See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/234720062

Design spectra and amplification factors for Eurocode 8

Article  in  Bulletin of Earthquake Engineering · October 2012

DOI: 10.1007/s10518-012-9367-6

CITATIONS READS

70 1,416

3 authors:

Kyriazis Pitilakis Evi Riga

Aristotle University of Thessaloniki Aristotle University of Thessaloniki
348 PUBLICATIONS   5,333 CITATIONS    36 PUBLICATIONS   305 CITATIONS   

SEE PROFILE SEE PROFILE

Anastasios Anastasiadis
Aristotle University of Thessaloniki
89 PUBLICATIONS   1,528 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

SAFELAND View project

GRATIS - Ground Real-time Active Tremor Inhibition System View project

All content following this page was uploaded by Evi Riga on 24 September 2015.

The user has requested enhancement of the downloaded file.

This is the authors’ accepted manuscript. The final publication
is available at Springer via http://dx.doi.org/10.1007/s10518-
012-9367-6

Design spectra and amplification factors for

Eurocode 8
Kyriazis Pitilakis, Evi Riga, Anastasios Anastasiadis
Aristotle University of Thessaloniki, P.O.B. 424, 54124, Thessaloniki, Greece
00302310995842
kpitilak@civil.auth.gr, eviriga@civil.auth.gr, anas@civil.auth.gr
Abstract
Eurocode 8-Part 1 accounts for site effects through the suggestion of appropriate elastic design
spectra and amplification factors based on different soil classes. The influence of local site
conditions is reflected in Eurocode 8 (EC8) spectra with the shape of the PGA-normalized
response spectra and the so-called “soil factor” S, which represents ground motion amplification
with respect to outcrop conditions. In the present work we use a global strong-motion database,
which was compiled in the framework of SHARE FP7 Collaborative Project of the European
Union, to evaluate the present EC8 normalized spectra and associated amplification factors for
different soil conditions. Uncertainties in the estimation of improved soil amplification factors are
captured using a logic-tree approach, which allows the use of alternative models and methods in an
effective way. The influence of PGA range of the strong motion records included in the dataset on
spectral shape and soil factors is also examined. The results indicate that for all datasets examined,
EC8 normalized elastic acceleration spectra match the empirical data to a satisfactory extent, while
estimated soil factors for classes C and D are significantly higher than the ones proposed in
Eurocode 8. The proposed design spectra and the new amplification factors are validated with a
smaller dataset, which includes records only from stations with well documented soil profiles
regarding the dynamic soil properties.
Keywords: Eurocode 8, Design spectra, Soil amplification factors, Logic tree

1
Introduction
Based on the results of extended research studies and valuable records, after
several recent earthquakes like 1989 Loma Prieta, 1994 Northridge, 1995 Kobe
and many others all over the world, current seismic code provisions have largely
accepted the significant role of local site conditions. Their influence is described
through appropriate elastic design spectra based on different soil categories and
levels of shaking intensity. These provisions, namely IBC and Eurocode 8,
recognizing that spectra amplification is a nonlinear function of intensity of
expected motion and of site conditions, attempt to categorize soil conditions using
rational quantitative criteria widely used in engineering practice. The main
parameter proposed for representing depth-dependent velocity profiles and soil
categorization is the Vs,30, i.e. the time-based average value of shear wave velocity
in the upper 30m of the soil profile. This parameter, initially introduced by
Borcherdt and Glassmoyer (1992) and Borcherdt (1994) as a means for
classifying sites for building codes, despite several shortcomings, which are
widely discussed, seems to overcome other difficulties in current practice of the
quarter-wavelength velocity approach (Joyner et al. 1981), which produces
frequency-dependent values. EC8 (CEN 2004) in Europe and the International
Building Code (ICC 2012) in USA both share basically the same principles and
approaches regarding soil classification and design procedures.

In EC8 the Vs,30 parameter is used along with NSPT blow count, plasticity index PI
and undrained shear strength Cu to define five soil types (A to E), while two extra
special ground types (S1 and S2) are also proposed for special soils (Table 1).
Linear response spectra are proposed for two different levels of seismic action,
Type 1 and Type 2. Type 1 spectra have more energy in long-period motions and
are proposed for use in regions having high seismic activity and stronger
earthquakes; it is estimated that Type 1 spectra should be used when the
earthquakes that contribute to the seismic hazard have a surface wave magnitude
Ms greater than 5.5. Type 2 spectra are recommended for Ms≤5.5, having larger
normalized spectral amplitudes at short periods. Ground motion amplification to
account for local soil and site effects is established through a constant soil factor
S, which increases uniformly the normalized elastic response spectra in all
periods.

The design response spectra for different soil classes may deviate, sometimes
seriously, from recorded data (Stewart et al. 2001; Choi and Stewart 2005; Pousse
et al 2005).

The aim of the present work is to examine the accuracy of EC8 normalized
response spectra and soil amplification S factors for different soil categories
proposed in EC8 using a very large set of ground motion records from a
worldwide strong motion database recently compiled in the framework of the
European research project called ‘Seismic Hazard Harmonization in Europe’
(SHARE) (Yenier et al. 2010).The final goal is to propose improved parameters
and normalized shapes for EC8 elastic response spectra, which could potentially
be included in an EC8 revision. Grouping the empirical normalized spectra based
on soil class and seismicity type, the shape of the PGA-normalized response
spectra proposed in EC8 was validated. The corresponding soil amplification
factors were estimated using a logic tree approach, which allows the use of

2
alternative models, capturing in this way effectively the epistemic uncertainties of
the methodologies used for this purpose.

The analyses were conducted using three different sub-sets of data having
different levels of PGA, all extracted from the main database. The criteria for
selecting the different datasets are described in the following section. In order to
confirm the validity of the results obtained from the three datasets, a fourth dataset
was also used, extracted from a smaller database compiled within SHARE
(Pitilakis et al. 2012), which includes records only from sites with very well
known and documented soil profiles concerning their dynamic properties and
depth of the bedrock.

Table 1 Ground types defined in EC8 (CEN 2004)

Ground Description of stratigraphic profile Parameters
type
Vs,30 (m/s) NSPT Cu
(blows/30cm) (kPa)
A Rock or other rock-like geological >800 - -
formation, including at most 5 m of weaker
material at the surface.
B Deposits of very dense sand, gravel, or very 360-800 >50 >250
stiff clay, at least several tens of meters in
thickness, characterized by a gradual
increase of mechanical properties with
depth.
C Deep deposits of dense or medium-dense 180-360 15-50 70-250
sand, gravel or stiff clay with thickness
from several tens to many hundreds of
meters.
D Deposits of loose-to-medium cohesionless <180 <15 <70
soil (with or without some soft cohesive
layers), or of predominantly soft-to-firm
cohesive soil.
E A soil profile consisting of a surface
alluvium layer with Vs values of type C or
D and thickness varying between
about 5 m and 20 m, underlain by stiffer
material with Vs>800 m/s.
S1 Deposits consisting, or containing a layer at <100 - 10-20
least 10 m thick, of soft clays/silts with a (indicative)
high plasticity index (PI>40) and high
water content
S2 Deposits of liquefiable soils, of sensitive
clays, or any other soil profile not included
in types A – E or S1.

