CHARACTERISTICS OF GROUND MOTION RESPONSE SPECTRA FROM RECENT
LARGE EARTHQUAKES AND THEIR COMPARISON WITH IEEE STANDARD 693
F. Su1, J. G. Anderson2 and Y. Zeng3
ABSTRACT
We have examined ground motion response spectra from the recent large
earthquakes in California, Japan, Turkey and Taiwan. We have studied the
distance, site, and magnitude effect on ground motion response spectra and
compared these spectra with the required response spectrum (RRS) specified by
the IEEE Standard 693, “Recommended Practice for Seismic Design of
Substations”. The shapes of the observed response spectra show strong
magnitude and site dependency, but weak distance dependency. A larger
earthquake generates much more long-period seismic energy than a smaller
earthquake. A soft soil site shows higher amplification of the long-period seismic
signal than a rock or stiff soil site. These combined effects of a large earthquake
acting on a soft soil site are to significantly increase seismic energy at long
periods. Thus, the normalized spectra from a large earthquake recorded on a soft
soil site have the highest likelihood of exceeding the IEEE 693 spectrum at long
periods. By comparing observed acceleration response spectra with the required
response spectrum given in IEEE 693 for high seismic performance level, we
found the stations that exceed the absolute level of IEEE 693 spectrum are
generally located: 1) very close to the fault rupture trace; 2) in a location that
experienced a strong directivity effect; 3) on the hanging wall adjacent to the
primary fault trace; or 4) at a site that experiences large site amplification. In the
low frequency range (f<1 Hz), the exceedance is usually associated with a larger
magnitude earthquake at soft soil sites at near source distance (D<50 km). For
larger earthquakes (M>7.5), exceedance at low frequencies may be common.
Introduction
Ground motion, and the corresponding response spectrum, is significantly influenced by
the size of the earthquake, the travel path between source and receiver, and the local site effect.
Sorting out the influences of these various factors is challenging. Past studies on response
1
Research Associate Professor, Dept. of Geophysics, Colorado School of Mines, Golden, CO 80401
Professor, Seismological Laboratory., University of Nevada-Reno, Reno, NV 89557
3
Research Geophysicist, US Geological Survey, Golden, CO 80401
2
1
spectra have been limited by the number of strong motion data available. Most studies (e.g.
Seed et al., 1976; Mohraz, 1976; Seed and Idriss, 1979; Singh, 1985; Atkinson and Boore,
1990; Crouse and McGuire, 1996; Sabetta and Pugliese, 1996) were based on combined data
sets that come from different sizes of earthquakes and recorded from different regions. Thus,
the isolation of individual factors from various influences is limited by the nature of the data.
Recently, as a result of the great improvements in strong motion instrumentation and
deployment, as well as the occurrence of several large destructive earthquakes in California,
Japan, Turkey and Taiwan, the number of strong motion recordings increased dramatically. Just
two of the earthquakes, the Northridge, California earthquake (1994, Mw=6.7) and the Chi-Chi,
Taiwan earthquake (1999, Ms=7.6), provided over a thousand strong motion records. They are
also excellent strong motion data sets in terms of data quality, the range of recorded ground
motion, and the range of travel distances and site conditions. These new data provide a great
opportunity to study strong ground motion and examine various factors influencing their
response spectra.
In this paper, we have investigated the influence on response spectra from earthquake
magnitude, propagation path, and local site effects. Our intention is to study the influence of
each of these factors separately. This requires event specific analyses, separating data of each
event into different groups, and making statistical analysis accordingly. Large numbers of
ground-motion data made this possible. We have also compared our results with the recent
IEEE (the Institute of Electrical and Electronics Engineers) standards.
Data
We have collected strong motion data from the Central Weather Bureau of Taiwan,
which operates the strong motion array in Taiwan, the Guerrero strong motion accelerographs
array jointly operated by the Seismological Laboratory at the University of Nevada, Reno and
the Instituto de Ingenieria, UNAM, and from the uniformly processed strong motion database at
the Pacific Earthquake Engineering Research Center (web site: http://peer.berkeley.edu/smcat)
and through Dr. Walter Silva of Pacific Engineering and Analysis who has also built this
database.
To study individual factors, such as earthquake magnitude, travel distance and site
response independently, strong motion data from different events were analyzed separately and
data from each event were divided into groups according to their source-site distances and
station site conditions. The source-site distance (D), defined as the closest distance between the
station and source, is divided into four ranges, D<20 km, 20<D<50 km, 50<D<100 km, and
D>100 km. Both Chi-Chi and Northridge earthquakes have been extensively recorded, and the
large numbers of ground-motion data made event specific statistical analysis possible. Only
horizontal components were used in this analysis.
