Short Time Response Spectrum Ground Motion Analysis

Journal of Earthquake Engineering
ISSN: 1363-2469 (Print) 1559-808X (Online) Journal homepage: http://www.tandfonline.com/loi/ueqe20
Ground Motion Analysis with the Use of the Short-

Time-Response-Spectrum
Juan F. Perri & Juan M. Pestana
To cite this article: Juan F. Perri & Juan M. Pestana (2016): Ground Motion Analysis with
the Use of the Short-Time-Response-Spectrum, Journal of Earthquake Engineering, DOI:
10.1080/13632469.2016.1174752
To link to this article: http://dx.doi.org/10.1080/13632469.2016.1174752
Accepted author version posted online: 29

Jun 2016.
Published online: 29 Jun 2016.
Submit your article to this journal
Article views: 1
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=ueqe20
Download by: [Library Services City University London] Date: 30 June 2016, At: 23:20
Ground Motion Analysis with the Use of the Short-
Time-Response-Spectrum
Juan F. Perri1 and Juan M. Pestana2
1
Exponent, Failure Analysis Associates, Oakland, California, USA
t
ip
2
Department of Civil and Environmental Engineering, University of California, Berkeley,
California, USA
Downloaded by [Library Services City University London] at 23:20 30 June 2016
cr
Address correspondence to Juan F. Perri, Exponent, Failure Analysis Associates, 475 14th Street,
Suite 400, Oakland, CA 94612, USA. E-mail: jperri@exponent.com
us
ABSTRACT
an
The elastic response spectrum provides the engineer with two important elements of the seismic
structural response: intensity and frequency content. However, this approach does not provide
M
information on the phasing, number of occurrences of the maxima, and duration. This paper uses
the Short Time Response Spectrum (STRS) methodology to evaluate ground motion time history
ed
characteristics and their implications to engineering analyses. The STRS provides a Time-
Period-Intensity representation of structural response to ground motions and compatible

pt
measures of earthquake duration. Thus, the STRS captures not only the maximum amplitude and
ce
frequency content of the earthquake motion but also its time evolution.
KEYWORDS: response spectrum, ground motion selection, time-frequency, time-period,

Ac
spectral matching, ground motion duration, site response.
1
Introduction
The selection of input earthquake ground motions for numerical analysis is a key element in the
prediction of structural performance during a seismic event. Due to the complex nature of the
t
ip
problem, it is generally difficult to select one or more parameters to accurately describe all of the
important ground motion characteristics [e.g., Joyner and Boore, 1988]. State-of-the-Practice
cr
design methods rely heavily on the use of the elastic response spectrum for assessing the seismic
us
demand on structural elements. The elastic response spectrum provides the maximum response
(e.g., maximum acceleration) for a suite of single degree of freedom (SDOF) systems, but gives
an
no indication of the time at which the maximum value occurs or the number of occurrences of a
given seismic demand. This shortcoming leads to some important practical questions, such as: a)
M
how many times does the response exceed a given threshold during the entire time history?; b)
when does the maximum response occur?; c) if the maximum response occurs at the beginning
ed
of the record and the structure is damaged (i.e., softened), will longer periods exhibit their
maximum values at later times and negatively impact structural performance?; and d) why do
pt
ground motion time histories with relatively similar response spectra result in very different
ce
predicted or measured structural responses?

Ac
These and other similar questions cannot be answered by studying the response spectrum alone.
Some questions could be addressed by analyzing the problem directly in the time domain, but
require significant expertise. In particular, the last question showcases the significant dilemma
regarding the use of synthetic time histories matching a “target” response spectrum
recommended by the code, the use of a suite of “scaled” measured ground motions time histories
2
to match “on average” the target response spectrum, or the use of hybrid approaches for the
selection of design time histories [e.g., ASCE/SEI, 2010].
The non-stationary nature of the ground motions may have a significant effect on the non-linear
response of structures and soils and, as a result, a number of Time-Period-Intensity (TPI)
t
ip
representations have been developed to characterize it.
cr
This paper focuses on the use of the Short Time Response Spectrum (STRS) methodology to
us
extract the temporal evolution of the frequency-intensity content of earthquake ground motions
[Perri et al., 2005] and allow the development of compatible measures of earthquake duration
[Perri and Pestana, 2007]. an

Time-Period-Intensity representation of the ground motion
M
Numerous tools have been developed for the analyses of the non-stationary signals, including the
ed
Short Time Fourier Transform (STFT), the Wavelet Transform (WT), and the Wigner-Ville
Distribution (WVD) [e.g., Cohen, 1995; Auger et al., 1996; Hubbard, 1998]. Similar efforts
pt
have taken place in the field of seismology to analyze earthquake ground motions. Early works
ce
include the Moving Window Analysis [Landisman et al., 1969] and the Multiple Filter
Technique [Dziewonski et al., 1969] for the study of surface waves’ characteristics: group
Ac
velocities, dispersion, and particle motion. Other early applications included the study of seismic
energy release by the application of a Moving Windows Technique [e.g., Trifunac and Brune,
1970]. Trifunac [1971] introduced the Response Envelope Spectrum (RES) to evaluate the non-
stationary response of a single degree of freedom system subjected to a seismic input motion.
3
This methodology was later used by Perez [1980] to develop spectra of amplitudes sustained for
a given number of cycles. Examples of the application of time-frequency analyses of seismic
ground motions include the development of meaningful frequency-dependent duration [e.g.,
Novikova and Trifunac, 1993]. Other developments have focused on the modeling of the non-
t
ip
stationary amplitude and frequency evolution of synthetic ground motions [e.g., Der Kiureghian
and Crempien, 1989; Conte and Peng, 1997]. More recently, Trifunac et al. [2001] studied the
cr
“apparent instantaneous frequency” of buildings, Argoul et al. [2002] studied the “instantaneous”
us
structural damping using Time–Frequency procedures, and Wang et al. [2002] studied the effect
of non-stationary motions on the nonlinear response of a SDOF system and a frame structure.
an
TPI representations have been used to evaluate non-stationary characteristics of ground
amplification and the non-linear transient behavior for soft cohesive deposits [Perri, 2007] and
M
more recently for liquefied soil deposits [Kamagata and Takewaki, 2015a, 2015b]. All these
studies describe a signal’s frequency and amplitude evolution in the time-frequency (or time-
ed
period) domain.
pt
Short Time Response Spectrum (STRS)

ce
Perri et al. [2005] proposed the STRS to obtain a TPI representation of the non-stationary
structural response of a single degree of freedom system for a given ground motion. The
Ac
equation describing the relative motion x(t) of a SDOF oscillator subjected to base acceleration
xg (t ) is given by Equation (1).
x(t ) + 2 ⋅ ζ ⋅ ω o ⋅ x (t ) + ω o2 ⋅ x(t ) = − xg (t )

(1)
4
where ¶ is the fraction of critical damping and É o is the natural frequency of the SDOF
oscillator. There are several methods available for the solution of this differential equation both
in the frequency and time domains, and they can be readily found in the literature [e.g., Nigam
and Jennings 1968, Chopra 2001].
t
ip
Given the response time history of a SDOF oscillator, it is possible to obtain a “local” maximum
cr
by windowing the response time history, x(τ ; T , ζ ) , around a particular instant (t) with a short
time window, h(t − τ ) . Using this procedure, it is possible to determine the maximum of the
us
windowed signal at a given time. By shifting the location of the short time window along the
an
SDOF response time history, the “local” maxima can be recorded, and the time evolution of
these maxima in the overall response is obtained (Equation 2).

