Engineering Structures 78 (2014) 1727

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Displacement damping modication factors for pulse-like and ordinary

records
Fabrizio Mollaioli , Laura Liberatore, Andrea Lucchini
Department of Structural Engineering and Geotechnics, Sapienza University of Rome, Via Gramsci, 53, 00197 Rome, Italy

a r t i c l e

i n f o

Article history:
Available online 15 August 2014
Keywords:
Displacement spectra
Damping modication factor
Prediction equation
Pulse-like records
Far-fault records

a b s t r a c t
In seismic codes, elastic response spectra are usually dened by assuming a conventional value for the
critical damping ratio equal to 5%. damping modication factors (DMFs), i.e. scaling factors, are then
applied to account for the effect of damping values higher or lower than the nominal 5%. Usually,
code-mandated DMFs depend neither on ground motion characteristics nor on structural properties.
However, the inuence of such factors on the DMF was highlighted by different studies.
In this paper, records from 110 near-fault pulse-like ground motions and 224 ordinary ground motions
are used to calculate elastic displacements and DMF spectra corresponding to different values of the
damping ratio ranging from 2% to 50%. The effect on DMFs of pulse period of the ground motion,
earthquake magnitude, site-to-source distance, and period of vibration of the structure is discussed. By
rotating the pulse-like records according to different directions with respect to the fault, including the
fault-normal and the fault-parallel one, the inuence of the angle of rotation is also investigated. Based
on results of regression analyses, equations for the prediction of the DMF for near-fault pulse-like ground
motions are nally proposed.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
The development of performance-based seismic engineering
brought to a growing interest in the denition of displacement
response spectra (e.g. see [14]). In the performance-based philosophy, in fact, the design criteria are expressed in terms of achieving
different performance objectives for different levels of seismic
hazard [5,6]. Such objectives may be related to damage levels
which in turn may be associated to displacement demands. In this
case, differently from conventional force-based seismic design
procedures, seismic actions are dened using displacement spectra
rather than pseudo-acceleration spectra. If an equivalent linear
system is used to model the structure (e.g. see [79]), then an
equivalent damping value and effective period of vibration must
be identied and the displacement demand calculated with a simple
elastic spectrum. Moreover, in case of structures protected with
base isolation systems or supplementary damping devices, response
spectra corresponding to high damping levels have to be dened.
Even though the energy dissipation characteristics of isolation
and damping devices may not be ideally viscous, also for this type
of structures equivalent linear models can be used to evaluate with
different degrees of accuracy the seismic demand [10].
Corresponding author. Tel.: +39 06 49919186.
E-mail address: fabrizio.mollaioli@uniroma1.it (F. Mollaioli).
http://dx.doi.org/10.1016/j.engstruct.2014.07.046
0141-0296/ 2014 Elsevier Ltd. All rights reserved.

Elastic spectra with damping ratios different from 5% are usually derived from the conventional 5%-damped response spectrum
by applying a simple scaling factor that is usually named damping
modication factor (DMF). Starting from the 80s, many different
equations have been proposed for the DMF (e.g. [1113]), and some
of them have been also adopted in seismic code provisions and
guidelines, as highlighted by Lin et al. [14] and Cardone et al.
[15]. Often, the DMF is given by codes as a function of the damping
ratio only. However, various studies showed that different parameters, e.g. period of vibration, magnitude of the earthquake, site
conditions and distance from the fault, may affect, to different
extents, the DMF [1618].
It is well known that at locations close to the fault, forward
directivity may produce large-amplitude velocity pulses, which
may affect the response of structures [1925]. In particular,
Priestley [26] observed that in the presence of velocity pulses the
effectiveness of damping might be reduced. On the other hand,
Hatzigeorgiou [27] found that DMF evaluations using near-fault
and far-fault ground motions lead to similar results. In the present
study, a review of the state of the art on DMF is made and the main
parameters inuencing its value are identied. Then, records from
110 near-fault pulse-like ground motions and 224 ordinary records
are used to calculate elastic displacement spectra and DMFs corresponding to seven different values of damping ratio equal to 2%,
5%, 10%, 20%, 30%, 40% and 50%. The inuence on DMF spectra of

18

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

pulse period of the ground motion (for pulse-like records), earthquake magnitude and site-to-source distance (for ordinary
records), and period of vibration is discussed. For the case of the
pulse-like records, components corresponding to different angles
with respect to the fault direction are considered. Finally, based
on results of regression analyses, an equation for the estimation
of the DMF for pulse-like ground motions is proposed, and predictions are compared with those obtained using models developed
by other researchers.

where D is the duration in seconds estimated as a function of magnitude and focal distance.
More recently Bommer and Mendis [17] observed that the inuence of magnitude and distance may be taken into account by considering the effect of the duration (or of the number of cycles).
Based on this observation Stafford et al. [30] developed the following model to evaluate in the period range 1.53.0 s the DMF as a
function of duration or number of cycles:

DMFn; x 1 
2. Factors inuencing the DMF
Reduction of spectral ordinates due to damping is inuenced by
various factors [28]. In this section, a state of the art review on the
main parameters that have been found to affect the DMF is
reported. The inuence on DMF of period of vibration of the structure, earthquake magnitude, ground motion duration and number
of cycles, distance to the fault, site condition and near-fault condition is presented and discussed here.
In the studies cited below, the DMF is estimated based on displacement or pseudo-acceleration spectra, with the exception of
the study of Lin and Chang [16] in which both pseudo-acceleration
and acceleration spectra are considered, and that of Hatzigeorgiou
[27] where DMFs are evaluated from acceleration, velocity and displacement spectra. More insight into this issue is provided at the
beginning of Section 4.
2.1. Period of vibration
Due to the specic properties of the elastic response spectra, at
very short and very long periods the effect of viscous damping is
not signicant. At very short periods, in fact, the pseudo-acceleration response of an elastic SDOF system tends to the Peak Ground
Acceleration (PGA) whereas the displacement response points
toward zero; at long periods, instead, the displacement response
tends to the Peak Ground Displacement (PGD) while the pseudoacceleration reduces almost to zero. Therefore, it is expected that
in the range of very short and very long periods the DMF value
tends to unity. On the contrary, the most signicant inuence of
damping is in the intermediate period range.
Several studies (e.g. [17]) report that at periods ranging between
approximately 1 s and 3 s the DMF varies slightly. In general, the
range within which the DMF value strongly depends on the period
extends with the increase of the damping ratio. Cameron and Green
[18] estimated that for rock sites, magnitude in the range of 56,
and period of vibration equal to 0.5 s, the mean value of the DMF
is equal to 0.961 and 0.808 for n = 7% and 30%, respectively, whereas
it is equal to 0.935 and 0.643 for a period of 1.5 s. This nding indicates that for small to medium magnitudes the inuence of period
on the DMF should be accounted for, at least for higher damping
values. For larger earthquakes, the effect of period is less marked
especially in the intermediate period range.

