Ground Motion Record Selection Based on Broadband Spectral Compatibility

Abstract

The increasing interest in performance-based earthquake engineering has promoted research on the improvement of hazard-consistent seismic input definition and on advanced criteria for strong motion record selection to perform nonlinear time history analyses. Within the ongoing research activities to improve the representation of seismic actions and to develop tools as a support for engineering practice, this study addresses the selection of displacement-spectrum-compatible real ground motions, with special reference to Italy. This involved (1) the definition of specific target displacement spectra for Italian sites, constrained-both at long and short periods-by results of probabilistic seismic hazard analyses; (2) the compilation of a high-quality strong ground motion database; and (3) the development of a software tool for computer-aided displacement-based record selection. Application examples show that sets of unscaled, or lightly scaled, accelerograms with limited record-to-record spectral variability can also easily be obtained when a broadband spectral compatibility is required.

Full text

Ground Motion Record Selection Based on
Broadband Spectral Compatibility
Chiara Smerzini,a) Carmine Galasso,b) M.EERI, Iunio Iervolino,b) and
Roberto Paoluccia)
The increasing interest in performance-based earthquake engineering has pro-
moted research on the improvement of hazard-consistent seismic input definition
and on advanced criteria for strong motion record selection to perform nonlinear
time history analyses. Within the ongoing research activities to improve the repre-
sentation of seismic actions and to develop tools as a support for engineering
practice, this study addresses the selection of displacement-spectrum-compatible
real ground motions, with special reference to Italy. This involved (1) the defini-
tion of specific target displacement spectra for Italian sites, constrained—both at
long and short periods—by results of probabilistic seismic hazard analyses;
(2) the compilation of a high-quality strong ground motion database; and (3)
the development of a software tool for computer-aided displacement-based record
selection. Application examples show that sets of unscaled, or lightly scaled,
accelerograms with limited record-to-record spectral variability can also easily
be obtained when a broadband spectral compatibility is required. [DOI:
10.1193/052312EQS197M]
INTRODUCTION
Performance-based approaches to seismic design and assessment, together with the ever-
increasing availability of high-quality digital strong ground motion (GM) records, have
raised research interest in improved criteria for the selection and scaling of waveforms to
perform nonlinear dynamic analyses of structures.
For example, the ASCE/SEI 7-10 Standard (ASCE 2010), recommends that input ground
motions for linear and nonlinear seismic analyses of structures shall be selected from actual
recorded events “having magnitude, fault distance, and source mechanism that are consistent
with those that control the maximum considered earthquake :::. The ground motions shall be
scaled such that the average value of the 5%damped response spectra for the suite of motions
is not less than the design response spectrum for the site for periods ranging from 0.2 Tto
1.5 Twhere Tis the fundamental period of the structure in the fundamental mode for the
direction of response being analyzed.”The latter requirement recognizes that spectral com-
patibility should be enforced in a period range large enough to account for the sensitivity of
nonlinear structural response to spectral ordinates at periods larger than the fundamental one,
and to the higher modes’contribution at shorter periods. Particularly, the importance to con-
strain not only the spectral ordinate at the fundamental period of the structure, but the
Earthquake Spectra, Volume 30, No. 4, pages 1427–1448, November 2014; © 2014, Earthquake Engineering Research Institute
a)
Dipartimento di Ingegneria Strutturale, Politecnico di Milano, Italy
b)
Dipartimento di Ingegneria Strutturale, Università degli Studi di Napoli Federico II, Italy
1427

response spectral shape itself in a sufficiently large period range has been remarked by sev-
eral authors (e.g., Baker and Cornell 2006,Haselton et al. 2009,Bojorquez and Iervolino
2011). Particularly, ground motion selection based on conditional spectra—that is, the con-
ditional mean spectrum (Baker 2011) and the conditional spectrum (Jayaram et al. 2011)—
has gained a growing relevance, although its applicability is still disputed for cases where
large nonlinear effects and/or participation of higher modes in structural response may affect
the choice of the conditioning period. Recently, the U.S. National Institute of Standards and
Technology (NIST) funded a project to improve guidance to the earthquake engineering pro-
fession for selecting and scaling earthquake ground motions for the purpose of nonlinear
response-history analyses in performance-based seismic engineering (NIST 2011).
Requirement for broadband compatibility, including long periods, may be especially
important for specific classes of earthquake engineering applications among which: (1) seis-
mic design approaches relying on a proper definition of spectral displacements at long per-
iods, such as the direct displacement-based design method (e.g., Priestley et al. 2007),
(2) dynamic analyses involving nonlinear soil-structure interaction effects, such as for the
seismic response of underground structures, or for soil stability assessment. In the latter appli-
cations, the response is not dominated by the inertial properties of the structure around its
natural vibration periods, but by the spectral characteristics of the ground motion itself in a
wide frequency range. Furthermore, results of numerical simulations of the dynamic response
may be significantly affected by the long-period noise of input ground motion, especially
when dealing with records from analog instruments, with potential non-physical drifts of
the velocity and displacement time series (e.g., Foti and Paolucci 2012). As a matter of
fact, long-period noise on recorded ground motions has traditionally limited the reliability
of response spectral ordinates for structural periods up to a maximum of 3–4s.
To address these limitations, in the last decade a significant effort has been made toward a
better characterization of long-period ground motion for seismic design (Faccioli et al. 2004),
by taking advantage of the ever-increasing number of high quality digital GM records from
worldwide earthquakes. A common objection to the determination of long-period spectral
ordinates from digital GM records, is that appropriate displacement traces cannot be gener-
ally retrieved upon simple double integration of the uncorrected accelerations and that high-
pass filtering is typically required. An answer to this objection has been given by a growing
number of recent studies, starting with Boore (2001), which have shown that long-period
spectral ordinates of digital accelerograms depend only weakly on the adopted baseline cor-
rection procedures, in opposition to the dramatic effect of those operations on the displace-
ment waveforms. Even more pervasive, with respect to this issue, is the evidence, pointed out
by Wang et al. (2007) and by Paolucci et al. (2008), that spectral ordinates, calculated from
co-located GM and broadband records, coincide up to at least 10 s.
Advances to assess the reliability of long-period response spectral ordinates have sup-
ported the calibration of up-to-date empirical ground motion prediction equations (GMPEs)
extending to long periods (Cauzzi and Faccioli 2008), the improved quantification of site
effects (Figini and Paolucci 2009), and the formulation of new seismic hazard maps at
long periods in Italy (Faccioli and Villani 2009).
Research activities for an improved determination of seismic input at long periods con-
tinued recently with a collaboration among the authors of this study, to combine, on one side,
1428 SMERZINI ET AL.

