 0.05 53.30 42.34 49.93 24.37 12.25 25.16 10.45 28.13 3.00 5.48 48.02 25.56 20.96 30.46 47.64 47.73 10.55 50.12 56.25 4.53 52.58 49.15 20.90 57.87 41.70 55.75 2.55 20.34 3.72 6.40 52.23 45.62 12.76 29.97 10.06 10.11 1.10 39.63 18.65 54.44 11.34 9.33 45.38 48.56 10.55 7,42

-0.05 0.52 1.18 1.95 1.75 1.32 1.59 1.98 0.45 0.17 0.82 1.70 0.45 0.65 1.01 1.77 1.69 1.64 0.65 0.05 0.10 0.77 0.64 0.34 0.91 0.99 0.45 0.57 1.00 0.41 0.56 0.94 -0.09 0.35 -0.54 0.94 -0.45 0.78 0.47 -0.79 0.32 0.42 0.09 0.99 0.23 0.10 -0.28 -0.11 0.65 -0.53

0.18 0.98 0.57 2.04 1.36 1.07 1.35 1.78 0.91 0.31 0.55 0.95 0.24 0.60 0.68 1.02 1.24 0.89 0.79 0.26 -0.37 0.76 0.35 0.21 0.75 0.98 0.40 1.03 0.53 0.25 1.22 0.63 -0.66 -0.15 -0.56 0.43 -0.38 0.70 0.26 -0.60 0.23 0.42 -0.07 0.65 0.15 0.17 -0.26 0.11 0.37 -0.21

0.44 1.01 0.46 1.77 0.97 0.76 1.04 1.56 1.15 0.62 0.47 0.76 0.28 0.57 0.11 1.11 0.99 0.53 0.80 0.69 0.00 0.44 -0.03 0.14 0.58 1.45 0.19 0.94 0.49 0.38 1.01 0.53 -0.78 -0.36 -0.28 0.10 -0.52 0.32 0.42 -0.44 0.03 0.35 -0.04 0.40 -0.20 0.09 -0.38 -0.04 0.02 -0.06

0.50 0.55 0.65 1.51 1.31 0.70 0.83 1.37 0.80 0.50 0.34 0.71 0.18 0.73 0.32 0.94 0.87 0.66 0.71 0.96 0.51 0.38 -0.33 -0.08 0.95 1.25 -0.13 1.04 0.41 0.80 0.85 0.41 -0.80 -0.13 -0.75 0.90 -0.90 0.03 -0.02 -0.47 0.42 -0.18 -0.44 0,69 -0.35 0.07 -0.35 -0.46 0.11 -0.21

Qal, Pml Qf Q Qt, Pml

Qc Qc
Qp Qp Qp, Pmlc Qp Qp Qp Qp

Qpc
Qp Pc Pc P P Pvr Pmlc Pmlc Pmlc Pml Pml Pml Pml Pml Pml Pml Pml Mu Mu Mu Mu Mv Mv Mvb Mvr Mmc Mm, Pml Mm, Pml, Qal Mm Mm Mm Mm, Mu Mm, Pml Mm M1, Kjf

SITE A M P L I F I C A T I O N F A C T O R AND S U R F I C I A L G E O L O G Y TABLE 1--Continued


Site Amplification Ln (A / A) Station Location 1.5 Hz 3 Hz 6 Hz 12 Hz

587

Geological Symbol

JPPV HBTV JBCV JHPV NGVV NTYV NSPV CPLV HCAV HSPV CMOV BMSV NAPV BBGV CMPV HCPV NLHV HSLV HFEV CLCV HJSV CBSV HLTV JHLV BRMV JALV CMRV HGSV JSSV JBMV HGWV COSV CAOV ADWV CSCV JLXV NLNV BPFV NTAV HPLV BEMV PAPV CMLV PHRV CMMV JRRV BAVV CALV CAIV JMGV CCYV HCRV JSAV JEGV MHDV

37 15.81 122 12,78 36 51.01 121 33.04 37 9.62 122 1.57 37 26.65 122 18.09 38 16.84 122 12.89 38 23.37 122 39.70 38 10.96 122 27.20 37 38.25 121 57.64 37 1.52 121 29.02 37 6.91 121 30.94 37 48.68 121 48.15 36 39.78 120 47.51 38 26.34 122 14.99 36 34.70 121 2.31 37 21.45 121 18.51 37 11.67 121 11.08 38 7.19 122 8.87 37 1.16 121 5.13 36 59.00 121 24.09 37 44.28 122 3.83 36 48,99 121 17.92 37 49.06 121 38.43 36 53.07 121 18.49 37 6.54 121 49.99 36 50.70 120 49.40 37 9.50 121 50.82 37 35.68 121 38.22 37 5.75 121 26.83 37 10.17 121 55.84 37 19.09 122 9.16 37 1.02 121 39.20 37 30.51 121 22.44 37 20.96 121 31.96 38 26.35 120 50.89 37 17.11 121 46.35 37 12.11 121 59.17 38 9.15 122 42.75 36 13.82 121 46.32 37 55.43 122 35.70 37 3.13 121 17.40 36 39.68 121 5.76 35 54.77 121 21.70 37 28.64 121 39.09 36 22.38 120 49.10 37 27.34 121 29.62 37 3.27 121 43.61 36 38.75 121 1.79 37 27.07 121 47.95 37 51.68 122 25.77 37 38.22 122 28.43 37 33.10 122 5.45 36 57.46 121 35.01 37 34.95 122 25.03 37 30.84 122 27.74 37 7.36 119 53.60

