SHEAR-WAVE VELOCITY PROFILE AND SEISMIC INPUT DERIVED FROM

AMBIENT VIBRATION ARRAY MEASUREMENTS: THE CASE STUDY OF

DOWNTOWN LAQUILA
Giuseppe Di Giulio1,3, Iolanda Gaudiosi2,3, Fabrizio Cara2,3, Giuliano Milana2 and Marco Tallini3

(1) Istituto Nazionale di Geofisica e Vulcanologia, LAquila, Via dellArcivescovado 8, 67100, ITALY

(2) Istituto Nazionale di Geofisica e Vulcanologia, Roma, Via di Vigna Murata 605, 00145, ITALY

(3) Dipartimento di Ingegneria Civile, Edile-Architettura e Ambientale Centro di Ricerca e Formazione in Ingegneria Sismica,
Universit dellAquila, Via Giovanni Gronchi, 18, 67100, LAquila, ITALY

SUMMARY

Downtown LAquila suffered severe damage (VIII-IX EMS98 intensity) during the April 6th 2009 Mw 6.3 earthquake. Downtown
L'Aquila shows distinctive features in the geological succession. Settled on a top flat hill , its velocity structure is characterized by a
reversal of velocity at a depth of the order of 50-100 m, which corresponds to the contact between calcareous breccia and lacustrine
deposits. In the southern sector of downtown, a thin unit of superficial red soils causes a further shallow impedance contrast that may
have influenced the damage distribution during the 2009 earthquake. Consistently, the data of the 2009 seismic sequence and the
damage patterns have revealed a heterogeneous seismic behavior between the southern and northern sectors of the downtown. Hence,
we carry out a study the main features of ambient seismic vibrations in the entire city center by using array measurements. We
deployed six 2-D arrays of seismic stations and 1-D array of vertical geophones. The 2-D arrays recorded ambient noise, whereas the
1-D array recorded signals produced by active sources. Surface-wave dispersion curves have been measured by array methods and
have been inverted through a neighborhood algorithm jointly with the microtremor ellipticity. We obtain shear-wave velocity (Vs)
profiles representative of the southern and northern sectors of downtown L'Aquila; the obtained Vs profiles well evidence the reversal
of velocity between the stiff calcareous breccia and the lacustrine succession.

The theoretical 1-D transfer functions for the estimated Vs profiles have been compared to the available empirical transfer functions
computed from aftershock data analysis, revealing a general good agreement. Then, the Vs profiles have been used as input for
deconvolution analysis aimed at deriving the ground motion at bedrock level. The deconvolution has been performed by means of
EERA and STRATA codes, two tools commonly employed in the geotechnical engineering community to perform equivalent-linear
site response studies. The waveform at the bedrock level has been obtained deconvolving the 2009 mainshock recorded at a strong
motion station installed in downtown. The assessed waveform at the bedrock has been finally used as seismic input for evaluating
synthetic time-histories in a strong-motion target site in the middle Aterno valley, called AQV, 5 km far from downtown LAquila.
AQV recorded the LAquila 2009 mainshock and its surface stratigraphy is adequately known. The comparison between modeled and
observed strong ground motion shows reasonable agreement.

Key words: Site effects, Earthquake ground motions, LAquila 2009 mainshock, seismic arrays.

1
1 INTRODUCTION

On April 6th 2009 a Mw 6.3 earthquake hit the historical city of LAquila (Central Italy; Fig.1a) causing 308 causalities, more than
40000 homeless and strong damage (VIII-IX EMS98 intensity) in downtown and surrounding minor villages (Tertulliani et al. 2011
and 2012). Heavy damage and collapses were concentrated in the unreinforced masonry buildings including historical churches of
downtown LAquila and of many surrounding municipalities. Starting from May 2009, the Italian Civil Defense Department (DPC)
promoted a microzoning study of the epicentral area for characterizing the surface geology and the areas prone to local seismic
amplification (MS-AQ Working Group 2010). The microzonation study was based on geological, geophysical and geotechnical
surveys, and documented a clear correlation between the geological conditions and the damage distribution in the epicentral area
(Martelli et al. 2012). For downtown L'Aquila, the microzoning studies defined the geological structure of the area. A stiff 50-100 m-
thick Pleistocene calcareous breccias unit (Br), outcropping in the northern part of historical city, lies abruptly on a 200-300 m-thick
Pleistocene lacustrine/fluvial deposit (L). The latter layer is in contact with a Meso-Cenozoic limestone bedrock (Fig. 1).

The LAquila Mw 6.3 mainshock and the strongest events of the sequence were recorded by 4 near-fault strong-motion stations within
the upper Aterno river valley at about 5 km NW from downtown LAquila; the highest peak ground acceleration (PGA) value was of
about 0.65 g (elebi et al. 2010) recorded on the horizontal component for site AQV (Lanzo & Pagliaroli 2012) in the middle of the
valley. The Mw 6.3 mainshock was also recorded by two strong-motion stations (AQU-AQK in Fig. 1) situated in the northern and
southern sectors of downtown LAquila, respectively. For these two sites, the PGA values were of about 300-350 cm/s2 both for
horizontal and vertical components, whereas the peak ground velocity (PGV) values ranged in the 35-36 cm/s interval (elebi et al.
2010).

One of main feature evidenced by seismic data recorded in the city (De Luca et al. 2005; Milana et al. 2011) is the presence of a
strong low-frequency resonance, very diffuse in the area and centered at about 0.5-0.6 Hz. This fundamental resonance (f0) is likely
related to the geological setting of the area, and more in detail to the deep impedance contrast between surface soils (quaternary Br
and L deposits) and stiff limestone assumed as seismic bedrock (MS-AQ Working Group 2010; Fig. 1).
For detecting the deepest part of the subsoil structure, classical geotechnical and geophysical approaches (such as refraction tests,
electrical resistivity tomography, active surface wave experiment) are in general difficult to be applied in urban areas where the
logistic is in general constrained. In these context, methods using ambient seismic noise are often successful in providing reliable
estimates of the subsoil structure (Panou et al. 2005; Bonnefoy-Claudet et al. 2009; Gosar et al. 2010). These techniques have also the
advantage to be noninvasive and low-cost. Noise measurements can be acquired by single-stations or by 2-D seismic arrays. Single-
station noise technique computes the horizontal-to-vertical spectral ratio (H/V; Nogoshi & Igarashi 1971; Nakamura 1989). The peak
of the H/V ratio identifies the resonance frequency (f0) of the subsoil, which is related to the thickness and to the shear-wave velocity
of the soft deposits. Array techniques are based mainly on surface wave theory (Tokimatsu 1997 among many others), and are very
suited for seismic site characterization (Foti et al. 2011). Because the energy content of ambient vibration is mostly in the low-
frequency range (approximately below 5 Hz) and the 2-D geometry can be adapted changing the inter-station distance, the method in
principle allows investigating depths more than 50 m with a good resolution. For superficial soil characterization, good results are
achieved using linear geometries and active sources (MASW survey).
Many studies used the 2-D arrays technique jointly with the H/V curves to obtain the Vs profile of soils (Bonnefoy-Claudet et al.
2006; Bard et al. 2010). The two main processing techniques of array data are the frequency-wavenumber (F-K) and the spatial
autocorrelation (SPAC) methods. The F-K method provides the propagation characteristics (i.e. apparent surface-wave phase velocity
and azimuth) of waves travelling across the seismic array, and it includes conventional beamforming (Lacoss et al. 1969; Kvaerna &
Ringdahl, 1986) and high-resolution F-K (Capon, 1969). The SPAC method is based on computation among receivers of spatially
averaged autocorrelation curves, which are related to a Bessel function of zero order (Aki 1957; Asten 2006).
The scientific literature agree that array noise experiments are suitable to obtain information on resonance modes that develop in deep
valleys, and therefore ambient vibrations can contribute to derive the velocity structure of complex buried structure. Recent

2
applications for deriving geotechnical models can be found for the Alpine valleys (Steimen et al. 2003; Cornou et al. 2003; Roten &
Fh 2007), for the Osaka basin (Uebayashi 2003), and the Tamar Valley (Claprood et al. 2011) among many other case studies.

In this paper, we investigated the properties of ambient seismic vibrations recorded by six 2-D arrays of seismological stations
deployed in downtown L'Aquila. We first compared the resonance frequencies and the H/V curves obtained in the city. Then we
performed a surface-wave analysis using noise array data. The main goal was to characterize thicknesses and shear-wave velocities of
the two main units present in the LAquila area; breccias (Br) and lacustrine/fluvial deposits (L). The maximum array aperture ranged
from 100 m to about 1 km allowing a reliable surface-wave dispersion curve in a wide frequency band (1-8 Hz). Additionally, the
high-frequency part (~ 8-40 Hz) of surface-wave dispersion was investigated through a MASW survey.
The joint inversion of the obtained dispersion curves and the microtremor ellipticities was performed by a conditional neighborhood
algorithm, and benefited by a-priori information on LAquila subsoil. Indeed, the microzonation studies provided lots of information
about the properties of the main geological units, the velocity-depth functions and the expected bedrock depth.
The Vs profiles resulting from surface-wave analysis were finally used for deconvolving the mainshock recording available at AQK
site. We used EERA and STRATA codes to perform 1-D equivalent-linear site response analysis, taking into the account the
uncertainties on the Vs profiles. To validate the ground motion assessed at the bedrock level, we further applied a convolution analysis
to the AQV site in the Aterno river valley, comparing the waveform resulting at the outcrop level with the actual L'Aquila mainshock
recorded at this site. The soil profile at AQV was furnished by a cross-hole survey. Our strategy based on a deconvolution/convolution
approach achieves a twofold purpose: on one side, we verify the reliability of the Vs model derived from array analysis and proposed
for the southern sector of downtown, on the other side we validate our estimate of the bedrock ground motion of the Mw 6.3 main
event of the 6th April 2009. The comparison between modeled and observed strong ground motion is good, although some
discrepancies occur because our strategy does not include near-source effects and complex site effects developing in the Aterno river
basin.