the expertise on long-period ground motion characterization with earthquake record selection
for engineering applications on the other side. The starting point was a software tool, REXEL
(Iervolino et al. 2010), which enables the selection of suites of multi-component ground
motions compatible with either code-based or user-defined pseudo-acceleration response
spectra. Spectrum compatibility will be referred to hereafter as the condition for which
the average response spectrum of the selected set of records approaches the target spectrum
within a prescribed tolerance level, and scatter of individual spectra with respect to the target
spectrum is limited as much as possible1. To achieve the goal of spectral compatibility at long
periods, two basic ingredients were introduced: (1) a target spectrum suitable up to long
periods, constrained by results of short-period and long-period probabilistic seismic hazard
analyses (PSHA) in Italy; (2) a strong GM database consisting of high-quality accelerograms,
covering, in a way as homogeneous as possible, the magnitude, distance range, and site con-
ditions of interest for Italian sites. In fact, the psuedo-spectral rule, that is, the simple multi-
plication of the acceleration spectrum, Sa,byðT∕2πÞ2, where Tis the vibration period, to
obtain displacement spectral ordinates, Sd, may not provide accurate results if the target spec-
trum has not been properly constrained at long periods. On the other hand, the use of accel-
erograms with an insufficient signal-to-noise ratio at long periods precludes the success of the
search.
The first part of this paper illustrates the substantial differences among several seismic
regulations, in terms of elastic design spectra at long periods, which may have consequences
when such spectra are used as targets for GM selection. Afterward, the following topics are
addressed: a hazard-based target displacement spectrum for Italian sites; a specifically devel-
oped GM database; and the REXEL-DISP software, in which the previous two ingredients
are implemented together with a search engine. Finally, the application examples provide a
key to the second objective of this research, that is, showing that the combination of an
appropriate target spectrum both at short and long periods, together with a high-quality data-
base, allows one to achieve consistent spectrum compatibility of selected GM records in a
broad period range, either in terms of accelerations or displacements. Although such results
are obtained for the Italian context, the approach presented in this paper has a general value
and may easily be applied to other parts of the world, provided that PSHA studies at long
periods are available.
RESPONSE SPECTRA AT LONG PERIODS
DISPLACEMENT SPECTRA FROM INTERNATIONAL SEISMIC CODES
This section reviews the specifications of seismic actions at long periods according to
four international seismic codes: Eurocode, or EC8 (CEN 2004); the Italian Building code, or
NTC08 (CS.LL.PP. 2008); the U.S. seismic provisions ASCE 7-10 (ASCE 2010), and the
New Zealand seismic standards, or NZS 1170 (NZS 2004). Results will be compared in terms
of displacement spectra, which are more suitable to highlight differences at long periods. For
all these codes, displacement spectra are defined (at least up to a certain corner period) by
1This is because two different sets, equivalent in terms of average spectral fit, can still provide very different results
in the estimate of seismic response, depending on the record-to-record variability with respect to the target
spectrum (e.g., Iervolino et al. 2010).
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1429

converting the corresponding elastic acceleration design spectra through the pseudo-spectral
relationship mentioned in the introduction.
The key feature that controls the shape and amplitude of the spectrum at long periods is
the corner period, TD, denoting the beginning of the maximum spectral displacement plateau
(MSD; see Figure 1), which was determined in the previously mentioned codes according to
different criteria. In particular, in NZS 1170, a constant value TD¼3.0 s is defined, regard-
less of the seismic hazard associated with the selected site. Similar to NZS 1170, EC8 defines
a constant value of TD, but an indirect dependence on magnitude is introduced by assigning
TD¼2.0 s for the Type 1 spectrum (i.e., seismic hazard associated with earthquakes of
surface-wave magnitude MS≥5.5) and TD¼1.2 s for Type 2 (MS<5.5). In NTC08,
the corner period TD(equal to ag∕g·4þ1.6) is made dependent on the maximum ground
acceleration at rock (ag), the latter being related to the return period under consideration
through the PSHA results (Montaldo et al. 2007). Finally, the ASCE 7-10 guidelines define
the long-period branch of the response spectra by providing maps of the long-period transi-
tion period TL(equivalent to TD), as illustrated in Figure 1b, as a function of the modal
magnitude (Md) of each region, determined by disaggregation of PSHA for probability
of exceedence of 2%in 50 years at T¼2s (Crouse et al. 2006). Mapped values of TL
for the conterminous United States show a large variability, ranging from 4 s (for Mdin
the 6.0–6.5 range) up to 16 s (Mdin the 8.0–8.5 range). As a final remark, it is worth noting
that in all these codes, TD(or TL) does not depend on site conditions.
With regard to the features of the displacement spectrum for periods beyond TD(or TL),
while NZS 1170 and ASCE 7-10 assume the spectral displacement to be constant with period
(see Figure 1b), understanding that beyond TDthe spectral ordinate approaches the peak
ground displacement, EC8 and NTC08 limit the application of the pseudo-spectral rule
up to a corner period, TE, ranging between 4.5 s and 6.0 s, as a function of the ground cate-
gory. Beyond TE, a linear decreasing branch of the spectrum is defined, up to the control
period TF¼10 s, beyond which the spectral displacement tends to have a constant value
(dmax, as depicted in Figure 1a). The latter is computed in both EC8 and NTC08 by the
following relationship:
EQ-TARGET;temp:intralink-;e1;41;274dmax ¼0.025agSTCTD¼0.025amaxTCTD(1)
Figure 1. Elastic displacement design spectral shape according to (a) the Italian Building Code
and Eurocode 8 and (b) the New Zealand and U.S. standards.
1430 SMERZINI ET AL.