0.29 -0.05 0.40 0.25 -0.09 0.86 -0.02 -0.44 -0.07 -0.10 0.33 -0.26 0.75 0.88 -0.69 -0.47 0.21 -0.08 -0.02 0.25 0.07 0.44 -0.34 -0.87 0.01 -0.83 -0.60 -0.78 -0.64 -0.52 -0.76 -0.87 -0.74 -0.69 0.34 -0.63 -0.48 -0.51 -0.63 -0.60 -0.56 -1.05 -0.90 -0.05 -0.75 -0.97 -0.67 -0.71 -0.67 - 0.46 -0.98 -0.65 -0.30 -0.52 -0.20

0.34 -0.22 0.44 0.43 -0.14 0.15 -0.10 -0.63 -0.34 0.08 0.15 0.29 -0.16 0.82 -0.61 0.27 0.46 0.49 -0.25 0.47 -0.32 0.87 -0.62 -0.82 0.56 -0.53 -0.73 -0.78 -0.41 -0.52 -0.78 -0.69 -0.63 -0.38 0.18 -0.39 -0.20 -0.20 -0.70 -0.46 0.31 -0.55 -0.90 -0.29 -0.87 -0.86 -0.86 -0.65 -0.52 - 0.42 -1.04 -0.92 -0.23 -0.21 0.06

-0.01 -0.58 0.38 0.12 -0.15 -0.09 -0.05 -0.55 -0.48 -0.20 0.08 0.15 -0.52 0.59 -0.74 0.50 0.39 0.39 -0.33 0.36 -0.50 0.77 -0.71 -0.40 0.72 -0.05 -0.79 -0.79 -0.61 -0.51 -0.82 -0.66 -0.76 -0.14 -0.02 -0.15 0.05 -0.12 -0.61 -0.32 -0.13 -0.41 -0.81 -0.51 -1.07 -0.17 -0.88 -0.61 0.17 - 0.28 -0.76 -0.70 -0.06 0.10 0.42

0.15 -0.32 0.07 0.08 -0.26 -0.35 -0.26 -0.40 -0.42 -0.46 -0.22 -0.36 -0.50 0.19 -0.75 -0.09 -0.13 0.09 -0.42 0.30 -0.45 0.59 -1.02 0.05 0.25 -0.05 -0.81 -0.85 -0.47 -0.47 -0.91 -0.55 -0.96 -0.69 -0.14 -0.02 0.39 0.04 0.10 -0.59 -0~27 -0.53 -0.26 -0,54 -0.92 0.03 -1.19 -0.77 0.12 0.44 -0.27 -0.22 0.26 0.37 0.29

M1, E &, gr, Qt E E Tv Tv Tv K K K K Ku Ku, Tv, U m Ku Ku Ku Ku Ku Ku Ku Ku Ku Ku Ku Ku

Kjf gjf Kjf gjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf Kjf
Kjf

Kjf Kjf
Kjf

Kjf Kjf Kjfv Kjfv Kjfv Kjfv Kjfv


gr gr

588

F. S U E T AL. TABLE 1--Continued Site AmplificationLn (A/A)

Station BPCV BVYV HMOV HAZV BAPV BJCV BPPV BPRV BJOV HDLV BSCV BSGV JBLV BCGV HJGV BSRV CRPV CADV JCBV JSTV CSHV CCOV BHSV BCWV BSMV JUCV HFPV 36 36 36 36 36 36 36 36 36 36 36 36 37 36 36 36 37 37 37 37 37 37 36 36 36 37 36

Location 34.32 44.96 36.03 53.08 10.55 32.82 10.12 24.42 36.65 50.12 38.50 24.83 7.69 42.55 47.88 39.99 54.75 9.83 6.71 12.41 38.88 15.46 21.35 18.40 23.03 0.07 45.22 121 121 121 121 121 121 121 121 121 121 121 121 122 121 121 121 121 121 121 121 122 121 121 121 121 122 121 37.56 24.80 55.06 35.45 38.56 23.53 22.68 43.77 18.81 38.64 15.59 15.22 10.08 20.60 34.43 31.12 54.33 37.55 41.33 47.84 2.57 40.35 32.41 33.96 25.63 2.91 29.43

