Ground Motion Evaluation Procedures of Performance Based Design

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Soil Dynamics and Earthquake Engineering 22 (2002) 765–772

www.elsevier.com/locate/soildyn

Ground motion evaluation procedures for performance-based design


Jonathan P. Stewarta,*, Shyh-Jeng Chioub, Jonathan D. Brayc, Robert W. Gravesd,
Paul G. Somervilled, Norman A. Abrahamsone
a
University of California, 5731 Boelter Hall, Los Angeles, CA 90095, USA
b
California Department of Transportation, Sacramento, CA, USA
c
University of California, Berkeley, CA, USA
d
URS Corporation, Pasadena, CA, USA
e
Pacific Gas and Electric Company, San Francisco, CA, USA

Abstract
The objective of performance-based earthquake engineering (PBEE) is the analysis of performance objectives with a specified annual
probability of exceedance. Increasingly undesirable performance is caused by increasing levels of strong ground motion having decreasing
annual probabilities of exceedance. Accordingly, the evaluation of ground motion intensity measures is a vital component of PBEE. This
paper provides a brief synthesis of ground motion analysis procedures within a performance-based design framework, and is a summary of a
recent report to which the reader is referred for details. The principal topics addressed are probabilistic characterizations of source, path, and
site effects, with the discussion of these effects focusing principally on applications in active regions such as California.
q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Performance-based earthquake engineering; Intensity measure; Probabilistic seismic hazard analyses

1. Introduction appropriate DMs. DVs may include human or collateral


loss, post-earthquake repair time, or other parameters of
The development of a scientific basis for performance- interest to an owner.
based earthquake engineering (PBEE) includes four vital
steps: These four steps in the performance-based design
methodology are linked through the theorem of total
1. Evaluation of the distribution of ground motion intensity probability, and the outcome of the analysis is no better
measures (IMs) at a site, given certain seismological than the weakest (or most uncertain) link in the process.
variables (i.e. fault characteristics, position of site relative Accordingly, high-quality ground motion characterization,
to faults, etc.). IMs may consist of traditional parameters which strives to develop unbiased estimates of IMs with
such as spectral acceleration or duration, or newly defined minimal uncertainty, is a vital component of PBEE.
parameters found to be useful for particular applications. The topics associated with a probabilistic evaluation of
2. Evaluation of the distribution of engineering demand earthquake ground motions in seismically active regions
parameters (EDPs) given a particular set of IMs. These consist of the following:
parameters describe the performance of a structure and
can be calculated from response models (e.g. inter-storey Source characterization: Minimum and maximum mag-
drift for buildings). nitude (m 0 and m u, respectively), magnitude-recurrence
3. Evaluation of the distribution of damage measures (DMs) relations.
given EDPs. DMs describe the physical condition of a Attenuation relations: describe the variation with mag-
damaged element, with information on how the element nitude (m ), distance (r ), and other seismological
would need to be repaired, and any life safety con- variables of spectral acceleration and other IMs averaged
sequences of the damage. over broad site categories (e.g. rock or soil).
4. Evaluation of the probability of exceeding decision Near fault ground motions: rupture-directivity and fling-
variables (DVs) within a given time period, given step effects.
Site effects: shallow ground response, basin response, and
* Corresponding author. topographic effects.
0267-7261/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 7 - 7 2 6 1 ( 0 2 ) 0 0 0 9 7 - 0
766 J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772