2 Data selection and site classification

In the framework of SHARE, an extended strong motion database was compiled,
by unifying existing national and international databases, such as the European
strong motion database, the Turkish national strong-motion database, the Next
Generation Attenuation database, the Kik-Net database, the global worldwide
database compiled by Cauzzi and Faccioli, the Internet Site for European Strong-
motion Data and the Italian Accelerometric Archive database (Yenier et al. 2010).
The database covers earthquakes dating back to 1930s and contains a total of
14193 recordings from 2448 earthquakes recorded at 3708 stations. Further

3
information on the compilation and characteristics of the database can be found in
SHARE (http://www.share-eu.org/). The version of the database used in the
present study is v3.1 (March 2010) and contains 13500 records.

For the purpose of this study, only recordings corresponding to sites with
available classification according to EC8 classification scheme were used. An
event magnitude criterion was set, which allowed the use of records with surface
wave magnitude Ms≥4, excluding in this way small magnitude events. Strong
motion recordings with usable spectral period less than 2.5s were also discarded,
leading to a dataset of 7161 3-component accelerograms, hereinafter called DS1.
Magnitude M and peak ground acceleration PGA distributions with respect to the
epicentral distance R, for the records in DS1 are shown in Figures 1a and 1b
respectively.

Fig. 1 Data coverage of the DS1 database records (7161 accelerograms) in terms of (a) Magnitude
M - Epicentral distance R and (b) Peak Ground Acceleration (PGA) – Epicentral distance R, used
to evaluate acceleration response spectra and soil amplification factors according to EC8 site
categories

It is observed that there are significantly fewer records with PGA values
exceeding 200 cm/sec2 and many weak motion records with peak values less than
20cm/sec2. For this reason, and considering the fact that design spectra should be
derived basically from records of strong earthquakes, all the analyses were
performed not only for the whole dataset (DS1), but also for two smaller datasets,
one with records with PGA greater or equal to 20cm/sec 2 (hereinafter DS2), and
one with records with PGA greater or equal to 150cm/sec 2 (hereinafter DS3),
which may be considered as representative for high seismicity regions.

In order to confirm the validity of the results obtained from the three
aforementioned datasets, a fourth dataset was also used, hereinafter called DS4,
which includes strong motion records only from sites with well known and
documented soil profiles concerning dynamic properties and the depth of the rock
basement. These records were extracted from a complementary database compiled
within SHARE (hereinafter SHARE-AUTH database, described in detail in
Pitilakis et al. 2012), which contains 3666 records from 536 recording stations
from Greece, Italy, Japan, Turkey and USA.

All sites in SHARE-AUTH database dispose accurately measured shear wave

velocity profiles, which were carefully collected from a number of sources, such
4
as the Italian Accelerometric Archive (Working Group ITACA 2010,
http://itaca.mi.ingv.it), Kik-Net (http://www.kik.bosai.go.jp/), K-Net
(http://www.k-net.bosai.go.jp/), the Turkish national strong-motion database
(Akkar et al. 2010), ROSRINE program (http://gees.usc.edu/ROSRINE/), Dr.
David Boore’s personal webpage (http://www.daveboore.com/) and USGS open-
file reports. The aforementioned criteria regarding the surface magnitude and
spectral period of the previous three datasets (DS1, DS2 and DS3) were also
applied to DS4 dataset, while the range of PGA values is equivalent to DS2
dataset (PGA≥20cm/sec2). The main features of the four different datasets used
are presented in Table 2.

Table 2 Main features of datasets used

Dataset Database of Number of Ms range PGA range
origin records (cm/s2)
DS1 SHARE 7161 4-7.9 ≤1400
DS2 SHARE 3500 4-7.9 20<PGA≤1400
DS3 SHARE 559 4-7.9 150<PGA≤1400
DS4 SHARE- 715 4-7.5 20<PGA≤1302
AUTH

The Vs profiles of the selected stations have been measured with a variety of
surveying methods; for most of the sites (70% of the total sample) Vs profiles have
been obtained from borehole measurements (e.g. cross-hole, down-hole), while
for the rest of the sites, for example the Turkish stations (24% of the total sample),
the Vs profiles have been evaluated using surface geophysical surveys. As a result,
SHARE-AUTH database constitutes a very reliable set of empirical measurements
for estimation of influence of local site conditions.

For these sites Vs,30 were calculated using Eq. (1):

Vs ,30  30 / h
i 1, N
i Vi
(1)
where hi and Vi denote respectively the thickness and shear-wave velocity of the i-
th layer, in a total of N, existing in the top 30m. However, for some sites (mostly
K-net sites) the available shear-wave velocity models did not extend down to
30m. In that case, Vs,30 was estimated using Eq. (2), which assumes a constant
shear wave velocity from the bottom of the available velocity model to the depth
of 30m:
Vs ,30  30 / tt (d )  (30  d ) / Vb 
(2)
where d is the depth to the bottom of the velocity model, tt(d) is the travel time to
depth d and Vb is the velocity at the bottom of the model (Boore 2004).

The number of available strong motion records per EC8 soil class and seismicity
type in each one of the four datasets is shown in Table 3. It should be noted that
the majority of the strong motion records come from sites which are classified as
B and C, while few data are actually available for soil classes D and E. The
records in DS3 dataset, although having relatively high PGA values, do not come
only from Type 1 earthquakes, but also from Type 2 earthquakes. All records of
datasets DS1, DS2 and DS3 given in Table 3 were utilized for the validation of

5
EC8 spectral shapes. However, as far as the estimation of soil amplification
factors is concerned, it was not feasible to use all the records of Table 3, since the
methodology that was applied sets some extra limitations to the available
metadata of the records that can actually be used. These limitations are explained
in detail in section 3.