2.3. Site conditions

The effect of site conditions was deeply investigated by
Cameron and Green [18]. They considered two different databases:
one representative of earthquake motions in active seismic regions
and one representative of motions in stable continental regions.
The accelerograms used in the study were grouped according to
the site classication (rock or deep soil). For motions in active seismic regions little inuence of site condition on DMF was found at
all the investigated periods (ranging from 0 s to 10 s), whereas a
greater effect was noted for motions in stable continental regions
but only at periods T 6 0.2 s. In the latter case, the effect of soil
conditions increases with increasing magnitude.
According to Hatzigeorgiou [27], the DMF values for soil types B,
C and D (corresponding to very dense, stiff and soft soil, respectively, in accordance with the USGS site classication system) are
very similar, whereas for soil type A (rock) they are different.

1.4
= 2%
1.2
Rosenblueth (1980), T=1.5 s
1.0

DMFn; T; D

DMF

The strong inuence of magnitude on the DMF was recognized

by different researchers. In general, the effect of damping is more
pronounced for large earthquakes [17]. However, this trend is
well-dened only for periods greater than 0.5 s, whereas for
shorter periods the opposite may occur [18]. An indirect correlation between DMF, magnitude and focal distance is given by the
following equation suggested by Rosenblueth [29]:

In Eq. (2), the variable x is either the signicant duration or the

number of cycles, and bi are coefcients whose value depends on
the used predictor variable. A comparison between the DMF calculated with Eq. (1) (for T = 1.5 s) and Eq. (2) (with x equal to the signicant duration from 5% to 75% of the Arias Intensity) is presented
in Fig. 1. Even though the denition of duration is different in the
two models, a very good match can be noted when the damping
ratio is equal to 10%; differences between the two models increase
with increasing damping. In both cases the inuence of duration
seems to be negligible when the duration is greater than about 20 s.
As observed by Stafford et al. [30], the duration of motion (or
the number of cycles) is an important parameter affecting the
DMF. However, the direct inclusion of duration in a prediction
model is not practical because duration is not generally specied
in earthquake design scenario. Therefore it seems more convenient
to capture the effect of duration implicitly, through the use of other
predictor variables such as magnitude and distance [31].

2.2. Magnitude, duration and distance from the fault

0:41
1 4:93n DT
0:41
4:93  0:05 DT

b1 b2 lnn b3 lnn2
h
i
4
1 exp  lnxb
b5

Stafford et al. (2008)

= 10%

0.8
= 30%

0.6
0.4
0.2
0

10

20

30

40

50

Duration (s)

Fig. 1. Inuence of duration on DMF, comparison between Eq. (1) (Rosenblueth

[29]) and Eq. (2) (Stafford et al. [30]).

19

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

Specically, in the short periods range the DMF value on rock soil is
lower compared to that observed for the other types of soil, while
at higher periods the trend is the opposite. Different results were
obtained by Lin and Chang [16]. Using a database of 1037 accelerograms classied in accordance to NEHRP (FEMA [32]) for the soil
conditions, they found that DMFs for site classes AB (rock) and D
(stiff soil) are very similar, whereas DMF values for site class C
(very dense soil) are generally slightly higher. A negligible dependence on site conditions was found by Rezaeian et al. [31] for
T = 1 s. Differences between the ndings of the mentioned studies
may be probably explained by the different site classications and
the different criteria used to select the ground motions. However,
based on these results it is reasonable to conclude that the effect
of site conditions is in general signicant only for rock sites. Finally,
based on the DMF values reported in Lin and Chang [16] it seems
that the effect of site conditions is less signicant than that of magnitude and period of vibration when the DMF is calculated for displacement spectra.
2.4. Pulse-like effects
Large-amplitude velocity pulses, which may be observed at
sites close to the fault rupture, are expected to decrease the effectiveness of damping in reducing the structural response. For this
reason Priestley [26] proposed the following expression for the
DMF in near-fault regions:


DMFn

10
5n

0:25
3

which modies the equation included in the Eurocode 8 (EC8, [33]):

DMFn

10
5n

0:5

P 0:55

Bommer and Mendis [17] showed, using a single pulse-like

record, that Eq. (3) may be conservative for periods close to the
duration of the pulse but it is adequate for the remaining part of
the spectrum.
Recently, Hatzigeorgiou [27] focused on DMFs for SDOF systems
subjected to near-fault records analyzing both far-fault and nearfault records (not necessarily pulse-like) from different types of
fault. The prediction equation for the DMF in near-fault regions
proposed in Hatzigeorgiou [27] is derived from a database of 110
accelerograms with closest distance to the fault less than or equal
to 10 km recorded during earthquakes of magnitude ranging
between 5.5 and 7.8. In such equation, the DMF is dened as a
function of both damping ratio and period of vibration. For the
equation coefcients different (but very similar) sets of values corresponding to different types of soil are reported. The effect of
magnitude is not accounted for.
The effect of pulse period, Tp, on DMF has not been studied yet.
Different studies, e.g. [3437], have shown that a ground motion
with a distinct velocity pulse tends to increase the elastic spectral
response in a narrow period range of the spectra close to the pulse
period. The inuence of Tp on inelastic displacement ratios has
been investigated by Ruiz-Garcia [38] and Iervolino et al. [39] concluding that an adequate characterization of such ratio should
explicitly take into account the period of the pulse.