Ground motion simulation: generation of synthetic given magnitude range, which enables the development of
seismograms using analytical models. magnitude-area scaling relationships from empirical data
Time history selection: de-aggregation of hazard, time [2,3]. These relationships can be used with known fault
history selection and scaling. dimensions to assign probable values of mu :
As shown in Fig. 2, several types of PDFs for earthquake
An overview of recent developments on these topics is magnitude ½f ðmÞ are used for seismic source zones,
provided here. Readers are referred to Ref. [1] for details. including truncated exponential (modified from Ref. [4]),
That report and this paper provide information on models characteristic earthquake [5], and maximum magnitude
and components of probabilistic seismic hazard analyses models. The choice of an appropriate model is often made
(PSHA). The overall PSHA computational formulation is on the basis of fault size and slip rate, or from observations
not presented here, but can be found in numerous references of historic seismicity. The rate of occurrence of earthquakes
including Ref. [1]. is derived by equating the rate of moment build-up on faults
(derived from slip rate and fault size) to the rate of moment
release [related to f ðmÞ]. Earthquakes can be assumed to
2. Source characterization occur at a fixed rate according to a Poisson process, or at a
time-varying rate that depends on the elapsed time since the
Faults are described as a series of segments (or a single previous event (time-predictable models).
segment) that can either rupture individually or in groups. A California hazard model [6,7] uses an A – B –C fault
A fault segment is characterized by an area (or length), a classification in which A-faults have relatively well-
probability density function (PDF) describing the relative constrained source parameters and large slip rates
likelihood of the fault producing earthquakes of different (s . 5 mm/yr). B-faults include most other faults in the
magnitudes ½f ðmÞ; and a long term slip rate. A fault state with s . 0.1 mm/yr, and have relatively poorly
segmentation model for the San Francisco Bay Area is constrained segmentation models and slip rates. C-faults
shown in Fig. 1. include poorly defined structures and background seismi-
Fault models are constructed so as to only allow moment city. A maximum magnitude model is used for A-faults with
release between m 0 and mu : Minimum magnitude is often dispersion set to zero and mode magnitude (mmax)
taken as m0 ¼ 5; although smaller values may be appropriate determined from magnitude-area scaling relations. B-faults
for stiff, brittle structures. Maximum magnitude is related to are modeled using a hybrid of: (1) a maximum magnitude
the stress drop that occurs on faults during earthquakes and model similar to that for A-faults and (2) a truncated
the size (area) of the fault segment. Stress drop is generally exponential model with m 0 ¼ 6.5 and m u ¼ mmax. Half of
observed to be reasonably consistent within a given tectonic the moment release is accommodated by the maximum
regime (e.g. active regions such as California) and over a magnitude model, and half by the truncated exponential
model. C-faults are modeled using a truncated exponential
distribution, with m0 ¼ 6 and mu # 7:3 for faults associated
with poorly defined structures (e.g. blind thrusts), and
m0 ¼ 5 and mu # 7 for background seismicity.

Fig. 1. Seismotectonic zones for San Francisco Bay region [2]. Fig. 2. PDFs for magnitude, f ðmÞ:
J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772 767

As discussed by Petersen et al. [8], approximately half


the moment release in the above model is associated with
A-faults, and most of the remaining portion with B-faults.
C-faults do not contribute significantly to the overall
moment release in California.