M
t
x max (t ; T , ζ , h) = ∫ max x(τ ; T , ζ ) ⋅ h(t − τ ) ⋅ dτ
τ
0 (2)
ed
If the analysis is repeated for all the desired SDOF oscillators with a fixed structural damping, ¶,
pt
and window length and shape, h, the displacement STRS is obtained and is referred to as STRS D .
Using this period proportional window (i.e., h(τ − t ; T ) ), the displacement, pseudo-velocity and
ce
pseudo-acceleration STRS become:

Ac
t
STRS D (t , T ; ζ , h) = ∫ max x(τ , T ; ζ ) ⋅ h(t − τ ; T ) ⋅ dτ
τ
0 (3a)
 2π 
STRSV (t , T ; ζ , h) = ω o ⋅ STRS D =   ⋅ STRS D
 T  (3b)
5
 2π 
2
STRS A (t , T ; ζ , h) = ω ⋅ STRS D = 
2
o  ⋅ STRS D
 T  (3c)
where STRS D , STRS V and STRS A , refer to the TPI representation of the displacement, pseudo-
velocity, and pseudo-acceleration time histories respectively ( t STRS V and t STRS A will be used
t
ip
for the STRS of the total velocity and total acceleration, respectively).
cr
Time resolution of the STRS is controlled by the length of the window used. When a relatively
us
long window is selected, the STRS approaches the response spectrum, resulting in significant
loss of time resolution. If, on the other hand, the window is of very short duration, the STRS will
an
approach the absolute value of the time history response at time t. Perri and Pestana [2007] and
Perri [2007] discussed the limitations of using a window of constant duration and proposed the
M
use of a period-proportional window length. Their results showed that a window length of 5T
gives the best overall response (where T is the natural period of interest), with lengths of 4T to
ed
6T providing very similar results. The authors also analyzed the influence of the window type
(Figure 1): Hanning, Hamming, Gaussian, Blackman, Triangular (i.e., Barlett) and Rectangular
pt
(a summary of their mathematical representation is given by Perri et al. [2005] and further
ce
description of their properties is available in the literature [e.g., Smith, 1997]). Their results
showed that the STRS is relatively insensitive to the window type, as long as the window is well
Ac
localized in time. The first five windows produced almost identical results, but the last one, the
Rectangular window, was not able to accurately localize in time the peaks of the response.
Similarly to the Short Time Fourier Transform (STFT), the Gaussian window slightly improves
the time resolution of the STRS with respect to the Hanning window, but requires the selection
6
of an additional parameter. For mathematical simplicity, the Hanning window was selected for
this work. All analyses presented in this paper use a 5% damping ratio for the response spectrum
and a Hanning window of length 5T for the computation of the STRS. Figure 1 shows
conceptually the response of a SDOF oscillator (i.e., signal) to an input ground motion and the
t
“windowed” signal [ x(τ ; T , ζ ) ⋅ h(t − τ ) ] using a Hanning window.
ip
Figure 2 shows the
representation of the STRS A for the Sunnyvale Colton Avenue recording of the 1989 Loma Prieta
cr
Earthquake. The projection of this three-dimensional figure into the spectral acceleration (S a ) –
us
period (T) vertical plane produces the widely used response spectrum. Therefore, the response
spectrum can be directly obtained from, and is contained within, the STRS. The projection in the
an
time (t) – period (T) horizontal plane produces the recommended representation of the STRS
(filled/color contour plot).

M
Perri [2007] and Perri and Pestana [2007] introduced two period-dependent duration measures
ed
which can be directly obtained from the STRS analyses. The Bracketed Duration Response
Spectrum (BDRS) gives the period-dependent time interval between the first and last
pt
exceedances of a given threshold (e.g., 0.05g), while the Cumulative Duration Response
Spectrum (CDRS) is obtained by computing the cumulative time over which a particular
ce
threshold is exceeded. Therefore, for a given threshold, the spectral value of the CDRS is always
Ac
less than or equal to the BDRS. For most structural analyses, the CDRS is more relevant since
the nonlinear response of a structure is directly related to the duration in which an intensity level
is exceeded. In contrast, the BDRS is more relevant for most site response analyses since the
excess pore pressures generated during shaking do not typically dissipate quickly enough to
consider several peaks in the motion as ‘separate’ events. Perri and Pestana [2007] describe the
7
detailed procedure for the calculation of the BDRS and CDRS and show several example
applications of the two measurements.
Applications of the STRS
t
ip
Several applications of the STRS are presented in this section. These examples, although not
exhaustive, describe some important advantages that Time-Period-Intensity analyses have over
cr
simpler methodologies which provide limited information or more complex analyses that are
us
better suited for research purposes and require significant expertise. As will be shown, the STRS
is a good tradeoff between limitative simplicity and excessive complexity. In addition, the
an
methodology provides a visualization tool to help gain insight into the non-stationary nature of
structural or site response.

M
Spectral Matching
ed
The selection of input ground motions is a key element in the estimation of seismic performance
pt
of engineered structures. In general, the input ground motions are selected from a database of
previously recorded time histories sharing similar characteristics with the design earthquake
ce
scenario(s) such as tectonic environment, source-site distance, earthquake magnitude, directivity
condition and site conditions, among others. In practice, it is very difficult to simultaneously
Ac
satisfy all these similarity requirements. Therefore, numerical simulations and adjustments may
be used, or some of the conditions must be relaxed. Carlton et al. [2015] presented a recent
literature review on the selection of target ground motion parameters for nonlinear site response
analyses, and summarized target measures used for three tectonic environments commonly
8
encountered in US practice: active plate margins, subduction zones and stable continental
regions. The selection of a target response spectrum may use Ground Motion Prediction
Equations (GMPEs) such as the ones recently developed for active plate margins [e.g.,
NGAWest2, Bozorgnia et al., 2014].
t
ip
In the past two decades, the use of spectrally matched records has become increasingly
cr
widespread. The basic idea is to generate a time history with a response spectrum as close as
desired to a target spectrum. Spectral matching requires the modification of the frequency
us
content and phase amplitude of the time history so that the target spectrum is ‘matched’ at all (or
most) spectral periods as recommended by the code for seismic analyses [e.g., ASCE/SEI, 2010].
an
Using spectrum compatible time series can reduce by nearly a factor of three the number of input
M
ground motions for numerical analysis as compared to using scaled time series for a given
variability of the mean of the nonlinear response of structures [Bazzurro and Luco, 2006].
ed
Spectral matching can be performed either in the time or frequency domain. Several algorithms
capable of performing spectral matching are available in the literature and have been
pt
implemented into computer codes: RSPMATCH [Abrahamson, 1998, Al Atik and Abrahamson,
ce
2010] in the time domain and SYNTH [Naumoski, 1985] and RASCAL [Silva and Lee, 1987] in
the frequency domain.

Ac
Currently, verification of the spectral matching procedure is an art requiring significant expertise
and judgment. It is common practice to visually compare the recorded and spectrally matched
time series. The final degree of alteration will be a function of the necessary spectral amplitude
modification, the numerical algorithm used in the matching technique, and the level of desired
9
‘fit’ to the target spectrum, among others. Several quantitative measures of ‘goodness of fit’ can
be used, including the bracketed duration [Bolt, 1969], significant duration [Husid, 1969], the
root-mean-square acceleration, and Arias intensity [Arias, 1970]. Each one of these measures
captures a particular characteristic of the ground motion, and, therefore, different analyses are
t
ip
required. Since all these measures are frequency independent, they are not able to capture the
degree of modification suffered by each component of the waveform as different components of

cr
the time history will be modified to a different degree in order to match the target spectrum.
us
The STRS can be used to quantify the goodness of fit of spectrally matched ground motion time
histories by considering the effect of the time of occurrence of the change introduced by the
an
matching procedure in each one of the waveform components [Perri and Pestana, 2007]. This is
M
important because shifting the time of occurrence of maxima for different structural periods has
significant implications in nonlinear analyses, but it cannot be captured or observed by matching

ed
the response spectrum alone.
To showcase this application, an example ground motion is modified to match a target response
pt
spectrum, and subsequently, the degree of alteration is measured. For this example, the selection
ce
of the original time histories was based on the similitude that the elastic response spectrum had
with the empirical Abrahamson and Silva [1997] attenuation relationship for the station’s
Ac
distance from the source, site conditions, fault type, and earthquake magnitude. Other GMPEs
(i.e., attenuation relationships) can potentially be used without loss of generality. The target
spectrum was specified for 5% damping at 250 periods ranging from 0.01 to 10 seconds. For
periods between 0.01 and 2 seconds, a probability level from the Abrahamson and Silva [1997]
10
attenuation relationship was selected and used as the target spectrum. For periods between 2 and
10 seconds, the values of the target spectrum were selected to be equal to the response spectrum
of the input signal. This was done to prevent the matching procedure from excessively
modifying the long period components, which dominate the velocity and displacement
t
ip
responses. An alternative option consists in performing the matching procedure for up to 2
seconds, without imposing target spectral amplitudes for longer periods. This second procedure
cr
results in significant smoothing of the spectrum of the matched signal for periods larger than 2
us
seconds and was not used. The spectral matching was obtained after three iterations: first by
spectrally matching period components below 1.0 seconds, followed by spectrally matching
an
period components up to 2.5 seconds, and finally by matching the complete range of spectral
values up to 10 seconds. The spectral matching was obtained with the application of the time
M
domain procedure described by Lilhanand and Tseng [1988] and using the computer code
RSPMATCH [Abrahamson, 1998, Al Atik and Abrahamson, 2010].