forward-directivity, and the other group is composed by ordinary

ground motions. The ensemble of selected near-fault ground
motions contains records identied in other studies [20,24,
4144] as characterized by forward-directivity. Based on what
reported in Rodriguez et al. [45] and in Gillie et al. [43], records
from the Bam earthquake (2003) and the Parkeld earthquake
(2004) are also added in the set. The near-fault forward directivity
pulse observed in Bam is fully described in Ghayamghamian and
Hisada [46], while the ground motion records from Parkeld that
exhibit forward-directivity effects are discussed in Gillie et al. [43].
A complete list of the earthquakes from where these near-fault
pulse-like ground motions were recorded is reported in Table 1.
The selected ground motions were recorded in regions prone to
forward directivity effects and satisfy the geometric conditions
for forward directivity dened in Somerville et al. [40]. In addition,
the velocity time histories of these ground motions show a polarization in the fault-normal direction and are characterized by a
clear pulse in the fault-normal direction. Due to the fact that the
DMF spectra are calculated up to 8 s and that the pulse-like effects
are signicant only at periods around the period of the pulse, only
records with a pulse period lower than about 9 s were selected.
About the soil conditions at recording stations, with the exception
of 4 ground motions recorded on soil type B (refer to the NEHRP
site classication based on the VS30 value), all the others ground
motions were recorded on soil type C or D. The two horizontal
components of the ground motions were projected in the faultparallel (FP) and the fault-normal (FN) direction. Even though the
radiation of shear dislocation on the fault causes the large directivity pulse to be oriented in the direction perpendicular to the fault
[35], for the sake of comparison the DMF has been evaluated also
for the fault-parallel component of the motion.
The set of ordinary ground motions consists of 224 ordinary
records, where the term ordinary indicates the lack of impulsive
characteristics independently on the distance to the fault. These
records are divided into two groups according to their site-tosource distance value. The ranges of values for the site-to-source
distance considered are the following: 2040 km (group FA, 114
records, Table 2) and 45100 km (group FB, 110 records, Table
Table 1
Info about the near-fault pulse-like records used in the study.

3. Ground motion database

The records used in this study to calculate the DMF are selected
from the NGA database (http://peer.berkeley.edu/nga/), and are
from shallow crustal earthquakes occurred in active tectonic
regions. They are organized into two main groups: one group
consists of near-fault pulse-like ground motions characterized by

Earthquake name

Year

Magnitude

Number of
ground motionsa

Yountville
Coyote Lake
Coalinga-05
San Salvador
Westmorland
Whittier Narrows-01
Parkeld
N. Palm Springs
Northwest China-03
Morgan Hill
Chi-Chi, Taiwan-03
Managua, Nicaragua-01
Bam, Iran
Imperial Valley-06
Superstition Hills-02
Erzican, Turkey
Northridge-01
Gazli, USSR
Irpinia, Italy-01
Kobe, Japan
Loma Prieta
Cape Mendocino
Duzce, Turkey
Landers
Kocaeli, Turkey
Chi-Chi, Taiwan

2000
1979
1983
1986
1981
1987
2004
1986
1997
1984
1999
1972
2003
1979
1987
1992
1994
1976
1980
1995
1989
1992
1999
1992
1999
1999

5.00
5.74
5.77
5.80
5.90
5.99
6.00
6.06
6.10
6.19
6.20
6.24
6.50
6.53
6.54
6.69
6.69
6.80
6.90
6.90
6.93
7.01
7.14
7.28
7.51
7.62

1
1
3
2
2
3
10
2
1
2
3
1
1
13
2
1
11
1
1
5
5
2
2
2
3
30

Each ground motion has a fault-normal and a fault-parallel component.

20

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

3). Given that near-fault pulse-like ground motions were recorded

mainly on soil type C or D, only records characterized by soil conditions type C or D are selected also for the ordinary records.
A nal summary of the considered sets of records is given
below.

Table 3
Info about the ordinary records group FB (45 6 Dfa 6 100) used in the study.

FN (110 records): near-fault pulse-like ground motions, with

impulsive characteristics due to forward directivity. In this
dataset, only the fault-normal component of the motions is
included.
FP (110 records): fault-parallel components of the ground
motions in the dataset FN.
FA (114 records): ordinary records (i.e. not presenting impulsive
characteristics) with distance to the fault ranging between 20
and 40 km.
FB (110 records): ordinary records with distance to the fault
ranging between 45 and 100 km.
4. Displacement damping modication factor
In this section of the paper, the effect of damping on singledegree-of-freedom systems subjected to the pulse-like and the
ordinary records described in Section 3 is investigated. Elastic displacement spectra Sd(T, n) are estimated for different values of the
damping ratio n (namely, 2%, 5%, 10%, 20%, 30%, 40% and 50%), and
the DMF spectra are calculated as follows:

DMF

Sd T; n
Sd T; 0:05

As a matter of fact, DMFs obtained from displacement spectra

are identical to those derived from pseudo-velocity and pseudoacceleration spectra. The pseudo-acceleration spectrum, instead,
is equal to the acceleration spectrum only if the damping ratio is
zero, while the pseudo-velocity spectrum is never equal to the
velocity spectrum. If DMFs obtained from displacements and DMFs
derived from accelerations are compared, it can be observed that
up to a 10% damping ratio the differences are negligible. On the
contrary, for higher values of the damping ratio the use of displacement (or pseudo-acceleration) instead of acceleration can lead to

Table 2
Info about the ordinary records group FA (20 6 Dfa 6 40) used in the study.

Earthquake name

Year

Magnitude

Number of records

Whittier Narrows-02
Northridge-06
Sierra Madre
Coyote Lake
Chi-Chi, Taiwan-02
Whittier Narrows-01
N. Palm Springs
Morgan Hill
Chalfant Valley-02
Chi-Chi, Taiwan-03
Chi-Chi, Taiwan-04
Chi-Chi, Taiwan-05
Chi-Chi, Taiwan-06
Coalinga-01
Friuli, Italy-01
Imperial Valley-06
Superstition Hills-02
San Fernando
Northridge-01
Kobe, Japan
Loma Prieta
Landers
Kocaeli, Turkey
Chi-Chi, Taiwan

1987
1994
1991
1979
1999
1987
1986
1984
1986
1999
1999
1999
1999
1983
1976
1979
1987
1971
1994
1995
1989
1992
1999
1999

5.27
5.28
5.61
5.74
5.90
5.99
6.06
6.19
6.19
6.20
6.20
6.20
6.30
6.36
6.50
6.53
6.54
6.61
6.69
6.90
6.93
7.28
7.51
7.62

3
1
2
1
2
9
8
3
4
5
1
2
3
3
1
5
1
5
26
2
12
4
1
10

Df = closest distance from the fault projection in km.