3. Ground motion attenuation relations

The evaluation of seismic hazard requires the use of


probabilistic distributions of IMs conditioned on the
occurrence of an earthquake with magnitude m at distance
r. The PDF for a single IM is written as f ðIMlm; rÞ; and is
usually log – normal. Attenuation relationships define the
statistical moments of these PDFs (e.g. medians, standard
error terms) in terms of m, r, and other seismological
variables. Fig. 4. Attenuation for rock sites of equivalent number of uniform stress
A commonly used IM is spectral acceleration (Sa) at 5% cycles [15].
damping. Attenuation relations for Sa are available for
active regions, subduction zones, and intra-plate regions. earthquakes. These well-recorded events allow quantifi-
Most relations for active regions and subduction zones are cation of intra-event aleatory variability of ground motion
obtained from regression of empirical data. Fig. 3 shows the (random variability within an event), but inter-event
range of m and r sampled in a database for active regions. aleatory variability (random variability across events) is
There are two sampling problems with the inventory. First, less clearly defined. The presence of inter- and intra-event
there are only 82 recordings of large magnitude earthquakes variability affects data weighting for attenuation relations.
(m . 7) at close distance (r , 20 km), and 59 of these are Negligible inter-event variability would allow attenuation
from a single event (m ¼ 7.6 in 1999 at Chi Chi, Taiwan). relations to be developed with equal weight given to each
This range of m and r is critical for many regions, and the data point. Negligible intra-event variability would allow
lack of data leads to significant epistemic uncertainty (i.e. the collective data from each event to be weighted equally.
uncertainty about the proper form of attenuation functions). Since neither source of variability is small, statistical
The second data sampling problem is associated with the regression procedures used in developing attenuation
data set being dominated by a few well-recorded events. Of models weigh data from sparsely and well-recorded events
the 1800 recordings in Fig. 3, 1055 are from eight differently, and the manner in which this is done is a
distinguishing feature of several well-known models
[9 – 13]. Other characteristics distinguishing attenuation
models for Sa include the measure of site –source distance,
and the models for focal mechanism and site condition [14].
Attenuation relationships are also available for other IMs
such as peak horizontal velocity, vertical spectral accelera-
tion, Arias intensity, duration-related parameters, and mean
period. Example results for one of these models is provided
in Fig. 4.
It should be noted that in PSHA computations, aleatory
uncertainty from attenuation models is incorporated into the
solution of the hazard integral. The effects on PSHA
calculations of epistemic uncertainty in attenuation models
can be quantified through the use of logic trees.

4. Near-fault ground motions characteristics

Ground motions in close proximity to the seismic source


can be significantly influenced by rupture-directivity and
fling-step effects, which are shown in Fig. 5.
Fig. 3. Inventory of strong motion recordings from active tectonic regions
Forward directivity occurs when fault rupture propa-
(PEER strong motion database). gates toward a site and the direction of slip on the fault
768 J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772

Fig. 5. Schematic diagram of displacement time histories for strike-slip and


dip-slip faulting in which the fling-step and directivity pulse are shown
together and separately.

is also toward the site. This occurs because the velocity Fig. 7. Comparison of attenuation models for PHV in near fault, forward-
directivity regions, after Ref. [18].
of fault rupture is close to (generally slightly less than)
the shear wave velocity of the rock near the source.
Rupture-directivity affects the duration and long-period best described in the time domain. Models have been
energy content of ground motions, with the amplitude developed that describe time domain features of these
effect being polarized principally in the fault-normal velocity waveforms such as peak velocity, pulse period,
direction as shown in Fig. 6. Motions with forward and number of significant pulses [17 – 19]. Median
rupture-directivity typically exhibit a short-duration, large predictions of peak velocity based on these models are
amplitude pulse in velocity and displacement waveforms, shown in Fig. 7.
whereas motions with neutral or backward directivity Fling-step affects the peak velocity and displacement of
ground motions, and is best described in terms of time-
have long durations and relatively low amplitude. These
domain parameters. Statistical models for the amount of
effects are observed for r , 20– 60 km. Fling-step is
fling-step displacement have been compiled by Wells and
associated with permanent displacement across a ruptured
Coppersmith [3]. Models for the displacement rise time and
fault. Fling-step waveforms are characterized by offset
waveform shape are under development (Abrahamson,
displacements in the slip-parallel direction (Fig. 5), and
personal communication).
large, unidirectional velocity pulses.
Rupture-directivity affects the duration and long-period
energy content of ground motions. Engineering models for
spectral acceleration and duration have been developed by 5. Site effects
Somerville et al. [16] that adjust the results of attenuation
relations given near-fault geometric parameters (e.g. Ground motion attenuation relationships provide esti-
percentage of fault rupturing towards site, site-epicenter mates of IMs that typically apply for broadly defined site
azimuth). However, nonlinear structural response studies conditions such as rock or soil. Actual conditions at strong
have suggested that the most damaging feature of forward motion recording sites are highly variable with respect to
directivity ground motions is the velocity pulse, which is local ground conditions, possible basin effects, and surface
topography, and hence estimates from attenuation relation-
ships represent averaged values across the range of
possible site conditions. Analyses of site effects seek to
reduce the bias and dispersion of ground motion predic-
tions using information about site conditions. Bias is the
average misfit between observed ground motions and
model predictions. Dispersion is a measure of the degree to
which the analysis procedure can capture site-to-site
variations in ground motion.
A fundamental question related to the use of site effects
Fig. 6. Schematic diagram showing the orientations of fling-step and in hazard calculations is the impact that models incorpor-
directivity pulse for strike-slip and dip-slip faulting. ating increasingly detailed site characterizations have on
J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772 769