Table 3 Number of strong motion records for each dataset

Soil
DS1 DS2 DS3 DS4
Class
Type 2 Type 1 Type 2 Type 1 Type 2 Type 1 Type 2 Type 1
A 402 264 105 125 9 23 15 19
B 1508 1896 419 1151 38 214 172 163
C 1133 1775 353 1261 44 219 128 132
D 10 4 3 1 - - 3 1
E 73 96 33 49 5 7 33 49
Total 3126 4035 913 2587 96 463 351 364

3 Methodology and example results

3.1 Normalized acceleration response spectra
Datasets DS1, DS2 and DS3 were at first used to validate the spectral shapes
proposed in EC8. Datasets DS1 and DS2 were utilized for both seismicity types,
while DS3 was considered as more representative for Type 1 seismicity. The 5%
damped spectral values of the two horizontal components were included in the
original database. In order to combine the two horizontal components into one
single earthquake intensity measure, the geometric mean (GM) of the response
spectra for the two orthogonal horizontal components of motion was calculated
for each strong motion record. Each GM spectrum was then normalized to the GM
PGA of the record. All available records were grouped for the EC8 soil classes
and seismicity type. For each soil class and seismicity type the median of these
normalized spectra was calculated, along with the 16th and 84th percentiles, which
represent the average minus and plus one standard deviation respectively. The
normalized empirical response spectra were then compared with the normalized
EC8 design spectra.

3.2 Soil amplification factors

To account for the epistemic uncertainties, amplification factors were estimated
for all EC8 soil classes using a logic tree approach, which combines different
methods. Each of the methods is assigned a weighting factor that is interpreted as
the relative likelihood of that method being correct. In this way, the uncertainties
of the different methods can be captured in a more effective way. The logic tree
that was implemented, which is shown in Figure 2, combines two state-of-the-art
methods with equal weights. The first method (Approach 1, Choi and Stewart
2005) calculates period-dependent amplification factors using Ground Motion
Prediction Equations (GMPEs) for the estimation of reference acceleration
spectral values, while the second one (Approach 2, Rey et al. 2002) calculates a
constant period-independent amplification factor with respect to the rock sites of
the database. Reference rock site conditions are defined through Vs,30=800m/s.

6
Fig. 2 Logic tree used for estimation of soil amplification factors

3.2.1 Approach 1
In this approach amplification is evaluated by normalizing the spectra of recorded
motions to the reference outcrop spectrum obtained from GMPEs. It constitutes a
non-reference site approach, since it does not depend on the availability of a
nearby adequate reference site (i.e. a site with negligible site response, Field and
Jacob 1995), having the advantage of allowing the use of a large amount of strong
motion data. Similar non-reference site approaches have been used among others
by Sokolov (1997) and Sokolov et al. (2000), who used locally derived
attenuation functions for Fourier amplitude spectra in order to estimate site
amplification for Racha Seismogenic Zone in Caucasus and Taipei basin
respectively. Field (2000), Lee and Anderson (2000) and Steidl (2000) used
different GMPEs for spectral acceleration to evaluate amplification factors for the
southern California region from differences between observed and predicted
ground-motion levels (residuals). Stewart et al. (2003) developed discrete
nonlinear amplification factors as a function of site category for several soil
classification schemes relative to a modified attenuation relationship, while Choi
and Stewart (2005) developed nonlinear amplification factors as a continuous
function of Vs,30 using three modified rock attenuation relationships for active
regions.

The amplification factor for ground motion j within site class i, S ij, at spectral
period T, was evaluated from the geometric mean of 5% damped acceleration
response spectra for the two horizontal components of shaking, GMij, and the
reference ground motion for the site, (GMr)ij, using Eq. (3) (Choi and Stewart
2005):
Sij (T )  GM ij / (GM r )ij
(3)
GMij and (GMr)ij were computed at the same spectral period, which was varied
from 0 to 2.5s.

For each dataset the reference motion parameter (GMr)ij was estimated as the
weighted average of the rock predictions of the four GMPEs proposed in SHARE
for active shallow crustal regions (Delavaud et al. 2012), since the majority of the
SHARE database stations are in active regions. The proposed models and
corresponding weights are given in Table 4.

7
Table 4 SHARE logic tree for ground motion prediction for active shallow crustal regions.
Selected models and corresponding weights (Delavaud et al. 2012)
Model Abbreviation Proposed weight
Akkar and Bommer (2010) A&B 0.35
Cauzzi and Faccioli (2008) C&F 0.35
Zhao et al. (2006) Zh 0.10
Chiou and Youngs (2008) C&Y 0.20

The definition of rock-site conditions differs among the four GMPEs. In the
Akkar and Bommer (2010) GMPE, rock sites are defined as having V s,30>750m/s,
which is very close to the selected reference V s,30 of 800 m/s. Cauzzi and Faccioli
(2008) use the EC8 classification for the estimation of the site term in their
GMPE, thus rock sites are the ones with a V s,30>800m/s, which is compatible with
the selected reference site conditions. Zhao et al. (2006) use 4 soil classes based
on the natural period of the site. The site term that corresponds to soil class I (rock
sites with a Vs,30 between 600 and 1100 m/s) was used to estimate the reference
ground motion. In Chiou and Youngs (2008) sites effects are represented by
smooth functions of V s,30. The reference site in C&Y has a Vs,30 equal to 1130m/s,
which differentiates a little from the prescribed reference site conditions. The
relatively small weight (0.2) of this GMPE in the logic tree, however, encouraged
us to use the GMPE without any modification.

The four GMPEs require the knowledge of different parameters for the prediction
of ground motion. For example, regarding the distance measures, A&B use
Joyner-Boore distance Rjb, C&F the hypocentral distance Rhyp and Zh and C&Y
the shortest distance to the rupture plane Rrup. These distance measures were not
available for all records in the database. As a result, each GMPE could be
implemented for a variable number of records, and in any case, only for those
records for which all necessary data were either available or could be estimated in
a reliable manner.

Taking into account all four GMPEs and using the weights of Table 4, the
reference ground motion (GMr)ij at a certain period was calculated for each ground
motion j within site class i with Eq. (4):
(GM r )ij (T)  0.35  (GM r ) ij,AB 0.35  (GM r ) ij,CF
0.10  (GM r ) ij,Zh 0.20  (GM r ) ij,CY
(4)
where (GMr)ij,AB, (GMr)ij,CF, (GMr)ij,Zh, (GMr)ij,CY, are the reference spectral
accelerations at period T, calculated using the A&B, C&F, Zh and C&Y GMPEs
respectively.

The computation of reference spectral acceleration using Eq. (4) requires that
reference spectral acceleration at period T can be estimated with all four GMPEs.
However, each GMPE could be implemented for a different subset of data, since
each GMPE requires the knowledge of different parameters. As a result, the
computation of reference spectral acceleration with all four GMPEs, and thus the
estimation of amplification factors with Approach 1, was not feasible for all
records of each dataset, but only for subsets of these datasets, hereinafter referred
to as subsets of common records (Table 5). A representative plot for DS1 dataset

8
showing the median amplification factors for soil class C-Type 2, estimated using
Eq. (4), is illustrated in Figure 3a.