Earthquake name

Year

Magnitude

Number of records

Little Skull Mtn, NV

Lazio-Abruzzo, Italy
Chi-Chi, Taiwan-02
Whittier Narrows-01
N. Palm Springs
Morgan Hill
Chalfant Valley-02
Chi-Chi, Taiwan-05
Chi-Chi, Taiwan-06
Coalinga-01
Dinar, Turkey
Imperial Valley-06
San Fernando
Northridge-01
Ierissos, Greece
Borah Peak, ID-01
Loma Prieta
Hector Mine
Landers
Kern County
Kocaeli, Turkey
Chi-Chi, Taiwan

1992
1984
1999
1987
1986
1984
1986
1999
1999
1983
1995
1979
1971
1994
1983
1983
1989
1999
1992
1952
1999
1999

5.65
5.80
5.90
5.99
6.06
6.19
6.19
6.20
6.30
6.36
6.40
6.53
6.61
6.69
6.70
6.88
6.93
7.13
7.28
7.36
7.51
7.62

2
1
6
10
8
3
1
13
7
2
1
1
3
12
1
2
9
11
6
1
2
8

Df = closest distance from the fault projection in km.

unconservative results in the evaluation of the DMF, with differences increasing with the increase of the period of vibration
[15,47]. On the other hand, as highlighted by Lin and Chang [47],
if the additional damping of structures derives from dissipation
devices the DMF obtained from displacement spectra should be
adopted. In those cases in which damping is produced by the structure itself (inherent damping) and the damping ratio is greater
than 10%, the DMF to be used should be derived directly from
acceleration spectra. In accordance with the purpose of this study,
which is to evaluate the effect of damping on displacement
demand, Eq. (5) is used in the following to calculate the DMF.
4.1. DMF for near-fault pulse-like records
For each near-fault ground motion, the pulse period of the
velocity time history is evaluated according to the methodology
proposed by Baker [24]. With the aim of highlighting the effect
of the pulse on the damping modication factor, the ground
motions are grouped into bins having different values of the pulse
period Tp (using a similar criterion than that used in Tothong et al.
[37] and Ruiz-Garcia [38]). In particular, the following seven intervals of Tp are considered: Tp < 1 s; 1 6 Tp < 2 s; 2 6 Tp < 3 s;

Fig. 2. Magnitude-pulse period plot for the fault-normal component of the pulselike ground motions used in this study. Observed and predicted values from Eq. (6).

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

21

Fig. 3. Mean spectra of the damping modication factor (DMF) calculated using the FN (fault-normal) and the FP (fault-parallel) records.

3 6 Tp < 4 s; 4 6 Tp < 5 s; 5 6 Tp < 6 s; 6 s 6 Tp. In order to compare

the results with the fault-parallel counterpart, the same division in
Tp bins used for the FN records is also used for the FP records.

In Fig. 2, the pulse period of the FN records is plotted against the

magnitude. Using a linear regression, the following relationship
between pulse-period and magnitude can be derived:

22

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

Fig. 4. Considered orientations of the pulse-like ground motions components.

LnT p 5:52 0:958M w

which is similar to the equation obtained by Baker [24].

Mean DMF spectra obtained for the FN and FP dataset are
shown in Fig. 3 for each Tp bin and for three selected values of
the damping ratio. It is worth noticing that the bin corresponding
to Tp P 6 s is strongly dominated by signals from the same event
(i.e. the Chi-Chi earthquake). However, the trend for the variation
of DMF with period and damping ratio observed for this group of
records seems consistent with those obtained for the other
groups.
The period-dependent nature of the DMF is evident in the plots
of Fig. 3. Such dependence is clearly inuenced by the pulse period
Tp. For low values of Tp the variability with the period is evident in
the whole spectrum, while for high Tp values the effect of the period
is signicant in the short period range only (T < 0.5 s). Differences in
the trends obtained for the DMF spectra calculated with the FN
records and the FP records are negligible for Tp < 1 s and Tp P 6 s.
By looking at the plots obtained for the other Tp intervals, it can
be observed that the presence of the pulse in the fault-normal component more signicantly affects the spectral shape of the DMF. In
particular, because of the pulse the fault-normal mean spectra are
characterized by a peak or a valley (depending on whether the
damping ratio is lower or higher than 5%, respectively) at a period
value that is about one second less than that of the pulse. This
difference in the spectra obtained with the two components is more
evident when Tp is between 3 s and 6 s.
In general, it is well known that the orientation of the ground
motion components has a non-negligible inuence on structural
demand of buildings [48,49], as the incident angle may signicantly affect the response not only of complex and irregular in-plan
structures, but also that of symmetric buildings. Because of this,
the effect on the DMF of the orientation of the ground motion components is investigated by rotating in a set of different directions
each pulse-like ground motion, i.e. each pair of fault-normal and
fault-parallel components. As shown in Fig. 4, for each pair four
directions are considered, including that producing the fault-normal and the fault-parallel components.
In Fig. 5, displacement spectra and DMF spectra obtained with
the FN and FP components of the ground motion recorded at Sylmar-Converter Station (1994 Northridge earthquake) and with
components rotated of an angle a = 45 (X3, Y3 components) are
shown. The displacement spectra obtained with the FN component
show a peak at a period close to the duration of the velocity pulse
(Tp = 3.48 s). The peak is smoothened by increasing the damping
ratio. The peak displacement obtained with the FP component is
lower than its FN counterpart. The DMF spectra of the two components (FN and FP) show some differences especially at periods
corresponding to local maxima or minima. As expected, with the

increase of the angle a the displacements in the Xi direction

decrease while those in the Yi direction increase. However, global
minimum and maximum values of the DMF are slightly affected
by the angle a, even though the DMF spectral shapes may vary
noticeably. As shown in the plots of Fig. 6, independently of the
considered period value the DMFs obtained using the components
rotated of the angle a are in general between the values obtained
with the FN and FP components.
In order to evaluate the record variability of the DMF values in
each Tp bin, the maximum standard deviations across the period
values (sd) and the maximum coefcient of variation (cov) are estimated. In Table 4 and Table 5, the sd and cov values are reported
for the fault-normal and the fault-parallel groups of records,
respectively. The cov value varies between 0.106 (FN records,
n = 10%, Tp < 1 s) and 0.466 (FP records, n = 50%, 3 6 Tp < 4 s). For
damping ratios greater than 10%, the coefcient of variation in
general increases with the increase of damping, showing heteroscedasticity in the data with respect to n.