where amax (equal to agtimes S) is the design peak ground acceleration, Sis the soil ampli-
fication factor and TCis the control period marking the start of the constant velocity branch of
the design spectrum. Thus, Equation 1explicitly, albeit rather arbitrarily, constrains the long-
period spectral ordinates to the short-period ones (i.e., amax and TC) and to the corner period
TD. A consequence of such constraint is that the ratio MSD∕dmax is equal to the ratio
MSA∕amax,MSA being the maximum response spectral acceleration. This ratio is 2.5 accord-
ing to EC8, while NTC08 denotes it by F0and provides its values, based on the results of
short-period PSHA at a national scale.
To illustrate the effect of the previous assumptions on design spectra, Figure 2shows
(a) the spectral elastic accelerations and (b) displacements for an arbitrary site on similar
ground types, either B (for NTC08 and EC8) or C (for NZS 1170 and ASCE 7-10). For
the purpose of comparison, the design spectra are anchored to the same value of maximum
ground acceleration on rock, ag¼0.30 g. This results in a hazard factor Z¼0.30 g for NZS
1170 and spectral accelerations SS¼1.125 g and S1¼0.4 g, at 0.2 s and 1.0 s, respectively,
for ASCE 7-10, which serve to define the design spectra in these codes. While spectral accel-
erations appear to be relatively close, differences are much clearer in terms of spectral dis-
placements at long periods, mainly because of the heterogeneous criteria for the definition of
the corner period TD. Even if such differences may not be relevant for specification of seismic
actions for design, unless periods larger than about 3 s are of interest, the consequences in
terms of ground motion selection may be more important, since it is clear that records
approaching the ASCE 7-10 displacement spectrum at long periods will correspond in gen-
eral to a different magnitude level than those approaching the EC8 or NTC08 spectra.
DISPLACEMENT SPECTRA FROM LONG-PERIOD SEISMIC HAZARD STUDIES:
THE ITALIAN EXPERIENCE
To provide a rationale framework for the characterization of long-period ground
motion for design, Faccioli and Villani (2009) illustrated a novel representation of
Figure 2. Elastic design acceleration (a) and displacement (b) spectra according to the Italian,
European, New Zealand and U.S. seismic codes for 475-yr return period on ground category B
(for NTC08 and EC8) and C (for NZ 1170 and ASCE 7-10). The spectra are anchored to the value
of peak ground acceleration on exposed bedrock agequal to 0.30 g.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1431

long-period seismic hazard for Italy, based on the findings of the S5-Project “Seismic
Input in Terms of Expected Spectral Displacements,”funded by the Department of
Civil Protection from 2005 to 2007. This study produced new probabilistic seismic hazard
maps for Italy in terms of horizontal displacement response spectral ordinates in a wide
range of vibration periods, from 0.05 s up to 20 s, taking advantage, on one side, of the
broadband ground motion attenuation relationship developed by Cauzzi and Faccioli
(2008) and, on the other side, of the criteria for the reliability of long-period response
spectral ordinates from digital data introduced by Paolucci et al. (2008).2
Specifically, this work provided, throughout the Italian territory and for all local muni-
cipalities, mapped values of the 5%damped displacement response spectral ordinates (16th,
50th, and 84th percentiles), for T¼2s, 5 s, 10 s and for the return periods: TR¼72 yr,
475 yr and 975 yr. Based on these results, Faccioli and Villani (2009) proposed a simplified
bilinear approximation for the displacement response spectra on exposed rock, defined as
follows:
EQ-TARGET;temp:intralink-;e2;41;463SdðTÞ¼(D10
TDTT≤TD
D10 T>TD
(2)
where D10 is the displacement response spectral ordinate at T¼10 s and TDis the corner
period separating the constant displacement branch of the spectrum from the constant velo-
city one. The latter is computed as follows:
EQ-TARGET;temp:intralink-;e3;41;374TD¼2πD10
max
TPSV (3)
where PSV denotes the pseudo-velocity response spectrum, obtained from the uniform hazard
(UH) displacement spectra using the pseudo-spectral approximation rule.
As underlined by the authors of the hazard assessment, the bilinear approximation of
Equation 2cannot be used to estimate the seismic actions in terms of acceleration at
short periods via the pseudo-spectral relationship, as it would not lead in general to spectral
ordinates compatible with those prescribed by the current seismic code.
A Target Displacement Spectrum for Italian Sites (TDSI)
To satisfy the short-period prescriptions of the NTC08, on one side, and, on the
other side, the results of long-period PSHA from Project S5, the following target
displacement spectrum for Italy, referred to as TDSI hereafter (see Figure 3), is
introduced:
2This hazard study is not yet implemented in NTC08, which is currently based on PSHA at short periods.
1432 SMERZINI ET AL.

EQ-TARGET;temp:intralink-;e4;62;441SdðTÞ¼
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
T≤TBagηSF0hT
TBþ1
ηF01T
TBiT2
4π2
TB<T≤TCagηSF0T2
4π2
TC<T≤TDagηSF0TC
TαT2
4π2
TD<T≤TED10ηSCc
TE<T≤TFD10ηFþD10 ηðSCcFÞTFT
TFTE
T>TFD10ηF
(4)
where:
•D10 and TDare defined in Equations 3and 4, based on the mapped values of long-
period PSHA (Faccioli and Villani 2009).
•agis the design ground acceleration on rock for a given return period;
F0¼MSA∕amax,Sand Ccare the soil factors as a function of ground type; TC,
TE,TFare corner periods; and ηis the damping correction factor (η¼1for 5%
viscous damping). All the previous parameters are defined according to NTC08.
•α¼logð4π2D10Cc
agF0T2
D
Þ∕logðTC∕TDÞis the factor introduced to ensure matching between
the short- and long-period branches of the proposed spectral shape (see dotted line in
Figure 3). Values of αfor Italy range approximately between 0.85 and 1.4, the lar-
gest values being typically found in low-seismicity regions, while regions of high
seismicity are characterized by values close or slightly less than unity.
•F¼ð800∕VS30 Þ0.375—VS30 being the site shear wave velocity averaged over the
top 30 meters—is the site factor for long periods, as obtained in the framework
of Project S5 (Cauzzi et al. 2007).
Figure 3. Target displacement spectrum for Italy (TDSI, see Equation 4) in the (a) short- and
(b) long-period range, based on the results of the PSHA at long periods developed by Faccioli and
Villani (2009). Representation in the short-period range (left side) is given in terms of spectral
acceleration.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1433