1. Distance Effect on Response Spectra
The effect of distance on response spectra can be examined by comparing response
spectra from the same site category at different distances. Figure 1 compares average
acceleration spectra in four different distance ranges for the Chi-Chi earthquake. From the
2
figure we can see that for the hard sites (Figure 1a), the shapes of the averaged response spectra
are generally similar among different distance groups. The spectral acceleration gradually
increases with the increasing frequency, reaches peak at about 1 Hz and slowly decreases
towards their averaged PGA at high frequencies. The spectral amplitude of each distance group
steadily decreases with increasing distance. For the medium and soft soil sites (Figure 1b and
1c), however, this attenuation pattern is not as uniform. While the spectral amplitude is steadily
decreasing with increasing distance for hard sites, it is flatter for soft soil sites. Figure 2
compares average acceleration response spectra for different distance groups from the
Northridge earthquake. The general features are similar to those shown in Figure 1, although
differences in the attenuation due to site condition are less pronounced on Figure 2. Other
earthquakes we studied display a range of behavior consistent with the distance dependence
shown by these two earthquakes.
Figure 1: Comparison of average acceleration spectra of the same site category at different distance ranges
for the Chi-Chi earthquake. The distance ranges are D<20km, 20<D<50km, 50<D<100km, and
D>100km for all three frames. Each line is obtained by averaging all the data in that site-distance
group. Although some lines nearly tangent in places, none of the lines cross.
Figure 2: Comparison of average
acceleration spectra of
the same site category at
different distance ranges
for the Northridge
earthquake. The
distance ranges are
D<20km, 20<D<50km,
and 50<D<100km for all
four frames.
3
2. Local Site Effect on Response Spectra
To examine local site effects on response spectra, we compared averaged acceleration
response spectra of different site categories within the same distance range for each event.
Figure 3 is the result from the Chi-Chi earthquake. It shows that the spectral response curves
from different site conditions differ in both amplitude and spectra shape.
In Figures 3a and 3b, which have data for distances within 50 km, the response spectra
are higher for soft soil sites than for medium or hard sites at frequencies from about 0.1 Hz to 1
Hz, but the situation is opposite at higher frequencies. The crossover occurs at about 1 Hz. At
lower frequencies, the apparent cause of soil amplitudes being larger than rock motions is large
site amplification at the softer soil sites. At higher frequencies, however, soil motions become
lower than rock motions. This characteristic is most easily explained by the nonlinear site
response of softer sites. In Figure 3d, which has data from distances over 100 km, the soil
motion is consistently higher than rock motion over the whole frequency range observed and
there is no crossover as in Figures 3a and 3b. At this distance range we generally do not expect
nonlinearity to occur. In Figure 3c for the distance range of 50 to 100 km, the soil motion is
greater than the rock motion at lower frequencies and they become close to each other for
frequencies over 1 Hz. This could be due to the result of decreasing soil nonlinearity within this
distance range. Figure 4 shows the result from Northridge earthquake. It has a similar soil-rock
crossover pattern as discussed for the Chi-Chi earthquake with the crossover also occurring at
about 1 Hz (Figure 4a).
Figure 3: Comparison of average acceleration spectra for different site categories within the same distance
range for the Chi-Chi earthquake. Each line is obtained by averaging all the data in that sitedistance group. Different line style indicates different site conditions, as labeled.
4
Figure 4: Comparison of average acceleration spectra for different site categories within the same distance
range for the Northridge earthquake. Different line style indicates different site conditions, as
labeled.
3. Earthquake Magnitude Effect on Response Spectra
A good example for looking at the magnitude effect on the shape of response spectra
would be to compare response spectra from earthquakes of different size that occurred in the
same source area and were recorded at the same station. Such an approach is closely
represented in Figures 5 and 6. Figure 5 shows the epicenter of the Chi-Chi mainshock and an
aftershock that occurred next to the mainshock epicenter. Both earthquakes were very well
recorded. Figure 6 compared response spectra from these two events recorded by a station on a
soft soil site (TCU116) and a hard site (TCU089). The response spectra were normalized by
peak acceleration in order to focus on the comparison of the spectral shape. We can see the
normalized spectra of the mainshock are much higher than that of the aftershock at lower
frequencies, indicating that the mainshock generates
proportionately more long-period energy than the
aftershock. Figure 7 compares response spectra from six
earthquakes with a range of magnitudes from 3 to 8
recorded in Guerrero, Mexico (earthquake data are from
Figure 5: Map shows the epicenter
of the Chi-Chi mainshock
(star) and an aftershock
(circle) and the location of a
rock site (TCU089) and a
soil site (TCU116). The
ruptures fault trace is shown
as a thick solid line.
Figure 6: Comparison of response spectra from the Chi-Chi
mainshock and the aftershock recorded at: (a) the soft
soil site (TCU116), and (b) the rock site (TCU089).