4.2. DMF for ordinary records

In this subsection, the DMF spectra obtained with the pulse-like
records are compared with those obtained with the ordinary
records. In addition, the effect of site-to-source distance on ordinary ground motions is investigated by comparing the DMF obtained
with the FA and the FB records (characterized by a closest distance
from the fault projection equal to 2040 km and 45100 km,
respectively).
In order to compare the DMF of pulse-like records falling within
certain Tp bins with the DMF of ordinary motions, magnitude intervals related to pulse period intervals are dened. This was made by
means of Eq. (6). A measure of the suitability of the equation in
relating magnitude values and pulse periods is then given by the
correlation coefcient, R2 = 0.578, calculated in obtaining such
equation. In Fig. 7, the mean DMF spectra obtained with the FN,
FP, FA and FB records are plotted for damping ratios equal to 2%,
20% and 50%, and for three pulse period/magnitude intervals. For
magnitude values lower than 6.4, the inuence of distance for
the ordinary records is negligible. For higher magnitude values,
instead, the inuence increases with differences in the DMF values
obtained with the FA and the FB records not exceeding 20%. This
moderate dependence of DMF on distance may be probably
explained by the combined effect that duration and amplitude of
records have on DMF. With the increase of distance, in fact,
duration of the ground motion and number of cycles increases,
and consequently also the effect of damping increases [50].
Conversely, the increase of distance decreases the amplitude of
the excitation as well as the effect of damping.

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

23

Fig. 5. Ground motion recorded at Sylmar-Converter Station (1994 Northridge earthquake). Displacement spectra (Sd) and damping modication factor (DMF) spectra for the
FN-FP and the X3Y3 components (refer to Fig. 4). Bold lines are used for 5% damped spectra, while solid lines for damping ratios other than 5%, i.e. 2%, 10%, 20%, 30%, 40% and
50%.

With the exception of the spectral region around the period of

the pulse, the effect of damping is in general slightly greater for
ordinary records than for near-fault pulse-like records (FN dataset).
Finally, it is observed that for the ordinary records the trend of the
DMF spectra with magnitude is the same one as that observed for
the case of the pulse-like records with respect to Tp.

4.3. Prediction model of DMF for pulse-like ground motions

Based on the estimated DMF values, the prediction model given
in Eq. (7) is proposed for pulse-like ground motions. The dependence on damping ratio is given by a polynomial expression of
the natural logarithm of n, as already suggested by Stafford et al.

24

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

2

DMF 1  5  n1 a lnn blnn c d lnT

2

elnT 1  c1  fT

0:15

In Eq. (7) the variable c is 1 or 0 according to whether n < 0.05 or

n P 0.05. The parameters a, b, c, d, e and f are regression coefcients whose value depends on the considered pulse period interval (see Table 6). They are obtained with a nonlinear regression
that minimizes (using a standard LevenbergMarquardt algorithm)
the sum of squares of relative errors dened as follows:

r 1

Fig. 6. Mean spectra of the damping modication factor (DMF) for selected values
of pulse period and damping ratio, obtained with the FN-FP and the X3Y3
components (refer to Fig. 4).

[30] and Hatzigeorgiou [27]. The dependence on period is given by

two terms: the rst term is a quadratic expression of the logarithm
of T; the second term, which is a linear function of T0.15, is introduced to improve the predictive capability of the equation in the
case of damping ratio values smaller than 5%.

b
D MF
DMF

b and DMF denote the predicted and the observed values,

where D MF
respectively.
Due to the observed heteroscedasticity of the DMF with respect
to the variable n (as already evidenced in the discussion of Table 4),
least squares regression based on relative errors is used instead of
ordinary least squares for better tting the data. The coefcient of
determination values obtained in the regression analyses are
reported in Table 7. It is important to underline that the good t
obtained for the case of Tp P 6 s may be due to the fact that data
in this pulse period interval are dominated by a single event, i.e.
the Chi-Chi earthquake. The reliability of the predictive model
would improve in the case additional records from other earthquakes were available for deriving the regression coefcients.
In Fig. 8, the proposed prediction model (Eq. (7), Table 6) is
compared with the mean and the mean one standard deviation
DMF spectra of the pulse-like records (FN dataset); damping ratios
equal to 2%, 20% and 50%, and two different intervals of Tp are considered. Even if some differences between observed mean spectra
and predicted spectra may be noticed, the model ts the data well.
In the same gure the proposed model is compared with those
developed by Lin and Chang [16] and Hatzigeorgiou [27], in which
the DMF is expressed as a continuous function of both period and
damping, and with that by Priestley [26] (see Eq. (3)). Lin and
Chang [16] used a wide dataset composed of 1037 records from
102 earthquakes with magnitude ranging between 5.4 and 7.4.
The signals were recorded both on rock and rm sites, and have
a closest distance to the fault ranging from 0 to 180 km. The
records were grouped according to their site class (dened in
accordance with NEHRP site classication) in three groups (AB, C
and D). Both site-dependent and not site-dependent (valid for all

Table 4
Maximum standard deviation across the period values (sd) and coefcient of variation (cov) of the FN records.
n (%)

2
10
20
30
40
50

Tp < 1 s

1 6 Tp < 2 s

2 6 Tp < 3 s

3 6 Tp < 4 s

4 6 Tp < 5 s

5 6 Tp < 6 s

6 s 6 Tp

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

0.154
0.090
0.152
0.168
0.176
0.179

0.123
0.106
0.200
0.238
0.269
0.295

0.158
0.097
0.147
0.175
0.191
0.201

0.133
0.113
0.206
0.259
0.294
0.324

0.189
0.108
0.179
0.197
0.199
0.196

0.154
0.131
0.272
0.343
0.388
0.422

0.284
0.124
0.186
0.217
0.223
0.220

0.238
0.147
0.257
0.297
0.318
0.337

0.199
0.126
0.185
0.184
0.188
0.188

0.148
0.153
0.290
0.341
0.379
0.421

0.195
0.106
0.165
0.172
0.173
0.183

0.151
0.129
0.247
0.296
0.348
0.381

0.194
0.116
0.178
0.180
0.180
0.181

0.143
0.143
0.270
0.315
0.351
0.385

Table 5
Maximum standard deviation (sd) across the period values and coefcient of variation (cov) of the FP records.
n (%)