Fig. 3 DS1: Amplification factors estimated with Approach 1 for EC8 soil class C-Type 2
seismicity, using (a) subsets of common records and (b) different subsets

Table 5 Number of strong motion records for which reference spectral acceleration could be
estimated with all GMPEs (subsets of common records)
Soil
DS1 DS2 DS3 DS4
Class
Type 2 Type 1 Type 2 Type 1 Type 1 Type 2 Type 1
B 310 964 51 699 120 41 44
C 289 1130 50 869 142 47 47
D - - - - - - -
E - 4 - 4 1 - 4

It is obvious that the restriction of using only those strong motion records, for
which all GMPEs can be applied, limits the dataset significantly. As a result, for
soil classes B and C there are adequate data in all datasets, while for soil classes D
and E there are very few or even no available data. In order to overcome this
obstacle, we decided, in addition to estimating amplification factors using only the
subsets of common records, to calculate the median amplification factors by
implementing each one of the A&B, C&F, Zh and C&Y GMPEs separately to as
many records as possible and then to apply the weights of Table 4 to the these
median values. In this way, each GMPE is applied to a different subset and
amplification factors are given by the following equation:
1
S(T) 
0.35 0.35 0.10 0.20
  
SAB SCF SZh SCY (5)
where SAB , SCF SZh and SCY are the median amplification factors, calculated using
solely each one of the A&B, C&F, Zh and C&Y GMPEs respectively. Table 6
presents the number of strong motion records for which each one of the four
GMPEs could be implemented separately for the four datasets. The available
strong motion records, mainly for soil class D, and in certain extent for soil class
E, are still very limited, allowing however for an estimate for the corresponding
soil factors. A sample plot for DS1 dataset showing the median amplification
9
factors for soil class C-Type 2 calculated using each one of the four GMPEs
separately (different subsets), as well as the weighted average as derived from Eq.
(5), is illustrated in Figure 3b.

Table 6 Number of strong motion records for which each GMPE could be implemented (different
subsets)
Soil A&B C&F Zh C&Y
Class Type 2 Type 1 Type 2 Type 1 Type 2 Type 1 Type 2 Type 1
DS1
B 433 1053 774 1586 774 1586 311 1048
C 383 1191 746 1608 746 1608 289 1189
D 7 2 8 4 8 4 - -
E 10 4 42 58 42 58 - 7
DS2
B 114 755 205 1021 205 1021 52 778
C 100 908 228 1170 228 1170 50 925
D 3 1 3 1 3 1 - -
E 8 4 21 32 21 32 - 6
DS3
B 14 145 24 201 24 201 6 152
C 14 145 36 215 36 215 6 169
D - - - - - - - -
E 4 1 5 5 5 5 - 2
DS4
B 73 61 108 147 108 147 50 68
C 59 52 84 122 84 122 50 59
D 3 1 3 1 3 1 - -
E 8 4 21 32 21 32 - 6

Finally, in order to estimate a single period-independent amplification factor for

each soil class and each seismicity level, similar to the soil factor “S” proposed in
EC8, the amplification factors were averaged over a range of periods from T=0 to
T=2.0s, since Zh and C&Y GMPEs provide no intermediate values between 2.0
and 2.5s. The resulting amplification factors, however, do not represent only the
amplification related to the increase of ordinates of soil spectra with respect to
rock spectra, but also the amplification due to the change in shape of PGA-
normalized response spectra, since average spectra of softer soils tend to have a
larger and shifted towards longer periods plateau compared to rock spectra (Rey et
al. 2002). For this reason, the period-averaged amplification factors were divided
by the “spectral shape ratio” values (SR), which reflect the difference in the shape
of soil spectra with respect to rock spectra, when all spectra are constrained to
have the same ordinate at zero period spectra (Rey et al. 2002). SR factors for
EC8 soil classes B, C, D and E were calculated for the two seismicity types by
dividing the area under the PGA-normalized EC8 spectrum of each soil class by
the area under the corresponding (Type 1 or Type 2) PGA-normalized EC8
spectrum for soil class A (Table 7).

In order to validate Eq. (5), soil amplification factors for soil classes B and C,
where adequate data are available in both the subset of common records and the
different subsets, were estimated not only with Equations (3) and (4), but also
with Equation (5). The discrepancies in the resulting soil factors S, after averaging
10
and dividing with SR are less than 7.1%, which can be considered as acceptable.
A representative example for the amplification factors calculated from DS1
dataset for soil class C-Type 2 is illustrated in Figure 3. Figure 3a illustrates the
median weighted average amplification factors calculated using the subset of
common records, while Figure 3b illustrates the weighted average determined
using the different subsets. The corresponding EC8 acceleration response spectra
divided by the spectral values for soil class A are also depicted. The resulting soil
factors S, after averaging over the period range from T=0 to T=2.0s and dividing
these average values with SR, are equal to 1.93 for the subset of common records
and to 1.90 for the different subsets.

Table 7 Spectral shape ratios SR for EC8 classification scheme

Soil Class Type 2 (Ms≤5.5) Type 1 (Ms>5.5)
B 1.00 1.16
C 0.99 1.29
D 1.13 1.53
E 1.00 1.16

3.2.2 Approach 2

This approach was used by Rey et al. (2002) to recommend the soil factors
included in the current version of EC8. They used a dataset of strong motion
recordings provided by European Strong Motion Database (ESMD, Ambraseys et
al. 2000) to perform a systematic analysis of response spectra ordinates as a
function of magnitude and site conditions. In this method amplification is
quantified through a measure strictly related to the Spectrum Intensity definition
of Housner (1952). Period-independent soil amplification factors are calculated
for each soil class and for different magnitude intervals (M.I.) of Ms=0.5, with
respect to the rock sites of the database using Eq. (6):
S  ( I soil / I rock )  (1/ SR) (6)
where SR is the spectral shape ratio (see Table 7), Isoil and Irock are the spectrum
intensities for soil and rock respectively, originally defined by Housner (1952) for
spectral velocities and here adapted for spectral accelerations (Eq. 7):
2.5
I 
0.05
R  S a (T )dt
(7)

where R  Sa (T) denotes the log-average of distance-normalized 5% spectral

ordinates R  Sa (T) for each soil class and magnitude interval. Isoil/Irock ratios
provide a scaling factor for site effect that represents an average amplification
globally affecting the whole spectrum (Rey et al. 2002).

The same approach was also applied to a different dataset of ESMD by Schott and
Schwarz (2004) to estimate soil factors for EC8 soil classes B and C.