1.5 Hz 0.30 -0.66 -0.86 -0.26 -0.79 -1.56 -1.22 -0.72 -0.96 -0.68 -0.11 -1.05 -0.88 0.07 -0.52 -0.92 0.20 -0.21 -0.49 -0.26 -0.54 -0.08 -0.68 -1.01 -0.68 -0.43 -1.14

3 Hz 0.16 -0.80 -0.89 -0.71 -0.08 -1.43 -0.61 -0.38 -0.73 -0.55 -0.35 -0.80 -0.89 -0.06 -0.74 -0.58 0.79 -0.79 0.17 -0.40 -0.54 -0.25 0.28 -0.66 -0.24 -0.53 -1.22

6 Hz 0.16 -0.73 -0.62 -1.05 0.36 -1.15 -0.49 -0.44 0.44 -0.24 -0.63 -0.18 -0.72 0.13 -0.66 0.20 0.41 -0.35 0.05 -0.38 -0.44 -0.58 0.66 -0.11 -0.06 -0.32 -1.24

12 Hz 0.19 -0.44 -0.07 -0.87 0.12 -0.69 -0.76 -0.25 0.62 -0.20 -0.60 0.63 -0.36 0.07 -0.42 0.63 0.07 0.03 -0.28 -0.13 0.30 -0.72 0.58 -0.31 -0.31 -0.05 -1.14

Geological Symbol gr gr gr gr gr, m gr gr gr, m gr gr gr gr gr gr gr gr ub, ku, gr ub ub ub, Kjf ub, Jk Jk m, Mm, gr m m ms 1S

Diablo range of the Franciscan formation has low amplification at all frequency bands studied (Fig. 2a to d). The Gabilan range and the southwest segments of the Nacimiento faults have the lowest site amplification at 1.5 Hz, but their site amplification gradually increases with increasing frequency. These areas are mostly composed of competent Mesozoic granitic rocks and Pre-Cretaceous metamorphic rocks. The site amplification along the San Andreas fault is less systematic since the geologic conditions along the fault zone changes dramatically. In general, we found that the site amplification of a station is controlled by its underlying surface geology.

The Relation of Coda Site Amplification Factors with Surficial Geology


To investigate how the site geology affects the station site amplification, we have obtained the surface geology underling each station from CDMG 1:250,000 scale geologic maps. The surficial geology at each station is listed in Table 1. From Table 1, we can see that most stations fall in six categories of surficial geology: Quaternary sediments, Pliocene sediments, Miocene sediments, Cretaceous Marine sediments, Franciscan formation, and Mesozoic granitic rocks. Figure 3 shows a plot of site amplification as a function of frequency for each of the six geologic categories. Each line in the figure represents site amplification at one station and the average is shown by the dotted line. From Figure 3, we can see that the site amplification decreases with increasing geologic age and different geologic settings show different frequency dependencies. The Quater-

SITE AMPLIFICATION FACTOR AND SURFICIAL GEOLOGY

589

[]

.38

0 0 0 0 0 0 []

~
37

\_

*$k' ~ D

./....

\'

t;o

[]

:%.oo

`35 - -

~, - 2 . 0 + o [] -1.0+ -0.4+ 0.2+

0.8+ 1.4+

<_
122 121 o 120

55 --

20

KM

{a)

FIG. 2. Geographical distribution of the natural log of site amplification with respect to the average station for octave frequency bands centered at frequency (a) 1.5 Hz, Co) 3.0 Hz, (c) 6.0 Hz, and (d) 12,0 Hz. The six different symbols represent six different site amplification ranges (in natural log) as indicated above.

590

F.

SUET

AL.
i Ii I L I J I ,
[]

I i

I i
[]

I i

I ~ I ~l

I ,I

I r 1 ,

I ,

,
58

~,
57 --

[] 0 )

[]
0

nD

0 0

[]

56 --

' -2.0+ o -1 .0+ [] - 0 . 4 + ,$, 0.2+ 0.8+


1.4+

X . "~

-q
~ I I I I i
!22

55 --

20 KM
I I i I

121
(b~
FIG. 2.

120

(Continued).

SITE AMPLIFICATION

FACTOR AND SURFICIAL GEOLOGY

591

JliJlll
o

f f l l l J l l r l l l l l l l l l l
[] []

fr

fllll

- ~

% 0 \ _

[]

4~

~.

k2

[] \ e

[], ~,

[]

#-

~i~

t
122

~ i

E j

J ]

121

120

(c)
FIG. 2.
(Continued).