Tp < 1 s

1 6 Tp < 2 s

2 6 Tp < 3 s

3 6 Tp < 4 s

4 6 Tp < 5 s

5 6 Tp < 6 s

6 s 6 Tp

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

sd

cov

2
10
20
30
40
50

0.157
0.097
0.150
0.161
0.168
0.177

0.129
0.113
0.233
0.305
0.359
0.407

0.191
0.100
0.161
0.189
0.208
0.215

0.150
0.124
0.220
0.283
0.333
0.365

0.181
0.111
0.164
0.177
0.181
0.183

0.154
0.127
0.231
0.300
0.340
0.362

0.275
0.122
0.197
0.255
0.266
0.269

0.229
0.138
0.258
0.369
0.429
0.466

0.208
0.105
0.159
0.165
0.161
0.157

0.159
0.133
0.244
0.295
0.332
0.360

0.207
0.105
0.159
0.182
0.192
0.194

0.156
0.127
0.225
0.289
0.330
0.358

0.188
0.100
0.137
0.156
0.163
0.167

0.146
0.122
0.214
0.259
0.284
0.309

25

DMF

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

1.6

1.6

1.4

1.4

1.2

1.2

1.0

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

FN
FP
FA
FB

0.8
0.6
0.4

1 Tp < 2 s
5.7 M < 6.4
= 2%

0.2

DMF

= 20%

1.2

1.6

1.4

1.4

1.2

1.2

1.0

1.0

1.0

0.8

0.8

0.8

0.6

0.6

0.4

0.4

2 Tp < 3 s
6.4 M < 6.9
= 2%

0.4
0.2
1

2 Tp < 3 s
6.4 M < 6.9
= 20%

1.2

1.0

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

3 Tp < 4 s
6.9 M < 7.2
= 20%

T (s)

3 Tp < 4 s
6.9 M < 7.2
= 50%

1.4
1.2

0.2

0.2
0

1.6

1.2

0.2

2 Tp < 3 s
6.4 M < 6.9
= 50%

1.4

0.4

0.2
0

1.4

0.6

1.2

1.6

3 Tp < 4 s
6.9 M < 7.2
= 2%

1.4

1.6

0.8

1.6

0.2
0

= 50%

0.2
0

1.6

0.6

1 Tp < 2 s
5.7 M < 6.4

1.4

0.2
0

DMF

1.6

1 Tp < 2 s
5.7 M < 6.4

T (s)

T (s)

Fig. 7. Mean spectra for groups of records FN (fault-normal pulse-like component), FP (fault-parallel pulse-like component), FA (ordinary, 2040 km) and FB (ordinary,
45100 km). From left to right damping ratio increases. From top row to bottom row pulse period and magnitude increase.

Table 6
Model parameters of the prediction equation for DMF obtained for different values of the pulse period Tp.

a
b
c
d
e
f

Tp < 1 s

1 6 Tp < 2 s

2 6 Tp < 3 s

3 6 Tp < 4 s

4 6 Tp < 5 s

5 6 Tp < 6 s

6 s 6 Tp

0.27513
0.01793
0.06640
0.01025
0.00822
0.87902

0.28068
0.01895
0.07427
0.00748
0.01001
0.92861

0.31371
0.02523
0.09171
0.00224
0.00754
0.88061

0.29172
0.02032
0.08348
0.00344
0.00497
0.94676

0.32238
0.02702
0.09131
0.00072
0.00208
0.86887

0.33944
0.03036
0.10365
0.00033
0.00258
0.87211

0.35714
0.03379
0.11760
0.00652
0.00444
0.92982

Table 7
Coefcient of determination values obtained in the regression analyses.

R2

Tp < 1 s

1 6 Tp < 2 s

2 6 Tp < 3 s

3 6 Tp < 4 s

4 6 Tp < 5 s

5 6 Tp < 6 s

6 s 6 Tp

0.807

0.862

0.866

0.866

0.859

0.858

0.878

site classes) relationships were proposed for the DMF. Hatzigeorgiou [27] developed a DMF prediction model with coefcients calibrated using different sets of records consisting in 100 far-fault
records (with magnitude ranging between 5.2 and 7.6), 110 nearfault records (with magnitude ranging from 5.5 to 7.8, and closest
distance to the fault less than or equal to 10 km) and 100 articial

accelerograms. Both Lin and Chang [16] and Hatzigeorgiou [27] did
not take the pulse period (or the magnitude) as a variable in the
prediction of the DMF.
In the short period range the DMF predicted with the different
models are very similar, even though the models are derived from
different databases. Differences increase with the increase of

26

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

Fig. 8. Comparison between: proposed model of damping modication factor (DMF) for pulse-like records (red line); mean DMF spectra (black line), and mean one standard
deviation DMF spectra (grey line) of the pulse-like records (FN dataset) used in this study; model proposed by Lin and Chang [16] (green lines), by Hatzigeorgiou [27] (cyan
line) and by Priestley [26] (blue dotted line). In the left side plots, spectra predicted with Lin and Chang [16] are missing because the model was proposed only for n greater
than 5%. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

period, pulse period and damping ratio and are due to the fact that
both Lin and Chang [16] and Hatzigeorgiou [27] did not take the
magnitude as a variable in the estimation of DMF. As a matter of
fact, the predictive model by Hatzigeorgiou [27] and that proposed
in the present study are very similar for values of the pulse period
up to 2 s, while produces different predictions at higher values of
Tp, with the former underestimating the effect of damping. The
equation proposed by Priestley tends to overestimate the DMF in
the long period range of the spectra at low values of the pulse period, while it underestimates the effect of damping in the other
cases.
5. Conclusions
In this study the effect of damping on the displacement demand
of SDOF systems subjected to pulse-like and ordinary seismic excitations is studied. The damping modication factor DMF, dened as
the ratio between the displacement demand of a SDOF system with
damping ratio other than 5% and that of the corresponding 5%damped SDOF is calculated using near-fault pulse-like records
and ordinary records from different earthquakes. Different values
of the damping ratio, ranging from 2% to 50%, are considered. With
the aim of highlighting the effect of the pulse, the pulse-like
records were grouped into bins corresponding to different values
of the pulse period Tp. A prediction model was then proposed to
estimate the DMF for this type of records, and the model parameters calibrated for each considered Tp bin. The main results
obtained in the study may be summarized as follows.
The period of vibration of the structure affects the DMF noticeably. In particular, it was observed that in the case of near-fault
pulse-like ground motions the inuence of period on the DMF
becomes more signicant with the decrease of the pulse period
value.