All strong-motion records of Table 3 were utilized for this approach. A sample
plot for DS2 dataset, showing the log-average of distance-normalized response
spectra for soil class C with respect to soil class A, for the M.I. covering a
magnitude range from 4 to 8, is illustrated in Figure 4. The log-average curves for
11
soil type C have a general trend to lie above the corresponding curves for soil A,
so the estimation of an average amplification through a single period-independent
factor is justified (Rey et al. 2002). Table 8 presents the calculated IC/IA ratios for
the different M.I. The table also contains the number of available strong motion
records (third and seventh columns) and the range of PGA (fourth and eighth
columns, with the number in the parenthesis representing the median PGA for soil
class C records) for each M.I. It should be noted that for Type 2 seismicity there
are many strong motion records for both soil class C and A in all M.I. Taking this
into consideration, and given the lack of apparent trend of IC/IA values with
respect to magnitude, an average IC/IA ratio was calculated from all M.I. equal to
2.10. The resulting S factor, as derived from Eq. (6) is equal to 2.12.

Fig. 4 DS2: Log-average, distance-normalized acceleration response spectra for sites of soil class
C (grey lines) and rock sites (black lines) for (a) Type 2 and (b) Type 1 magnitude intervals

12
Table 8 DS2: Isoil/IA ratios for EC8 soil class C obtained for all Magnitude Intervals [M.I.] (M.I.
with a satisfactory number of available strong motion records for both soil class C and A are
depicted in bold)
Type 2 Type 1
M.I. IC/IA n(C)/n(A) PGA (cm/s2) M.I. IC/IA n(C)/n(A) PGA (cm/s2)
4.0-4.5 1.93 100/33 ≤447 (41) 5.5-6.0 1.62 176/28 ≤630 (55)
4.5-5.0 2.35 126/36 ≤347 (55) 6.0-6.5 2.26 520/47 ≤560 (50)
5.0-5.5 2.02 127/36 ≤415 (47) 6.5-7.0 1.78 207/39 ≤1302 (128)
7.0-7.5 2.85 158/5 ≤775 (52)
7.5-8.0 1.65 200/6 ≤687 (70)
Average 2.1 Average 1.88

For Type 1 seismicity and soil class A only five and six records are available for
the 7-7.5 and 7.5-8 M.I. respectively. The corresponding IC/IA ratios could not be
considered as reliable, so the average IC/IA ratio for Type 1 seismicity was
calculated from the remaining three M.I. (5.5-6.0, 6.0-6.5, 6.5-7) and was found
equal to 1.88. The resulting S factor, as derived from Eq. (6) is equal to 1.46.

4 Results and Discussion

4.1 Normalized acceleration response spectra
In this section we present some representative results for the normalized response
spectra of each soil class according to EC8 classification scheme, obtained from
the procedure described in section 3.1.

Figure 5 illustrates the median spectra derived from DS1 and DS2 datasets with
the 16th and 84th percentiles and the normalized acceleration spectra recommended
in EC8 provisions for soil class A. Figure 6 illustrates the normalized spectra for
soil classes B and C derived from DS1, DS2 and DS3 datasets for Type 1
seismicity. For soil classes A, B and C, EC8 spectra match the empirical data to a
satisfactory extent for all datasets examined, lying between the median and the
84th percentile of the empirical normalized spectra.

Spectral values of soil classes B and C present a very wide range of normalized
values, which becomes more constrained as weaker strong-motion records are
removed from the dataset, remaining however remarkably wide. The removal of
weaker strong-motion records also leads to a decrease of the amplitudes of the
median spectra. For DS1 dataset, EC8 spectra lie close to the median empirical
normalized spectra, for DS2 dataset EC8 spectra lie between the median and the
84th percentile, while for DS3 dataset they are closer to the 84 th percentile. As a
result, EC8 spectra tend to become more conservative while moving towards
datasets with larger PGA values. EC8 spectra also seem to be better representing
short periods than long period motions. For example, concerning DS1 and DS2
datasets, the proposed spectrum for soil class C-Type 1 is well below the 84th
percentile for spectral periods greater than 0.7s, while it is closer to the 84 th
percentile for shorter periods. The overall impression for soil classes B and C and
Type 2 seismicity is similar as for Type 1 seismicity.

Figure 7 illustrates the normalized spectra for soil class D records derived from
DS1 dataset, which has only ten records for Type 2 and four records for Type 1
seismicity. The available soil class D records in DS2 dataset are even more
13
limited, while in DS3 dataset there are none. Keeping in mind that the empirical
spectra have been derived from such a poor sample, the ordinates of both Type 1
and 2 EC8 spectra do not provide a satisfactory fit to the median empirical
spectra. For Type 2 seismicity, the empirical spectra are shifted towards much
longer periods than the proposed ones, while for Type 1 seismicity the plateau of
the empirical spectra reaches much higher values than the spectrum proposed by
EC8. The lack of sufficient data for soil class D does not justify any change in
spectral shapes. However, the need for more strong-motion records from class D
sites is more than evident.

Figure 8 illustrates the normalized spectra for soil class E records derived only
from DS1 and DS2 datasets, since the available soil class E records in DS3 dataset
are very limited. EC8 design spectra are on the conservative side for periods
greater than 0.3s, lying close to or even above the 84 th percentiles. For smaller
periods EC8 spectra are unable to capture the wide range of normalized spectral
acceleration values, which is observed in the empirical spectra. A potential need
to increase the plateau for both Type 1 and Type 2 spectra is identified.

Fig. 5 Normalized elastic acceleration response spectra for soil class A, for Type 2 (left) and Type
1 seismicity (right), obtained from DS1 and DS2 datasets. Dashed lines correspond to EC8
proposed spectra, shaded areas correspond to the region between the 16 th and 84th percentile and
solid lines correspond to median empirical spectra

14
Fig. 6 Normalized elastic acceleration response spectra for Type 1 seismicity obtained from DS1,
DS2 and DS3 datasets for soil class B (left) and soil class C (right). Dashed lines correspond to
EC8 proposed spectra, shaded areas correspond to the region between the 16th and 84th percentile
and solid lines correspond to median empirical spectra

15
Fig. 7 Normalized elastic acceleration response spectra for soil class D, for Type 2 (left) and Type
1 seismicity (right), obtained from DS1 dataset. Dashed lines correspond to EC8 proposed spectra,
shaded areas correspond to the region between the 16th and 84th percentile and solid lines
correspond to median empirical spectra

Fig. 8 Normalized elastic acceleration response spectra for soil class E, for Type 2 (left) and Type
1 seismicity (right), obtained from DS1 and DS2 datasets. Dashed lines correspond to EC8
proposed spectra, shaded areas correspond to the region between the 16 th and 84th percentile and
solid lines correspond to median empirical spectra

A general remark depicted from the above figures is that although EC8 design
spectra generally match well the empirical ones, they do not seem to have been
derived based on a common rationale for all soil classes. For example, in some
16
cases (e.g. soil classes A, B and C) EC8 spectra lie between the median and the
84th percentile, while in other cases (e.g. soil class E) EC8 spectra lie closer to or
even above the 84th percentile of the empirical normalized spectra.