2 GEOLOGICAL SETTING

In the framework of microzonation activities, several studies have been performed for downtown LAquila and the surrounding area
(Amoroso et al. 2010; MS-AQ Working Group 2010; Lanzo et al. 2011; Martelli et al. 2012; Monaco et al. 2012; Tallini et al. 2010
and 2012). Also, a series of 2-D geological cross-sections have been produced (Fig. 1d) with the meso-cenozoic limestone being the
seismic bedrock. LAquila historical city center is settled on the flat part of a terrace (De Luca et al. 2005), which is composed of
relatively stiff Pleistocene calcareous breccias unit (Br). LAquila terrace is bounded toward south by the Aterno river valley along a
steep scarp whereas in the north direction it is in contact with outcropping limestone. Br is over-imposed to ancient fluvial and
lacustrine sediments (L) consisting mainly of silty/sandy layers with thickness of some hundreds of meters. Br is outcropping in the
northern-part of downtown and along the terrace flanks, whereas in the southern part of the city very thin layers (in general thickness
below 20 m and Vs < 400 m/s) of colluvial red soil (Limi Rossi; Lr) superficially overlay the Br unit (Fig. 1). A deep borehole drilled
in the city center (S2 survey in Piazza Duomo, see Fig. 1) documents the contact between Br and L at a depth of about 80 m, and a
thickness of more than 200 meters of L (Amoroso et al. 2010; Del Monaco et al. 2013). A second deep borehole, drilled at the base of
terrace along the Aterno River valley (S3 survey, Fig. 1), finds a homogeneous sequence of L with a thickness of 200 meters overlying
the limestone bedrock. A cross-hole survey available at the same position of S3 shows for L a shear-wave velocity profile increasing
with depth: from 400 m/s in the very uppermost layer up to 800 m/s at a depth of about 80 m (Cardarelli & Cercato 2010). A
downhole survey situated in southern-eastern downtown (S8 survey, Fig. 1) shows relatively high velocities for Br with Vs around
1000 m/s (MS-AQ Working Group 2010).

Damage within the city center was severe during the 2009 seismic sequence. Recent papers (Milana et al. 2011; Bordoni et al. 2011;

3
Del Monaco et al. 2013) suggested a correlation between site effects and the damage distribution. The buildings in downtown were
characterized by a vulnerability of class B (about 65%) and C (about 25%) following the EMS98 scale (Grnthal 1998). In particular
in the southern part, some reinforced concrete (RC) buildings without anti-seismic design (vulnerability of class C), collapsed during
the L'Aquila 2009 mainshock (Tertulliani et al. 2011 and 2012). Milana et al. (2011) used the aftershocks analysis to estimate the
amplification factors (Fa) in the low-periods range (0.1-0.5 s), where Fa was computed as the ratio between the integrals of the
pseudo-velocity response spectra at the analyzed sites and at a reference site. They found Fa values of about 2 for the southern part
and about 1 for the northern sector of downtown. The spatial distribution of Fa in the low-periods range, which includes the resonance
frequencies of RC buildings present in L'Aquila, is well correlated with the presence of soft deposits of red silts (Lr) in the southern
part of the city (Fig. 1). Lr deposit has a thickness up to 20 m and Vs of about 300-400 m/s (MS-AQ Working Group 2010). Milana et
al. (2011) evaluated Fa values also in the high-periods range (0.5-2 s). They still found larger Fa factors in the southern sector of the
city respect to the northern part. However, in this case to search a connection between damage distribution and Fa pattern is not
meaningful, because the natural frequencies of buildings in L'Aquila (not more than six stories tall) are outside the high-periods range.
Del Monaco et al. (2013) have recently published the results of a very dense survey of single-stations seismic noise recordings in
downtown. Apart from the low-frequency resonance at 0.5 Hz present everywhere in downtown, these authors focused on a secondary
peak (f1) observed at frequencies ranging from 3 up to 15 Hz. This secondary peak is likely caused by a very shallower impedance
contrasts in the uppermost 20 m, between red silts or weathered breccias and more consistent breccias. Del Monaco et al. (2013)
suggested a spatial correlation between the most severe damages on buildings and the f1 distribution (Fig. 1c), especially evident in
the north-western part of downtown. Fig. 1c refers mostly to residential buildings of vulnerability class B. Taking into account only
RC buildings, the maximum damage grade was concentrated along the southern-western border of downtown where the morphologic
slope of L'Aquila terrace increases abruptly toward the Aterno river valley (Fig. 2). Tertulliani et al. (2011) stressed a possible role of
the topography on the RC damage pattern. In proximity of the slope, a seven-story RC building hosting a students dorm (Mulas et al.
2013) also partially collapsed during the 2009 earthquake causing eight casualties.

3 DATA AND ARRAY ANALYSIS

The six 2-D arrays of seismic stations were installed in different time-periods in downtown LAquila (Table 1 and Fig. 2). We used
different array apertures with a number of remote stations between 10 and 15. The seismological stations, deployed approximately in a
circular configuration, were composed of high resolution digitizer (Reftek130 or Lennartz_MarsLite) connected to three-component
sensor (Lennartz Le3D 5s; eigenfrequency of 0.2 Hz). The sampling rate in data acquisition was fixed to 250 Hz. The time-
synchronism was supplied by a GPS antenna connected to each seismic station. To reduce the bias introduced by errors in sensors
location, the absolute position of the stations was determined after differential GPS measurements obtained performing a real time
kinematic survey with a Leica Systems 1200 GNSS GPS instrument. The differential corrections were assured by a network of
reference stations (www.italpos.it) connected through GPRS. With this procedure the error in stations positioning was always lower
than few tens of centimeters.

The location of the six 2D-arrays is shown in Fig. 2. For its central position in the city center, we assumed the square Piazza
Duomo, which corresponds to the position of S2 survey (Fig. 2), as limit between the southern and northern sector of downtown.
Taking into account this spatial division, the arrays SMP, PZZ, AN and DUO are located in the northern part of town, whereas the
arrays AS, AL and the MASW experiments are in the southern part (arrows delimiting northern and southern part of the city are also
shown in Fig. 2). The two largest arrays (AS and AN in Table 1, with maximum aperture of about 1 km) were dedicated to investigate
the low-frequency range of the two parts of the city (Fig. 2). We successively arranged a 2-D array with maximum aperture of 170 m
(AL in Table 1) in the southern area where Lr outcrops. In order to investigate the high-frequency range, we also performed a MASW
experiment using a 1-D linear array of 72 vertical 4.5 Hz geophones equally spaced of 1.5 m. The linear array recorded data from
shots with different offset using a mini-gun as active source. The three remaining arrays of Table 1 (SMP, PZZ and DUO) were
deployed in the northern sector of downtown (Fig. 2) where Br unit is mostly outcropping. The data were processed through the

4
package geopsy (http://www.geopsy.org) developed during the SESAME (European Site Effects Assessment using Ambient
Excitations project; http://sesame-fp5.obs.ujf-grenoble.fr/) and the NERIES-JRA4 projects (Network of Research Infrastructures for
European Seismology; www.neries-eu.org; Bard et al. 2010).

3.1 HVNSR

The horizontal-to-vertical spectral ratios of ambient vibrations (HVNSR or H/V curves) has been largely used in the microzoning
activities of LAquila epicentral area, providing in general reliable indication on site effects (Martelli et al. 2012). In the present paper,
we focus on downtown L'Aquila. Recently, Del Monaco et al. (2013) presented H/V results of many single-station noise
measurements in downtown, but their discussion was concentrated on the presence of a secondary resonant peak at frequency 3-15 Hz,
and on the correlation between such peak and the uppermost soil properties. Milana et al. (2011) computed HVNSR curves at several
seismic stations installed in downtown after the earthquakes, observing differences in the low-frequency H/V peak (0.5 Hz) between
the northern and southern sector of the city. In this study, we used more seismic stations and in array configuration in comparison to
the work of Milana et al. (2011). Hence, we can follow better the variation on f0 within the entire historical center. HVNSR curves
were computed at each site using a moving window, 60 sec long, and an antitrigger software to remove short transients in the
recordings. For each accepted time window, we removed the mean, the linear trend and applied a 5% cosine taper to each end of the
signal. Then, we calculated the Fourier amplitude spectra (FAS) and smoothed the amplitudes following the method proposed by
Konno & Omachi (1998) using a coefficient of 40 for the bandwidth. The two horizontal spectra were combined with the geometrical
mean and finally divided by the vertical spectrum to get the H/V ratios.