ed
Figure 3 shows the target acceleration response spectrum together with the response spectra of
pt
the recorded and spectrally matched signals for the Corralitos station record (CLS220) of the
1984 Morgan Hill Earthquake (M w 6.2). The figure shows an excellent matching of the target
ce
spectrum. In this example, the target spectrum was selected as the median minus one-half
standard deviation of the Abrahamson and Silva [1997] attenuation relationship to maintain the
Ac
recorded Peak Ground Acceleration (PGA) of the original record. To have an estimate of the
modification level introduced by the matching procedure, the lower plot in Figure 3 shows the
ratio of the spectral acceleration of the recorded ground motion to the spectral value of the
11
matched time series. To further assess the changes that the spectral matching procedure caused in
the signal, the time series were analyzed directly in the time domain.
Acceleration, velocity, and displacement time histories of the recorded ground motion and
spectrally matched time series are shown in Figure 4. The spectral matching procedure did
t
ip
maintain the general long period (displacement) non-stationary characteristics of the ground
cr
motion. Nevertheless, it is difficult to assess the influence that the matching procedure had on
individual periods and how these changes may affect structural response.
us
The STRS can be used to evaluate the effects of the ground motion modification and its overall
an
effect on the response of the structure. Figure 5 shows the STRS A for the same CLS220 record
and for the spectrally matched time series (plotted in the same Time-Period-Intensity scale).
M
Analysis of these two STRS A shows the goodness of the spectral matching. The overall ‘shape’
(or signature) of the ground motion is maintained in the spectrally matched time series. In fact,
ed
the strong motion is concentrated in the initial portions of the record (between 6 and 12 seconds
from the beginning of the record) as was the case for the original time history. It is also possible
pt
to note that the strong component present in the original record at a period of about T = 0.5 sec
ce
was highly concentrated in time (at about 9 seconds in the time scale) and how its removal from
the matched time history did not significantly affect the non-stationarity characteristics of other
Ac
frequencies in the record. Another interesting point that can be noticed by the use of the STRS,
and which was not possible to observe in the time or frequency domains separately, is that the
matching of very short periods (T < 0.1sec), which in this example consisted of increasing their
intensity, occurs early in the record and tends to coincide with the arrival time of longer periods
12
(T = 0.1 – 0.3sec) which were present in the original ground motion. Projection of these three
dimensional STRS A into the spectral acceleration (S a ) – period (T) plane produces the commonly
used response spectrum (shown in the upper portion of Figure 3 for the recorded and matched
signals for the CLS220 record of the 1984 Morgan Hill Earthquake).
t
ip
Assessment of the Modification Level Caused by the Spectral
cr
Matching Procedure
us
Perri and Pestana [2007] proposed a period dependent intensity measure to quantify the
an
information contained in the STRS. This Cumulative Intensity Spectrum, C I , is defined by the
integration over time of the TPI representation as expressed in Equation 4.