Since the results of the long-period PSHA are given for three return periods alone (i.e.,
TR¼72 yr, 475 yr, 975 yr) values of D10 and TDas a function of the return period TRare
computed through a linear interpolation of logarithms. Note that due to the poor dependence
of TDon the return period, as emerged from the seismic hazard studies, a constant value of TD
was assumed at all sites for simplicity’s sake. For the computation of the long-period site
factor F, the average values of VS30 within the variability range of each site class have been
considered, namely: VS30 ¼800 m∕sfor site class A (F¼1.0); VS30 ¼580 m∕sfor class B
(F¼1.13); VS30 ¼270 m∕sfor classes C and E (F¼1.5); VS30 ¼140 m∕sfor class
D(F¼1.9).
The key features of the proposed TSDI target spectrum can be summarized as follows:
•The maximum spectral displacement plateau MSD for periods TD≤T≤TE—the
latter being inherited from NTC08 (i.e., TE¼4.5 s for ground type A, TE¼5.0 s
for ground type B, and TE¼6.0 s for other soil classes)—and the constant spectral
displacement dmax, reached at periods T≥TF¼10 s, are calculated based on the
long-period PSHA of Project S5.
•To match the NTC08 spectra when site factors are accounted for, we have consid-
ered the same site factor SCc, as in the NTC08, between TCand TE, while the
long-period site factor Ffrom Project S5 was adopted for periods T≥TF; for inter-
mediate periods, a linear variation of the site factor was assumed.
•The MSD∕dmax ratio is equal to SCc∕F, thus leading rock sites (soil class A) to a
constant displacement branch (SCc∕F¼1) for periods beyond TD, while leading
other soil classes to a more gentle decay of spectral displacements between TEand
TFwith respect to NTC08 (since in general, SCc∕F<F0).
To highlight the effects of different assumptions on site factors, Figure 4shows the ratio
between the spectral ordinates for soil class C and those for soil class A, computed according
to the NTC08 design spectrum (black thick line) and to the TDSI of Equation 4(gray dotted
line). Up to TE, the spectral amplification factors coming from NTC08 and from Equation 4
are the same, while for T>TE, the spectral amplification factor in Equation 4decreases
linearly until it reaches the value of Fat the corner period TF.
Figure 5shows the comparison between the NTC08 design displacement spectrum (black
line) and the TDSI (dotted gray line) for the return period TR¼475 yr both on ground type A
Figure 4. Ratio of the spectral ordinates for soil class C with respect to soil class A, according to
NTC08 (black line) and to the proposed representation, TDSI (gray dashed line).
1434 SMERZINI ET AL.

(Figure 5a) and C (Figure 5b), at L’Aquila, central Italy (top panel) and at Udine, northeastern
Italy (bottom panel). A general feature highlighted by this example is that for high seismicity
regions in Italy, such as L’Aquila, the spectral ordinates at long periods are generally
underestimated by the NTC08 design spectrum with respect to the TDSI, while, for low-
to-medium seismic regions, such as Udine, they may be significantly overestimated by
the NTC08, especially in terms of MSD. As remarked previously, the decay of spectral ordi-
nates for periods between TEand TFis much more pronounced in the code spectra than
observed in the TDSI.
DATABASE FOR ENGINEERING ANALYSES OF LONG-PERIOD
GROUND MOTION
DATA SELECTION
After introducing the target spectrum for displacement-based seismic analyses in Italy,
the second step is the compilation of the strong GM database, mainly consisting of digital
recordings suitable for displacement-based applications.
The SIMBAD database (Selected Input Motions for displacement-Based Assessment and
Design) was created by assembling records from different strong ground motion databases
worldwide, with the main objective of providing records of engineering relevance for the
Figure 5. Comparison of the displacement response spectra coming from NTC08 (black
line) and from this work, TDSI (gray dashed line), for 475-yr return period, at two represen-
tative sites, namely L’Aquila (top panel) and Udine (bottom panel), on ground types (a) A
and (b) C.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1435

most frequent design conditions in Italy. For this reason, only records from shallow crustal
earthquakes, at epicentral distance Repi approximately less than 35 km, with moment
magnitude, MW, ranging from 5 to 7.3, were considered. These are the conditions generally
governing seismic hazard throughout Italy, for most return periods of practical interest.
For the scope of this study, the selected records should be accurate at long periods. In fact,
most records (about 90%) included in the database are from digital instruments, while a lim-
ited number of analog records was retained, typically from large magnitude earthquakes, for
which a good signal to noise ratio at long periods could be achieved.
In general, raw acceleration time histories were processed according to the procedure
devised by Paolucci et al. (2011) and applied to the ITalian ACcelerometric Archive
(ITACA, http://itaca.mi.ingv.it). One of the features of the aforementioned procedure is
that single and double integration of processed acceleration records provides velocity and
displacement waveforms without non-physical baseline trends, and no further correction
is required. For each record, the same filter band was selected and applied to the three spatial
components. Except for a few exceptions, records were included in SIMBAD only if the
high-pass filter frequency was not larger than 0.15 Hz.
Only for the ground motions derived from ITACA or from U.S. providers (the PEER,
CESMD, and NSMP databases; see Table 1), processed records were included in SIMBAD
as disseminated by the data provider, without reprocessing raw records.
Table 1. Source of strong ground motion records included in the SIMBAD database
Country # records Data provider
Japan 220 K-NETa
KiK-neta
Italy 83 ITalian ACcelerometric Archive ITACAb1
Department of Civil Protectionb2
New Zealand 77 Institute of Geological and Nuclear Sciences: GNSc
United States 44 Center for Engineering Strong Ground Motion Data: CESMDd
PEER Strong Motion Databasee
U.S. Geological Survey National Strong Motion Project: NSMPf
Europe 18 European Strong-Motion Data Base: ESMDg
Turkey 15 Turkish National Strong Motion Project: T-NSMPh
Greece 7 Institute of Engineering Seismology and Earthquake Engineeringi
Iran 3 Iran Strong Motion Network ISMNj
ahttp://www.kyoshin.bosai.go.jp/.
b1http://itaca.mi.ingv.it/.
b2http://www.protezionecivile.gov.it/.
chttp://www.geonet.org.nz.
dhttp://strongmotioncenter.org/.
epeer.berkeley.edu/products/strong_ground_motion_db.html.
fhttp://nsmp.wr.usgs.gov/.
ghttp://www.isesd.hi.is/.
hhttp://daphne.deprem.gov.tr.
ihttp://www.itsak.gr/en/head.
jhttp://www.bhrc.ac.ir/.
1436 SMERZINI ET AL.

A further requirement for records to be included in SIMBAD is the availability of VS30 at
the recording station. We retained records only for some cases where the ground classifica-
tion according to the EC8 (or, equivalently, NTC08) was available, even without direct VS30
measurements.
Table 1provides a list of the worldwide ground motion networks used for assembling the
SIMBAD database. Most records come from the Japanese strong motion networks K-NET
and KiK-net of the National Research Institute for Earth Science and Disaster Prevention
(NIED), which provide a valuable source of high-quality digital data, with detailed informa-
tion on subsoil conditions.
DATABASE ORGANIZATION
The metadata associated to the waveforms included in the SIMBAD database are
organized into three main sections related to: (1) earthquake: event name and date, area,
MW, focal mechanism, event latitude and longitude, focal depth, fault solutions (when
available); (2) station: station code and name, station latitude and longitude, elevation,
site class according to EC8, VS30 measurements, type of instrument, analog/digital recorder,
data source; (3) record: Repi (other source-to-site distance metrics are included when
available), low- and high-cut filter values and type of filter used for record processing.
Sites were classified into five ground categories according to the European and Italian seis-
mic norms: A (VS30 ≥800 m∕s), B (360 ≤VS30 <800 m∕s), C (180 ≤VS30 <360 m∕s), D
(VS30 <180 m∕s) and E (site C or D with thickness smaller than 20 m over rigid rock).
The SIMBAD database presently consists of 467 three-component acceleration time his-
tories from 130 earthquakes worldwide. Most records come from Japan (47%), Italy (18%),
New Zealand (17%), and United States (9%), with minor contributions from Greece, Turkey,
Iran, and other European countries (9%), as shown in Figure 6. In the selection of records
for the database, priority was given to achieve an approximately uniform distribution in the
Figure 6. Distribution MW,Repi (left panel) and MW,Rhy (right panel), with indication of the
geographical origin of the records included in SIMBAD. Repi and Rhy denote the epicentral and
hypocentral distance, respectively.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1437