Fig. 3a shows the H/V ratios obtained at each station for each array, for clarity of the figure we plotted the mean H/V curve computed
by averaging the horizontal components. Fourier amplitude spectra (FAS) at the three components of motion, and the directional noise
analysis computed both at the central station of each array are illustrated in Fig. 3b and 3c, respectively. The HVNSR curves show a
good agreement in terms of shape within each single array. A clear low-frequency resonance (0.5-0.6 Hz) is evident especially in the
southern part of the city (Fig. 3 2). f0 is caused by the largest energy at low frequencies of the horizontal ground motion compared to
the vertical one, as clearly indicated by the FAS of Fig. 3b. The resonance frequency f0 (Fig. 3) tends to lower frequency values
moving from South to North (from 0.6 to 0.5 Hz). This frequency shift of f0 is also true for the shift in frequency of the H/V trough;
the latter occurs at 1 Hz for south arrays, and at 0.8 Hz for north arrays. A part from the shift in frequency, it is also evident a variation
of the amplitude levels. The amplitudes of the H/V peak around f0 is larger in the south (mean amplitude of about 7) rather than the in
north part of downtown (mean amplitude 4). A large scatter in the amplitude of H/V peak around f0 is shown in Fig. ?? by array AN
and AS. This is consistent with the large spatial extension of these arrays (maximum aperture of about 1 km; Table 1). The amplitudes
of the H/V peak around f0 is larger in the south (mean amplitude of about 7) rather than the in north part of downtown (mean
amplitude 4).
Assuming that the Rayleigh surface wave dominate in the H/V curve, modeling studies show that leading parameters in determining
the H/V peaks and troughs are the velocity contrast and the Poissons ratio of the soft sediment (Bonnefoy-Claudet et al. 2006;
Lunedei & Albarello 2010; Tuan et al. 2012). The change in the position of f0 and in the amplitudes of the H/V peaks, moving from
south to north, could be theoretically connected both to a thickness variation of soft-to-bedrock interface (L-to-limestone in Fig. 1d),
or to a modification of velocities and Poissons ratio of soft deposits between the two sectors of the city.
The H/V spectral ratios show minimum value between 1 and 2 Hz, where the spectra of horizontal and vertical components have
comparable energy (Fig.s 3a and 3b). In this frequency range, the H/V curves show minimum amplitude levels varying from 0.8 to 1
(Fig. 3a). Castellaro & Mulargia (2009) indicated that H/V amplitudes below 1 in a large frequency interval could be associated to the
presence of velocity inversion in the subsoil. A moderate de-amplification of H/V curves between 1 and 2 Hz could be consistent with
the geology of LAquila where velocity inversion between Br and L can occur at a depth of the order of 100 m (De Luca et al. 2005;
Bordoni et al. 2011). The mean Vs is of the order of 1000 m/s and of 800 m/s for Br and for deep L deposits (S3 cross-hole survey in
Fig. 1), respectively (MS-AQ Working Group 2010). Two explanations are possible for the troughs of the H/V around 1-2 Hz; the
minimum amplitudes of the H/V around 1-2 Hz could be connected to the reversal of Vs between Br and L according to the results of
Castellaro & Mulargia (2009), or to the vertical polarization of Rayleigh waves in correspondence of 2f0 (Fh et al. 2001; Tuan et al.

5
2011).
Fig. 3 points out also the presence of secondary peaks in the high-frequency range (from 3 to 9 Hz depending on the array). For these
peaks, the mean amplitude of the H/V varies between 3 and 5 (Fig. 2), with the larger amplitudes observed for the arrays in the south
(AS and AL) where also the resonant frequencies are higher than the arrays in the north. The resonances of this secondary peak is
strongest in amplitude and centered toward highest frequencies in the southern arrays (AS and AL). Although we cannot exclude that
the secondary high-frequency H/V peaks are related to human-induced activity during the time recording (Parolai et al. 2004; Cara et
al. 2010), they are more likely due to to the presence of thin deposits of colluvial Lr soil at the top of Br (especially for the southern
part), or also to the weathering of the uppermost breccia, in good agreement with the interpretation of Del Monaco et al. (2013).

Fig. 3c illustrates also the directional H/V noise analysis considering the central station of each array. The low-frequency resonance is
particularly polarized in a direction between N+120 and N+140 in the southern part of the city (arrays AS and AL), where the
geometrical complexity increases in proximity of the boundary between LAquila terrace and the Aterno river (see Fig. 2). In the
northern part of the city, the polarization is still present but less evident (Fig. 3). For SMP array, the directional noise analysis points
out that there is more than one peak in the low-frequency range (see Fig. 3c where two close peaks at 0.46 and 0.56 Hz appear with a
different polarization). Anyhow, the cause of the low-frequency polarization and of multiple close peaks is not clear and could be
related to 2-D or 3-D resonance behavior of the L'aquila terrace (Roten & Fh 2007; Claprood et al. 2011). The N+120 and N+140
polarization in the southern part is approximately parallel to the mean strike of the Aterno river valley (N+130), but also nearly
perpendicular to the direction of the main elongation of the hill (N+20) where downtown L'Aquila is settled (see Fig. 2). A
polarization transverse to the main topography elongation is commonly observed in field experiments considering stations installed at
top of ridges (Marzorati et al. 2011; Pischiutta et al. 2010).
The directional dependence of f0 is also discussed in Matsushima et al. (2014), where the H/V ratio is alternately interpreted on the
base of the diffuse wavefield theory (Snchez-Sesma et al. 2011). Under this assumption, H/V noise ratio corresponds to the square
root of the ratio between the imaginary part of horizontal displacement when is applied a unit horizontally harmonic load, and the
imaginary part of vertical displacement for a vertically applied unit load. Matsushima et al. (2014) followed this approach for
analyzing noise polarization observed at the campus of Kyoto University in Japan characterized by a basin where the bedrock depth
vaies from 250 m up to 420 m. Using 3-D numerical simulations, they calculated the imaginary part of the Green's function applying
unit loads to the surface of their model. Their modeling returned directional dependency of H/V noise curves when the point of
simulation approached the basin edge. They concluded that the directional dependence of H/V could be a proxy of non 1-D subsurface
geology.

It would be interesting to verify if the strong fundamental resonance of L'Aquila terrace can be regarded as a topographic effect. It is
indeed well known that together with local soil conditions, geometrical morphological irregularities can amplify the seismic ground
motion during earthquakes. Numerical studies on topographic site effects used very simple 2-D elastic homogeneous models and SH,
P and SV waves as input (Boore 1972; Bouchon 1973; Geli et. al. 1988; Snchez-Sesma & Campillo 1993 among many others). A
common finding of theoretical and experimental papers is that sites at the top show ground motion amplification respect to those at the
base. The most significant effects are for incident wavelengths comparable to the horizontal dimension of the topographic feature (e.g.
mountain width). For longer wavelengths, the topographic effect is negligible (Kamalian et al. 2008). Ashford et al. (1997) modeled a
steep slope of elevation H. These authors inferred that the topographic frequency is equal to Vs/5H, with a maximum effect for H/ =
0.2, where is the incident wavelength. The elevation H for L'Aquila hill is about 50 m, and Vs is of the order of 800-1000 m/s (MS-
AQ Working Group 2010). Hence, following the formula proposed by Ashford et al. (1997), the topographic frequency should be 3-4
Hz that is far from the actual f0 at 0.5 Hz. Paolucci (2002) suggested to regard the topographic effect as the resonance of the whole
hill, similar to the 2-D resonance occurring in deep basins (Roten & Fh 2007). In this case, Paolucci (2002) furnished the 2-D
resonant frequency (f2D) of the hill for typical value of geometrical shape ratio as

f2D= (0.7-1) Vs/L

6
where L is the total width at the base of the relief. Assuming for L values 300-600 m in the southern part of L'Aquila terrace, the above
formula returns a f2D values larger than 0.9 Hz, still far from the f0 of L'Aquila.

3.2 DISPERSION CURVES

Dispersion curves derived from the FK analysis of vertical component recorded by arrays installed at L'Aquila are shown in Fig. 4.
The analysis is performed through a grid search in the wavenumber plane (kx, ky) using a sliding time-window of filtered data in
narrow frequency bands. The maximum of the spectral estimator in the wavenumber plane returns the apparent phase velocity.
Resolution and aliasing limits were defined according to Wathelet et al. (2008) computing the theoretical transfer function based on
the geometrical configuration of each array (Di Giulio et al. 2006).

The arrays arranged in the southern sector of the city (AS, AL and the 1-D MASW survey) provide consistent dispersion curves with
similar surface phase velocities in the overlapping frequencies. Therefore, the combination of the surface-wave dispersion curves
obtained for AS, AL and MASW arrays, produces a comprehensive dispersion curve (Fig. 4c) representative of the southern part of
LAquila. The shape of this combined dispersion curve can be discussed considering three main branches (Fig. 4c). At very low
frequencies (below 1.5 Hz; AS array) the surface-wave dispersion curve tends to high velocities (about 1500 m/s). For frequencies up
to 3 Hz the apparent phase velocity is quite stable at about 600 m/s. Within intermediate frequencies (from 5 to 12 Hz; AL array), we
observe an increase of apparent velocity with frequencies (from about 600 up to 800 m/s, respectively). This is in agreement with an
effect of velocity reversal in the subsoil profile, corresponding to the contact between Br and the softer lacustrine deposits. At high
frequencies (from 10 to 40 Hz as obtained from MASW survey), the surface-wave velocity decreases to about 300 m/s likely caused
by the presence of Lr.

