Anderson 1991
Anderson 1991
Anderson 1991
2186 2193,December1991
BY JOHN G. ANDERSON
ABSTRACT
The spectral decay parameter, K, defined by Anderson and Hough (1984) is
described as a function of distance, R, and site, S, as K(R, S) = Ko(S ) + E(R).
The terms E(R) and K0(S) are found by a method that is unbiased by prior
assumptions about the nature of the distance dependence. Variance using this
model is substantially smaller than in simpler models that do not incorporate
both site effects and a distance effect. For data gathered near Anza, California,
and for distances less than 100 km, ~T(R) is very similar for both P- and S-wave
windows, but K0(S) is smaller for P than for S on average by a factor of about 2.
INTRODUCTION
The spectral decay parameter, K, is an observational parameter of the spec-
trum of earthquake ground motion. The basic observation (Anderson and
Hough, 1984) is that, at high frequencies, the spectrum of ground acceleration
falls off exponentially with frequency, f, i.e.,
A ( f ) c¢ A o e x p ( - w K f ) . (1)
In its action on the seismic spectrum, K may be compared with t*, but is only
equal to t* under specialized assumptions (e.g., Hough et al., 1988).
The basic characteristics of the spectral decay parameter in California, as
described by Anderson and Hough (1984), Anderson (1986), and Hough et al.
(1988) are that (1) for a single site, K has a slight, but meaningful, increase as
the epicentral distance increases and (2) the curve K(R) moves upward or
downward as a whole from site to site. A clear illustration of these observations
is presented in Anderson (1986) and Hough et al. (1988), where K(R, S) follows
approximately parallel trends for stations in Imperial Valley, on deep sedi-
ments, and at Pifion Flat Observatory, on granite with a thin weathered layer.
The numerical values of K at Pifion Flat are about 0.06 sec smaller than in
Imperial Valley at all distances.
A mathematical formulation of this is to regard K(R, S) as a function of
epicentral distance, R, and a categorical variable to identify the site, called S
here, as follows:
K(R,S):Ko(S)+E(R), (2)
where Ko(S) takes a unique value for each site. The distance dependence is
entirely described by ~(R), which is constrained to equal zero at zero distance.
There are several practical motivations for obtaining a description of ~(R).
First, such a description will facilitate comparisons among regions, where ~(R)
might differ. Second, it may be interesting to know Ko(S) for a new site (for
earthquake engineering or geophysical investigations) and a reliable ~(R)
2186
SPECTRAL DECAY PARAMETER IN SOUTHERN CALIFORNIA 2187
NUMERICAL RESULTS
For data, this paper uses observations of c for both P and S waves, deter-
m i n e d by Hough (1987) and Hough et al. (1988), from 10 digital seismic stations
in the Anza network in southern California. It also uses S-wave observations
from Anderson and Hough (1984) for the Imperial Valley. Maps of events and
stations are found in Hough et al. (1988) and Anderson and Hough (1984).
Table 1 lists ~(R) for P and S waves, and Figures 1 and 2 show the fit of
these curves to the data for each station. Table 2 lists co(S) for every station.
Finally Figure 3 compares £(R) derived for P and S waves. Figures 1 and 2
demonstrate that the data match the trend of the average curve without
exception. Table 1 also lists the number of data in each distance range which
contribute to the determination of £(R). Six S-wave data and 11 P-wave data
at distances beyond 200 km further constrain the results. The number of data
drop off quickly for distances beyond 100 km, implying that the numerical
estimates for ~(R) beyond 100 km are less reliable than at shorter distances.
The model is truncated at 200 km because the number of data beyond 200 km is
very small and the smoothness constraints influence the fit to the data there.
The model in Figures 1 and 2 is a rather complex description for the
parameter c. Using analysis of variance (Mason et al., 1989), it can be shown
that the model given in Tables 1 and 2 for both P and S waves is statistically
better than each of the following, given in order of increasing numbers of
2188 J. G. ANDERSON
TABLE 1
PRELIMINARYNUMERICALVALUESFOR K(R) IN
SOUTHERNCALIFORNIA
S Waves P Waves
0 0.0 13 0.0 20
10 3.5 57 2.7 28
20 5.7 62 4.8 32
30 6.9 51 6.3 27
40 6.9 40 7.7 18
50 6.5 80 9.6 29
60 7.9 70 12.4 23
70 11.1 58 16.1 26
80 15.5 36 19.5 12
90 21.9 52 21.9 26
100 28.7 22 23.7 6
110 34.0 8 25.6 7
120 37.3 10 27.5 6
130 39.1 9 29.0 3
140 40.6 3 29.8 3
150 42.5 8 29.9 3
160 44.3 2 29.6 2
170 45.5 4 29.4 2
180 46.4 4 29.0 1
190 48.5 6 28.2 3
200 53.3 0 27.0 1
unknown parameters:
Model 1: K(R, S) = c o n s t a n t .
Model 2: K ( R , S ) = K o + m R .