592
~ I I I I I t ~ I I I I

F. S U E T A L .
i I I I J..I I I I I I I I I i I I I I I I

[]

58
-q

37

56

~ o
_

-2.0+ -1.0+
-0.4+
0.2+

[]
4)

0.8+ 1.4.+

.55
i

20 KM I,,,~,l~,,,,,l
I i i i t I
i

122

12!

120

(d} FIG. 2. (Continued).

SITE A M P L I F I C A T I O N
' ' ' ' I

FACTOR AND SURFICIAL

GEOLOGY
' ' ' ' I

593

q(N
I
I

Quaternary Sediments
I 1 I I I ,,

Pliocene Sediments
I I I I I I I I t

(N

<
E

qCN
I

Miocene Sediments
I I I I I I I

Cretaceous Sediments
T I I I i I I I

C'-I

<
E

7
c--I

:ranciscan
I I I I I I I I I

Granite
I I I I I I I I

10 FREQUENCY (Hz)

10 FREQUENCY (Hz)

FIG. 3. Coda site amplification factors versus frequency for six groups of station geology sites as indicated in the figure. Each line in the figure represents site amplification at one station and the average is shown by the dotted line.

nary sediments have high amplification at low frequency and the amplification factor gradually decreases with increasing frequency, whereas granitic rocks have low amplification at low frequency and the amplification factor gradually increases with increasing frequency. To quantify how the frequency-dependent coda site amplification factor related with surficial geology, we divided all the station sites into several groups. From Figure 3, we can see that the average site amplification of Miocene and Cretaceous sediments are nearly the same, so we assign these sites to one group. In the same way, the Franciscan and granite sites were assigned to one group. A few of our stations lie on Pre-Cretaceous metamorphic rock. Since the

594

F. S U E T A L .

site amplification factors at these stations are relatively low as compared to the other stations, they were assigned to one group. Thus, in our final analysis, station sites were divided into five groups: (1) Quaternary sediments, (2) Pliocene sediments, (3) Miocene sediments through Cretaceous Marine sediments, (4) Franciscan formation and Mesozoic granitic rocks, and (5) Pre-Cretaceous metamorphic rocks. The station site amplification factors in each group were logarithmically averaged and the mean value was assigned to the median geologic age of that group. Figure 4 shows the mean values versus their median geologic age. The standard error of individual measurement as well as the standard error of the mean are shown. From Figure 4 we see that the site amplification is high for young, Quaternary sediments at all frequencies up to

F=1.5 Hz
.... ~ ........ __' standard ........ errorJ ......
ean

F=3.0 Hz

u3
d

" '"l

........

:ta:d::d' :~'or

' ' '


can

u3

,111

. . . . . . . .

. . . . . . . .

. . . . . . .

F=6.0 Hz
--

F=12.0 Hz
standard error
san

standar~d

error

- -

I
eon

c~

~o
Y
T

~*
T

,,.[
I

....

. . . . . . . .

. . . . . . .

. . . . . . . .

.......
1000

10

I00

1000

10

100

AGF" (million years)

AGE (million years)

FIG. 4. Coda site amplification factors versus geologic age. The station sites were divided into five groups: (1) Quaternary sediments, (2) Pliocene sediments, (3) Miocene sediments through Cretaceous Marine sediments, (4) Franciscan formation and Mesozoic granitic rocks, and (5) Pre-Cretaceous metamorphic rocks. The station site amplification factors in each group were logarithmically averaged and the mean value was plotted against the median geologic age of that group. The standard error as well as standard error of the mean are also shown.

SITE A M P L I F I C A T I O N FACTOR AND SURFICIAL GEOLOGY

595

12 Hz. The amplification decreases with increasing geology age. The rate of decrease varies for different frequencies. Site amplification at low frequencies show a faster rate of decrease with age than that at high frequencies. From Table 1, we see that, although the site amplification is mainly determined by the surficial geology, there is variation within the same geologic unit. These could be due to the complex geologic structure around a particular station, such as thickness of the underlying geologic unit, surface topography, some small scale geologic features, and attenuation properties around that station. The Correlation between Coda Site Amplification and Station Magnitude Residuals Our results show that the site amplification factors in central California vary dramatically between young sediments and old rocks. The site effect should affect the measurement of earthquake source parameters such as earthquake magnitude. Eaton (1990) used over 1000 earthquakes to determine the duration magnitude (FMAG) and peak amplitude magnitude (XMAG) station residuals for the USGS central California seismic network. Our site amplification factors were compared with Eaton's FMAG and XMAG station residuals in Figures 5 and 6. A remarkable linear correlation was found between the logarithmic amplification factors and the magnitude station residuals, particularly for the FMAG station residuals with our site amplification factor at lower frequencies. As shown in Figure 5, the correlation coefficient between our results and FMAG station residuals is 0.90 for frequency 1.5 Hz and 0.89 for 3 Hz. Since the duration magnitude is determined by the duration of relatively low-frequency waves, the high correlation between FMAG residuals and amplification factors at low frequency is expected. T b strong correlation between our results and the FMAG station residuals demonstrates that a long duration is caused by large site amplification, consistent with the existence of the common decay curve, that is path(u[ t) in equation (1). In Figure 6, we can see that the correlation between coda site amplification factors with XMAG residual values is highest at frequencies 3 to 6 Hz, which implies that the predominant frequency of peak motion used in the determination of amplitude magnitude lies in this frequency range.
COMPARISON OF WEAK- AND STRONG-MOTION SITE RESPONSE