In the northern sector of downtown, the dispersion curves provided by the four arrays (AN, PZZ, SMP and DUO) show persistently
larger apparent phase velocities than those measured in the southern area (Fig. 4a). In contrast to what observed in the southern part,
the dispersion curves related to the three deployed arrays are not so clearly overlapping (Fig. 4b), but the trend related to a velocity
reversal is still evident. The arrays with largest aperture (AN and AS) provides, below 1 Hz, similar apparent velocities in the two
sectors (Fig. 4a) suggesting comparable velocity for the seismic bedrock. On the other hand, from 1 to 3 Hz the velocities derived by
AN and AS arrays tend to 900 and 600 m/s, respectively. The very high velocities measured by PZZ and SMP arrays cannot be fully
merged with the dispersion curves obtained by the AN and DUO arrays (Fig. 4b), suggesting that local heterogeneities and possible
presence of very stiff Br layer, play an important role in the northern part of the city. It is worthy to note that a reflection survey
performed in the downtown using a minitruck as vibratory source found a strong and superficial refractor in correspondence of the
position of SMP and PZZ array (Tallini et al. 2010).

3.3 VELOCITY PROFILE FROM INVERSION

The surface-wave dispersion curves of Fig. 4 are inverted through an improved neighborhood algorithm (Wathelet 2008), with the aim
to derive the velocity profile of the subsoil. The geotechnical information collected during the microzoning activities helped in
defining the range of variability of free parameters during the inversion of the dispersion curves; e.g. we allowed the possibility of
velocity reversal among layers during the inversion step, and for the lacustrine L deposit we assumed a Vs increasing with depth
following a linear gradient as indicated by the S3 cross-hole survey carried out at the base of the L'Aquila hill (Cardarelli & Cercato
2010; Fig.s 1 and 2). Although we selected the linear gradient in the parameterization of the lacustrine layer, we allowed during the
inversion a variation of the free Vs parameter around the mean value, from 400 to 900 m/s depending on the depth as indicated by the
S3 cross-hole survey (Fig. 2).

The velocities profiles assessed by the joint inversion, including dispersion and spatial autocorrelation curves and H/V ellipticity, are
reported in Figs 5 and 6 for the southern and northern downtown, respectively. The surface-wave dispersion curves have been obtained

7
combining the results of arrays deployed in southern (AS, AL and MASW) and northern sector (AN, DUO, SMP and PZZ), as shown
in Fig. 4. The H/V used as target for the inversion has been computed averaging the H/V of the corresponding arrays up to 1 Hz; the
mean H/V curve at the northern sector was computed by taking into account the stations of SMP, PZZ, AN and DUO arrays, whereas
the mean H/V curve at the southern sector included AS and AL arrays. In the computation of the mean H/V curve, we excluded few
stations approaching the city center (Fig. 3) with the f0 peak deviating from the predominant behavior of the two sectors. Spatial
autocorrelation curves of the largest arrays (AS and AN) deployed in the southern and northern sector were also used during the
inversion; indeed the SPAC method has generally a better resolution at long periods than F-K methods (Ohori et al. 2002). Because
the combined arrays have a large spatial extension, the inverted velocity profiles of Figs. 5 and 6 can be considered representative of
the mean geotechnical properties of the respective areas.

Assuming the fundamental mode of Rayleigh wave, the surface-wave inversion for the southern sector (Fig. 5) reproduces fairly well
the experimental curves. The velocity profiles of the inverted models of Fig. 5d show from top to down: i) a very thin layer of
covering soil (thickness less than 10 m and Vs about 250 m/s corresponding to Lr or landfill); ii) a layer with depth ranging from 50 to
100 m, and with mean Vs of about 800-900 m/s corresponding to Br unit; iii) a lacustrine L deposit with Vs linearly increasing with
depth (up to about 300 m). The choice of a linear velocity-depth function in the lacustrine deposits was a-priori choice based on the
trend of the S3 survey, but the goodness of inverted models validates this initial selection in the model parameterization. The seismic
bedrock is found at a depth of about 300 m, in acceptable agreement with the geological AA section based on independent gravimetry
data derived from the microzonating activities (MS-AQ Working Group 2010; see Fig. 2).

In the northern part of the city, the selection of a combined dispersion curve is difficult (Fig. 4b) and the results of the inversion are
more questionable. To fit simultaneously the experimental curves (AN, DUO, SMP and PZZ arrays), we have to take in account higher
modes of Rayleigh waves during the inversion. Fig. 6 shows the inversion models considering the fundamental mode and three higher
Rayleigh modes. Major difficulties of the inversion are in fitting properly: i) the abrupt variation of the experimental dispersion curve,
ii) the autocorrelation curves computed on the largest spatial ring, and iii) the left branch of the H/V curve. The inverted models of
Fig. 6d still indicate a velocity reversal at a depth of about 60-80 m. In addition, L layer as derived from inversion in the northern part
of the city shows larger velocities and greater thickness in comparison to the southern sector. The seismic basement in the northern
sector is found at a depth of 550 m, which is larger than the basement depth provided by gravimetry information (less than 300 m; see
geological cross-sections in Fig. 1).

Fig. 7 shows a comparison between the best Vs models obtained in the two parts of downtown. To summarize, the inversion in both
sectors is able to resolve the velocity reversal related to the contact between Br and L deposits. The transition between Br and L occurs
in both sectors at a depth between 50-80 m. A largest depth of the seismic bedrock, and a largest Vs of the lacustrine L deposit is found
in the northern downtown (Fig. 7). The Vs of the best model in the uppermost 5m is about 300 m/s in the southern part of the city, in
agreement with the velocities expected for the red silts. The Vs of the topsoil found in the northern part is about 600 m/s, which could
be related to outcropping Br. However the Vs of the topsoil are related to the high-frequency values of the dispersion curves, which
have been assumed as higher modes of Rayleigh waves in the northern part. Fixing the frequency, higher modes are faster than the
fundamental mode, and this explains the larger values of Vs found in the northern sector (Fig. 7). The presence of higher modes was
necessary during the inversion step, aimed at model the not perfectly matching experimental dispersion curves estimated by array data
in the northern sector (Fig. 4b). Such difficulties can suggest the presence of some shallow high-velocity body that could affect the
resolution of surface methods in recognizing deeper layers. To test this possibility, we considered the SMP and PZZ dispersion curves
without account for the AN array (Fig. 8). In this case, the experimental dispersion curves can be interpreted by invoking a role of the
first higher mode of Rayleigh wave from 15 to 25 Hz, assuming a 50 m thick layer of Br , with Vs of 1000 m/s, over-imposed to an
high-velocity body (Fig. 8).

To confirm the reliability of our assumptions within the surface-wave analysis performed up to now, we compare the theoretical
transfer functions of our inverted 1-D models with the empirical transfer functions computed from aftershock data (Fig. 9). We
computed the 1-D transfer functions of SH waves (vertical incidence) on the best 300 Vs models in Fig. 5d and 6d. The experimental

8
amplification functions were derived by Milana et al. (2011) from aftershock data analysis using the standard spectral ratio respect to a
rock reference site (SSR; Borcherdt 1970). SSR method computes the ratio of Fourier amplitude spectra (FAS) of consistent
components (horizontal-to-horizontal or vertical-to-vertical ratio) between a target site and a bedrock site. SSR method implicitly
assumes that the ground motion at a given site can be deconvolved (of the source and propagation terms) by the spectral ratio
considering the same event recorded at a reference site. It is well known that the main limitation of SSR method is the selection of a
good reference site that should not be affected by any seismic local amplification (Steidl et al. 1996). SSR was computed at two
seismic stations (FAQ2 and MARG; Fig. 2) situated in the southern and northern sector of downtown, respectively. SSR at these two
stations (Fig. 9a and Fig. 9b) were computed on about 200 local aftershocks (M ranging from 0.7 to 3.6) collected during LAquila
2009 seismic sequence; the reference station was a rock site (called AQ12) in proximity of the village Poggio di Roio 2 km SW
from LAquila (see Milana et al. 2011). To check the quality of the reference site, we also computed the H/V spectral ratio on noise
and on earthquake data (Lermo & Chvez-Garca 1993). For an ideal rock site, these spectral ratios should be unitary in the entire
frequency band of analysis. The H/V ratios at the reference site (Fig. 9c) show that the reference site is characterized by a weak
seismic amplification at about 3 Hz, with a mean amplitude around 2. This local amplification on supposed bedrock site could be
related to the weathering and cracking of rocks (Steidl et al. 1996) outcropping at Poggio di Roio village. The SSRs at both stations
(Fig. 9a and Fig. 9b) depict unit value around 3 Hz, underestimating likely the real amplification because the reference site shows a
weak amplification at this frequency (see Fig. 9c). The SSR method assumes a similar seismic path between the reference site and the
analyzed site. This cannot be fully verified in our analysis because we are in near-fault conditions. Hence, the reference-to-site
distance can be comparable to the earthquake-site distance especially for very local earthquakes. As discussed in Milana et al. 2011,
the seismic sequence was characterized by several sources distributed around the city with many events at longer epicentral distances.
Therefore, the average operation on a large number of earthquakes reduces the bias introduced by very local earthquakes.

Although the limitations induced by the bedrock site and despite the fact that the velocity models obtained by inversion reproduced
average properties over wide areas, the agreement between 1-D transfer functions and SSR curves is acceptable for both stations
(FAQ2 and MARG), at least in terms of resonance frequency and general trend. We believe that the behavior of the seismic
amplification is sufficiently reproduced at both stations considering all the uncertainties of the surface-wave inversion and of the 1-D
modeling. At station FAQ2 (Fig. 9a), the SSR computed on earthquakes shows a mean amplification of about 6 at f0, and of 4 for
frequencies between 6-9 Hz. This latter high-frequency amplification agrees with the presence of soft top soils (such as red soils), as
documented by the recent microzoning activities in the southern downtown (Fig. 1a). The 1-D SH transfer functions evidence multiple
peaks well separated in frequency, these peaks are not recognized on the experimental SSR results. In order to explain such differences
between theoretical and experimental transfer functions, the effect of Q, or complexities of the seismic propagation for real
earthquakes not included in the 1-D modeling can be invoked.