Response spectra were normalized to a common peak
5
Anderson and Quaas, 1988). All these earthquakes were recorded on hard rock sites with S-P
travel time of about 3 seconds. Thus, the
recording site conditions and the travel paths
are quite similar among all these records.
We can see from the figure that the
magnitude of the earthquake has a strong
effect on the shape of the response spectra.
The larger earthquake generates much more
long-period energy than the smaller ones.
In Figure 8 we have compared
averaged acceleration spectra from the ChiChi, Kocaeli, Landers, Kobe, Loma Prieta,
and Northridge earthquakes. Although these
earthquakes are from different tectonic
Figure 7: Comparison of response spectra for
regions, we compare them together in order
earthquakes with magnitude 3 to 8
to gain the general picture of the response
recorded on rock sites in Guerrero,
spectra from these recent earthquakes. Each
Mexico. Response spectra were
curve in the figure is obtained by averaging
normalized to a common peak
all available data within 50 km of the event.
Figure 8a compared average acceleration spectral shape for soil sites and Figure 8b for rock
sites. In general, the shape of the response spectra, in spite of regional differences, has
characteristics similar to those discussed above. Larger sized earthquakes have proportionately
more low frequency energy and less high frequency energy relative to smaller ones.
(a)
(b)
Figure 8: Comparison of average acceleration spectra among Chi-Chi, Kocaeli, Landers, Kobe, Loma
Prieta and Northridge earthquakes. Each curve represents one earthquake with the spectrum
averaged from data within distance of 50 km of the earthquake. Response spectra were
normalized to a common peak acceleration for the comparison. (a) For soil sites (the curve for
Chi-Chi earthquake is from soft soil sites, others are from site D sites). (b) For rock sites (the
curve for Chi-Chi earthquake is from hard sites, others are from site A sites).
6
Comparison of Our Results with Recent IEEE Standards
Since individual response spectra of actual earthquakes are highly irregular, for the
purpose of describing ground motion for design, a smoothed response spectrum representing
possible variability in ground motion (a design spectrum) is used. The design spectrum is
usually determined by enveloping those of individual response spectra at a particular level, such
as a pth percentile of the spectra. An 84 percentile of the spectra is commonly used by engineers,
which is approximately mean-plus-one standard deviation of the spectra.
Figure 9: Plots of 84 percentile acceleration spectra (approximately mean-plus-one standard deviation of
the spectra) of different site categories within the same distance range for Chi-Chi earthquake
compared with the standard response spectrum (thin solid line) specified by the IEEE 693-1997
for the high seismic performance level. Response spectra were normalized to a common peak
acceleration.
Figure 9 plots 84 percentile acceleration spectra of different site categories for the ChiChi earthquake compared with the standard response spectrum specified by the IEEE-693 for
high level of seismic qualification of electric substations (which include all electric equipments).
Significant differences in spectral shape between the data from the Chi-Chi earthquake and the
IEEE-693 standard occurred at frequencies between about 0.1 and 2 Hz for soft soil sites within
distance of 50 km. The average response spectra from the Chi-Chi earthquake at these distances
are significantly higher than the standard spectrum defined by IEEE-693. This suggests that the
current IEEE standard is not protecting those electric equipments if catastrophic earthquakes
such as the Chi-Chi event should happen and if the affected equipment is vulnerable at this
frequency range. Figure 10 compared 84 percentile acceleration spectra of different site
categories with distance of less than 20 km for Northridge earthquake. Comparison shows that
7
the standard response spectrum specified by the IEEE-693 has effectively represented the input
from the Northridge earthquake.
We have also compared individual
response spectra of the recent strong motion
acceleration recordings with the standard
response of IEEE 693 for high seismic
performance level. Every event we have
examined (see Figure 8 for the listing of the
events) has some cases that the recorded
acceleration response spectra have exceeded
the IEEE standard for high seismic
performance level. We found that those
stations that recorded exceedances are
generally located at sites that are: 1) very
close to the fault rupture trace; 2) near the
edge of the fault that suffer strong directivity
effect; 3) on the hanging wall adjacent to the
primary fault trace; 4) at the site that suffers
large site amplification. In the low frequency
range, the exceedance is usually associated
with a larger magnitude earthquake at soft
soil sites at near source distance (D<50 km).
Due to the magnitude effect on response
spectra, it is to be expected that for the rare
earthquakes with magnitude significantly
larger than Chi-Chi (M7.6), exceedances at
long periods will be more common.
Figure 10: Plot of 84 percentile acceleration
spectra of different site categories within
distance of 20 km for Northridge
earthquake, compared with the standard
response spectrum (thin solid line)
specified by the IEEE 693-1997 for the
high seismic performance level.
Response spectra were normalized to a
common peak acceleration.
Discussion and Conclusion
In this article, we have investigated the influence of earthquake magnitude, propagation
path, and local site effects on response spectra. Through event-specific analysis and comparing
data among different site and distance groups, we are able to obtain a more clear and realistic
perspective on the influence of these individual parameters on the response spectra.