M
∞
C I [TPI ] = ∫ TPI (T , t ) ⋅ dt (4)
0
ed
C I , creates a spectrum able to incorporate the effects of earthquake duration and intensity,
pt
therefore, capturing the three fundamental ground motion characteristics that most engineers are
concerned with: intensity, frequency content, and duration. C I can be calculated integrating the
ce
TPI representation above any given threshold. For certain analyses, it may be more appropriate
to consider the intensity measure above a certain threshold spectral acceleration (e.g., above a
Ac
structural yielding level).
13
This intensity measure can be used to quantify the degree of modification to the original time
series caused by the spectral matching procedure. The normalized difference between the STRS A
of the recorded and the spectrally matched time series can be computed as:
STRS Match (T , t ) STRS Re cord (T , t )
t
∆STRS N (T , t ) = − (5)
ip
PGAMatch PGARe cord
cr
where STRS Match (T,t) and STRS Record (T,t) are the acceleration STRS of the spectrally matched and
us
recorded time series, respectively.
Through the calculation of the Cumulative Intensity Spectrum, C I , for ”STRS N in Equation 5, a
an
spectrum capable of measuring the effect that the matching procedure caused in the recorded
ground motion is obtained ( C I [∆STRS N ] ). Unlike the difference (or the ratio) between the
M
recorded and matched response spectra (shown in the bottom plot of Figure 3), C I [∆STRS N ] is
ed
able to capture the non-stationary (time) effect of the matching procedure.
Figure 6 shows ”STRS N and C I [∆STRS N ] for the example CLS220 record. As expected from
pt
the examination of Figure 5 for this CLS220 record, the wave components which suffered the
ce
largest modifications are those close to periods of T = 0.5 seconds (evident in the C I [∆STRS N ]
plot). It is important to note that components of periods between T = 1.0 – 1.5 seconds also
Ac
suffered important modifications as shown by C I [∆STRS N ] in Figure 6.
To further quantify the effects of the spectral matching procedure, one can compare the time of
occurrence of the maxima for different periods, the Cumulative Duration Response Spectra
14
(CDRS), and the number of cycles above a selected threshold (e.g., 0.05g) for the recorded and
the spectrally matched time series, as shown in Figure 7. The figure shows that for periods
shorter than about T < 0.6sec, the matching procedure maintained all the principal characteristics
of the recorded ground motion. Times of occurrence of the maxima as well as spectral response
t
ip
duration were not significantly different from the original value. This is remarkable given the
significant intensity modification that was required for periods between T = 0.1 and 0.7 seconds.
cr
As shown on Figure 6 for C I [∆STRS N ] , periods longer than T > 0.7 seconds also suffered
us
significant modifications. From Figure 7a it is clear that the time of occurrence of the maxima
for periods between T = 1.0 and 1.5 seconds changed significantly (by up to 10 seconds).
an
Furthermore, the CDRS and number of cycles for this period range were substantially decreased.
M
Ground Motions with Significant Long Period Content
An example response spectrum for a ground motion with a strong long period effect is shown in
ed
Figure 8. The seismogram was recorded at the Gilroy Array #2 station during the 1989 Loma
Prieta Earthquake. The closest distance from the station to the rupture surface (r rup ) is estimated
pt
to be 12.7 km. The station is located on a USGS site type C and at the edge of a steeply-dipping
ce
bedrock interface. The Abrahamson and Silva [1997] attenuation relationship is also shown for a
‘soil’ site (dotted lines are ± one standard deviation). The spectrum shows the large
Ac
amplification of periods of T ≈ 1 to 2 seconds. Large long period components are common in
records with directivity effect, surface waves, and/or site amplification. Although these
phenomena could produce similar characteristic shapes in the response spectrum, the effect that
each one of them has on nonlinear structures or liquefaction analyses are significantly different.
15
Figure 9 shows the STRS A of the same ground motion capturing not only the large peak
observed in the response spectrum, but also shows the period dependent duration of the strong
motion. It is clear from this figure that the strong motion at long periods is concentrated at the
beginning of the record (i.e., 5-10 seconds into the record).
t
ip
To show the importance of the added information captured by the STRS, Figure 10 compares the
cr
pseudo-acceleration and pseudo-velocity response spectra of the G02090 record with the spectra
of the Sunnyvale Colton Avenue (SVL360) record for the same 1989 Loma Prieta Earthquake.
us
The SVL360 station is also located on a USGS site C, but at a distance of 28.8 km. Figure 10
clearly shows the presence of strong long period components in both records with the Sunnyvale
an
record showing strong components between 2 and 4 seconds. Both records have a similar
M
maximum in the pseudo-velocity spectrum of approximately 140cm/sec (although they occur at
different frequencies). By only comparing the two response spectra, one may be lead to believe
ed
that it is the same phenomena causing such large long period amplitudes at both stations.
Figure 11 shows the STRS A of the G02090 and SVL360 records. The difference between the
pt
two ground motions is readily apparent in this representation. The most evident feature is the
ce
later arrival (t > 15 sec) and longer permanence of the long period components in the SVL360
record. This wave trend cannot be attributed to directivity, as is the case in the G02090 record,
Ac
but to the effect of surface waves. The implications for engineering analyses are multiple.
Signals characterized by directivity concentrate most of the energy in a few pulses at the
beginning of the record, while surface waves arrive later and persist for a longer period of time.
16
Because the velocity response spectra amplifies the long period motions and make their effect
more evident, this representation is particularly useful for the study of long period motions. As
shown in Figure 10, both pseudo-velocity response spectra have almost the same maximum
spectral velocity of approximately 140cm/sec, but they occur at different periods; ~1.2sec and
t
ip
~3.0sec for the G02090 and SVL360 records, respectively. On the other hand, their STRS V are
significantly different, as shown on Figure 12. The G02090 record concentrates all the strong
cr
long period energy in a few seconds at the beginning of the record, while the long period
us
component of the SVL360 record is present in the middle and late portions of the record. Similar
to the STRS A , projection of the STRS V (shown in Figure 12) into the spectral velocity (S v ) –
an
period (T) plane produces the response pseudo-velocity spectra, as shown in the lower portion of
Figure 10.
M
Sequence of Component Arrivals
ed
It has been observed that ground motions with similar response spectra can cause significantly
different structural responses. An important influence on the overall response is caused by the
pt
phasing, duration, and arrival order of the different wave components present in the ground
ce
motion as well as the number of high intensity cycles each of these component presents.
The STRS can be used as a tool for understanding and assessing these ground motions
Ac
characteristics. Figure 13 shows the STRS A and response spectra (which can be obtained as the
projection of the STRS A into the S a – T plane) of two different ground motions in which the
arrival order of several components has been highlighted. The two ground motions were
17
recorded at the Gebze station (GBZ270) during the 1999 Kocaeli Earthquake and at the KJM
station (KJM090) during the 1995 Kobe Earthquake.
For both records, the strong ground motion is concentrated over a short period of time (less than
about 10 seconds), and most importantly, the peak arrivals go from longer to shorter periods as
t
ip
time passes. In both cases, the high intensity peaks observed in the response spectra for short
cr
periods occur after the high intensity peaks for longer periods. Figure 14 shows the pseudo-
acceleration response spectrum as well as the time of occurrence of the maxima for different
us
periods and the Cumulative and Bracketed Duration Response Spectra (CDRS and BDRS,
respectively) for the KJM090 recording of the 1995 Kobe Earthquake. Two major peaks are
an
evident in the pseudo-acceleration response spectrum (Figure 14a): one at periods T ≈ 0.40sec
M
and another at T ≈ 0.75sec. Figure 14b shows that the maximum for the longer period peak (T ≈
0.75sec) in the pseudo-acceleration response spectrum occurs more than four seconds before the
ed
occurrence of the shorter period peak (T ≈ 0.40sec).
The implications of the sequence of arrivals on the nonlinear response of a structure are
pt
significant. For a structure with fundamental period in coincidence with an early peak in the
ce
response spectrum, an earthquake with late arrivals of the high-intensity longer period motions
can cause significant damage if the structure goes into the nonlinear state, and the structure’s
Ac
fundamental period increases while the high-intensity longer period motions reach the structure.
Although it is commonly observed that longer period components arrive later in the record, this
is not always the case, as clearly evidenced by the two examples in Figure 13.
18
Figure 15 shows the STRS A for two ground motions recorded during the 1989 Loma Prieta
Earthquake. The ground motion recorded at the Anderson Dam (ADL250) presents
characteristics that are more damaging (to structures in the period range of the maximum) than
the one recorded at the Sago South Surface station (SG3261). The latter presents characteristics
t
ip
similar to the records shown in Figure 13 (GBZ270 and KJM090) where the strong ground
motion is concentrated over a short period of time, and the peak arrivals go from longer to
cr
shorter periods as time passes. On the other hand, the peaks of strong ground motion recorded at
us
the Anderson Dam (ADL250) excite almost the same natural frequency (slightly increasing over
time) for a long time.
an
The ADL250 and SG3261 ground motions can be compared to the one recorded at the Camarillo
M
station (CMR270) during the 1994 Northridge Earthquake. Figure 16 shows the pseudo-
acceleration and pseudo-velocity STRS for this record. The difference in the phasing among all
ed
three records is evident. In particular, the Camarillo station (CMR270) record presents the most
damaging characteristics (this section is concerned with the normalized amplitude response so
pt
that the non-stationary portion of the records can be studied). These damaging characteristics are
evidenced by the longer lasting strong shaking at longer periods of this record (CMR270) when
ce
compared to the previous two 1989 Loma Prieta records (ADL250 and SG3261). Furthermore,
the period of the strong motion peaks tends to increase with increasing time for the CMR270
Ac
record. This is an important factor that would especially influence the response of nonlinear
structures. As a structure is damaged, the degraded period increases and ‘moves towards’ or
‘remains within’ the high intensity portion of the record.
19
It is of paramount importance to determine and define the range of periods of interest in which to
concentrate the study of the ground motion. For site response and structural response analyses,
the range should include the elastic (undamaged) and the degraded (damaged) periods. As
discussed earlier, a complete analysis cannot be limited to the selection and evaluation of the
t
ip
spectral amplitudes for the periods of interest, but it must include the order in which the different
wave components arrive to the site. This information is not contained in the elastic response
cr
spectrum, and, therefore, a more complete description of the ground motion is necessary. This
us
can be achieved, without creating significant complexity, by using the STRS.
The behavior in which longer period components arrive later in the time history is commonly
an
observed in ground motions recorded far from the source and is mainly caused by the slower
M
traveling long period surface waves, dispersion of surface waves, and other phenomena related to
the heterogeneity of the earth. It is not possible to observe this characteristic by analyzing the
ed
elastic response spectrum, and it is not always easy to appreciate this behavior in the time
domain. On the other hand, the combined time-frequency (or time-period) analysis, such as the
pt
STRS, allows recognizing these phenomena in a straightforward manner.

ce
Site Degradation Analyses
Biot presented seminal work describing the propagation of elastic waves in a fluid-saturated
Ac
porous medium [Biot, 1956a, 1956b]. This work is consistently used for the analysis of the
propagation of P-waves in rocks, for which the compressibility of the fluid is in the same order
of magnitude as that of the porous medium. Illustration of recent applications include the
modeling of reservoirs [e.g., Santos et al., 2010], modeling in the mining industry for coal
20
bearing deposits [e.g., Zou at al., 2011] or geophysical interpretation and monitoring of carbon
sequestration [e.g., Moradi and Lawton, 2014]. Carcione et al. [2010] provide a recent review of
the application of poroelasticity for the analysis of waves propagating in a porous medium. The
analysis of vertically polarized shear propagating waves (i.e., commonly referred to as site
t
ip
response analyses) in soil materials, such as sands and clays, has been traditionally assumed to
be through a single medium. These materials show a significantly higher compressibility of the
cr
porous medium as compared with the fluid in the porous space (i.e., high B –value). The
us
simplification of a single medium has been justified as these materials exhibit complex
characteristics such as significant strain dependent nonlinearity, stress induced and evolving
an
anisotropic strength properties among others [e.g., Pestana et al., 2000].
M
In recent years, significant efforts have been directed to the development of advanced
constitutive soil models for seismic applications. Despite these efforts, few of these soil models
ed
are consistently used in practice. The State-of-the-Practice continues to rely heavily on site
response predictions using equivalent linear models, such as those provided by the program
pt
SHAKE [Schnabel et al., 1972]. These one-dimensional site response analyses are typically
performed with simple models empirically derived to fit experimental data.