magnitude and distance ranges defined in the previous section, while an extension of the
database to cover a wider range of magnitude and source-to-site distance, as well as specific
near-fault conditions, such as directivity and fault mechanisms, was considered secondary
and then was not explicitly addressed. Inclusion of new records in the database aims presently
at filling in the less represented regions, in terms of magnitude, distance, and site conditions.
The latter point is probably the most critical issue, since most records are representative of
soils B (44%) and C (43%), while only a few of them are registered on rock (8%), or soft soils
D(4%) and E (1%), as depicted in Figure 7.
DEPENDENCE OF TDON MAGNITUDE AND SITE CONDITIONS
As pointed out in the previous section, significant discrepancies exist among different
seismic codes worldwide in the criteria used to define the corner period TDand its depen-
dence on magnitude and site conditions. The main differences are summarized in Table 2.
Because this is relevant to the selection of displacement spectrum–compatible GM record
sets, the corner periods, as estimated on the SIMBAD database, are investigated in this sec-
tion. To this end, the median horizontal displacement response spectra of the SIMBAD
records have been computed for different equally spaced magnitude ranges, namely
MW¼5.05.4, 5.5–5.9, 6.0–6.4, 6.5–6.9. Records have been grouped according only
to ground categories B and C of the EC8, due to the scarcity of records belonging to
soil classes A, D, and E (see Figure 7). For each magnitude range, the corner period TD
marking the beginning of the constant displacement branch of the spectrum is evaluated
using Equation 3, where D10 is approximated here by the average spectral displacement
in the period range 5–10 s.
Figure 8illustrates the median horizontal (geometric mean of the two horizontal
components) displacement response spectra of the SIMBAD records as a function of the
soil conditions (soil class B, left-hand side, and C, right-hand side) for the four magnitude
ranges under consideration. On each plot of Figure 8, the bilinear approximation for the
median displacement spectrum, computed according to Equation 3, is also superimposed
Figure 7. Distribution MW,Repi (left panel), and MW,Rhy (right panel) with indication of the
ground category (EC8-NTC08 soil classification) of the records included in SIMBAD.
1438 SMERZINI ET AL.

(thin line). TDis found to vary from about 0.60 s in the 5.0–5.4 magnitude range, up to about
2.1 s for MW¼6.5 6.9, with a slight dependence on site conditions.
Based on the summary of Table 2, it is interesting to clarify the reasons, on one side, of
the relatively good agreement of these results with the EC8 provisions, that is, TD¼1.2 s for
low-seismicity regions in Europe (M<5.5) and TD¼2.0 s for high-seismicity regions
(M≥5.5), and, on the other side, the mismatch with respect to the Italian code NTC08
and the proposed target spectrum TDSI (Equation 4). As a matter of fact, while the EC8
spectra are mainly based on the envelope of observed spectral shapes from GM records,
the Italian spectra derive either from the UH spectra proposed for the Italian territory
(Montaldo et al. 2007), in the case of NTC08, or from the PSHA at long periods introduced
by Faccioli and Villani (2009), in the case of the TDSI spectrum. Therefore, since the UH
spectra are a combination of different earthquake sources—either small-magnitude and short-
distance or large-magnitude and long-distance—typically in the short- and long-period
ranges, respectively, TDfrom UH spectra typically tends to be larger than the corresponding
value calculated from GM records.
Finally, it emerges from Figure 8(bottom panel) that the median spectral displacements
of SIMBAD records are in good agreement with the corresponding median from the GMPE
by Cauzzi and Faccioli (2008), confirming that the database is unbiased.
REXEL-DISP: COMPUTER-AIDED DISPLACEMENT-BASED RECORD
SELECTION
The availability of a high-quality digital strong motions database, along with target spec-
tra constrained both at short and long periods, may allow a more rational GM record selection
for engineering applications. To this aim, a user-friendly software, REXEL-DISP (Figure 9),
based on the same core algorithms of REXEL (Iervolino et al. 2010) and REXELite
(Iervolino et al. 2011), was developed. Although the basic features of REXEL are described
in the aforementioned papers, a brief discussion herein is still worthwhile because significant
enhancements in the interface and options are now available with respect to what was
described in former publications.
REXEL-DISP has been freely available at http://www.reluis.it since January 2012. The
software allows one to select combinations of (multi-component) horizontal accelerograms
Table 2. Range of variability of the corner period TDaccording to
selected seismic norms worldwide and the proposed TDSI (Equation 4)
Displacement spectrum Range of variability of TD
EC8 1.2 s, for M<5.5; 2.0 s, for M≥5.5
NTC08 Ranging from about 1.8 s (for ag∕g¼0.05) to 2.8 s
(for ag∕g¼0.30)
NZS 1170 3.0 s
ASCE 7-10 Ranging from about 4 s (for M¼6.06.5) to 16 s
(for M¼8.08.5)
TDSI 3.7 s median value, σ¼1.4s
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1439

whose average response spectrum is compatible with a target displacement spectrum in
an arbitrary period range. The record search is carried out such that the response spectral
shape of individual records is as similar as possible to the target one in the same period
interval.
Spectrum-compatible record set selection is based on four main steps:
1. The first is the definition the target spectrum; the latter may be: TDSI for Italy, the
design spectra according to NTC08 and EC8, or an arbitrarily defined one (typically
a UH spectrum entered by the user).
2. The second step, to be followed only in case of the TDSI and the NTC08, is the
definition of additional input parameters, namely, geographical coordinates of the
site, latitude and longitude in decimal degrees, Site Class (according to NTC08/EC8
classification), Topography Category (as in NTC08/EC8), Nominal Life,Functional
Type, and Limit State. These parameters are required to calculate the elastic design
spectrum according to these codes.
3. The third step involves the definition of the bins in which the software will search
for spectrum-compatible sets. In fact, the software allows one to search for records
Figure 8. Median horizontal (geometric mean) displacement response spectra of the SIMBAD
records as a function of soil conditions according to EC8-NTC08 ((a): soil type B; (b): soil type C)
for four different magnitude ranges. Top panel: the superimposed thin lines denote the bilinear
approximation used for the determination of the corner period TD(see labels indicated by the
arrows). Bottom panel: comparison with the median spectrum as predicted by the GMPE of
Cauzzi and Faccioli (2008) for representative values of magnitude and hypocentral distance
(Rhy ¼18 km).
1440 SMERZINI ET AL.