At station MARG in the northern part of the city (Fig. 9b), the amplitude level assessed by SSR is lower than the southern part. This is
true up to 10 Hz. The SSR differences emerging between FAQ2 (south) and MARG station (north) suggest that the southern part of
the city has suffered larger seismic amplification than the northern part (Fig.s 9a and 9b). This is true both at low- and at high-
frequencies. The larger amplification around f0, found consistently by noise and SSR methods in the southern downtown, is probably
connected to complex site effects that cause a not-isotropic resonance of the geological structure of L'Aquila terrace. The low-
frequency resonance of L'Aquila is strongest approaching the Aterno river valley (Fig. 3).

4 BEDROCK MOTION EVALUATION

Because no time history of acceleration was recorded during the main Mw 6.3 event of the 2009 sequence on bedrock outcropping, we
deconvolved the LAquila mainshock recorded at the AQK strong-motion site (Fig.s 1 and 2) using the velocity model obtained from
the array analysis carried out in the southern downtown (Fig. 5). The reason of a deconvolution at AQK station (instead of at AQU)
arises from the lesser uncertainties on the velocity profile we have found from the inversion analysis in the southern rather than in the

9
northern part of the city. In fact, the inverted models obtained for the northern sector (AQU station) did fit simultaneously all the
fundamental and the higher modes observed on the experimental curves only invoking the presence of three higher modes (Fig. 6).
Furthermore, the bedrock depth derived from surface-wave inversion is of the order of 600 m (Fig.s 6d and 7), which is approximately
the double obtained for the bedrock depth found in the southern sector. A bedrock interface at a depth of about 600 m, in the northern
part of the city, does not agree with the bedrock depth provided by gravimetric survey within the microzoning activities ( MS-AQ
Working Group 2010). Deeper understanding of the missed accord between the bedrock depth found and the models provided by the
microzonation studies is necessary to apply the deconvolution at AQU station also.

The procedure based on a deconvolution approach has a twofold aim: the reconstruction of the bedrock motion and the validation of
the proposed velocity model. Estimating the ground motion actually experienced on the bedrock site during the mainshock is very
useful in determining eventual causes of damage in structural design verifications. Indeed, seismic response analyses need a reliable
bedrock motion to be applied as input motion at the base of the geotechnical models.

We retrieved the bedrock motion of the 6th of April 2009 earthquake by means of a 1-D deconvolution performed using both the EERA
(Bardet et al. 2000) and the STRATA code (Kottke & Rathje 2008). EERA and STRATA are well-known software used by the
engineering and geotechnical community for performing 1-D site response analysis. Both codes allow to take into account the non-
linearity in the calculation based on an equivalent linear approach: iterations are run in order to follow the variation of shear modulus
and damping ratio with shear strain. The inputs required by these codes are the soil layering and the soil properties (i.e. the unit
weight, velocity, modulus reduction and damping ratio curves as function of strain at each layer), and the seismic input motion. The
possible outputs provided by the codes are the time-series and the 1-D transfer functions. The codes also compute the shear-strain or
shear-stress evolution, and the response acceleration (Sa) or Fourier amplitude spectra (FAS). The main advantage of STRATA respect
to EERA is the possibility to include uncertainties of site properties in the computation. STRATA allows variation at each layer of Vs,
thickness, damping and modulus curves. STRATA faces the equivalent-linear site-response analysis with two alternative approaches:
time series (TS) and random vibration theory (RVT). The latter does not require an input-time series but only the Fourier amplitude
spectrum and the duration of ground-motion. Discussion on TS and RVT approaches can be found in Kottke & Rathje (2013) and
Graizer (2014). In this paper we applied STRATA using the more consolidated TS approach.

The standard and most diffuse practice in applying EERA and STRATA codes is to place the seismic input at the bedrock, and to
compute the time-series at the surface, or the transfer function between the uppermost layer and the bedrock. These codes give the
possibility to specify the location of both the seismic input and output time series. Hence, they can also be alternatively used for
estimating the motion at the bedrock once the ground motion at the outcropping surface and the properties of the soil profile are
known. The latter strategy corresponds to a 1-D deconvolution at a bedrock level (within) starting from a waveform recorded at the
surface (outcropping), and this kind of applications can be found in Figini & Paolucci (2009) or in Mulas et al. (2013). We follow a
similar strategy for deriving the bedrock motion, using both the EERA and STRATA codes. The bedrock-input-motion obtained by the
deconvolution approach is successively used for determining surface strong-motion synthetics at a target site. We consider as target
site the strong-motion site of AQV in the Aterno river valley; AQV recorded the Mw 6.3 LAquila mainshock showing the maximum
PGA in the epicentral area (PGA of 646 cm/s2; see elebi et al. 2010). We first illustrate the deconvolution at AQK site, and then the
convolution approach for the target site AQV.

The soil profile obtained by surface-wave analysis, representative of the southern part of downtown, has been used to deconvolve at
the bedrock level the mainshock recording at AQK. The Vs profile used in EERA has been obtained computing the mean of the best
Vs inverted profiles (black curve in Fig. 5d). The starting Vs profile adopted by STRATA was the same, but we included the variation
of Vs and thickness at each layer following the statistical models of Toro (1995). Using STRATA, we preferred to set the bedrock
depth to the one obtained by the surface-wave inversion, and we did not introduced variation of modulus and damping curves. The
constrain of a constant bedrock depth is necessary in this case to reduce the number of models produced by STRATA. Moreover, it
contributes in general at improving the search within the parameters space (Renalier et al. 2010).

10
One of the main difficulties of site response analysis occurs in selecting the curves that describe the variation of modulus and
damping with shear strain since laboratory tests are missing. Further, also in presence of laboratory tests on small samples, the results
cannot fully describe the real behavior of soils (Beresnev & Wen 1996). Thus, we considered stiffness decay curves G/G0 and damping
ratio ratio curves D/D0 used in the literature or suggested by the recent microzoning investigations for similar soils. In detail, for the
deconvolution at AQK site, we used modulus and damping curves obtained by microzonation studies at Roio Piano for
fluvial/lacustrine deposits (MS-AQ Working Group 2010), whereas Br and seismic limestone bedrock were assumed as materials with
linear behavior and small value of damping ratio. Following the rule of thumb of assigning quality factors equal to Vs/10, we use a
damping value (D0) of 0.5% for Br (with Vs of 800-1000 m/s) and a D 0 equal to 0.2% for limestone (Vs of 2000-2500 m/s; see Fig.
5d). For the first meters of anthropic filling behaving like sands, we used the shear-modulus curves proposed by Seed & Idriss (1970)
and the damping curves proposed by Idriss (1990). The non-linear characteristics of stiffness and damping curves used for the
deconvolution of AQK are reported in Fig. 10a.

By means of the STRATA code, we computed the bedrock motion for 100 realizations of site property variation, discarding the
models (about 10) that provided a PGA at the bedrock level larger than the actual PGA recorded at AQK (0.33 g). The deconvolution
at the bedrock level of the Mw 6.3 earthquake recorded at AQK is shown in Fig. 11. We used the time-series recorded on the EW
component, which is very similar in the spectral content to the NS component (recordings are available at:
http://itaca.mi.ingv.it/ItacaNet/; Pacor et al. 2011). The black thick curves in Fig. 11 are the median, and the median+/- 1 standard
deviations computed on the STRATA models, the red curves refer to the seismogram recorded at AQK. The deconvolved seismograms
obtained by EERA and by STRATA are shown in Fig. 11e. The seismogram in Fig. 11e by STRATA is one of the possible solutions; it
corresponds to a model that approaches the median value (black curve) over the ensemble of possible models. The two deconvolved
seismograms (Fig. 11e) look very similar each other as expected because the two codes used the same soil properties and input
motion. The deconvolved seismograms have a lower PGA than the recorded one; from 0.33 to about 0.2 g (Fig. 11e). After the
deconvolution at the bedrock level, an energy decrease mainly around the f0 is evidenced by the plots of response spectra (Fig. 11c),
of FAS (Fig. 11d) and by the time series (Fig. 11e). A similar effect is also present at high-frequency (about 10 Hz) since the very thin
surface layer effect is removed. The role of soil non-linearity is connect to the level of strain. The maximum strain level is about
0.05% reached at a depth of 100 m (Fig. 11b), and the curves in Fig.10a indicate a major role of the G/G0 decrease of the lacustrine
deposits (Fig. 10a) and a weak role of the damping.