the bias and dispersion of ground motion predictions. The sediments generally having lower dispersion than older
following sections present recent work on models of sediments
increasing sophistication requiring increasing amounts of
site data to implement. The simplest of these models are 5.2. One-dimensional ground response
amplification factors that require only the identification of
an appropriate site category. More complex, site-specific One-dimensional ground response analyses are a means
models are discussed for 1D ground response and basin by which to adjust ground motion estimates at soil sites using
response. Surface topographic effects are not discussed. detailed information on shallow soil conditions (i.e. these
analyses are seldom performed for sediment depths . 100–
5.1. Amplification factors 200 m). All one-dimensional ground response calculation
methods assume vertical propagation of seismic waves and
Amplification factors represent the ratio of an IM for a horizontal soil layering. What distinguishes different anal-
specified site condition to the value expected for a reference ysis methods is the model for soil behavior, which ranges
site condition. Reference site conditions are typically intact from relatively simple models requiring few input par-
rock or a regional rock average. Amplification factors can be ameters (i.e. equivalent-linear) to more complex, fully
derived from observational data or numerical analyses. nonlinear analyses requiring a greater number of input
Amplification factors have been developed as a function of parameters.
detailed surface geology [20,21], shear wave velocity (Vs) in Ground response analyses are performed with the
the upper 30 m [21 – 23], and geotechnical-based classifi- expectation that accounting for nonlinear sediment response
cation systems [21,24]. Example amplification factors for reduces bias and dispersion in estimated motions at soil sites.
Quaternary alluvium are shown in Fig. 8. Insights into the ‘benefits’ of performing site-specific
Nonlinearity in amplification factors is demonstrated by characterization and analysis work have been developed by
statistically significant dependence on reference site peak Roblee et al. [26] through simulation exercises that evaluated
horizontal acceleration (PHA), and is more pronounced at the variability in soil motions arising from parametric
short than at long periods. Statistically significant variations variability in source, path, and site effects. A stochastic
in amplification factors have been observed at small periods model was used that incorporates a finite source, a range of
between categories such as Mesozoic and Tertiary rock, source-to-site distances, and an equivalent-linear analysis of
Quaternary alluvium, and Holocene lacustrine/marine 1D ground response. Repeated analyses performed with
deposits. Amplification factors have also been found to be randomization of relevant parameters showed that variability
dependent on depth to basement rock, which has been in ground motions arising from variability in soil properties
defined as having Vs ¼ 2.5 km/s [20,23,25]. can overwhelm the parametric variability of source and path
While the above models can significantly affect median for T , , 1 s. These results provide insight into the relative
IM predictions, dispersion of residuals calculated from importance of having information on shallow soil conditions
amplification factors (as measured by standard error terms) when estimating ground motions with simulations.
is not reduced significantly from the dispersion of Stewart and Baturay [27] compared the ability of a soil
attenuation models. However, there are variations in attenuation relation [9] and equivalent-linear ground
dispersion across site categories, with younger and softer response analyses to predict the response spectra of
recorded motions at soil sites. The ground response analyses
utilized large suites of control motions that were scaled to
match a modified rock attenuation median, but retain the
inherent randomness of the suite. By retaining this
variability, ground motions estimated from these response
analyses incorporate the variability in source/path effects
that is present in design situations for a fixed m and r. By
comparing the results to observation, bias and dispersion of
the predictions were evaluated. Dispersion of the residuals
from ground response ðsg Þ and soil attenuation ðsa Þ models
are compared in Fig. 9 for soft clay sites, and indicate
significant benefit of ground response for T , 1 s. This
benefit is moderate to negligible at longer periods and for
stiffer soil conditions.