4.2 Soil amplification factors

4.2.1 Results from DS1, DS2 and DS3 datasets

Period-dependent amplification factors obtained from DS2 dataset with Approach
1 are presented in Figures 9-12. Following the rationale that was described in
detail in paragraph 3.2.1, for soil classes B and C the median amplification factors
derived from the subsets of common records are shown, along with the 16th and
84th percentiles (Figures 9 and 10). The range of the amplification between these
two percentiles, which is equivalent to the ±one standard deviation, is remarkably
wide. For soil classes D and E the results from the different subsets are presented,
illustrating the medians of the amplification factors derived by using each one of
the four GMPEs separately for the estimation of reference ground motion, as well
as the weighted average as derived with Eq. (5) (Figures 11 and 12). In all cases
the corresponding EC8 acceleration response spectra divided by the spectral
values for soil class A are also provided for direct comparison.

The median amplification factors for soil class B and both types of seismicity
(Figure 9) do not exhibit a strong period-dependency, ranging between one and
two for a broad period range between 0 and 2s. Moreover they are in good
agreement with the EC8 proposed spectra, with the latter lying very close to the
median for Type 2 and slightly below the median for Type 1 seismicity. For soil
class C on the other hand, amplification factors are influenced by the spectral
period, especially for Type 1 seismicity (Figure 10). In particular, for Type 2
seismicity the peak amplification is observed for spectral periods ranging from 0.7
to 1.4s, while for Type 1 amplification is constantly increasing for all periods
greater than 0.1s. EC8 spectra seem unable to represent accurately the
amplification potential of this soil class, since they are very close to the 16th
percentile for almost the whole period range. A much stronger tendency for
increasing amplification with period is observed for soil class D-Type 2, with
peak amplification exceeding the value of 4.5, while for Type 1 the only available
record presents a plateau between 0.5 and 0.9s with amplification levels around
4.5 (Figure 11). For soil class E-Type 2, amplification peaks at very low periods
(around 0.15s) reaching a value of 4.5 for the weighted average, and then
decreases sharply, remaining around 1.0 for periods greater than 0.4s. For soil
class E-Type 1, amplification ranges between 1 and 1.5 regardless of period value
(Figure 12).

17
Fig. 9 DS2: Amplification factors estimated with Approach 1 for soil class B, for Type 2 (left) and
Type 1 seismicity (right), compared to the corresponding EC8 spectra

Fig. 10 DS2: Amplification factors estimated with Approach 1 for soil class C, for Type 2 (left)
and Type 1 seismicity (right), compared to the corresponding EC8 spectra

Fig. 11 DS2: Median amplification factors estimated with Approach 1 for soil class D and PSA rock
estimated with all GMPEs, for Type 2 (left) and Type 1 seismicity (right)

18
Fig. 12 DS2: Median amplification factors estimated with Approach 1 for soil class E and PSA rock
estimated with all GMPEs, for Type 2 (left) and Type 1 seismicity (right)

Isoil/IA ratios obtained from DS1, DS2 and DS3 datasets with Approach 2 are
presented in Table 9. These ratios were calculated as the mean values of the ratios
corresponding to those magnitude intervals (M.I.) considered as reliable based on
the number of available strong motion records, as already explained in section
3.2.2. Dataset DS3 was utilized for derivation of soil factors for Type 1 seismicity
only.

Table 9 Isoil/IA ratios for EC8 soil classes and both seismicity contexts
Soil Type 2 (Ms≤5.5) Type 1 (Ms>5.5)
Class DS1 DS2 DS3 DS1 DS2 DS3
B 1.55 1.37 - 1.56 1.25 1.09
C 2.51 2.10 - 2.89 1.88 1.48
D 3.47 2.26 - 2.18 1.41 -
E 1.79 1.96 - 1.24 0.96 0.90

Tables 10 and 11 summarize the soil factors obtained for EC8 soil classes with the
different approaches and datasets, for Type 2 and Type 1 seismicity, respectively.
For Approach 1, the soil factors resulted after averaging the period-dependent
amplification factors (presented in Figures 9-12 for DS2 dataset) and dividing
them with SR, while for Approach 2 they resulted after dividing the I soil/IA ratios
presented in Table 9 with SR. The weighted average soil factors are the mean
values calculated from Approach 1 and Approach 2, since the two approaches
have been given equal weights of 0.5. Dataset DS3 was utilized only for
derivation of soil factors for Type 1 seismicity.

Table 10 EC8 factors and soil factors obtained with the different approaches and datasets and their
weighted average for Type 2 seismicity
Ms≤5.5
Soil SHARE-DS1 SHARE-DS2 SHARE-DS3 EC8
Class Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A.
B 0.90 1.55 1.23 1.51 1.37 1.44 - - - 1.35
C 1.93 2.54 2.23 2.19 2.12 2.16 - - - 1.50
D 3.36 3.07 3.22 2.92 2.00 2.46 - - - 1.80
E 0.98 1.79 1.39 1.30 1.96 1.63 - - - 1.60

19
Table 11 EC8 factors and soil factors obtained with the different approaches and datasets and their
weighted average for Type 1 seismicity
Ms>5.5
Soil SHARE-DS1 SHARE-DS2 SHARE-DS3 EC8
Class Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A.
B 1.47 1.34 1.41 1.53 1.08 1.31 1.49 0.94 1.22 1.20
C 2.09 2.24 2.16 2.06 1.46 1.76 1.82 1.15 1.48 1.15
D 1.74 1.42 1.58 1.56 0.92 1.24 - - 1.35
E 0.91 1.07 0.99 0.97 0.83 0.90 0.93 0.78 0.85 1.40

The average S factors derived from DS1 dataset, which contains all records
regardless of PGA, generally leads to higher values compared to DS2 dataset,
which makes use only of records with PGA≥20cm/s2. For Type 1 seismicity, even
lower S factors are obtained when using only records with large values of PGA
(DS3 dataset with PGA≥150cm/s2). However, when examining the two
approaches separately, we observe that the removal of weaker ground motion
records from the database leads to a clear reduction of soil factors S as far as
Approach 2 is concerned, while in Approach 1 there is no apparent trend.