The bedrock motion is then used for the convolution at AQV, which is situated 5 km NW from downtown LAquila in the middle
Aterno river valley; the Vs profile at AQV is provided by cross-hole data showing an average Vs of the topmost 30 m (Vs30) of about
475m/s (see http://itaca.mi.ingv.it/ItacaNet/). We also performed a noise measurement in correspondence of AQV, and the H/V curve
indicates a resonance frequency of 3 Hz (Fig. 12). The near-surface geology at AQV is composed of silty clay soils in the uppermost 5
m with Vs of 400 m/s, then a gravel layer about 25 m thick, characterized by velocity reversal at about 25 meters depth. This gravel
layer overlays silty sands and gravels. Limestone bedrock is found at a depth of 50 m (Fig. 12); the f0 at 3 Hz is in good agreement
with the Vs/4H formulation using a mean Vs of 600 m/s. The secondary HVNSR peak around 10 Hz is connected to the uppermost
velocity contrast in the first 8 m between silty clay and gravel layers. For silty clays, we adopted the Vucetic & Dobry curves (1991)
with a plasticity index (PI) of 15 as suggested in Lanzo & Pagliaroli (2012). For gravelly soils, we adopted curves proposed by San
in the MS-AQ Working Group 2010. For deepest layers of silty sands, sandy silts and gravels, we used the curves proposed by Seed &
Idriss (1970) for modulus and by Idriss (1990) for damping. The soil properties used for the convolution at AQV site are indicated in
Fig. 10b. The convolution of the bedrock motion with the soil column of AQV is shown in Fig. 13, using both the EERA and the
STRATA codes. We obtained a reasonable agreement between simulated and recorded seismograms at AQV; the observed Sa and FAS
are within the standard deviation of the STRATA computation in the frequency band between 0.5 and 10 Hz (Fig. 13c and Fig. 13d).
The EERA and STRATA time-series provide still very similar time-series showing a PGA of 0.4 g, with an underestimation of the real
PGA that was of 0.66 g. Station AQV is in the middle of the Aterno river basin, 2-D effects are expected (Lanzo & Pagliaroli 2012,
Gaudiosi et al. 2013), but cannot be captured by our 1-D analysis. EERA and STRATA code propagates vertically-incident shear
waves in a simple 1-D model, hence 2-D effects and propagating lateral surface waves are not taken into account in the computation of

11
time histories at AQV. An indication that 2-D effects are important at AQV is in the difference of the spectral content and PGA values
of the two horizontal components of the 2009 mainshock recordings (available at http://itaca.mi.ingv.it/ItacaNet/; PGA of 0.55g for the
NS, and 0.66 g for the EW record). The maximum strain level is about 0.3%, larger than the one found in the AQK analysis, indicating
a significant role of soil nonlinearities in the Aterno river valley (Fig. 10b) respect to the downtown L'Aquila. Gaudiosi et al. (2013),
comparing 1-D and 2-D simulations, observed a relevant amplification at AQV related to the generation of Rayleigh waves at the
edges of the Aterno river basin. A strong discrepancy between observed and modeled waveforms is at 0.3 Hz, where our models
exhibit spectral troughs (Fig.s 13c and 13d). The divergence between modeled and recorded data at AQV may be attributed to different
causes, such as the difficulty related to the numerical nature of the deconvolution problem. In fact, if a system produces a null output
at a particular frequency f* it is not possible to retrieve the input signal that has contribution at the same frequency f*. Around 0.3 Hz
the deconvolved signal that is the input of convolution has not energy (Fig. 11), and consequently the convolution at AQV shows the
trough at the same frequency. The discrepancy at 0.3 Hz could be also related to a near-source effect or to a deeper velocity contrast in
the Vs site profile of AQV, which is not recognized by the relatively shallow profile of Fig. 12.

Ameri et al. (2012) performed ground motion simulations in the 0.1-10 Hz frequency band at many strong-motion sites in LAquila
area using a hybrid integral-composite approach. These authors included the source kinematic rupture model and site effects
calculating the Greens function by means of the available 1-D velocity models. They provided the simulated ground motion also for
AQV, considering the 1-D soil profile shown in Fig. 12. The strong-motion synthetics of Ameri et al. (2012) is illustrated in Fig. 13e,
and the comparison between simulated and recorded waveform shows significant differences suggesting the presence of more
complex effects not reproduced by their simulation. Smerzini & Villani (2012) performed a 3-D numerical modeling by a spectral
element code in L'Aquila Aterno valley using four different kinematic fault models. They integrated the low-frequency (< 2.5 Hz)
obtained by 3-D simulation up to frequencies of engineering interest by an hybrid approach. These authors, similarly to the results of
Ameri et al. (2012), detects an underestimation of the simulated motion at AQV with respect to the observed data. In terms of PGA,
Smerzini & Villani (2012) underestimated the AQV motion by a factor of 3.

Although near-source and 2-D/3-D effects are not considered by our analysis, a generally good agreement is observed between the
ground shaking recorded at AQV and the solutions provided by the equivalent-linear model including soil nonlinearities. Finally, it is
interesting to investigate if the fit between convolved and recorded time-history at AQV may be improved allowing a variation of
damping and shear modulus curves as a function of the strain. The role of soil non-linearity in the Aterno River Valley has been
discussed in several previous studies (Puglia et al. 2011; Lanzo & Pagliaroli 2012; Gaudiosi et al. 2013), with a general agreement that
non-linear effects are possible in the soft deposits of the Aterno river valley. However, uncertainties on non-linearity of soils of Aterno
valley remain in general large, because real information are available only on few discrete sites. We performed a new run of STRATA
allowing, additionally to variation of Vs and thickness within the layers, also variation of damping and modulus curves. STRATA code
varies the nonlinear soil properties by means of a truncated normal distribution where G/G0 and D curves have a negative correlations
(see Kottke & Rathje 2008). The STRATA results including variation of nonlinear curves (third synthetic in Fig. 13e), are similar to
the models obtained by varying only the Vs and the thickness (second synthetic in Fig. 13e).

12
5 CONCLUSION

This paper shows a case history of surface-wave analysis based on array techniques in downtown LAquila, which is characterized by
a low-frequency resonance (0.5-0.6 Hz). This resonance is polarized, strong and shifted toward higher frequency in the southern sector
of downtown. By analyzing data from several 2-D arrays with different apertures and 1-D array of geophones, we derived surface-
wave dispersion curves in a large frequency band (0.8-40 Hz). Inverting the dispersion curves through a neighbourhood algorithm, we
provided velocity models representative of the subsoil structure of LAquila, in agreement with the geological models produced during
the microzoning activities.

For both the southern and northern part of the city, the results show clearly the effect of a velocity inversion between relatively stiff
calcareous breccias (Br) and ancient lacustrine deposit (L). In the northern part of the city, the surface-wave dispersion curves indicate
a complex situation with effect of higher modes, and very high-velocity bodies that likely could mask the presence of the lacustrine
deposit. Further, the depth of seismic bedrock is not consistent with independent information provided by microzoning activities. In
the southern part of the city, the inversion based on the fundamental Rayleigh mode reproduces well the experimental curves.

The Vs profile, obtained by surface-wave inversion for the southern sector of the city, has been used as input for deriving the seismic
input at the bedrock level by deconvolving the Mw 6.3 mainshock recorded at LAquila. Then, this seismic input has been convolved
with the soil profiles of a strong-motion target site; AQV in the middle Aterno valley that recorded the maximum PGA during the
mainshock. Deconvolution with EERA and STRATA codes is a simple practice that, in our case, is successful in returning a reasonable
match between simulated and recorded data. It also confirms the reliability of the proposed geotechnical model and suggests that a 1-
D modelling is able to reproduce the main features of the seismic amplification in downtown. The final comparison at AQV between
synthetics and observed strong-motion waveform, shows that our synthetic reproduces better the real ground motion than the
synthetics found at the same site by 2-D or 3-D modeling of other authors, even when they use hybrid approaches to enlarge the
frequency band of analysis up to 10 Hz. The main advantage of a deconvolution/convolution approach is to start from a real
earthquake using available codes for performing 1-D site nonlinear response analysis. The drawbacks are to not consider near-source
or 2-D/3-D effects. This works empathizes that a seismic characterization at strong-motion site is fundamental to have a detailed
knowledge of the site soil properties.

ACKNOWLEDGEMENTS

Preliminary results of this work were shown at the 4th IASPEI/IAEE International Symposium (Effects of Surface Geology on
Seismic Motion; August 2326, 2011; University of California Santa Barbara). We thank Gaetano De Luca, Francesco Del Monaco
and Federica Durante for their help in field operation. We thank Sara Amoroso for fruitful discussions on geotechnical aspects, and
Gabriele Ameri for providing us synthetic data at AQV station.

13
REFERENCES

Aki, K., 1957. Space and time spectra of stationary stochastic waves, with special reference to microtremors, Bull. Earthq. Res. Inst.,
35, 415456.

Ameri, G., Gallovi, F. & Pacor, F., 2012. Complexity of the Mw 6.3 2009 LAquila (central Italy) earthquake: 2. Broadband strong
motion modeling, J. geophys. Res., 117, B04308, doi:10.1029/2011JB008729.

Amoroso, S., Del Monaco, F., Di Eusebio, F., Monaco, P., Taddei, B., Tallini, M., Totani, F. & Totani, G., 2010. Campagna di indagini
geologiche, geotecniche e geofisiche per lo studio della risposta sismica locale della citt dell'Aquila: la stratigrafia dei sondaggi
(Giugno-Agosto 2010), Report CERFIS, 1, 1-51, available at http://www.cerfis.it/it/attivita/microzonazione.html, in Italian.

Ashford, S. A., Sitar, N., Lysmer, J., & Deng, N., 1997. Topographic effects on the seismic response of steep slopes, Bull. seism. Soc.
Am., 87(3), 701-709.

Asten, M. W., 2006. On bias and noise in passive seismic data from finite circular array data processed using SPAC methods:
Geophysics, 71, V153-V162, doi: 10.1190/1.2345054.

Bard, P. Y., et al., 2010. From non-invasive site characterization to site amplification: recent advances in the use of ambient vibration
measurements, in Earthquake Engineering in Europe, Geotechnical, Geological and Earthquake Engineering, 17, doi: 10.1007/978-
90-481-9544-2_5, M. Garevski and A. Ansal eds.