ce
Two of the main disadvantages of the simplified models are that 1) the predicted stress-strain
Ac
response is decoupled from the pore pressure development and 2) because their predictions are
based on an ‘equivalent linear’ methodology, a unique set of degraded soil conditions is selected
throughout the seismic event. This leads to important shortcomings in the predicted site
response. One of these shortcomings is the inability of equivalent linear models of capturing the
21
evolution of soil stiffness reduction as these models predict instantaneous soil degradation of the
site. As a result, the high frequency components of the ground motions are ‘damped-out’ of the
output surface response predictions.
In order to provide a realistic assessment of the effect that seismic excitation causes in the soil
t
ip
mass, it is necessary to accurately describe the stress-strain-strength relationship under irregular
cr
cyclic loading. To achieve this goal and overcome the limitations of equivalent linear models,
several methodologies and advanced constitutive soil models have been developed and are
us
discussed extensively in the literature (e.g., Hashash et al. [2010]). Advanced soil constitutive
models can be effectively implemented when detailed information on soil behavior is available.
an
When this information is not available, simplified models are used, even in nonlinear time
M
domain site response analysis. Widely used time-domain codes include DESRA [Lee and Finn,
1978], DMOD [Matasovic, 1993], and DEEPSOIL [Hashash et al., 2015].

ed
At the University of California, Berkeley, significant research has been devoted to the
development of advanced soil models for lightly overconsolidated clayey soils [Pestana et al.,
pt
2000; Anantanavanich, 2006; Anantanavanich et al., 2012a; Carlton and Pestana, 2012]. These
ce
models are able to capture the stress-strain-strength characteristics and pore pressure generation
of soils consolidated under level and sloping ground when subjected to multidirectional seismic
Ac
loading with particular application to submarine slopes [Pestana and Nadim, 2000;
Anantanavanich et al., 2012b].
Comparisons of model predictions are generally carried out in the time domain or through
comparisons of the predicted response spectra (frequency domain). Figure 17 shows the ground
22
motion and response spectrum of the Rincon Hill (RC5090) scaled record of the 1989 Loma
Prieta Earthquake. This record is one of the ground motions studied by Anantanavanich [2006]
to analyze the nonlinear effects caused in a soft soil profile by strong shaking. The original PGA
of the recorded motion was nearly PGA ≈ 0.1g, but to study the nonlinear effects of the site
t
ip
response analyses the ground motion was scaled to PGA = 0.2g. Figure 17 also shows a site
response prediction to this input motion for a deep soft soil site composed of 20m of normally
cr
consolidated clay overlaying bedrock. The predictions were computed by two different
us
methodologies: an effective stress constitutive soil model (MSIMPLE DSS) implemented in the
finite element code AMPLE2D [Anantanavanich, 2006] and the linear equivalent code SHAKE
an
[Schnabel et al., 1972]. Detailed site characteristics as well as the complete description of the
analyses are given by Anantanavanich [2006].

M
As shown in Figure 17, both models predict major amplification of the input rock motion at the
ed
site surface for periods longer than T > 0.7sec. There is a larger amount of high frequency
content throughout the time series for the nonlinear soil model (AMPLE2D) predictions when
pt
compared to the surface predictions of the linearly equivalent model (SHAKE) as shown by the
acceleration time series shown in Figure 17. This was noted earlier as one of the limitations of
ce
the equivalent linear models, and is particularly evident in the acceleration prediction given by
Ac
SHAKE, where the initial (for time, t < 7.5sec) high frequency, low amplitude portion of the
signal, which was present in the input ground motion, is absent in the output surface prediction.
Although the early arrival of these low amplitude, high frequency components in the input record
do not significantly degrade the soil, the adoption of an equivalent linear analysis implies an
immediate degradation of the soil, and these components, having a shorter period than the
23
degraded site period, tend to be ‘damped-out’. On the other hand, the fully nonlinear model
implemented in AMPLE2D shows significant amplification of the earlier portion of the record
when compared to the SHAKE predictions. This is partially due to the overestimation of high
frequency components often observed for hysteretic constitutive soil models.
t
ip
Because the initial portion of the input motion has relatively low amplitude, it does not
cr
significantly degrade the soil, and the site period is close to the ‘elastic’ site period (T site ≈ 4H/V S
where H is the depth to bedrock and V S is the average shear wave velocity of the site). As shown
us
in Figure 17, AMPLE2D is able to capture this intensity dependence in the response by
an
predicting an amplification of this frequency in the early portions of the record.
By looking at the time histories in Figure 17, it is evident from both site predictions that the
M
strongest motion arrives between 10 and 14 seconds after the beginning of the record, but it is
not possible to discern the frequency content of the various components in the signal. The
ed
STRS A of the recorded input motion and for both site predictions (SHAKE and AMPLE2D) are
shown in Figure 18 using the same TPI scale. The evolution of the major amplification has been
pt
highlighted. This type of analysis helps to understand the degradation evolution of the site as
ce
time passes.
From the STRS representation of the SHAKE prediction, it is evident that the largest amplitude
Ac
occurs at the degraded site period (T = 1.3sec). Furthermore, this occurs for all time instants. In
contrast, the STRS representation for the AMPLE2D prediction clearly shows the site
degradation evolution. Initially, when high amplitude components have not reached the site,
short periods are amplified. As the stronger seismic waves reach the site, more stiffness
24
degradation of the soil takes place. This causes the degraded site period to increase as
highlighted by the significant amplification of longer periods in the middle portions of the
record. Similar TPI representations have been used recently to describe the transient behavior
during soil liquefaction [Kamagata and Takewaki, 2015a, 2015b].
t
ip
Figure 19 shows the Bracketed Duration Response Spectrum (BDRS) for the recorded input
cr
motion and for both site predictions (SHAKE and AMPLE2D). For comparison purposes,
Figure 19 also shows the Significant Duration, D 5-95 , [Husid, 1969] for all three records (Bolt’s
us
Bracketed Duration [Bolt, 1973] is the value of the BDRS at very short periods). Several
comments can be made regarding these results: AMPLE2D predictions increase both, D 5-95 and
an
Bolt’s Bracketed Duration, when compared to the recorded input motion. SHAKE also predicts
M
an increase in Bolt’s Bracketed Duration (similarly to AMPLE2D), but opposite to the nonlinear
model, it predicts a decrease in D 5-95 when compared to the recorded input motion. This is
ed
particularly interesting because the Arias intensity (I a ) for the SHAKE’s prediction is about 15%
larger than the Arias intensity for prediction by AMPLE2D: I aSHAKE = 0.80 cm/sec and I aAMPLE2D
pt
= 0.67 cm/sec. Therefore, the strong motion predicted by SHAKE is highly concentrated – a
situation that was clearly observed in the STRS representation shown in Figure 18.
ce
When this study is extended to the analysis of the duration spectrum (e.g., BDRS), further
Ac
insights are gained. SHAKE predicts an increase in the duration response for periods longer than
T > 0.8sec. On the other hand, AMPLE2D predicts an increase in the duration for all periods
shorter than T < 4.2sec. In fact, the AMPLE2D time history presents significantly longer
25
duration than the SHAKE prediction for periods below T >1.0sec, while for periods longer than
T >1.0sec, both models predict similar duration (except for T H 2sec).
The STRS can also be used to interpret ground motion records from arrays where bedrock and
soft soil time histories are recorded from the same event. For example, several studies have used
t
ip
bedrock ground motions recorded at Yerba Buena Island (YBI) as input to estimate the
cr
earthquake site response at Treasure Island (TI) in California [Jarpe et al. 1989; Idriss 1990;
Darragh and Shakal 1991; Seed et al. 1991; Rollins et al. 1994]. In such studies it is common to
us
observe longer period, high intensity components on soft soil sites, when compared to the ground
motions recorded at the adjacent bedrock sites. It is, however, often difficult to differentiate
an
between site amplification effects and surface wave effects. Studying the YBI and TI records,
M
Baise et al. [2003] used time domain correlation analyses to show that the rock and soil ground
motions were not highly coherent between the two sites and concluded that YBI was an
ed
inappropriate reference site for TI. The STRS methodology could be used to interpret similar
bedrock/soil sites in an attempt to evaluate the influence of surface waves and site amplification.
pt
Conclusions
ce
This paper used the Short Time Response Spectrum (STRS) to obtain a complete picture of the
Ac
Time-Period-Amplitude structural response to input ground motions. In contrast with the
response spectrum, the STRS includes important information on the phasing, number of
occurrences of the maxima, and duration of the ground motion. Therefore, it is able to capture
and describe the non-stationary characteristics of the ground motion, which are manifested by
26
variations of intensity and frequency content of the signal as a function of time. Furthermore,
because a projection of the three dimensional STRS A into the spectral acceleration (S a ) – period
(T) plane produces the commonly used response spectrum, the response spectrum can be directly
obtained from, and is contained within, the STRS A . Since the time-frequency resolution of the
t
ip
STRS is controlled by the length of the window used, the use of a window length of 5T is
recommended. Similarly, a window with shape that is well localized in time improves the time
cr
resolution of the TPI representation. The Hanning window provides a good compromise
us
between mathematical simplicity and time resolution.
Several example applications of the STRS as a tool for the selection of ground motions for
an
engineering analyses were presented. In the first example, the STRS was used to analyze the
M
effect that the spectral matching procedure has on the ground motion. A methodology which
takes into consideration the effect of the time of occurrence of the change introduced by the
ed
matching procedure in each one of the waveform components was proposed. Furthermore, an
intensity spectrum was introduced and used to quantify the modifications made by the spectral
pt
matching procedure. In the second example, it was shown how the STRS can be used to assess
the type of strong long period ground motions present in the record (e.g., directivity effect,
ce
surface waves, and/or site amplification). The third example application showcased the
importance of the sequence in which various phases of the ground motion are observed over
Ac
time. In fact, significantly different nonlinear responses can be expected from a time series in
which the strong long period components reach the structure later than the short period
components as compared to those time series in which short period components arrive after the
strong long period phases.
27
The last example application involved the analysis of site response and soil degradation. A
comparison between the predictions of equivalent linear and fully nonlinear analyses was made
with the use of the STRS. When the response spectra of the two predictions are compared, small
differences are observed. However, the STRS representation showed that, for the fully nonlinear
t
ip
analyses, the instantaneous soil degradation and the ‘site period’ evolved as the intensity of the
input ground motion changed during the time series, whereas this evolution was not captured by
cr
the equivalent linear analyses. These differences in the predictions, although not readily
us
obtained from the elastic response spectrum, can have significant influence in the nonlinear
response of a structure located at a site experiencing soil degradation.
an
Through the use of the STRS, this paper has put in evidence important insights gained by its use
M
and has shown that the increased amount of information can be presented in a simple and
practical manner. In fact, the STRS can help the analyst decide which ground motions may be
ed
more appropriate for the analyses of a given structure, and help to interpret the results of a
dynamic analysis. By evaluating the time series with the STRS, it is possible to anticipate which
pt
ground motion would cause more damage to a structure and better evaluate the effect of the non-
stationarity nature of ground motions on structural response.