The DMF spectra calculated with pulse-like records are characterized by a pronounced peak or valley (depending on whether
the damping ratio is lower or higher than 5%, respectively) that
is located at a period value which is about one second less than
that of the pulse. This effect was observed to be more signicant
when Tp is between 3 s and 6 s. At periods away from the pulse
period a reduction of the effect of damping compared with the
fault-parallel counterpart is observed.
By rotating the pulse-like records according to different directions with respect to the fault, the inuence of the angle of rotation was investigated. The observed trend is that the DMF
values obtained using the rotated components are between
the values obtained with the fault-normal and fault-parallel
components.
The maximum reduction or amplication of the displacement
response produced by damping is in general slightly higher
for ordinary records than for pulse-like records. However, for
damping ratios greater than 5%, in the spectral region around
the period of the pulse, lower values of the DMF were observed
in the case of pulse-like records. An analogous trend of the
ordinary records DMF spectra with magnitude was observed
for the case of the pulse-like records but with respect to Tp.
Neglecting the inuence of the pulse period Tp (or that of the
magnitude) on DMF can lead to signicant overestimation of
the effect of damping in the long period range at low values
of Tp (or magnitude).
For ordinary records, the inuence of distance to the fault on
DMF is negligible for magnitudes smaller than 6.4. For larger
magnitudes, instead, the inuence increases but remains less
signicant than that of magnitude.
For the case of pulse-like records, a predictive model that is a
function of period T and damping ratio n of the SDOF system
is proposed for the evaluation of the DMF (Eq. (7)). The parameters of the model are calibrated through regression analyses

F. Mollaioli et al. / Engineering Structures 78 (2014) 1727

for each interval of pulse period considered. The model is built

based on pulse-like records currently available from the literature. Due to the limited number of data used in the regression
analyses, it is recommended to use the predictive equations
for expected earthquakes and site conditions consistent with
the properties of the records used to calibrate the model. In particular, the proposed model is suitable for soil types B, C and D,
magnitude values ranging from 5.0 to 7.6 and Tp values ranging
between 0.4 and 9 s. In addition, the model can be used to predict the response of structures with period up to 8.0 s.

Acknowledgments
This work has been partially carried out under the program
Dipartimento di Protezione Civile-Consorzio ReLUIS, signed on
2009-09-24 (no. 823), Research Line RS2. The nancial support of
the Italian Ministry of the Instruction, University and Research
(MIUR) is acknowledged. The authors are grateful to Prof. Luis
Decanini for his valuable suggestions and encouragement. Finally,
the contribution of the anonymous reviewers of the article is
greatly acknowledged.

References
[1] Bommer JJ, Elnashai AS. Displacement spectra for seismic design. J Earthquake
Eng 1999;3(1):132.
[2] Decanini LD, Liberatore L, Mollaioli F. Characterization of displacement
demand for elastic and inelastic SDOF systems. Soil Dynam Earthquake Eng
2003;23:45571.
[3] Cauzzi C, Faccioli E. Broadband (0.05 to 20 s) prediction of displacement
response spectra based on worldwide digital records. J Seismolog
2008;12(4):45375.
[4] Mollaioli F, Bruno S. Inuence of site effects on inelastic displacement ratios for
SDOF and MDOF systems. Comput Math Appl 2008;55(2):184207.
[5] Ghobarah A. Performance-based design in earthquake engineering: state of
development. Eng Struct 2011;23:87884.
[6] Chandler AM, Lam NTK. Performance-based design in earthquake engineering:
a multidisciplinary review. Eng Struct 2001;23:152543.
[7] Priestley MJN, Calvi GM. Concepts and procedures for direct displacementbased design. In: Fajfar P, Krawinkler H, editors. Seismic Design Methodologies
for the Next Generation of Codes. Balkema: Rotterdam; 1997. p. 17181.
[8] Priestley MJN, Calvi GM, Kowalsky MJ. Displacement based seismic design of
structures. Pavia, Italy: IUSS Press; 2007.
[9] Franchin P, Pinto PE. Method for probabilistic displacement-based design of RC
structures. J Struct Eng 2012;138(5):58591.
[10] Kelly JM. Base isolation: linear theory and design. Earthquake Spect
1990;6(2):22344.
[11] Newmark NM, Hall WJ. Earthquake spectra and design. Earthquake
Engineering Research Institute, Monograph Series, Berkeley, California;
1982.
[12] Ashour SA. Elastic seismic response of buildings with supplemental damping.
Ph.D. Dissertation, Department of Civil Engineering, University of Michigan;
1987.
[13] Wu J, Hanson RD. Study of inelastic spectra with high damping. J Struct Eng
1989;115(6):141231.
[14] Lin YY, Miranda E, Chang KC. Evaluation of damping reduction factors for
estimating elastic response of structures with high damping. Earthquake Eng
Struct Dynam 2005;34:142743.
[15] Cardone D, Dolce M, Rivelli M. Evaluation of reduction factors for highdamping design response spectra. J Struct Eng 2009;130(11):166775.
[16] Lin YY, Chang KC. Effects of site classes on damping reduction factors. J Struct
Eng 2004;130(11):166775.
[17] Bommer JJ, Mendis R. Scaling of spectral displacement ordinates with damping
ratios. Earthquake Eng Struct Dynam 2005;34:14565.
[18] Cameron WI, Green RA. Damping correction factors for horizontal groundmotion response spectra. Bull Seismol Soc Am 2007;97(3):93460.
[19] Hall JF, Heaton TH, Halling MW, Wald DJ. Near source ground motion and its
effects on exible buildings. Earthquake Spect 1995;11:569605.
[20] Bray JD, Rodriguez-Marek A. Characterization of forward-directivity ground
motions in the near-fault region. Soil Dynam Earthquake Eng 2004;24:
81528.
[21] Mavroeidis GP, Dong G, Papageorgiou AS. Near fault ground motions, and the
response of elastic and inelastic single-degree-of-freedom (SDOF) systems.
Earthquake Eng Struct Dynam 2004;33:102349.