Bardet, J. P., Ichii, K. & Lin, C.H. 2000. EERA, a computer program for equivalent-linear earthquake site response analyses of layered
soil deposits, Department of Civil Engineering, University of Southern California, available at http://gees.usc.edu/GEES/.

Beresnev, I. A., & Wen, K. L., 1996. Nonlinear soil responsea reality?. Bull. seism. Soc. Am., 86(6), 1964-1978.

Bettig, B., Bard, P.Y., Scherbaum, F., Riepl, J., Cotton, F., Cornou, C. & Hatzfeld, D., 2001. Analysis of dense array noise
measurements using the modified spatial autocorrelation method (SPAC). Application to the Grenoble area, Boll. Geof. Teor. Appl., 42,
281-304.

Bonnefoy-Claudet, S., Cornou, C., Bard, P.Y., Cotton, F., Moczo, P., Kristek, J. & Fh, D., 2006. H/V ratio: a tool for site effects
evaluation. Results from 1-D noise simulations, Geophys. J. Int., 167, 827837, doi:10.1111/j.1365-246X.2006.03154.x.

Bonnefoy-Claudet, S., Baize, S., Bonilla, L. F., Berge-Thierry, C., Pasten, C., Campos, J., Volant, P. & Verdugo, R., 2009. Site effect
evaluation in the basin of Santiago de Chile using ambient noise measurements. Geophys. J. Int., 176, 925937. doi: 10.1111/j.1365-
246X.2008.04020.x.

Borcherdt, R. D. 1970. Effects of local geology on ground motion near San Francisco Bay. Bull. seism. Soc. Am., 60(1), 29-61.

Bouchon, M., 1973. Effect of topography on surface motion. Bull. seism. Soc. Am., 63, 615632.

Boore, D. M., 1972. A note on the effect of simple topography on seismic SH waves. Bull. seism. Soc. Am., 62, 275284.

Bordoni, P., Haines, J., Milana, G., Marcucci, S., Cara, F., & Di Giulio, G., 2011. Seismic response of LAquila downtown from
comparison between 2D synthetics spectral ratios of SH, P-SV and Rayleigh waves and observations of the 2009 earthquake sequence,
Bull. Earthq. Eng., 9, 761781, doi 10.1007/s10518-011-9247-5.

Capon, J., 1969. High-resolution frequency-wavenumber spectrum analysis, Proc. of the IEEE, 57, 14081418.

14
Cara, F., Di Giulio, G., Milana, G., Bordoni, P., Haines, J., & Rovelli, A., 2010. On the stability and reproducibility of the horizontal-
to-vertical spectral ratios on ambient noise: case study of Cavola, Northern Italy, Bull. seism. Soc. Am., 100(3), 1263-1275.

Cardarelli, E. & Cercato, M., 2010. Relazione sulla campagna d'indagine geofisica per lo studio della risposta sismica locale della citt
dell'Aquila, Prova Crosshole Sondaggi S3-S4, Report DICEA, pp. 1-13, available at http://www.cerfis.it/download/file/62-relazione-
sulla-campagna-d-indagine-geofisica-per-lo-studio-della-risposta-sismica-locale-della-citta-dell-aquila-prova-crosshole-sondaggi-s3-
s4, in Italian.

Castellaro, S., & Mulargia, F., 2009. The Effect of Velocity Inversions on H/V, Pure appl. Geophys., 166, 567592, 0033
4553/09/04056726, doi: 10.1007/s00024-009-0474-5.

elebi, M., Bazzurro, P. Chiaraluce, L., Clemente, P., Decanini, L., De Sortis, A., Ellsworth, W., Gorini, A., Kalkan, E., Marcucci, S.,
Milana, G., Mollaioli, F., Olivieri, M., Paolucci, R., Rinaldis, D., Rovelli, A., Sabetta, F. & Stephens, C., 2010. Recorded Motions of
the 6 April 2009 Mw 6.3 LAquila, Italy, Earthquake and Implications for Building Structural Damage: Overview, Earthquake
Spectra, 26, 651-684, doi:10.1193/1.3450317.

Claprood, M., Asten, M.W. & Kristek, J., 2011. Using the SPAC microtremor method to identify 2D effects and evaluate 1D shear-
wave Velocity profile in valleys, Bull. seism. Soc. Am., 101, 826847, doi: 10.1785/0120090232.

Cornou, C., Bard, P.Y. & Dietrich, M., 2003. Contribution of dense array analysis to the identification and quantification of basin-
edge-induced waves, Part II: Application to Grenoble basin (French Alps), Bull. seism. Soc. Am., 93, 26242648.

De Luca, G., Marcucci, S., Milana, G. & San, T., 2005. Evidence of Low-Frequency Amplification in the City of LAquila, Central
Italy, through a Multidisciplinary Approach Including Strong- and Weak-Motion Data, Ambient Noise, and Numerical Modeling, Bull.
seism. Soc. Am., 95, 1469 - 1481.

Del Monaco, F., Tallini, M., De Rose, C. & Durante, F., 2013. HVNSR survey in historical downtown L'Aquila (central Italy): site
resonance properties vs. subsoil model, Engineering Geology, 158, 34-47, doi: 10.1016/j.enggeo.2013.03.008.

Di Giulio, G., Cornou, C., Ohrnberger, M., Wathelet, M. & Rovelli, A. 2006. Deriving wavefield characteristics and shear-wave
velocity profiles from two-dimensional small-aperture arrays analysis of ambient vibrations in a small-size alluvial basin, Colfiorito,
Italy, Bull. seism. Soc. Am., 96, 1915-1933, doi: 10.1785/0120060119.

Fh, D., Kind, F. & Giardini, D., 2001. A theoretical investigation of average H/V ratios , Geophys. J. Int., 145, 535-549.

Figini, R. & Paolucci, R., 2009. Site effects at long periods from digital strong motion records of the Kik-net, Japan, Journal of
Earthquake Engineering, 13, 567584.

Foti, S., Parolai, S., Bergamo, P., Di Giulio, G., Maraschini, M., Milana, G., Picozzi, M. & Puglia, R., 2011. Surface wave surveys for
seismic site characterization of accelerometric stations in ITACA, Bull. Earthq. Eng., 9, 1797-1820, doi: 10.1007/s10518-011-9306-y.

Gaudiosi, I., Giuliano, M., & Marco, T., 2013. Site effects in the Aterno River Valley (LAquila, Italy): comparison between empirical
and 2D numerical modelling starting from April 6th 2009 Mw 6.3 earthquake. Bull. Earthquake Eng., 1-20, doi10.1007/s10518-013-
9540-6.

Gli, L., Bard, P. Y., & Jullien, B. 1988. The effect of topography on earthquake ground motion: A review and new results, Bull. seism.
Soc. Am., 78(1), 42-63.

Gosar, A., Roer, J., Motnikar, B. ., & Zupani, P. 2010. Microtremor study of site effects and soil-structure resonance in the city of

15
Ljubljana (central Slovenia), Bull. Earthquake Eng., 8(3), 571-592.

Graizer, V.. 2014. Comment on Comparison of time series and random vibration theory site response methods by Albert R. Kottke
and Ellen M. Rathje. Bull. seism. Soc. Am., 104(1), doi: 10.1785/0120130176.

Grnthal, G., 1998. European macroseismic scale, EMS-98, vol. 15. Luxembourg: Centre Europen de Godynamique et de
Sismologie.

Kamalian, M., Jafari, M. K., Sohrabi-Bidar A. & Razmkhah, A. 2008. Seismic response of 2-D semi-sine shaped hills to vertically
propagating incident waves: amplification patterns and engineering applications. Earthquake Spectra, 24(2), 405-430.

Konno, K., & Ohmachi, T., 1998. Ground-motion characteristics estimated from spectral ratio between horizontal and vertical
components of microtremor, Bull. seism. Soc. Am., 88, 228-241.

Kottke, A. R., & Rathje, E. M., 2008. Technical Manual for Strata, PEER Report 2008/10, Pacific Earthquake Engineering Research
Center College of Engineering, University of California, Berkeley. (available
http://peer.berkeley.edu/publications/peer_reports/reports_2008/web_PEER810_KOTTKE_Rathje.pdf)

Kottke, A. R., & Rathje, E. M., 2013. Comparison of Time Series and Random Vibration Theory Site Response Methods, Bull. seism.
Soc. Am., 103(3), 2111-2127.

Idriss, I. M., 1990. Response of Soft Soil Sites during Earthquakes, Proceedings, Memorial Symposium to honor Professor Harry
Bolton Seed, Berkeley, California, Vol. II, May.

Lacoss, R.T., Kelly, E.J. & Toksoz, M.N., 1969. Estimation of seismic noise structure using arrays, Geophysics, 34, pp. 21-38.

Lanzo, G., Tallini, M., Milana, G., Di Capua, G., Del Monaco, F., Pagliaroli, A. & Peppoloni, S., 2011. The Aterno Valley strong-
motion array: seismic characterization and determination of subsoil model, Bull. Earthq. Eng., 9, 18551875, doi: 10.1007/s10518-
011-9301-3.

Lanzo, G. & Pagliaroli, A., 2012. Seismic site effects at near-fault strong-motion stations along the Aterno River Valley during the M
w 6.3 2009 LAquila earthquake, Soil Dyn. Earth. Eng., 40, 1-14, doi: 10.1016/j.soildyn.2012.04.004.

Lermo J., & Chvez-Garca, F. J. 1993. Site effect evaluation using spectral ratios with only one station, Bull. seism. Soc. Am., 83(5),
1574-1594.