Fig. 8. Amplification of short- and mid-period spectral acceleration for 5.3. Basin response
Quaternary alluvial sediments as function of peak horizontal acceleration
(PHA) on reference site condition. Results shown include median from
regression analysis and standard deviation of the median (Steidl) and At frequencies around 1 Hz and less, seismic waves can
^95% confidence intervals on the regression (Stewart et al.) have wavelengths longer than typical profile depths for
770 J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772

Fig. 9. Prediction dispersion for soft clay sites from 1D ground response
ðsg Þ and soil attenuation ðsa Þ: After Ref. [27].

ground response calculations, and their amplitudes are


controlled by geological structure having depths of
hundreds or thousands of meters which in many cases,
such as in sedimentary basins, is not horizontally layered.
Basin response analyses provide a means by which to
account for the effects of 2D or 3D deep basin structure on
ground motion characteristics.
It is important when discussing basin response to
separate the amplification associated with 1D wave Fig. 10. Schematic diagram showing that seismic waves entering a
propagation from that associated with the more complex sedimentary layer from below will resonate within the layer and escape if
the layer is flat (left) but become trapped in the layer if it has varying
2D or 3D basin structure. As shown on the on the left side of thickness and the wave enters the layer through its edge (right). Graves [33].
Fig. 10, in the case of 1D wave propagation, the wave that
enters the layer may resonate in the layer, but cannot
become trapped. In the 2D case, if the wave is propagating
in the direction in which the basin is thickening and enters 6. Ground motion simulation
the basin through its edge, it can become trapped within the
basin if post-critical incidence angles develop. The resulting Available data resources are inadequate to constrain
total internal reflection at the base of the layer is illustrated models for a number of important problems such as ground
at the top right of Fig. 10. Similar amplification effects are motions from very large magnitude earthquakes, near-fault
shown for the 1D and 2D cases, however, the trapping of ground motions, and basin effects, as well as ground
body waves in the 2D case generates a surface wave that motions in intra-plate regions. Ground motion simulation
propagates across the basin. These trapped waves have been refers to seismological methods for numerically simulating
observed in recordings from the San Fernando earthquake a rupture process, path effects, and site response effects
[28,29], the 1994 Northridge earthquake [30], and the 1995 within a unified framework. These types of analyses are
Kobe earthquake [31]. attractive because they can be used to generate ground
Recent weak motion studies by Hartzel et al. [32] have motion time histories for the aforementioned situations
shown that in some cases 1D modeling is capable of where actual recordings are sparse.
accounting for observed amplification at intermediate to low Models of the seismic source, wave propagation, and site
frequencies ( f , 2 Hz), whereas in other cases 2D and 3D effects are essential features of a simulation procedure.
models are necessary, particularly when the measure of Several procedures that have been widely used in practice
amplification is sensitive to duration. Use of the 2D and 3D and which have undergone extensive calibration against
models was necessary for locations near a steeply sloping data include the Stochastic Ground Motion model [34], the
basin edge. More studies comparing 1D and 2D/3D composite source method [35,36], and the hybrid green’s
predictions in this manner are needed to clearly quantify function method [37,38]. These procedures have some
the benefit of basin response modeling for ground motion common features. For example, the ground motion from a
prediction. large-magnitude earthquake is computed by summing the
J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772 771