Comparing the determined weighted average S factors with those indicated in

EC8 provisions, EC8 S factors for soil class B are well compared for both
seismicity Types, regardless of method or dataset. For soil class C, however, the
values proposed in EC8 are much lower than the estimated ones for all datasets.
For soil class D-Type 2, the estimated soil factors are also higher than the ones
proposed in EC8, while for Type 2 they are close to the present EC8 S factors.
However, as they have been derived from a limited number of records, they
should be interpreted with caution. For soil class E-Type 2, the obtained empirical
S factors are close to the soil factors proposed in EC8, while for Type 1, the
obtained S factors are much lower, and in fact below unity regardless of dataset.
This is due on one hand to the poor seismological documentation of soil class E
stations of the database, which led to a scarce number of records that could be
utilized for the estimation of amplification on soil class E stations, and on the
other hand on the averaging process that both methods use in order to compute a
period-independent S factor, which can be applied to the whole spectrum. The
latter is evident when looking at the period-dependent soil amplification factors
obtained with Approach 1 (e.g. Figure 12 for DS2 dataset), where we observe that
amplification peaks in very low periods, reaching high values ranging from 3.5 to
5.2, and then decreases abruptly to values close to or even less than unity for a
long spectral period range.

4.2.2 Validation with a well-documented dataset (DS4)

In order to check the accuracy of the above soil factors using the three datasets
extracted from the general SHARE database (DS1, DS2 and DS3), amplification
soil factors were also estimated with the last DS4 dataset, which includes strong
motion records stemming from sites with very well-documented soil profiles
concerning their dynamic properties and depth to the rock basement. Tables 12
and 13 compare the soil factors obtained from DS4 dataset with those from DS2
dataset, since the two datasets are equivalent in terms of PGA, containing records
with PGA≥20cm/s2. The differences between the soil factors obtained with DS2
and DS4 datasets are quite low (<14%), proving the accuracy of the results
obtained with DS1, DS2 and DS3 datasets.
20
Table 12 Soil factors obtained with DS2 and DS4 for Type 2 seismicity
Ms≤5.5
Soil DS2 DS4 W.A. Difference
Class Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A. (%)
B 1.51 1.37 1.44 1.39 1.36 1.38 4.17
C 2.19 2.12 2.16 2.23 2.06 2.15 0.46
D 2.92 2.00 2.46 2.92 2.64 2.78 13.01
E 1.30 1.96 1.63 1.30 1.66 1.48 9.20
Table 13 Soil factors obtained with DS2 and DS4 for Type 1 seismicity
Ms>5.5
Soil DS2 DS4 W.A. Difference
Class Ap.1 Ap.2 W.A. Ap.1 Ap.2 W.A. (%)
B 1.53 1.08 1.31 1.14 1.13 1.13 13.74
C 2.06 1.46 1.76 1.79 1.73 1.76 0.00
D 1.56 0.92 1.24 1.56 - - -
E 0.97 0.83 0.90 0.97 0.98 0.98 8.89

5 Conclusions and Recommendations for EC8

The validity and accuracy of the elastic response spectra defined in Eurocode 8
provisions is checked against a large worldwide dataset of strong motion records
in terms of both PGA-normalized spectra and soil factors. All analyses were
performed for three sub datasets having different threshold values for PGA.

Shape of normalized elastic response spectra: For soil classes A, B and C the
spectral shapes provided by EC8 are in good agreement with the derived empirical
data for both seismicity types (1 and 2) prescribed in EC8. For soil classes D and
E the sample of data is not as rich as for the other soil classes and hence the
results may not be as convincing as for soil classes A, B and C. However, we
found a clear tendency in soil class D spectral shapes to differ substantially from
the EC8 shapes. Equally important differences are found in soil class E where the
EC8 spectra seem to be conservative enough for medium and high periods, but
probably in short period the plateau should be somehow increased. In conclusion,
considering all the above analyses, observations and comments, it is believed that
on average the EC8 spectral shapes are in good agreement with the empirical data
and the proposition of improved shapes is not really justified for the moment.
However, given the difficulty of establishing rich datasets of soil class D records
in the near future, it is suggested that Eurocode 8 should prompt for site-specific
ground response analyses for the definition of seismic action in class D sites. The
same suggestion might be necessary also for soil class E sites.

Amplification factors: Two approaches were used and combined with equal
weighting factors to estimate soil factors for the different EC8 soil classes. EC8
factors for class B are in good comparison to the empirical data for both
seismicity types, while for soil class C the S factors derived from all datasets are
higher compared to EC8. For soil class D the estimated soil factors, which were
derived from a limited dataset, are also higher than the EC8 factors for Type 2
seismicity but relatively close to the EC8 factors for Type 1 seismicity. For soil
class E, the weighted average soil factors are found quite low for Type 1

21
seismicity, a result, which is attributed to both the limited data and the averaging
process.

Based on the results derived from this comprehensive study, improved soil factors
S are proposed for potential use in an EC8 update, supposing that no further
changes are made in the definition of soil classes and seismicity types and the
shape of normalized design spectra. The PGA ranges of DS2 and DS3 datasets
can be considered as more representative for Type 2 and Type 1 seismicity,
respectively. However, since the soil factors obtained from DS2 are more
conservative than the ones obtained from DS3, we decided to keep the soil factors
derived from DS2 as more appropriate for both seismicity types. For soil classes B
and C the proposed soil factors resulted from rounding the weighted average of
the soil factors estimated with Approaches 1 and 2 using the DS2 dataset. For soil
classes D and E, due to the insufficient datasets, it was decided to keep the same S
period independent amplification factors as in EC8, knowing however from the
few available records that most probably the EC8 amplification factors might be
inadequate. Again, site-specific ground response analyses should be
recommended by EC8 for the estimation of soil amplification for soil class D and
in some cases for soil class E sites. The proposed soil factors S for EC8 soil
classes are given in Table 14.

Table 14 Soil factors for EC8 soil classes

Type 2 (Ms≤5.5) Type 1 (Ms>5.5)
Soil Class
Proposed EC8 Proposed EC8
B 1.40 1.35 1.30 1.20
C 2.10 1.50 1.70 1.15
a a
D 1.80 1.80 1.35 1.35
E 1.60 a 1.60 1.40 a 1.40
(a) site specific ground response analysis required

Data and resources

All records used in this study were selected from the worldwide strong motion
database compiled in the framework of SHARE (Yenier et al. 2010). For the
compilation of SHARE-AUTH database, shear wave velocity profiles were
gathered from the Italian Accelerometric Archive (Working Group ITACA 2010,
http://itaca.mi.ingv.it), Kik-Net (http://www.kik.bosai.go.jp/), K-Net
(http://www.k-net.bosai.go.jp/), the Turkish national strong-motion database
(Akkar et al. 2010, in DVD format), ROSRINE program website
(http://gees.usc.edu/ROSRINE/), Dr. David Boore’s personal webpage
(http://www.daveboore.com/), selected USGS open-file reports and the shear
wave velocity archives of the Research Unit of Soil Dynamics and Geotechnical
Earthquake Engineering of AUTH.