The 18 October 1989 Loma Prieta earthquake triggered over 130 strongmotion instruments operated by the USGS (Maley et al., 1989) and CSMIP (Shakal et al., 1989), providing a large number of data for strong-motion study in central California. It is interesting to compare our weak-motion site amplification results obtained by analyzing local earthquake coda waves in this area with that of strong ground motion results obtained from the Loma Prieta earthquake. It was found that the peak accelerations caused by the Loma Prieta earthquake were relatively independent of surface geology within the epicentral source region. However, outside of this region (at distances beyond about 50 km), surface geology appears to strongly influence the amplitude of ground motion. The peak acceleration is lowest for crystalline rock and rocks of the Franciscan complex, intermediate on soft rock and alluvium, and highest on artificial fill and Bay mud (Boore et al., 1989; EERI, 1990). This surface geology

596
F = 1.5 Hz

F. S U E T

AL.
F = 3 . 0 Hz

,--o.',o

. . . .

'

. . . .

'

" ;

r=0.89

*t~

'

. ,'..:
ej ~

. .:..:" ::'

:..~ ".
i ~1,

o ~e

-2

F:

6 . 0 Hz
. . . . I . . .

F=12.0
. I .

Hz
. . . I . . . .

r=0.80

r=0.71

ie=

al

~ ea

-
a ~==l =1~ r~o el

o,,'..:e"
ou ==,=e era==

I:

.t,;

:..'.

J
-2
, , , I . . . . t . . . . I . . . . , , , , r . . . . t . . . . I i = n i

-0.5

0 F'MAG

0.5

-1

-0.5

0.5

F~G

Fro. 5. Comparison of coda site amplification factors obtained in this study with FMAG site residuals determined by Eaton (1990) for the stations used in this study. The correlation coefficients for each frequency are also shown.

dependence outside of the epicentral source region is similar to our results from weak-motion study. For the epicentral region within 50 km, Chin and Aki (1991) show that the weak-motion amplification factors determined by the coda method do not apply to strong ground motion, suggesting a nonlinear site effect at sediment sites. Similar studies of comparing weak and strong ground motion site response were also made in the past. Hudson (1972) studied the strong ground motion of the 1971 San Fernando earthquake recorded in the Pasadena area (within epicentral distance of 50 km) and found that the strong-motion amplification factors did not correlate with local geology nor with the amplification factors estimated by Gutenberg (1957) using small local earthquakes. However, strong-motion site amplifications analyzed from this earthquake in Los Angeles and Long Beach correlated well with the weak-motion site amplification results

SITE A M P L I F I C A T I O N

FACTOR AND SURFICIAL GEOLOGY

597

F= 1.5 Hz
. . . . I . . . . I . . . . I ~ ' '

F= 3.0 Hz . - . . . . . r=0.79' . . . .

r==0.69

e~l*

*l

:: ". ;,~"
f*

' ....

" ~ / : .' /. " ..

*=

i~

.o ~..;,.

~/'Lo

IQ

-2

F=
. . . . I . . . .

6.0
I

Hz
. . . . I . . . . .

F=12.0

Hz

r-0.78

'

'

'r-O.'2 7= . . . . . . . . .

- ,., ""/~
to *t ol =

'

. . . .

~.
*%=~ == tl

=_.= .g .

t " t~"

i**

-2

z"

le

. . . .

,
I i i =

-0.5

0
XMAG

0.5

-0.5

0.5

XMAG

residuals determined by Eaton (1990) for the stations used in this study. The correlation coefficients for each frequency are also shown.

FIG, 6. Comparison of coda site amplification factors obtained in this study with XMAG site

estimated from nuclear explosions at the Nevada Test Site (Rogers et al., 1979, 1984). Although weak-motion site amplification has been difficult to connect with strong ground motion site response in the epicentral source region due to the nonlinear effect, it may be useful for earthquake damage estimation. Borcherdt (1970) found that there was a consistent correlation for the weak-motion site amplification factors determined from NTS data with intensities from the 1906 San Francisco earthquake and spectral amplification curves for the 1957 San Francisco earthquake. He suggested that the areas of high amplification determined from weak motion may coincide with areas of high intensity for future earthquakes. Since then, many researchers consistently showed that there was a strong correlation between the weak motion site response and the site-dependent part of variation in the distribution of earthquake intensity (Borcherdt and Gibbs, 1976; Chavez-Garcia et al., 1990; King et al., 1990).