ground motions from many smaller-magnitude earthquakes cal conditions, and that the selected time histories be
called subevents. The models differ in how they model the appropriately scaled.
subevents and how they sum them together (size, location,
and timing of subevents). Comparisons and tests of different
simulation procedures are described in Refs. [39,40]. 8. Conclusions and research needs
Long period motions are more-or-less deterministic in
the sense that theoretical seismology can be used to describe Calibrated procedures for developing probabilistic esti-
the waveforms and spectral content if source, path, and site mates of ground motion IMs are essential to the implemen-
characteristics are known. Features that can be modeled tation of PBEE. This paper and a recent PEER report [1]
include spatial variation of long-period ground motion provide an overview of contemporary procedures that
amplitude, waveforms near the ruptured fault, and basin- enable such calculations.
induced surface waves. These procedures are fairly well Despite recent progress, there remain significant
validated and can provide estimates of ground motions from research needs related to the probabilistic characterization
future earthquakes that are independent of empirical ground of source and path effects. With regard to source effects,
motion models. for many faults in California and elsewhere, critical
At high frequencies, theoretical predictions of ground parameters remain poorly constrained by existing geologic
motion levels are biased with respect to observed data. To data that contribute significant epistemic uncertainty to
produce unbiased estimates of the high-frequency ground hazard calculations. These include maximum magnitude
motions, numerical simulation procedures must add some ðmu Þ; slip rate, and time since the most recent large
form of randomness into the high-frequency ground earthquake. With regard to path effects, two particularly
motions. Different procedures use different approaches for critical needs include (1) data to constrain ground motion
incorporating the randomness. In general, some randomness levels from very large magnitude earthquakes (e.g. from
is added to the process or empirical recordings (which precarious rocks or simulation), and (2) simple models for
include the effects of randomness) are used for high- basin effects that enable such effects to be incorporated
frequency ground motions. into hazard calculations. Research on these issues is being
Since all simulation procedures require random or conducted within the Lifelines and core Programs of the
empirical model components at high frequencies, these PEER center.
aspects of the model require calibration against recorded Taking a broader view of ground motion characteriz-
motions. For this reason, numerical simulations typically do ation for use in PBEE, there is a need to develop IMs that
not do a better job of predicting high-frequency spectral provide optimal correlation to damage in structures.
accelerations than empirical attenuation models for magni- Moreover, since the nonlinear response of many structures
tudes and distances for which there are empirical data. While is correlated to multiple IMs, there is also a need to
simulations may do a better job of extrapolating to develop vector hazard capabilities. These research needs
magnitudes and distances not well represented in the are also being addressed within the PEER center.
empirical database since they have more physics behind
them, they nonetheless suffer from limitations similar to
those of empirical models. Namely, since the high-frequency
Acknowledgments
prediction is based on calibration to empirical data, how do
we know that the randomness in the numerical simulation is
This work was supported by the Pacific Earthquake
appropriate for magnitude-distance pairs outside of the
Engineering Research Center through the Earthquake
empirical data range? Additional calibration of simulation
Engineering Research Centers Program of the National
methodologies against recorded data is needed to answer
Science Foundation under Award number EEC-9701568.
these and other important questions.

7. Time history selection References

[1] Stewart JP, Chiou S-J, Bray JD, Graves RW, Somerville PG,
PSHA produces estimates of IMs for specified annual
Abrahamson NA. Ground motion evaluation procedures for perform-
probabilities of exceedance. For many applications in ance based design. Rpt. No. PEER-2001/09. PEER Center; 2001.
performance-based engineering, the ground motions may [2] Working Group on California Earthquake Probabilities Earthquake
need to be specified not only as IMs, but also by suites of probabilities in the San Francisco Bay Region: 2000 –2030—a
time histories for input into time-domain nonlinear analyses summary of findings. Rpt. OFR 99-517, USGS; 1999.
[3] Wells DL, Coppersmith KJ. New empirical relationships among
of structures. Development of these time histories requires
magnitude, rupture length, rupture width, rupture area, and surface
that the seismic hazard first be de-aggregated to identify the displacement. Bull Seismol Soc Am 1994;84:974– 1002.
most critical ranges of magnitude and distance, that [4] Gutenberg B, Richter CF. Frequency of earthquakes in California.
appropriate time histories be selected for these seismologi- Bull Seismol Soc Am 1944;34:1985–8.
772 J.P. Stewart et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 765–772