Acknowledgements
This research has been mainly funded by the European Community's Seventh Framework Program
[FP7/2007-2013] under grant agreement n° 226967 (Seismic Hazard Harmonization in Europe,
http://www.share-eu.org/). We thank Ezio Faccioli for his constructive comments during the
progress of this work, Sinan Akkar, Abdullah Sandikkaya and their coworkers for providing the
SHARE strong-motion database, John Douglas for his advice on seismological issues, Roberto
Paolucci, Chiara Smerzini and Francesca Pacor for checking the SHARE-AUTH data on the

22
Italian strong motion stations, and everyone who helped with the development of the SHARE-
AUTH database. We also thank two anonymous reviewers for their comments, effort and time
allocated to improve the paper.

References
Akkar S, Bommer JJ (2010) Empirical equations for the prediction of PGA, PGV and spectral
accelerations in Europe, the Mediterranean region and the Middle East. Seismol Res Lett 81:195-
206. doi:10.1785/gssrl.81.2.195
Akkar S, Çağnan Z, Yenier E, Erdogan E, Sandıkkaya MA, Gülkan P (2010) The recently
compiled Turkish strong-motion database: preliminary investigation for seismological parameters.
J Seismol 14:457-479. doi: 10.1007/s10950-009-9176-9
Ambraseys N, Smit P, Berardi R, Rinaldis D, Cotton F, Berge C (2000) Dissemination of
European Strong-Motion Data, CD-ROM collection. European Council, Environment and Climate
Programme

Boore DM (2004) Estimating Vs,30 (or NEHRP Site Classes) from shallow velocity models
(Depths<30 m). B Seismol Soc Am 94:591–597. doi:10.1785/0120030105
Borcherdt RD (1994) Estimates of site-dependent response spectra for design (methodology and
justification). Earthq Spectra 10:617–653. doi:10.1193/1.1585791
Borcherdt RD, Glassmoyer G (1992) On the characteristics of local geology and their influence on
ground motions generated by the Loma Prieta earthquake in the San Francisco Bay region,
California. B Seismol Soc Am 82:603-641
Cauzzi C, Faccioli E (2008) Broadband (0.05 to 20 s) prediction of displacement response spectra
based on worldwide digital records, J Seismol 12:453–475. doi:10.1007/s10950-008-9098-y
CEN (European Committee for Standardization) (2004) Eurocode 8: Design of structures for
earthquake resistance, Part 1: General rules, seismic actions and rules for buildings. EN 1998-
1:2004. Brussels, Belgium
Chiou BSJ, Youngs RR (2008) An NGA model for the average horizontal component of peak
ground motion and response spectra, Earthq Spectra 24:173–215. doi:10.1193/1.2894832
Choi Y, Stewart JP (2005) Nonlinear site amplification as function of 30 m shear wave velocity,
Earthq Spectra 21:1-30. doi:10.1193/1.1856535
Delavaud E, Cotton F, Akkar S, Scherbaum F, Danciu L, Beauval C, Drouet S, Douglas J, Basili
R, Sandikkaya MA, Segou M, Faccioli E, Theodoulidis N (2012). Toward a ground-motion logic
tree for probabilistic seismic hazard assessment in Europe, J Seismol. doi:10.1007/s10950-012-
9281-z
Field EH (2000) A modified ground-motion attenuation relationship for southern California that
accounts for detailed site classification and a basin-depth effect. B Seismol Soc Am 90:S209-
S221. doi: 10.1785/0120000507
Field EH, Jacob KH (1995) A comparison and test of various site response estimate techniques,
including three that are not reference-site dependent. B Seismol Soc Am 85: 1127-1143.
Housner GW (1952) Spectrum intensities of strong-motion earthquakes. In Proceedings of
Symposium on Earthquakes and Blast Effects on Structures, Earthquake Engineering Research
Institute, pp 20-36
ICC (International Code Council) (2012) 2012 International Building Code. International Code
Council, Inc. www.iccsafe.org/Pages/default.aspx.
Joyner WB, Warrick RE, Fumal TE (1981) The effect of Quaternary alluvium on strong ground
motion in the Coyote Lake, California, earthquake of 1979, B Seismol Soc Am 71:1333– 1349
Lee Y, Anderson JG (2000) Potential for improving ground-motion relations in southern
California by incorporating various site parameters. B Seismol Soc Am 90:S170-S186.
doi:10.1785/0120000509
Pitilakis K, Anastasiadis A, Riga E (2012) New site classification scheme and associated site
amplification factors. SHARE Deliverable 4.3
Pousse G, Berge-Thierry C, Bonilla F, Bard P-Y (2005) Eurocode 8 design response spectra
evaluation using the K-net Japanese database. J Earthq Eng. 9:547 –574. doi:
10.1080/13632460509350555
Rey J, Faccioli E, Bommer JJ (2002) Derivation of design soil coefficients (S) and response
spectral shapes for Eurocode 8 using the European Strong-Motion Database. J Seismol 6:547-555.
doi:10.1023/A:1021169715992

23
 Schott C, Schwarz J (2004) Reliability of Eurocode 8 spectra and the problems of their application
to central European earthquake regions. In Proceedings of 13th World Conference on Earthquake
Engineering, paper n. 3403, Vancouver, B.C., Canada
Sokolov VY (1997) Empirical models for estimating Fourier-amplitude spectra of ground
acceleration in the Northern Caucasus (Racha seismogenic zone). B Seismol Soc Am 87:1401-
1412
Sokolov VY, Loh CH, Wen KL (2000) Empirical study of sediment filled basin response: the case
of Taipei city. Earthq Spectra 16:681-707. doi:10.1193/1.1586134
Steidl JH (2000) Site response in southern California for probabilistic seismic hazard analysis.B
Seismol Soc Am 90:S149–S169. doi:10.1785/0120000504
Stewart JP, Liu AH, Choi Y, Baturay MB (2001) Amplification factors for spectral acceleration in
active regions, Report No. PEER- 2001/10, Pacific Earthquake Engineering Research Center,
University of California, Berkeley.
http://peer.berkeley.edu/publications/peer_reports/reports_2001/0110.pdf.
Stewart JP, Liu AH, Choi Y (2003) Amplification factors for spectral acceleration in tectonically
active regions. B Seismol Soc Am 93:332–352. doi:10.1785/0120020049
Yenier E, Sandikkaya MA, Akkar S (2010) Report on the fundamental features of the extended
strong motion databank prepared for the SHARE project (v1.0)
Zhao JX, Zhang J, Asano A, Ohno Y, Oouchi T, Takahashi T, Ogawa H, Irikura K, Thio HK,
Somerville PG, Fukushima Y, Fukushima Y (2006) Attenuation relations of strong ground motion
in Japan using site classification based on predominant period. B Seismol Soc Am 96:898–913.
doi:10.1785/012005012

24

View publication stats