within SIMBAD belonging to the same site class of the target spectrum, or to any
site class (i.e., recordings from different site conditions may fit the target spectrum
and be selected). Moreover, the user may select records corresponding to prescribed
magnitude and source-to-site distance intervals, or to selected ranges of several GM
intensity measures. Once the selection options are defined, the software returns the
number of records, along with the corresponding number of events, available in the
SIMBAD database.
4. The spectra returned by the preliminary search in Step 3, are then used by REXEL-
DISP to find suites of records (of sizes 1, 7, or 30), whose average response spec-
trum is compatible with the target one in an arbitrary period range, ½T1;T2, between
0 s and 10 s. Spectral compatibility is ensured within a tolerance band defined by the
user (upper and lower tolerances to be provided); for example, 0%lower tolerance
should be set to ensure that the average spectrum is not lower than the target one, as
specified by the ASCE 7–10 recommendations quoted previously.
The sets of compatible records may consist either of single-component or of pairs of
horizontal component accelerograms. In the latter case, the software will return 14 (from
7 recording stations) or 60 (from 30 recording stations) records time histories to be used,
for example, for the analysis of three-dimensional structures.
It is worth underlining that to select spectrum-compatible records, the software analyzes
all possible combinations of spectra identified in the preliminary search (Step 3) and checks
whether each combination approaches—in an average sense and within the assigned
tolerances—the target spectrum.
Site definition and
reference
structural code
selection (Italian,
European, or S5).
Limit state
definition for
code-based
spectrum
construction,or be
user defined
spectrum input.
Target spectrum
and compatibility
bounds.
Definition of
range of periods
where the
average of the set
has to match the
target spectrum
and matching
tolerances.
Selection of
search for one set
only or for
multiple sets, and
search for original
or amplitude-
scaled records.
Choice of set
size: one, seven
or thirty muti-
component
records.
Choice of how many horizontal components should have each record in the set.
Definition of bins
of interest in
terms of source,
intensity, and site
parmeters.
Output management and secondary options
Figure 9. Graphic user interface of REXEL-DISP and main required information or record selec-
tion options.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1441

An especially important feature of REXEL and REXEL-DISP is that the list of records
determined in Step 3 is sorted in ascending order with respect to the following parameter:
EQ-TARGET;temp:intralink-;e5;41;615δj¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
NX
N
i¼1Sd;jðTiÞSd;targetðTiÞ
Sd;targetðTiÞ2
v
u
u
t(5)
where Sd;jðTiÞdenotes the spectral ordinate of record jat period Ti,Sd;target ðTiÞis the value of
the displacement target spectrum at the same period, and Nis the number of periods within
the considered range. The parameter δjin Equation 5 measures the deviation of the spectrum
of an individual record from the target one. Preliminary ordering ensures that the first com-
binations found are those with the smallest individual scattering with respect to the target
spectrum (i.e., Iervolino et al. 2008). In fact, the I’m feeling lucky option can be used to stop
the analysis as soon as the first combination is found, which is expected to be one of the best
record sets according to this criterion.
REXEL-DISP allows one to obtain sets of either unscaled or scaled accelerograms
compatible with the target spectrum. When the scaling option is chosen, a search for
spectrum-compatible scaled record sets is carried out in terms of non-dimensional values,
by normalizing both record and target spectra by the corresponding value at 10 s. When
a combination is found, the normalization factor is returned to the user as the scale factor
for the specific record in the set and is applied uniformly to all spectral ordinates (i.e., linear
scaling). In this case, it is also possible to specify the maximum mean scale factor (SF)
allowed, and the software will discard combinations with an average SF larger than that
assigned.
Quality measures for the output record sets may be derived from Equation 5. Indeed, δavg
is introduced as a measure of average spectrum matching by replacing the spectral ordinate of
individual spectra with that of the average spectrum of the output combination, while δmax is
defined as the largest value from Equation 5among those computed for all records in a given
combination. Therefore, the former is a measure of average spectral goodness of fit, while the
latter is a measure of record-to-record variability within a combination (Iervolino et al. 2010).
As for other REXEL software, complementing functions of REXEL-DISP are related to:
visualization of results; the return of selected waveforms, spectra, and metadata to the user;
repetition of the search excluding an undesired waveform from an output combination; or
visualization of acceleration spectra compatibility for the selected set. Applications in the
next section will show how, in fact, this software may render record displacement-based
records selection feasible and practice-ready.
APPLICATION EXAMPLES
Some applications for representative Italian sites are presented hereafter to clarify the
performance of ground motion selection, depending on the selected target spectrum, on
the seismicity level of the site of interest and on the period range bounds.
1442 SMERZINI ET AL.

EFFECT OF TARGET SPECTRAL SHAPE
As a first example, REXEL-DISP is used to select seven horizontal ground motions com-
patible either with the NTC08 design spectrum or with the TDSI (Equation 4). Reference is
made to the city of L’Aquila for a return period TR¼475 yr, soil class B, and no topography
special conditions. Selection criteria involve a broadband spectrum compatibility in the per-
iod range from T1¼0.5 s to T2¼8s, with 20%lower and 20%upper tolerance. It is worth
underlining that no scaling factor is introduced, and no specific magnitude and distance inter-
val, nor site class, have been specified.
With the above parameters, the software returns the best suites of seven accelerograms
compatible with the NTC08 spectrum and with the TDSI presented in Figures 10a and
10b, respectively. A reasonable agreement is found between the average displacement spec-
trum and the NTC08 spectrum, resulting in a maximum spectral deviation δmax of the indi-
vidual accelerograms limited to less than 30%and an average spectral deviation δavg of
around 10%, even in case no scaling factors are applied to the selected records. However,
the response spectral ordinates of selected records divert significantly from the decreasing
branch of the design spectrum, starting from 5 s.
The agreement with selected records improves when considering the TDSI (Figure 10b),
with a reduced scatter of the mean spectrum (δavg ∼6%). Furthermore, it is noted that the
target spectral shape fits the trend of spectral ordinates of GM records, especially beyond 5 s,
much better. This is a common finding for all considered Italian sites, confirming that the
sharply decreasing trend proposed both by the EC8 and by the NTC08 spectra is not suitable
for long periods, as already pointed out by Faccioli et al. (2004).
Figure 10. Displacement response spectra, at 5%damping, of the set of 7 unscaled horizontal
ground motions compatible with the NTC08 design spectrum (a) and the TDSI of Equation 4
(b) for the reference site of L’Aquila (soil class B) and for TR¼475 yr. The thin solid lines
represent the individual displacement response spectra of the selected accelerograms, the
thick black line denotes the average spectrum of the output combination, while the thick
gray line indicates the target displacement spectrum along with the corresponding lower and
upper tolerances (dashed gray lines). The vertical dashed lines denote the range of periods
where spectral compatibility of the average is ensured.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1443