ce
Funding
Ac
This material is based upon work partially supported by the National Science Foundation under
Grant No. CMMI 0201561. This support is gratefully acknowledged.
28
References
Abrahamson, N.A. [1998] “Non-stationary spectral matching program RSPMATCH,” PG&E
Internal Report, February.
t
Abrahamson, N.A., and Silva, W.J. [1997] “Empirical response spectral attenuation relations
ip
for shallow crustal earthquakes.” Seismological Research Letters 68, 94-127.
cr
Al Atik, L., and Abrahamson, N. [2010]. “An improved method for nonstationary spectral
us
matching.” Earthquake Spectra, 26(3), 601-617.
Anantanavanich, T. [2006] “Modeling the Rate Dependent Dynamic Response of Submarine
an
Soft Clay Deposits,” Ph.D. Dissertation, University of California, Berkeley.
Anantanavanich, T., Pestana, J.M., and Carlton, B.D. [2012a] “Development of a Constitutive
M
Model to Describe the Dynamic Response of Soft Submarine Clay Deposits”, Proceedings, 15th
World Conference on Earthquake Engineering, Lisbon, Paper 3823

ed
Anantanavanich, T., Pestana, J.M., and Carlton, B.D. [2012b] “Multidirectional Site Response
pt
Analysis of Submarine Slopes using the Constitutive Model MSimple DSS, Proceedings, 15th
World Conference on Earthquake Engineering, Lisbon, Paper 3825

ce
Argoul, P., Ceravolo, R., DeStefano, A., and Perri, J.F. [2002] “Instantaneous Estimators of
Ac
Structural Damping from Linear Time-Frequency Representations.” 3rd World Conference on
Structural Control, Como, Italy.
Arias, A. [1970] “A Measure of Earthquake Intensity,” R.J. Hansen, ed. Seismic Design for
Nuclear Power Plants, MIT Press, Cambridge, Massachusetts, pp. 438-483.
29
ASCE/SEI [2010] “Minimum Design Loads for Buildings and Other Structures (ASCE/SEI 7-
10).”American Society of Civil Engineers, Reston, Virginia.
Auger, F., Flandrin, A., Goncalves. P., and Lemoine, O. [1996] “Time Frequency Toolbox for
use with Matlab – Tutorial,” On-line publication.
t
ip
Baise, L. G., S. D. Glaser and D. Dreger [2003]. “Site Response at Treasure and Yerba Buena
cr
Islands, California”. Journal Geotechnical and Geoenvironmental Engineering, Vol. 129, No. 6,
pp. 415-426.
us
Bazzurro, P., and Luco, N., [2006]. Do scaled and spectrum-matched near-source records
an
produce biased nonlinear structural responses?, in Proc. of the 8th U. S. National Conf. on
Earthquake Engin., San Francisco, California.

M
Biot, M. A. [1956]. “Theory of propagation of elastic waves in a fluid-saturated porous solid. I.
Low-frequency range.” Journal of the Acoustical Society of America, 28(2), 168-178.

ed
Biot, M. A. [1956]. “Theory of propagation of elastic waves in a fluid-saturated porous solid. II.
Higher frequency range.” Journal of the Acoustical Society of America, 28(2), 179-191.
pt
Bolt, B.A. [1969] “Duration of strong motion.” Proceedings of the 4th World Conference on
ce
Earthquake Engineering, Santiago de Chile, 1304-1315.

Ac
Bozorgnia, Y., Abrahamson, N.A., Al Atik, L., Ancheta, T.D., Atkinson, G.M., Baker, J.W.,
Baltay, A. et al.[2014] "NGA-West2 research project." Earthquake Spectra 30 (3), 973-987.
Carcione J.M., Morency C., Santos J.E. [2010] Computational poroelasticity - A review.
Geophysics. 75(5): A229-A243.
30
Carlton, B.D., and Pestana, J.M. [2012] “Small Strain Shear Modulus of High and Low Plasticity
Clays and Silts,” Proc., 15th World Conference on Earthquake Engineering, Lisbon, Paper 3821.
Carlton, B.D., Pestana, J.M., and Bray, J. [2015] “Selection of Target Ground Motion Parameters
for Nonlinear Site Response Analysis,” Geoengineering Report No. UCB/GE/2015-01,
t
ip
University of California, Berkeley.
cr
Chopra, A.K. [2001] Dynamics of Structures: Theory and Applications to Earthquake
Engineering, 2nd edition, Prentice Hall, Englewood Cliffs, New Jersey, 844 pp.
us
Cohen, L. [1995] Time-Frequency Analysis. Prentice Hall PTR, Englewood Cliffs, NJ.
an
Conte, J.P., and Peng, B.F. [1997] “Fully Nonstationary Analytical Earthquake Ground Motion
Model.” Journal of Engineering Mechanics, ASCE, 123(1), 15-24.