27

[22] Mollaioli F, Bruno S, Decanini LD, Panza GF. Characterization of the dynamical
response of structures to damaging pulse-type near-fault ground motions.
Meccanica 2006;41:2346.
[23] Kalkan E, Kunnath SK. Effects of ing-step and forward directivity on the
seismic response of buildings. Earthquake Spect 2006;22:36790.
[24] Baker JW. Quantitative classication of near-fault ground motions using
wavelet analysis. Bull Seismol Soc Am 2007;97(5):1486501.
[25] Mollaioli F, Bosi A. Wavelet analysis for the characterization of forwarddirectivity pulse-like ground motions on energy basis. Meccanica
2012;47:20319.
[26] Priestley MJN. Myths and Fallacies in Earthquake Engineering Revisited. The
Mallet Milne Lecture. Pavia: IUSS Press; 2003.
[27] Hatzigeorgiou GD. Damping modication factors for SDOF systems subjected
to near-fault, far-fault and articial earthquakes. Earthquake Eng Struct
Dynam 2010;39:123958.
[28] Bozorgnia Y, Campbell KW. Engineering characterization of ground motion. In:
Bozorgnia Y, Bertero VV, editors. Earthquake engineering from engineering
seismology to performance-based engineering. Boca Raton (Florida): CRC
Press; 2004.
[29] Rosenblueth E. Characteristics of earthquakes. In: Rosenblueth E, editor.
Design of Earthquake Resistant Structures, Wiley; 1980 [Ch. 1].
[30] Stafford PJ, Mendis R, Bommer JJ. Dependence of damping correction factors
for response spectra on duration and number of cycles. J Struct Eng
2008;134(8):136473.
[31] Rezaeian S, Bozorgnia Y, Idriss IM, Campbell K, Abrahamson N, Silva W.
Spectral Damping Scaling Factors for Shallow Crustal Earthquakes in Active
Tectonic Regions. PEER Report 2012/01, Pacic Earthquake Engineering
Research Center, University of California, Berkeley; 2012.
[32] FEMA(Federal Emergency Management Agency). Prestandard and
commentary for the seismic rehabilitation of buildings. FEMA-356,
Washington, D.C.; 2000.
[33] Eurocode 8. Design of structures for earthquake resistance, Part 1: General
rules, seismic actions and rules for buildings. DRAFT No. 6, European
Committee for Standardization, January 2003.
[34] Alavi B, Krawinkler H. Effects of Near-Fault Ground Motions on Frame
Structures. John A Blume Earthquake Engineering Center Technical Report
138, 2001. Stanford Digital Repository, available at: <http://purl.stanford.edu/
cx534fy3768>.
[35] Somerville PG. Magnitude scaling of the near fault rupture directivity pulse.
Phys Earth Planet Inter 2003;137:20112.
[36] Fu Q, Menun C. Seismic-environment-based simulation of near-fault ground
motions. In: Proc. of the thirteenth world conference on earthquake
engineering, Vancouver; 2004. Paper No. 322.
[37] Tothong P, Cornell A, Baker JW. Explicit directivity-pulse inclusion in
probabilistic seismic hazard analysis. Earthquake Spect 2007;23(4):86791.
[38] Ruiz-Garca J. Inelastic displacement ratios for seismic assessment of
structures subjected to forward-directivity near-fault ground motions. J
Earthquake Eng 2011;15(3):44968.
[39] Iervolino I, Chioccarelli E, Baltzopoulos G. Inelastic displacement ratio of nearsource pulse-like ground motions. Earthquake Eng Struct Dynam
2012;41:23517.
[40] Somerville PG, Smith NF, Graves RW, Abrahamson NA. Modication of
empirical strong ground motion attenuation relations to include the
amplitude and duration effects of rupture directivity. Seismol Res Lett
1997;68(1):199222.
[41] Rodriguez-Marek A. Near-Fault Seismic Site Response, Ph.D. Dissertation.
University of California at Berkeley, Berkeley, CA; 2000.
[42] Mavroeidis GP, Papageorgiou AS. A mathematical representation of near-fault
ground motions. Bull Seismol Soc Am 2003;93(3):1099131.
[43] Gillie JL, Rodriguez-Marek A, McDaniel C. Strength reduction factors for nearfault forward-directivity ground motions. Eng Struct 2010;32:27385.
[44] Shahi SK, Baker JW. An empirically calibrated framework for including the
effects of near-fault directivity in probabilistic seismic hazard analysis. Bull
Seismol Soc Am 2011;101(2):74255.
[45] Rodriguez-Marek A, Cofer W. Dynamic response of bridges to near-fault,
forward directivity ground motions. Washington State Transportation Center
Report n. WA-RD 689.1; 2007.
[46] Ghayamghamian MR, Hisada Y. Near-fault strong motion complexity of the
2003 Bam earthquake (Iran) and low-frequency ground motion simulation.
Geophys J Int 2007;170:67986.
[47] Lin YY, Chang KC. Study on damping reduction factor for buildings under
earthquake ground motions. J Struct Eng 2003;129(2):20614.
[48] Tsourekas AG, Athanatopoulou AM, Avramidis IE. Effects of seismic incident
angle on response of structures under bi-directional recorded and articial
ground motion. In: Papadrakakis M, Lagaros ND, Fragiadakis M, editors.
ECCOMAS Thematic Conference on Computational Methods in Structural
Dynamics and Earthquake Engineering. COMPDYN 2009. Rhodes, Greece, 22
24 June 2009.
[49] Lucchini A, Mollaioli F, Monti G. Intensity measures for response prediction of
a torsional building subjected to bi-directional earthquake ground motion.
Bull Earthq Eng 2011;9(5):1499518.
[50] Atkinson GM, Pierre JR. Ground-motion response spectra in eastern North
America for different critical damping values. Seismol Res Lett
2004;75(4):5415.