EFFECT OF HAZARD LEVEL
To evaluate the influence of the level of seismicity on the ground motion selection pro-
cedure, the Udine site is considered, which is characterized by lower hazard with respect to
L’Aquila. The selected target spectrum is TDSI, while the other parameters are the same as in
the previous search. The output combination returned by the software at L’Aquila and Udine
are compared in Figures 11a and 11b, respectively. A good agreement is achieved in Udine as
well, leading to values of δavg and δmax of about 9%and 28%, respectively.
It is worth noting that the average moment magnitude and epicentral distance of the out-
put suite of strong motion records is about 6.6 and 12 km in L’Aquila and about 6.5 and
20 km in Udine. These values are in good agreement with the disaggregation results at both
sites obtained by the long-period PSHA analyses (M. Villani, pers. comm.), namely modal M
in the range 6.4–6.7 and Rof 10 km (L’Aquila), 6.2–6.5 and 20 km (Udine). Note that such
an agreement has been obtained with unscaled records and with no specification of the mag-
nitude and distance range. Therefore, it turns out, as a major achievement, that GM record
selection based on displacement-spectrum compatibility tends to naturally fit the proper mag-
nitude and distance ranges, provided that the target spectrum at long periods is constrained by
a suitable seismic hazard analysis.
EFFECT OF PERIOD RANGE FOR SPECTRUM COMPATIBILITY
In this example, the proposed approach is explored, considering different period ranges
where spectral compatibility is ensured. For this purpose, we consider the same example at
L’Aquila (class B) for the TDSI (TR¼475 yr) and compare the results using three different
period ranges, namely, (a) T¼[0.5, 3] s, (b) [0.5, 8] s, and (c) [0.15, 10] s, with a upper and
lower tolerance of 20%. The results are illustrated in Figure 12, not only in terms of dis-
placement response spectra (left-hand side), but in terms of acceleration spectra as well
(right-hand side). By examination of these results, some interesting comments can be derived
on the use of the software:
Figure 11. Output combinations of real ground motions compatible with the TDSI for
TR¼475 yr at L’Aquila (a) and at Udine (b).
1444 SMERZINI ET AL.

•Restricting the period range of spectrum compatibility improves the search perfor-
mance in the selected range, but returns a set of accelerograms with large variability
at longer (and shorter) periods, thus involving a reduced homogeneity of the suite in
terms of magnitude.
•Extending the period range to include short periods involves a moderate decrease of
search performance in terms of maximum error (δavg ¼8%and δmax ¼40%), but it
allows one to achieve a broadband spectrum compatibility, without scaling the
ground motion records.
CONCLUSIONS
In the framework of ongoing research activities to improve the representation of
the seismic action in Italy, this study addressed the selection of displacement-spectrum com-
patible earthquake ground motions from real accelerograms. The following tasks were
achieved:
Figure 12. Displacement (left-hand side) and acceleration (right-hand side) response spectra of
the record sets compatible with the TDSI (TR¼475 yr) for an ideal site in the city of L’Aquila
considering three different ranges of vibration periods: (a) T¼0.53s, (b) T¼0.58s, and
(c) T¼0.1510 s. For clarity in the right panel only the average spectrum of the selected accel-
erograms is shown along with the target spectrum.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1445

•After comparing the provisions from several seismic codes worldwide, target design
spectra for Italian sites (TDSI), tailored to fit results of PSHA both at short and long
periods, were proposed.
•A high-quality strong-motion database, SIMBAD, was set up, consisting of digital
recordings from shallow crustal earthquakes with epicentral distances of less than
about 30 km, with the purpose of covering as homogenously as possible the mag-
nitude and distance ranges of interest for seismic hazard at Italian sites.
•The REXEL-DISP software was developed, as a byproduct of REXEL, and spe-
cifically featured to search for displacement-spectrum compatible records from
the SIMBAD database.
Application examples at some representative Italian sites suggest that the combination of
(i) a target spectrum calibrated both at long and short periods from seismic hazard studies and
(ii) a digital strong-motion database covering in a relatively homogeneous way the ranges of
magnitude, distance and site conditions which dominate seismic hazard, together with the
search engine implemented in the REXEL-DISP software, allow in most cases to obtain sets
of unscaled, or lightly scaled, records sets closely approaching the target spectrum in a broad
period range. Furthermore, the magnitude and distance ranges of the selected records are
consistent with the seismicity levels deduced from the hazard disaggregation study. This
means that such selected records may be used with a similar level of confidence, either
in terms of acceleration or of displacement spectrum, for a variety of applications, where
constraining compatibility of records within a limited period range may not be satisfactory,
such as for nonlinear time history analyses of structures with large participation factors of
higher modes, for nonlinear dynamic soil-structure interaction problems, and for seismic soil
stability studies.
Finally, it is noted that this approach was applied to Italian sites; however, it can be easily
extended to any other site for which a reliable definition of long-period ground motion for
seismic design is available.
ACKNOWLEDGMENTS
This work was funded by the Italian Department of Civil Protection and by the Rete dei
Laboratori Universitari di Ingegneria Sismica (ReLUIS) in the framework of the DPC-
RELUIS Research Programme (2010–2013). The authors are grateful to Manuela Villani
for providing results in terms of disaggregation of the long-period PSHA for Italy. The
authors acknowledge the providers of the records used to construct the SIMBAD database,
listed in Table 1, for continuous support to maintain and disseminate to researchers world-
wide such a wealth of high-quality information, which is crucial for progress in earthquake
engineering. Remarks by three anonymous reviewers helped to significantly improve the
quality of the paper.
REFERENCES
American Society of Civil Engineering (ASCE), 2010. Minimum Design Loads for Buildings and
Other Structures (7–10), Standards ASCE/SEI 7-10, Reston, VA.
Baker, J. W., and Cornell, C. A., 2006. Spectral shape, epsilon and record selection, Earthquake
Engineering & Structural Dynamics 35, 1077–1095.
1446 SMERZINI ET AL.