598

F. SU E T A L .

Besides the comparison of site response between weak motion and strong motion of major destructive earthquakes, a comparative study between weakand strong-motion results from moderate size earthquakes were also made at various epicentral distances. Tucker and King (1984) compared middle/edge spectral ratios of weak-motion (10 -~ to 10 -3 g peak acceleration) and strongmotion (0.04 to 0.2 g peak acceleration) data at sediment-filled valleys in Garm, USSR, and concluded that there were no significant differences in site amplification between the weak-motion and strong-motion data groups. The magnitude of the largest earthquake they used was M L = 5.8 (located 100 km away from the recording sites). Their result is in agreement with our results (Fig. 5 and 6), which showed that our coda site amplifications are well correlated with Eaton's magnitude station residuals obtained using events with magnitude as large as M L = 5.5. Although our findings from weak-motion data show that the frequency dependence of site amplification for Quaternary sediments is opposite to that for hard-rock sites (Fig. 3), the site amplification for Quaternary sediments show higher amplification t h a n hard rocks at all frequencies studied (Fig 4). This is different from the crossover phenomenon consistently found in strong ground motion studies. As summarized by Aki (1988), strong ground motion studies show that soil sites have higher amplification than rock sites by a factor of 2 to 3 for periods longer than about 0.2 sec, but this relation is reversed for periods shorter than about 0.2 sec. Aki (1988) also recognized a similar crossover phenomena in the coda amplification factor determined by Phillips and Aki (1986) between granite sites and fault-zone sediment sites, but not between Franciscan rock sites and nonfault zone sediments. A close examination of Phillips and Aki's result yielded only a few sites with the crossover phenomena. The average of all stations on the same surficial geology using their results did not show the same trend. From Figure 3, we also find several stations on young sediments showing the crossover relative to some stations on granite. However, this phenomenon is eliminated as an average effect on weak motion. So, according to our results, the amplification due to low impedance of younger sediments still dominates over the deamplification due to high absorption at least up to 12 Hz on the average for central California, as far as the weak motion is concerned.
DISCUSSION

It is well known that the seismic velocity of near-surface materials is a critical factor in determining the site amplification. From Figure 3 and 4, we see that the site amplification factor at a station is well correlated with its surficial geology. Since the older rocks have been subjected to gravitational and tectonic loading over a longer period of time, they become more compact and competent; thus, they usually have higher seismic velocities than younger rocks. The relation shown in Figure 4 provides a simple w a y of estimating site effect at a specific site with known surficial geology. From Figure 4, we can also see that the relation of the site amplification versus geologic age has an interesting frequency dependent pattern. Site amplification at low frequencies show a faster rate of decrease than at higher frequencies. The range of site amplification from Quaternary sediments to hard rocks at 1.5 Hz is almost twice that of 12 Hz. Since high-frequency waves obey ray-theoretical law more t h a n low-frequency waves for a given inhomogeneity,

SITE AMPLIFICATION FACTOR AND SURFICIAL GEOLOGY

599

the site amplification may be more proportional to the inverse of the square root of impedance following the ray theory. The low frequencies may be influenced more by resonance for which peak amplification scales with the inverse of the impedance linearly. Thus, we can expect that the range of site amplification at lower frequencies be about twice that at higher frequencies. The kappa effect (Anderson and Hough, 1984) predicts the dominance of attenuation due to high absorption for younger sediments; however, at least on the average for central California, our results show that the dominant effect is the amplification due to low impedance for younger sediments. Our site amplifications are well correlated with Eaton's magnitude residuals, which confirms that our method effectively estimated weak-motion site amplification. Our results are also compared with Phillips and Aki's (1986) earlier results obtained using a singular value decomposition method. As shown in Figures 7 and 8, we can see that these two data sets are correlated well at lower
F= 1.5 Hz
. . . . l I"=0o83 . . . . I . . . . I . . . . .
. . . . i . . . .

F= 3.0 Hz
i . . . . i . . . .

. . . . /

ii~I

r=0.82
e =

s,,
I

' ; :~
I

!,%

II

"Z-..
.
%l

..

"r n

~".-.-,.."

'..

+J

,;.

+
-2

Oo ,%,,
Ir ~ea~e

e?le

i
. . . . i i i i i I i J I I I , , , ,

F = 6 . 0 Hz
. . . . i . . . . i . . . . i , , , , . . . . I . . . .

F= 12.0 Hz
I . . . . i . . . .

r=0.72

r-=0.48
I

{D

Q.