M
Darragh, R. B., and Shakal, A. F. [1991]. “The site response of two rock and soil station pairs to
strong and weak ground motion.” Bull. Seismol. Soc. Am., 81(5), 1885–1899
ed
Der Kiureghian, A. and Crempien, J. [1989] “An evolutionary model for earthquake ground
pt
motion.” Structural Safety, 6(2-4); 235-246.
Dziewonski, A., Bloch, S., and Landisman, M. [1969] A technique for the analysis of transient
ce
seismic signals, Bulletin of the Seismological Society of America, 59(1), 427-444.

Ac
Hashash, Y.M.A., Phillips, C. and Groholski, D. [2010] Recent advances in non-linear site
response analysis, Fifth International Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics, Paper no. OSP 4.
31
Hashash, Y.M.A., Musgrove, M.I., Harmon, J.A., Groholski, D.R., Phillips, C.A., and Park, D.
[2015] “DEEPSOIL 6.1, User Manual”. Urbana, IL, Board of Trustees of University of Illinois at
Urbana-Champaign
Hubbard, B.B. [1998] The world according to wavelets: the story of a mathematical technique in
t
ip
the making, A.K. Peters. Wellesley, MA.
cr
Husid, L.R. [1969] “Características de terremotos. Análisis general”. Revista del IDIEM, 8,
Santiago de Chile, 21-42.
us
Idriss, I. M. [1990]. “Response of soft soils during earthquakes.” H. Bolton seed volume 2
an
memorial symposium proceedings, J. Michael Duncan, ed., Bitech, Vancouver, B.C.
Jarpe, S. P., Hutchings, L. J., Hauk, T. F., and Shakal, A. F. [1989]. “Selected strong- and weak-
M
motion data from the Loma Prieta Sequence.” Seismol. Res. Lett., 60, 167–176.
Joyner, W.B. and Boore, D.M. [1998] “Measurement, characterization, and prediction of strong
ed
ground motion.” Earthquake Engineering and Soil Dynamics II- Recent Advances in Ground
Motion Evaluation Geotechnical Special Publication 20, ASCE 43-102.

pt
Kamagata, S., and Takewaki, I. [2015a]. ‘Non-linear transient behavior during soil liquefaction
ce
based on re-evaluation of seismic records.’ Soil Dynamics and Earthquake Engineering, 71, 163-
184.
Ac
Kamagata, S., and Takewaki, I. [2015b]. ‘Analysis of Ground Motion Amplification during Soil
Liquefaction via Nonstationary Fourier Spectra.’ International Journal of Geomechanics,
C4015002.
32
Landisman, M., Dziewonski, A. and Sato, Y. [1969] “Recent improvements in the analysis of
surface wave observations,” Geophys. J. Astron. Soc. 17, 369-403.
Lee, M. K., and Finn, W. D. L. [1978] DESRA-2, Dynamic effective stress response analysis of
soil deposits with energy transmitting boundary including assessment of liquefaction potential,
t
ip
Soil Mechanics Series, No. 36, Department of Civil Engineering, University of British
Columbia, Vancouver, Canada.
cr
Lilhanand, K. and Tseng, W.S. [1998] “Development and Application of Realistic Earthquake
us
Time Histories Comparable with Multiple Damping Design Spectra.” Proceedings of the Ninth
World Conference on Earthquake Engineering, Tokyo-Kyoto, Japan, Vol.2.

an
Matasovic, N. [1993] Seismic response of composite horizontally-layered soil deposits, Ph.D.
M
Dissertation, University of California, Los Angeles.
Moradi, S., and Lawton, D. [2014]. Model-based Assessment of Seismic Monitoring of CO2 in a
ed
CCS Project in Alberta, Canada, Including a Poroelastic Approach. Energy Procedia, 63, 4305-
4312.
pt
Naumoski, N. [1985] “SYNTH program: generation of artificial acceleration time history

ce
compatible with a target spectrum,” Department of Civil Engineering, McMaster University,
Hamilton, Ontario, Canada.

Ac
Nigam, N.C. and Jennings, P.C. [1968] “Digital Calculation of Response Spectra from Strong-
Motion Earthquake Records,” Earthquake Engineering Research Laboratory, California Institute
of Technology, Pasadena, CA.
33
Novikova, E.I. and Trifunac, M.D. [1993] “Duration of Strong Earthquake Ground Motion:
Physical Basis and Empirical Equations.” Report No. 93-02, Department of Civil Engineering,
University of Southern California.
Perez, V. [1980] “Spectra of amplitudes sustained for a given number of cycles; an interpretation
t
ip
of response duration for strong-motion earthquakes records”, Bulletin of the Seismological
Society of America, v. 70, 1943-1954.
cr
Perri, J.F. [2007] “Assessment of Capacity and Seismic Demand on Axially Loaded Piles in Soft
us
Clayey Deposits.” Ph.D. Dissertation, University of California, Berkeley.
an
Perri, J.F., and Pestana, J.M. [2006] “Analyses of Ground Motions for Seismic Studies using the
Short Time Response Spectrum”, GeoEngineering Research Report No. UCB/GE/2006-01,

M
University of California, Berkeley.
Perri, J.F. and Pestana, J.M. [2007] “Use of the short time response spectrum for selection of
ed
spectrally matched ground motions,” 4th International Conference on Earthquake Geotechnical
Engineering, Thessaloniki, Greece, June 25–28.

pt
Perri, J.F., Pestana, J.M. and Bea, R.B. [2005] “Short Time Response Spectrum,”
ce
GeoEngineering Research Report No. UCB/GE/2005-05, University of California, Berkeley.
Pestana, J.M., Biscontin, G., Nadim, F. and Andersen, K. [2000] "Modeling cyclic behavior of
Ac
lightly overconsolidated clays in simple shear," Soil Dynamics and Earthquake Engineering, 19,
501-519.
Pestana, J.M. and Nadim, F. [2000] “Nonlinear site response analyses of submerged slopes,”
Geotechnical Engng Report No. UCB/GT/2000-04, University of California, Berkeley.
34
Rollins, K. M., McHood, M. D., Hryciw, R. D., Homolka, M., and Shewbridge, S. E. [1994].
“Ground response on Treasure Island.” Loma Prieta earthquake of October 17, 1989: Strong
ground motion and ground failure, R. D. Borcherdt, ed., USGS Prof. Paper P 1551-A, A155-
A168
t
ip
Santos, C. A., Urdaneta, V., Jaimes, G., and Trujillo, L. [2010]. “Ultrasonic spectral and
complexity measurements on brine and oil saturated rocks.” Rock mechanics and rock
cr
engineering, 43(3), 351-359.
us
Schnabel, P.B., Lysmer, J. and Seed, H.B. [1972] “SHAKE -- A computer program for
earthquake response analyses of horizontally layered sites,” Earthquake Engineering Research

an
Center Report No. EERC 72-12, University of California, Berkeley, 104 p.
M
Seed, R. B., Dickenson, S. E., and Idriss, I. M. [1991]. “Principal geotechnical aspects of the
1989 Loma Prieta earthquake.” Soils and Foundations, 31(1), 1 26.