Baker, J. W., 2011. Conditional mean spectrum: Tool for ground-motion selection, Journal of
Structural Engineering, ASCE, 137,322–331.
Bojórquez, E., and Iervolino, I., 2011. Spectral shape proxies and nonlinear structural response,
Soil Dynamics and Earthquake Engineering 31, 996–1008.
Boore, D. M., 2001. Effect of baseline corrections on displacements and response spectra for
several recordings of the 1999 Chi-Chi, Taiwan, earthquake, Bulletin of Seismological Society
of America 91, 1199–1211.
Cauzzi, C., Faccioli, E., and Paolucci, R., 2007. A reference model for prediction of long-
period response spectral ordinates, Deliverable D2 (Task 1), Project S5 –Seismic Input in
terms of expected spectral displacements, 2005–2007 DPC-INGV Agreement, Politecnico
di Milano.
Cauzzi, C., and Faccioli, E., 2008. Broadband (0.05 s to 20 s) prediction of displacement
response spectra calibrated on worldwide digital records, Journal of Seismology 12,
453–475.
Consiglio Superiore dei Lavori Pubblici (CS.LL.PP.), 2008. DM 14 Gennaio 2008. Norme tec-
niche per le costruzioni, Gazzetta Ufficiale della Repubblica Italiana,29 (in Italian).
Crouse, C. B., Leyendecker, E. V., Somerville, P. G., Power, M., and Silva, W. J., 2006. Devel-
opment of seismic ground-motion criteria for the ASCE 7 standard, Proceedings of the 8th
U.S. National Conference on Earthquake Engineering, Paper no. 533, 18–22 April 2006,
San Francisco, CA.
European Committee for Standardization (CEN), 2004. Eurocode 8: Design Provisions for
Earthquake Resistance of Structures, Part 1.1: General Rules, Seismic Actions and Rules
for Buildings, Pren1998-1.
Faccioli, E., Paolucci, R., and Rey, J., 2004. Displacement spectra for long periods, Earthquake
Spectra 20, 347–376.
Faccioli, E., and Villani, M., 2009. Seismic hazard mapping for Italy in terms of broadband dis-
placement response spectra, Earthquake Spectra 25, 515–539.
Figini, R., and Paolucci, R., 2009. Site effects at long periods from digital strong motion records
of the Kik-net, Japan, Journal of Earthquake Engineering 13, 567–584.
Foti, S., and Paolucci, R., 2012. Influence of strong motion processing on numerical simulation of
soil-structure interaction, in Proc. 2nd Int. Conf. on Performance-Based Design in Earthquake
Geotechnical Engineering, Paper n.12.09, Taormina, Italy.
Haselton, C. B., Baker, J. W., Bozorgnia, Y., Goulet, C. A., Kalkan, E., Luco, N., Shantz, T.,
Shome, N., Stewart, J. P., Tothong, P., Watson-Lamprey, J., and Zareian, F., 2009. Evaluation
of Ground Motion Selection and Modification Methods: Predicting Median Interstory Drift
Response of Buildings, PEER Report 2009-01, prepared by the PEER Ground Motion Selec-
tion and Modification Working Group, edited by Curt B. Haselton, for the Pacific Earthquake
Engineering Research Center, University of California, Berkeley.
Iervolino, I., Maddaloni, G., and Cosenza, E., 2008. Eurocode 8 Compliant Real Record Sets for
Seismic Analysis of Structures, Journal of Earthquake Engineering 12,54–90.
Iervolino, I., Galasso, C., and Cosenza, E., 2010. REXEL: computer aided record selec-
tion for code-based seismic structural analysis, Bulletin of Earthquake Engineering 8,
339–362.
Iervolino, I., Galasso, C., Paolucci, R., and Pacor, F., 2011. Engineering ground motion record
selection in the ITalian ACcelerometric Archive, Bulletin of Earthquake Engineering 9,
1761–1778.
GROUND MOTION RECORD SELECTION BASED ON BROADBAND SPECTRAL COMPATIBILITY 1447

Jayaram, N., Lin, T., and Baker, J. W., 2011. A computationally efficient ground motion selection
algorithm for matching a target response spectrum mean and variance, Earthquake Spectra 27,
797–815.
Montaldo, V., Meletti, C., Martinelli, F., Stucchi, M., and Locati, M., 2007. On-Line seismic
hazard data for the new Italian building code, Journal of Earthquake Engineering 11,
119–132.
National Institute of Standards and Technology (NIST), 2011. Selecting and Scaling Earthquake
Ground Motions for Performing Response-History Analyses, Technical Report NIST GCR
11-917-15, prepared for the Engineering Laboratory of the National Institute of Standards
and Technology (NIST) under the National Earthquake Hazards Reduction Program
(NEHRP) Earthquake Structural and Engineering Research Contract SB134107CQ0019,
Task Order 69220, November 2011.
Paolucci, R., Rovelli, A., Faccioli, E., Cauzzi, C., Finazzi, D., Vanini, M., Di Alessandro, C., and
Calderoni, G., 2008. On the reliability of long-period response spectral ordinates from digital
accelerograms, Earthquake Engineering & Structural Dynamics 37, 697–710.
Paolucci, R., Pacor, F., Puglia, R., Ameri, G., Cauzzi, C., and Massa, M., 2011. Record proces-
sing in ITACA, the new Italian strong-motion database, in Earthquake Data in Engineering
Seismology, Geotechnical, Geological, and Earthquake Engineering (S. Akkar, P. Gülkan,
and T. van Eck, Eds.), 14,99–113.
Priestley, M. J. N., Calvi, G. M., and Kowalsky, M. J., 2007. Displacement-Based Seismic Design
of Structures, IUSS Press, Pavia, Italy.
Standards New Zealand (NZS), 2004. 1170.5:2004 Structural design actions, Part 5: Earthquake
actions –New Zealand, Standards New Zealand, Wellington, NZ.
Wang, G.-Q., Boore, D. M., Tang, G., and Zhou, X., 2007. Comparisons of ground motions from
colocated and closely spaced one-sample-per-second global positioning system and acceler-
ograph recordings of the 2003 M6.5 San Simeon, California, earthquake in the Parkfield
region. Bulletin of Seismological Society of America 97(1B), 76–90.
(Received 23 May 2012; accepted 5 March 2013)
1448 SMERZINI ET AL.