[5] Youngs RR, Coppersmith KJ. Implications of fault slip rates and The Loma Prieta, California Earthquake of October 17, 1989-Strong
earthquake recurrence models to probabilistic seismic hazard Ground Motion, USGS Prof. Paper 1551-A; 1994. A77–108.
estimates. Bull Seismol Soc Am 1985;75:939– 64. [23] Field EH. A modified ground motion attenuation relationship for
[6] Frankel AD, Mueller C, Barnhard T, Perkins D, Leyendecker EV, southern California that accounts for detailed site classification and a
Dickman N, Hanson S, Hopper M. National seismic hazards maps. basin depth effect. Bull Seismol Soc Am 2000;90:S209–21.
June documentation, Open File Report. 96-532, USGS; 1996. [24] Rodriguez-Marek A, Bray JD, Abrahamson NA. An empirical
[7] Petersen MD, Bryant WA, Cramer CH, Cao MS, Reichle MS, Frankel geotechnical seismic site response procedure. Earthquake Spectr
AD, Lienkaemper JJ, McCroroy PA, Schwartz DP. Probabilistic 2001;17(1):65–87.
seismic hazard assessment for the State of California, Open File [25] Anderson JG. Expected shape of regressions for ground motion
Report 96-08, CDMG, Open File Report 96-706, USGS; 1996. parameters on rock. Bull Seismol Soc Am 2000;90:S43–S52.
[8] Petersen MD, Cramer CH, Reichle MS, Frankel AD, Hanks TC. [26] Roblee CJ, Silva WJ, Toro GR, Abrahamson NA. Variability in site-
Discrepancy between earthquake rates implied by historic earth- specific seismic ground motion design predictions. Shakelford,
quakes and a consensus geologic source model for California. Bull Nelson, editors. ASCE Geotech Spec Publ No. 58 1996;2:1133.
Seismol Soc Am 2000;90:1117– 32. [27] Stewart JP, Baturay MB. Uncertainties and residuals in ground motion
[9] Abrahamson NA, Silva WJ. Empirical response spectral attenuation estimates at soil sites. Proceedings of Fourth International Conference
relations for shallow crustal earthquakes. Seismol Res Lett 1997; on Recent Advances in Geotechnical Earthquake Engineering and
68(1):94–127. Soil Dynamics, San Diego, CA; 2001. Paper 3.14.
[10] Boore DM, Joyner WB, Fumal TE. Equations for estimating [28] Hanks TC. Strong ground motion of the San Fernando, California
horizontal response spectra and peak acceleration from western earthquake: ground displacements. Bull Seismol Soc Am 1975;65:
North American earthquakes: a summary of recent work. Seismol Res 193 –225.
Lett 1997;68(1):128–53. [29] Liu HL, Heaton TH. Array analysis of the ground velocities and
[11] Campbell KW. Empirical near-source attenuation relations for accelerations from the 1971 San Fernando California earthquake. Bull
Seismol Soc Am 1984;74:1951 –68.
horizontal and vertical components of peak ground acceleration,
[30] Graves RW, Pitarka A, Somerville PG. Ground motion amplification
peak ground velocity, and pseudo-absolute acceleration response
in the Santa Monica area: effects of shallow basin edge structure. Bull
spectra. Seismol Res Lett 1997;68(1):154–79.
Seismol Soc Am 1998;88:1224 –42.
[12] Campbell KW. Erratum to Campbell, 1997. Seismol Res Lett 2000;
[31] Pitarka A, Irikura K, Iwata T, Sekiguchi H. Three-dimensional simu-
71(3):352– 4.
lation of the near-fault ground motion for the 1995 Hyogo-ken Nanbu