-'r'

;t

"%

~o

r,,-.~.-,,

..," ..2' ~" :..


o;~
u i. el So
" "

..:,, ...

"

-2

. . . .

. . . .

-0.5

0.5

-0.5

0.5

F~G

F~G

FIG. 7. Comparison of coda site amplification factors obtained by Phillips with FMAG site residuals determined by Eaton (1990). The correlation coefficients for each frequency are also shown.

600
F= 1.5 Hz
. . . . i . . . . i . . . . i '

F. S U E T A L .

F= 3.0 Hz
' L . . . .
i . . . . i . . . . I ' ' , i

r=0.88

,,.

r=0.87

.,,' :.
n me 0q "~.| ' ..
~,,* .

,.* . s % = #~

0
O

~II |

ii

-2

,'

. . . .

F = 6 . 0 Hz
' ' ' ' i . . . . I . . . . i . . . . i . . . . l . . .

F=12.0 Hz
. I . . . . I . . . .

r=0.87

r=0.76

,$'-

PC...'. . r"., .;,"


e==.

J..~'o

*0

-2

-2

-1

-2

. . . .

-1

I , , , r l i f l l l ~ = , ,

In(A/'~), BY PHILLIPS

In(A/A), BY PHILLIPS

FIG. 8. Comparison of coda site amplification factors obtained in this study with Phillips' results. The correlation coefficients for each frequency are also shown.

frequencies but not at higher frequencies. In general, our results show better correlation with Eaton's magnitude residuals. The standard errors of our results are also smaller t h a n those of their results. It is interesting to compare the site amplification of weak motion with that of strong motion and search for the possible connections and differences between them. Our results and analysis suggest that the weak-motion site amplification measurement is important in defining the in situ nonlinearity of the site effect and delineating its cause in relation to the geologic and geotechnical soil condition. The correlation between weak-motion site amplification and the site-dependent part of variation in earthquake intensity suggests that the weak-motion site amplification result could be directly useful in earthquake hazard analysis. Obviously, our analysis here is only a small step toward understanding the relation between weak-motion and strong-motion site amplification. Since the USGS seismic network station locations are selected to avoid

SITE A M P L I F I C A T I O N FACTOR AND SURFICIAL GEOLOGY

601

noisy soft-sediments sites, and hard-rock sites for strong-motion stations are under-represented due to engineering considerations, future efforts in co-locating weak-motion and strong-motion stations are needed. To extend our understanding of the nonlinearity and earthquake hazard, it is important to measure both the weak-motion and strong-motion amplification factor simultaneously at sites with a wide range of geologic conditions.
CONCLUSIONS

The recursive stochastic inversion method we used provides an efficient and stable site amplification inversion. The results show that the coda site amplification of a station is mainly controlled by the surface geology underlying that station. The site amplification is high for young, Quaternary sediments and decreases with increasing geologic age. The rate of decrease varies for different frequencies. Site amplification at low frequencies shows a faster rate of decrease than at higher frequencies. The relation between coda site amplification with geologic age provides a simple way of estimating site effect at a specific site with known surficial geology. A remarkable linear correlation between our site amplification factors and the magnitude station residuals determined by Eaton (1990) confirms that our method provides an effective means of weak-motion site amplification estimation. The comparison of our weak-motion site amplification results with that of strong motion obtained from the Loma Prieta earthquake suggests that weakand strong-motion site amplifications correlate well in the region outside the epicentral source region beyond epicentral distance of about 50 km. However, within the epicentral source region, the weak-motion amplification factors estimated from coda waves do not agree with the observed site effect on strong ground motion, suggesting a nonlinear site effect at sediment sites. These results together with studies on the correlation between weak-motion site amplification and earthquake intensity demonstrate the importance of in situ determination of weak-motion site amplification for seismic zonation.
ACKNOWLEDGMENTS We are grateful to Dr. Jerry Eaton for providing us with" calibration information and Dr. Rick Lester for his help in data collection. We also thank an anonymous reviewer and Dr. David Boore for their valuable comments. This research was supported by NSF grant BCS-8819988 and CUREe. REFERENCES Aki, K. (1969). Analysis of seismic coda of local earthquakes as scattered waves, J. Geophys. Res. 74, 615-631. Aki, K. (1988). Local site effect on ground motion, in Earthquake Eng. Soil Dynamics. II, Recent Advances in Ground-Motion Evaluation, J. Lawrence Von Thun (Editor), Am. Soc. Civil Eng./Geotechnical Special Publication 20, 103-155. Aki, K. and B. Chouet (1975). Origin of coda waves: source, attenuation and scattering effects, J. Geophys. Res. 80, 3322-3342. Anderson, J. and S. E. Hough (1984). A model for the shape of the Fourier amplitude spectrum of acceleration at high frequencies, Bull. Seism. Soc. Am. 74, 1969-1993. Boore, D. M., L. Seekins, and W. B. Joyner (1989). Peak accelerations from the 17 October 1989 Loma Prieta earthquake, Seism. Res. Lett. 60, 151-166. Borcherdt, R. D. (1970). Effects of local geology on ground motion near San Francisco Bay, Bull. Seism. Soc. Am. 60, 29-61. Borcherdt, R. D. and J. F. Gibbs (1976). Effects of local geological conditions in the San Francisco Bay region on ground motions and the intensities of the 1906 earthquake, Bull. Seism. Soc. Am. 66, 467-500.