ed
Silva, W.J. and Lee, K. [1987] “WES RASCAL Code for Synthesizing Earthquake Ground
Motions,” US Army Engineer Waterways Experiment Station, Report 24, Misc. Paper S-73-1.
pt
Smith, S.S. [1997] The Scientist and Engineer’s Guide to Digital Signal Processing, California
ce
Technical Publishing.
Trifunac, M.D. [1971] “Response Envelope Spectrum and Interpretation of Strong Earthquake
Ac
Ground Motion.” Bulletin of the Seismological Society of America, 61, 343-356.
Trifunac, M.D. and Brune, J.N. [1970] “Complexity of energy release during the Imperial
Valley, California, earthquake of 1940.” Bulletin of the Seismological Society of America, 60,
137-160.
35
Trifunac, M.D., Ivanovic, S.S. and Todorovska, M.I. [2001] “Apparent Periods of a Building II:
Time-Frequency Analysis,” Journal of Structural Engineering, ASCE, 127(5), 527-537.
Wang, J., Fan, L., Qian, S. and Zhou, J. [2002] “Simulations of non-stationary frequency content
and its importance to seismic assessment of structures.” Earthquake Engineering and Structural
t
ip
Dynamics, 31.4, 993-1005.
cr
Zou, G., Peng, S., Yin, C., Deng, X., Chen, F., and Xu, Y. [2011]. “Modeling seismic wave
propagation in a coal-bearing porous medium by a staggered-grid finite difference method.”
us
Mining Science and Technology (China), 21(5), 727-731.
an
M
ed
pt
ce
Ac
36
Figure 1. Windowing of the Ground Motion Signal: a) original and windowed signal, b)
Rectangular and Bartlett windows, c) Hanning, Hamming, Blackman and Gaussian windows.
t
ip
cr
us
an
M
ed
pt
ce
Ac
37
Figure 2. 3D and 2D Time- Period-Intensity representation of the STRSA (Sunnyvale Colton
Avenue 360, 1989 Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
38
Figure 3. Acceleration response spectra for the recorded and spectrally matched time series of
the CLS220 record (1984 Morgan Hill Earthquake) [after Perri and Pestana, 2007].
t
ip
cr
us
an
M
ed
pt
ce
Ac
39
Figure 4. Time histories for the recorded and spectrally matched time series of the CLS220
record (1984 Morgan Hill Earthquake) [after Perri and Pestana, 2007].
t
ip
cr
us
an
M
ed
pt
ce
Ac
40
Figure 5. Acceleration STRS for the recorded and spectrally matched time series for the CLS220
record (1984 Morgan Hill Earthquake) [after Perri and Pestana, 2007].
t
ip
cr
us
an
M
ed
pt
ce
Ac
41
Figure 6. Cumulative Intensity Spectrum and Time-Period-Intensity representation of the
?STRSN for the CLS220 record (1984 Morgan Hill Earthquake) [after Perri and Pestana, 2007].
t
ip
cr
us
an
M
ed
pt
ce
Ac
42
Figure 7. Acceleration Duration Response Spectra for the CLS220 record: a) Time of maxima
occurrence, b) Cumulative Duration Response Spectra (CDRS), c) Number of cycles above
threshold (1989 Loma Prieta Earthquake) [after Perri and Pestana, 2007].
t
ip
cr
us
an
M
ed
pt
ce
Ac
43
Figure 8. Computed response spectrum and attenuation relationship for the G02090 records
(1989 Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
44
Figure 9. Acceleration STRS for the G02090 record (1989 Loma Prieta Earthquake) [after Perri
et al., 2006].
t
ip
cr
us
an
M
ed
pt
ce
Ac
45
Figure 10. Acceleration and velocity response spectra for the G02090 and SVL360 records (1989
Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
46
Earthquake).
Ac
ce
pt
ed
47
M
an
us
cr
ip
Figure 11. Acceleration STRS for the G02090 and SVL360 records (1989 Loma Prieta
t
Ac
ce
pt
ed
48
M
an
us
cr
ip
Figure 12. Velocity STRS for the G02090 and SVL360 records (1989 Loma Prieta Earthquake).
t
Figure 13. Acceleration STRS for the GBZ270and KJM090 records. (1999 Kocaeli Earthquake
and 1995 Kobe Earthquake, respectively).
t
ip
cr
us
an
M
ed
pt
ce
Ac
49
Figure 14. Acceleration response and duration spectra for the KJM090 record (1995 Kobe
Earthquake): a) Acceleration response spectrumb) Time of maxima occurrence,c) Cumulative
and Bracketed Duration Response Spectrum (CDRS and BDRS).
t
ip
cr
us
an
M
ed
pt
ce
Ac
50
Figure 15. Acceleration STRS for the ADL250 (top) and SG3261 (bottom) records (1989 Loma
Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
51
Earthquake).
Ac
ce
pt
ed
52
M
an
us
cr
ip
Figure 16. Acceleration and velocity STRS for the CMR270 record (1994 Northridge
t
Figure 17. Acceleration response spectra and time histories for the RC5090 record and the
surface site predictions by AMPLE2D and SHAKE (1989 Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
53
Figure 18. Acceleration STRS for the RC5090 record and the surface site predictions by
AMPLE2D and SHAKE (1989 Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
54
Figure 19. Cumulative Duration Response Spectra (CDRS) for the RC5090 record and the
surface site predictions by AMPLE2D and SHAKE (1989 Loma Prieta Earthquake).
t
ip
cr
us
an
M
ed
pt
ce
Ac
55

Short Time Response Spectrum Ground Motion Analysis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Short Time Response Spectrum Ground Motion Analysis

Uploaded by

Copyright:

Available Formats

Journal of Earthquake Engineering

ISSN: 1363-2469 (Print) 1559-808X (Online) Journal homepage: http://www.tandfonline.com/loi/ueqe20

Ground Motion Analysis with the Use of the Short-

Juan F. Perri & Juan M. Pestana

To link to this article: http://dx.doi.org/10.1080/13632469.2016.1174752

Accepted author version posted online: 29

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

Period-Intensity representation of structural response to ground motions and compatible

KEYWORDS: response spectrum, ground motion selection, time-frequency, time-period,

spectral matching, ground motion duration, site response.

predicted or measured structural responses?

selection of design time histories [e.g., ASCE/SEI, 2010].

response of structures and soils and, as a result, a number of Time-Period-Intensity (TPI)

[Perri and Pestana, 2007]. an

a given number of cycles. Examples of the application of time-frequency analyses of seismic

ground motions include the development of meaningful frequency-dependent duration [e.g.,

Short Time Response Spectrum (STRS)

xg (t ) is given by Equation (1).

x(t ) + 2 ⋅ ζ ⋅ ω o ⋅ x (t ) + ω o2 ⋅ x(t ) = − xg (t )

and Jennings 1968, Chopra 2001].

these maxima in the overall response is obtained (Equation 2).

pseudo-acceleration STRS become:

(filled/color contour plot).

applications of the two measurements.

Applications of the STRS

structural or site response.

scenario(s) such as tectonic environment, source-site distance, earthquake magnitude, directivity

NGAWest2, Bozorgnia et al., 2014].

the frequency domain.

degree of modification suffered by each component of the waveform as different components of

significant implications in nonlinear analyses, but it cannot be captured or observed by matching

the response spectrum alone.

RSPMATCH [Abrahamson, 1998, Al Atik and Abrahamson, 2010].

integration over time of the TPI representation as expressed in Equation 4.

structural yielding level).

STRS Match (T , t ) STRS Re cord (T , t )

able to capture the non-stationary (time) effect of the matching procedure.

suffered important modifications as shown by C I [∆STRS N ] in Figure 6.

amplification of periods of T ≈ 1 to 2 seconds. Large long period components are common in

beginning of the record (i.e., 5-10 seconds into the record).

station (KJM090) during the 1995 Kobe Earthquake.

occurrence of the shorter period peak (T ≈ 0.40sec).

time) for a long time.

‘remains within’ the high intensity portion of the record.

STRS, allows recognizing these phenomena in a straightforward manner.

Site Degradation Analyses

performed with simple models empirically derived to fit experimental data.

output surface response predictions.

1978], DMOD [Matasovic, 1993], and DEEPSOIL [Hashash et al., 2015].

Anantanavanich et al., 2012b].

analyses are given by Anantanavanich [2006].

frequency components often observed for hysteretic constitutive soil models.

during soil liquefaction [Kamagata and Takewaki, 2015a, 2015b].

T >1.0sec, both models predict similar duration (except for T H 2sec).

Time-Period-Amplitude structural response to input ground motions. In contrast with the

strong long period phases.

response of a structure located at a site experiencing soil degradation.

stationarity nature of ground motions on structural response.

Grant No. CMMI 0201561. This support is gratefully acknowledged.

Abrahamson, N.A. [1998] “Non-stationary spectral matching program RSPMATCH,” PG&E

Internal Report, February.