Model 3: ~(R) = 0 a n d Ko(S) developed for each station.
Model 4: ~ ( R ) = m R a n d Ko(S) developed for each station.
In the context of t h e adopted f i t t i n g procedure, it would be easy to develop
a d d i t i o n a l models w i t h c o m p l e x i t y i n t e r m e d i a t e b e t w e e n model 4 a n d t h e model
g i v e n in Tables 1 a n d 2. This w o u l d be done by i n c r e a s i n g the s e p a r a t i o n w i t h
distance, A R, b e t w e e n points at w h i c h ~(R) is d e t e r m i n e d , u n t i l one finds the
least complex distance d e p e n d e n c e t h a t is d e m a n d e d by the data. The given
model, w i t h A R = 10 km, p r o b a b l y h a s more p a r a m e t e r s t h a n necessary.
However, even t h o u g h large v a l u e s for A R (e.g., 50 km) m i g h t be justified from
a strictly n u m e r i c a l point of view, t h e r e is no physical justification for i m p o s i n g
a model in w h i c h c h a n g e s in slope of ~(R) occur only at large distance
intervals. Thus, t h e s e models are not i n v e s t i g a t e d . The p r o c e d u r a l improve-
m e n t of a v o i d i n g such unjustified artifacts in ~(R) w a r r a n t s the additional
complexity in t h e p r e f e r r e d models. The RMS misfit in the p r e f e r r e d model is
n e a r l y a factor of 2 g r e a t e r t h a n a n e s t i m a t e for t h e s t a n d a r d deviation in
SPECTRAL DECAY PARAMETER IN SOUTHERN CALIFORNIA 2189
.14 .14
.12 BZN S-waves .12 CRY S-waves
,10 .I0
.08 o .08
o o o .06 o
.og
~ .04 o ~ ,04 o
ao
.02 ~ ~ .02
.00 .00
I I I t I I i I I
-.02 ~ ~ ~ ~ ~ ~ , r ~ -.02
40 80 120 160 200 40 80 120 160 200
.14 .1'
.121 ELC S-waves o .1~ FRD S-waves
.lOi .1(
"~" .08 .0~
v~ .06 .0(
o
.04 e ,0, o o ~ o °8 o
.02 .0",
.00 .0( o8
i i i ~ i i i r i
-.02 -.0:
4-0 80 120 160 200 40 80 120 160 200
.14 .14
.12 KNW S-waves .12 LVA S-waves
.10 .10 8
08
%" 08 .08
.o6 .06
~z .04 8 ~ ,04 8
.001 .OC
_ . 0 2 1 1 1 1 r [ 1 1 1 _ 0 2 1 1 1 1 1 1 [ 1 1
0 40 80 120 160 200 0 40 80 120 160 200
Epicentral Distance (km) EpicentrQI Distance (km)
(a)
.14 .14
.12 PFO S-waves ,12 RDM S-waves
.10 .10
.08 .08
'~ .06 .06
~¢ ,04 04
.02 .02
.00 .OC
_ . 0 2 ~ 1 1 1 1 1 q l l _ . 0 2 1 1 1 1 ~ 1 1 1 1
0 40 80 120 160 200 0 40 80 120 160 200
.14 ,14
,12 SND S-waves .12 TRO S-waves
.10 .10
o
%" .08 .08
,06 .06 o~--
~ .04 = Bo~, 8o0 8 .04
.02 .02
.00 .OO o
I ~ I I I I I I I
-.02 i I I I I I I I -.02
40 80 120 160 200 40 80 120 160 200
Epicentral Distance (kin)
.14
.12 WMC S-waves
.10
~" .08
~ .06
~ .04
.02
.00
l I l i i I I i
-.02
40 80 120 I BO 200
Epicentral Distance (kin)
(b)
F I G . 1. D a t a and model K(R, S) for each station for S waves.