[13] Campbell KW. Erratum to Campbell, 2000. Seismol Res Lett 2001;
(Kobe), Japan, earthquake. Bull Seismol Soc Am 1998;88:428 –40.
72(4):474.
[32] Hartzel SA, Carver D, Cranswick E, Frankel A. Variability of site res-
[14] Abrahamson NA, Shedlock KM. Overview (of modern attenuation
ponse in Seattle, Washington. Bull Seismol Soc Am 2000;90:1237 –50.
relationships). Seismol Res Lett 1997;68(1):9–23.
[33] Graves RW. Modeling three-dimensional site response effects in the
[15] Liu AH, Stewart JP, Abrahamson NA, Moriwaki Y. Equivalent
Marina District, San Francisco, California. Bull Seismol Soc Am
number of uniform stress cycles for soil liquefaction analysis.
1993;83:1042–63.
J Geotech Geoenviron Engng 2001;127(12):1017– 26. [34] Silva WJ, Abrahamson N, Toro G, Costantino C. Description and
[16] Somerville PG, Smith NF, Graves RW, Abrahamson NA. Modifi- validation of the stochastic ground motion model. Report to
cation of empirical strong ground motion attenuation relations to Brookhaven National Laboratory, Upton, NY: Associated Univer-
include the amplitude and duration effects of rupture directivity. sities, Inc.; 1997.
Seismol Res Lett 1997;68:199–222. [35] Zeng Y, Anderson JG, Yu G. A composite source model for
[17] Alavi B, Krawinkler H. Consideration of near-fault ground motion computing realistic synthetics strong ground motions. Geophys Res
effects in seismic design. Proceedings of 12th World Conference on Lett 1994;21:725–8.
Earthquake Engineering, New Zealand; 2000. 8 pp. [36] Zeng Y, Anderson JG, Su F. Sub-event rake and random scattering
[18] Rodriguez-Marek A. Near-fault seismic site response. PhD Disser- effects in realistic strong ground motion simulation. Geophys Res Lett
tation. U.C. Berkeley: Civil Engineering Department; 2000. 1995;22:17–20.
[19] Somerville PG. Development of an improved representation of near [37] Somerville PG. Engineering applications of strong ground motion
fault ground motions. Proceedings of SMIP98 Seminar on Utilization simulation. Tectonophysics 1993;218(1–3):195–219.
of Strong Motion Data. CSMIP: Sacramento, CA; 1998. p. 1– 20. [38] Somerville PG, Saikia C, Wald D, Graves R. Implications of the
[20] Steidl JH. Site response in southern California for probabi- Northridge earthquake for strong ground motions from thrust faults.
listic seismic hazard analysis. Bull Seismol Soc Am 2000;90: Bull Seismol Soc Am 1996;86:S115–25.
S149– 69. [39] Abrahamson NA, Becker A. Proceedings of the MCEER Workshop
[21] Stewart JP, Liu AH, Choi Y, Baturay MB. Amplification factors for on Ground Motion Methodologies for the Eastern United States.
spectral acceleration in active regions. Rpt. No. PEER-2001/10, PEER MCEER Technical Report MCEER-99-0016; 1999.
Center; 2001. [40] Hartzell S, Harmsen S, Frankel A, Larsen S. Calculation of
[22] Borcherdt RD, Glassmoyer G. Influences of local geology on strong broadband time histories of ground motion; Comparison of methods
and weak ground motions recorded in the San Francisco Bay region and validation using strong-ground motion from the 1994 Northridge
and their implications for site-specific building-code provisions. earthquake. Bull Seismol Soc Am 1999;89:1484 –504.

You might also like