602

F. SU E T AL.

Chavez-Garcia, F. J., J. G. Pedotti, D. Hatzfeld, and P. Y. Bard (1990). An experimental study of site effects near Thessaloniki (Northern Greece), Bull. Siesm. Soc. Am. 80, 784-806. Chin, B. and K. Aki (1991). Simultaneous determination of source, path and recording site effects on strong ground motion during the Loma Prieta earthquake: a preliminary result on pervasive non-linear site effect, Bull. Seism. Soc. Am. 81, 1859-1884. Eaton, J. P. (1980). Response arrays and sensitivity coefficients for standard configurations of the USGS short.period telemetered seismic system, U.S. Geol. Surv. Open.File Rept. 80.316. Eaton, J. P. (1990). Determination of amplitude and duration magnitude and site residuals from short period seismographs in northern California, Preprint. EERI (1990). Strong ground motion, Earthquake Spectra (suppl.) 6, 65-68. Gutenberg, B. (1957). Effects of ground on earthquake motion, Bull. Seism. Soc. Am. 47, 221-250. Hudson, D. E. (1972). Local distribution of strong earthquake ground motions, Bull. Seism. Soc. Am. 62, 1765-1786. King, K. W., A. C. Tart, D. L. Carver, R. A. Williams, and D. M. Worley (1990). Seismic ground-response studies in Olympia, Washington, and vicinity, Bull. Seism. Soc. Am. 80, 1057-1078. Maley, R., A. Acosta, F. Ellis, E. Etheridge, L. Foote, D. Johnson, R. Porcella, M. Salsman, and J. Switzer (1989). U.S. Geological Survey records from the northern California (Loma Prieta) earthquake of October 17, 1989, U.S. Geol. Surv. Open-File Rept. 89.568, 85 pp. Mayeda, K. S. Koyanagi, and K. Aki (1991). Site amplification from S-wave coda in the Long Valley Caldera region, California (submitted for publication). Phillips, W. S. and K. Aki (1986). Site amplification of coda waves from local earthquakes in central California, Bull. Seism. Soc. Am. 76, 627-648. Rautian, T. G. and V. I. Khalturin (1978). The use of the coda for determination of the earthquake source spectrum, Bull. Seism. Soc. Am. 68, 923-943. Rogers, A. M., R. D. Borcherdt, P. A. Covington, and D. M. Perkins (1984). A comparative ground response study near Los Angeles using recordings of Nevada nuclear tests and the 1971 San Fernando earthquake, Bull. Seism. Soc. Am. 74, 1925-1949. Rogers, A. M., J. C. Tinsley, W. W. Hays, and K. W. King (1979). Evaluation of the relation between near-surface geologic units and ground response in the vicinity of Long Beach, California, Bull. Seism. Soc. Am. 69, 1603-1622. Shakal, A., M. Huang, M. Reichle, C. Ventura, T. Cao, R. Sherburne, M. Savage, R. Darragh, and C. Petersen (1989). CSMIP strong-motion records from the Santa Cruz Mountains (Loma Prieta), California earthquake of 17 October 1989, California Strong Motion Instrumentation Program Report No. OSMS 89-06, 195 pp. Tsujiura, M. (1978). Spectral analysis of the coda waves from local earthquakes, Bull. Earthq. Res. Inst., Tokyo Univ. 53, 1-48. Tucker, B. E. and J. L. King (1984). Dependence of sediment-filled valley response on the input amplitude and the valley properties, Bull. Seism. Soc. Am. 74, 153-165. Su, F. and K. Aki (1990). Temporal and spatial variation on coda Q-1 associated with the North Palm Springs earthquake of July 8, 1986, Pure Appl. Geophys. 133, 23-52. Su, F., K. Aki, and N. N. Biswas (1991). Discriminating quarry blasts from earthquakes using coda waves, Bull. Seism. Soc. Am. 81, 162-178. Zeng, Y. (1990). A fast recursive stochastic inverse solution with various seismological applications (submitted for publication).
DEPARTMENT OF GEOLOGICAL SCIENCES UNIVERSITY OF SOUTHERN CALIFORNIA LOS ANGELES, CALIFORNIA 90089-0740

Manuscript received 5 July 1991

You might also like