We found response spectra to be strongly magnitude dependent. By comparing response
spectra of different sized earthquakes, we found there was a marked increase in spectral
amplitudes at low frequencies as magnitude increases. This is consistent with the seismological
observations and is expected based on well-established seismic scaling laws (Aki, 1967; Brune,
1970). At extremely long periods (e.g. T~20 seconds) the amplitude is expected to increase
proportional to 101.5 M W in the far field spectra. In the engineering community, however, design
spectra in the past have not generally taken advantage of the magnitude dependency. There is
some implicit dependency on the “maximum magnitude” in the IBC 2000 code. However, the
knowledge of the broadband spectra from large earthquakes will certainly help us to be more
8
aware of the threat to structures from large earthquake ground motions and help implement
corresponding measures in structural design to mitigate earthquake-related damages.
We found a significant influence of local site effects on response spectra in both the ChiChi and Northridge earthquakes. By comparing average response spectra of different site
categories within the same distance range for each event, we found soft soil sites showed
amplification of factors of 2 to 3 at lower frequencies in comparison to about 50% at higher
frequencies relative to rock sites. The characteristic of the crossover between response spectra
from soil sites and rock sites is best explained by the nonlinear site response of softer sites.
We have also compared our result with recent IEEE standards. The IEEE-693 spectrum
for the high seismic performance level is exceeded by the average spectra at long periods for soft
soil sites within distances of 50 km. The result is that it will be no surprise to see the low
frequency part of the spectra exceeding the IEEE spectra in soil site categories in a significantly
larger magnitude earthquake.
Acknowledgements
We thank Dr. Norm Abrahamson for his valuable suggestions to this work and Dr.
Walter Silva for his help in data collection. This material is based upon work supported in part
by the PEER grant SA2976JB, NSF grants CMS-0000033 and CMS-0000050, and SCEC
funding.
References
Abrahamson, N. A., and W.J. Silva (1997). Empirical response spectral attenuation relations for shallow
crustal earthquakes, Seismological Research Letters. 68, 94-127.
Aki, K (1967). Scaling law of seismic spectrum, Journal of Geophy. Res. 72, 1217-1231.
Anderson, J. G and R. Quaas (1988). The Mexico earthquake of September 19, 1985 – Effect of
magnitude on the character of strong ground motion: An example from the Guerrero, Mexico
strong motion network, Earthquake Spectra, Vol. 4, 635-646.
Atkinson, G. M. and D. M. Boore (1990). Recent trends in ground motion and spectral response relations
for North America, Earthquake Spectra, Vol. 6, 15-35.
Boore, D.M., W.B. Joyner and T.E. Fumal (1997). Equations for estimating horizontal response spectra
and peak acceleration from western North American earthquakes: a summary of recent work,
Seism. Res. Lett. 68, 128-153.
Brune, J. N. (1970). Tectonic stress and the spectra of seismic shear waves from earthquakes, J. Geophys.
Res. 75, 4997-5009.
Campbell, K. W. and Y. Bozorgnia (2003). Updated near-source ground-motion (attenuation) relations
for the horizontal and vertical components of peak ground acceleration and acceleration response
spectra, Bull. Seism. Soc. Am. 93, 314-331.
Crouse, C.B. and J.W. McGuire (1996). Site response studies for purpose of revising NEHRP seismic
provisions, Earthquake Spectra, Vol. 12, 407-430.
Field, E.H. (2000). A modified ground motion attenuation relationship for southern California that
accounts for detailed site classification and a basin-depth effects, Bull. Seism. Soc. Am. 90, s209s221.
9
Mohraz, B., 1976, a study of earthquake response spectra for different geological conditions, Bull. Seism.
Soc. Am. 66, 915-936.
Sabetta, F. and A. Pugliese (1996). Estimation of response spectra and simulation of nonstationary
earthquake ground motions, Bull. Seism. Soc. Am. 86, 337-352.
Sadigh, K., C.Y. Chang, J.A. Egan, F. Makdisi, and R.R. Youngs (1997). Attenuation relationships for
shallow crustal earthquakes based on California strong motion data, Seism. Res. Lett. 68, 180189.
Seed, H.B. and Idriss, I. M., 1979, Influence of soil conditions on ground motions during earthquakes,
Journal of the Soil Mechanics and Foundations Divisions, ASCE, 95, 99-137.
Seed, H.B., Ugas, C., and Lysmer, J., 1976. Site-dependent spectra for earthquake-resistant design, Bull.
Seism. Soc. Am. 66, 221-243.
Singh, J. P. (1985). Earthquake ground motions: implications for designing structures and reconciling
structural damage. Earthquake Spectra, Vol. 1, 239-270.
10