2190 J . G . ANDERSON
.08 .08
BZN P-waves CRY P-waves
.06 o .06
o .04 8 o
.04
~o o° a° o°
.02 . . 4 ~ . _ _ ~ ~ .02
.00 .DO
o
i ~ i i i i i i i -,02 I I I I i I I i
-.02
40 80 120 160 200 40 80 120 160 200
.08 .08
FRD P-waves KNW P-waves
.06 .06
o
.04 ,04
o % °o
~e .02 o o .02 o o °
.00 .00
~ o ~
o
-.02 r [ i I I I I I I -.02 i t i i r i i i
.08 .08
LVA P-waves PFO P-waves
.06 .06
o o
%" .04 o o ~ .04 o o
°~"~° ° o o
.02 o .02 o ~ o
o
.00 .00
o
-.02 t t t t r I , r i -.02 I t J , r i i q
0 40 80 120 160 200 0 40 80 120 160 200
Epicentral Distance (kin) Epicentral Distance (kin)
(a)
.08 .Ok
RDM P-waves SND P-waves o
.06 o ,06
o
.04 .04 ~ o o
o o a~ oo @o
.02 o Oa .02 o o °° ooe
.00 .OC
0
-.02 r l ~ l r l r l l -.02 ~ r ~ ~ ~ ~ r ~ r
40 80 120 160 200 0 40 80 120 160 200
.08 .08
TRO P-waves
.06 .06 WMC P-waves t
o
o o
.04 .04
.02 .02
o
.00 .00
q i [ i i r i i [ -,02 i i i i i i i i i
-.02
40 80 120 160 200 40 80 120 160 200
Epicentra/ Distance (km) Epicentral DTstance (km)
(b)
.08
.O6
69
.04
i
.02
J
J j
.00 I I I I I i I I I
previously attributed Ko(S) to attenuation below the stations, and these results
support t h a t hypothesis qualitatively. For KNW and PFO, Fletcher et al. (1990)
characterized the attenuation in the upper 130 m with values for t* equal to 4
and 3 msec, respectively, agreeing within uncertainties with the present esti-
mates of Ko(S) of 1.9 and 3.6 msec. Under the assumptions t h a t cause t* and K
to be the same (source spectrum has an ~-2 asymptote at high frequencies, a
freqt~ency-independent contribution to Q is separable, wave propagation and
site resonances are small), this limited comparison is consistent quantitatively
with the hypothesis t h a t Ko(S ) is caused by attenuation below the site.
Unlike the terms in ~(R) for P and S waves, the station terms are not
similar. Rather Ko(S) is smaller for P waves t h a n for S waves at all but one
station (CRY). Two stations yielded Ko(S) for P waves slightly less t h a n zero, a
result t h a t can be attributed to small numerical values and observational
uncertainty but might also result from a smaller t h a n average increase in K for
distant earthquakes recorded at those stations. The ratio of Ko(S) for S to P
waves from the other eight stations is 2.2 + 0.8. As observed by Hough and
Anderson (1988), this implies t h a t Qp is greater t h a n Qs, on the average, in the
weathered layer.
£(R) with other regions, estimating Ko(S) from earthquakes at arbitrary dis-
tances, constraining the shapes of spectra in synthetic seismograms, and study-
ing the depth dependence of the frequency-independent contribution to Q.
Finally, the model can be used as a starting point to better understand the
remaining scatter in the data, since the RMS residual in the final model is still
nearly twice the uncertainty in the data, implying that unmodeled signal
remains.
ACKNOWLEDGMENTS
The author thanks Sue Hough for supplying the data used in this study and for reviewing the
manuscript. D. Boore, W. Peppin, and an anonymous reviewer suggested several improvements to
the manuscript. The National Science Foundation research grant EAR 85 18647 supported the work
done in this study.
REFERENCES
Anderson, J. G. (1986). Implication of attenuation for studies of the earthquake source, in Earth-
quake Source Mechanics, Maurice Ewing Series 6, S. Das, J. Boatwright, and C. H. Scholz
(Editors), American Geophysical Union, Washington, D.C., 311-318.
Anderson, J. G. 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. (1983). Stochastic simulation of high frequency ground motions based on seismological
models of the radiated spectra, Bull. Seism. Soc. Am. 73, 1865-1894.
Brillinger, D. R, and H. K. Preisler (1984). An exploratory analysis of the Joyner-Boore attenuation
data, Bull. Seism. Soc. Am. 74, 1441-1450.
Castro, R. R., J. G. Anderson, and S. K. Singh (1990). Site response, attenuation and source spectra
of S waves along the Guerrero, Mexico, subduction zone, Bull. Seism. Soc. Am. 80, 1481-1503.
Fletcher, J. B., T. Fumal, H.-P. Liu, and L. C. Carroll (1990). Near-surface velocities and
attenuation at two boreholes near Anza, California from logging data, Bull. Seism. Soc. Am. 80,
807-831.
Hanks, T. C. (1982). fmax, Bull. Seisrn. Soc. Am. 72, 1867-1879.
Hough, S. E. (1987). The attenuation of high frequency seismic waves, Ph.D. Thesis, University of
California, San Diego, 157 pp.
Hough, S. E. and J. G. Anderson (1988). High-frequency spectra observed at Anza, California:
Implications of Q structure, Bull. Seism. Soc. Am. 78, 692-707.
Hough, S. E., J. G. Anderson, J. Brune, F. Vernon III, J. Berger, J. Fletcher, L. Haar, T. Hanks,
and L. Baker (1988). Attenuation near Anza, California, Bull. Seism. Soc. Am. 78, 672-691.
Mason, R. L., R. F. Gunst, and J. L. Hess (1989). Statistical design and analysis of experiments,
Wiley, New York, 692 pp.
Vernon, F. L. (1989). Analysis of data recorded on the ANZA seismic network, Ph.D. Thesis,
University of California, San Diego.
SEISMOLOGICALLABORATORYAND
DEPARTMENTOF GEOLOGICALSCIENCES
MACKAYSCHOOLOF MINES
UNIVERSITYOF NEVADA
RENO, NEVADA89557