Lunedei, E. & Albarello, D., 2010. Theoretical HVSR curves from full waveeld modeling of ambient vibrations in a weakly
dissipative layered Earth, Geophys. J. Int., 181, 10931108, doi:10.1111/j.1365-246X.2010.04560.x.

Martelli, L., Boncio, P., Baglione, M., Cavuoto, G., Mancini, M., Scarascia Mugnozza, G. & Tallini, M. 2012. Main geological factors
controlling site response during the 2009 LAquila earthquake, Ital. J. Geosci. (Boll. Soc. Geol. It.), 131, 423-439,
doi:10.3301/IJG.2012.12.

Marzorati, S., Ladinam C., Falcuccim E., Gorim S., Sarolim M., Amerim G. & Galadini, F., 2011. Site Effects On the Rock: the case
study of Castelvecchio Subequo (LAquila, Central Italy), Bull. Earthq. Eng., 9, 841-868, DOI:10.1007/s10518-011- 9263-5.

Matsushima, S., Hirokawa, T., De Martin, F., Kawase, H., & Snchez Sesma, F. J., 2014. The Effect of Lateral Heterogeneity on
HorizontaltoVertical Spectral Ratio of Microtremors Inferred from Observation and Synthetics. Bull. seism. Soc. Am., 104, 381-393,
doi:10.1785/0120120321.

16
Milana, G., Azzara, R.M., Bertrand, E., Bordoni, P., Cara, F., Cogliano, R., Cultrera, G., Di Giulio, G., Duval, A.M., Fodarella, A. et
al., 2011. The contribution of seismic data in microzonation studies for downtown LAquila, Bull. Earthq. Eng., 9, 741-759, doi:
10.1007/s10518-011-9246-6.

Monaco, P. et al., 2012. Geotechnical aspects of LAquila earthquake, in Special Topics in Earthquake Geotechnical Engineering, pp.
1-66, Springer Netherlands.

MS-AQ Working Group, 2010. Microzonazione sismica per la ricostruzione dellarea aquilana, Ed. Regione Abruzzo- Dipartimento
della Protezione Civile, pp. 1-796, 3 vol & CD-rom, in Italian.

Mucciarelli, M., Bianca, M., Ditommaso, R., Gallipoli, M. R., Masi, A., Milkereit, C., Parolai, S., Picozzi, m, & Vona, M., 2011. Far
field damage on RC buildings: the case study of Navelli during the LAquila (Italy) seismic sequence, 2009. Bull. Earthquake Eng.,
9(1), 263-283.

Mulas, M. G., Perotti, F., Coronelli, D., Martinelli, L., & Paolucci, R., 2013. The partial collapse of Casa dello Studente during
LAquila 2009 earthquake, Engineering Failure Analysis, 34, 566584.

Nakamura, Y., 1989. A method for dynamic characteristics estimation of subsurface using ambient noise on the ground surface, Q. R.
Rail. Tech. Res. Inst., 30, 1, 25-33.

Nogoshi, M. & Igarashi, T., 1971. On the Amplitude Characteristics of Ambient noise, Part 2, J. Seism. Soc. Japan, 24, 26-40.

Ohori, M., Nobata, A. & Wakamatsu, K., 2002. A comparison of ESAC and FK methods of estimating phase velocity using arbitrarily
shaped microtremor analysis, Bull. seism. Soc. Am., 92, 2323-2332.

Pacor, F., Paolucci, R., Luzi, L., Sabetta, F. Spinelli, A., Gorini, A., Nicoletti, M., Marcucci, S., Filippi, L. & Dolce, M. 2011.
Overview of the Italian strong motion database ITACA 1.0, 9Bull Earthquake Eng, 6, 17231739. doi: 10.1007/s10518-011-9327-6.

Panou, A. A., Theodulidis, N., Hatzidimitriou, P., Stylianidis, K., & Papazachos, C. B. 2005. Ambient noise horizontal-to-vertical
spectral ratio in site effects estimation and correlation with seismic damage distribution in urban environment: the case of the city of
Thessaloniki (Northern Greece), Soil Dynamics and Earthquake Engineering, 25(4), 261-274.

Paolucci, R., 2002. Amplification of earthquake ground motion by steep topographic irregularities, Earth. Eng. Struct. Dyn., 31,
1831-1853, doi:10.1002/eqe.192.

Parolai, S., Richwalski, S. M., Milkereit, C., & Bormann, P., 2004. Assessment of the stability of H/V spectral ratios from ambient
noise and comparison with earthquake data in the Cologne area (Germany), Tectonophysics, 390(1), 57-73.

Pischiutta, M., Cultrera, G., Caserta, A., Luzi, L. & Rovelli, A., 2010. Topographic effects on the hill of Nocera Umbra, central Italy,
Geophys. J. Int., 182, 2, 977987.

Puglia, R., Ditommaso, R., Pacor, F., Mucciarelli, M., Luzi, L., & Bianca, M., 2011. Frequency variation in site response as observed
from strong motion data of the LAquila (2009) seismic sequence. Bull. Earthquake Eng., 9(3), 869-892.

Renalier, F., Jongmans, D., Savvaidis, A., Wathelet, M., Endrun, B., and Cornou, C., 2010. Influence of parameterization on inversion
of surface wave dispersion curves and definition of an inversion strategy for sites with a strong VS contrast. Geophysics, 75(6), B197
B209. Doi: 10.1190/1.3506556.

Roten, D. & Fh, D., 2007. A combined inversion of Rayleigh wave dispersion and 2-D resonance frequencies, Geophys. J. Int., 168,

17
12611275, doi 10.1111/j.1365-246X.2006.03260.x.

Snchez-Sesma, F. J. & Campillo, M. 1993. Topographic effects for incident P, SV and Rayleigh wavesm Tectonophysics 218(1), 113-
125.

Snchez-Sesma, F. J., Rodrguez, M., Iturrarn-Viveros, U., Luzn, F., Campillo, M., Margerin, L., Garca-Jerez, A., Suarez, M.,
Santoyo, M. A. & Rodrguez-Castellanos, A., 2011. A theory for microtremor H/V spectral ratio: application for a layered medium,
Geophys. J. Int., 186, 221225. doi: 10.1111/j.1365-246X.2011.05064.x.

Seed, H.B. & Idriss, I.M., 1970. Soil Moduli and Damping Factors for Dynamic Response Analysis, Earthquake Engineering
Research Center, Report No. UCB/EERC-70/10, University of California, Berkeley, California (USA), December, 48p.

Seed, H.B., Wong, R.T., Idriss, I.M. &Tokimatsu, K., 1986. Moduli and damping factors for dynamic analyses of cohesionless soils, J.
Geotech. Eng. Div., ASCE, 112, 1016-1032.

Smerzini, C., & Villani, M., 2012. Broadband Numerical Simulations in Complex Near Field Geological Configurations: The Case of
the 2009 Mw 6.3 LAquila Earthquake. Bull. seismol. Soc. Am., 102(6), 2436-2451, doi: 10.1785/0120120002.

Steidl, J. H., Tumarkin, A. G., & Archuleta, R. J., 1996. What is a reference site? Bull. seismol. Soc. Am. 86, 17331748.

Steimen, S., Fh, D., Kind, F., Schmid, C. & Giardini, D., 2003. Identifying 2D resonance in microtremor wavefields, Bull. seismol.
Soc. Am., 93, 58359.

Tallini, M., Restaino, L., Berarducci, R., Del Monaco, F., Di Fiore, V., Bruno, P.P., Castello, G., Cavuoto, G., De Rosa, D., Iavarone,
M., Pelosi, N., Punzo, M., Di Settimo, P.S., Taralli, D. & Varriale, F., 2010. Indagini sismiche, Report CERFIS, 3, available at
http://www.cerfis.it/it/attivita/microzonazione.html, in Italian.

Tallini, M., Cavuoto, G., Del Monaco, F., Di Fiore, V., Mancini, M., Caielli, G., Cavinato, G.P., De Franco, R., Pelosi, N. & Rampolla
A., 2012. Seismic surveys integrated with geological data for in-depth investigation of Mt. Pettino active fault area (Western L'Aquila
Basin), Ital. J. Geosci. (Boll. Soc. Geol. It.), 131, 389-402, doi: 10.3301/IJG.2012.10.

Tertulliani, A., Arcoraci, L., Berardi, M., Bernardini, F., Camassi, R., Castellano, C., Del Mese, S., Ercolani, E., Graziani, L.,
Leschiutta, I., Rossi, A. & Vecchi, M., 2011. An application of EMS98 in a medium-sized city: The case of LAquila (Central Italy)
after the April 6, 2009 Mw 6.3earthquake, Bull. Earthq. Eng., 9, 6780, doi:10.1007/s10518-010-9188-4.

Tertulliani, A., Leschiutta, I., Bordoni, P. & Milana, G., 2012. Damage distribution in LAquila City (Central Italy) during the 6 April
2009 Earthquake, Bull. seism. Soc. Am., 102, 15431553 , doi:10.1785/0120110205.

Tokimatsu, K., 1997. Geotechnical site characterization using surface waves, in Earthquake Geotechnical Engineering: Proceedings
IS-Tokyo 95, the first International. Conf. on Earthq. Geotech. Eng., K. Ishihara (Editor), A. A. Balkema, Rotterdam, The Netherlands,
13331368.

Toro, G.R., 1995. Probabilistic models of site velocity profiles for generic and site-specific ground-motion amplification studies,
Technical Report 779574, Brookhaven National Laboratory, Upton, NY.

Tuan, TT., Scherbaum, F. & Malischewsky, P.P., 2011. On the relationship of peaks and troughs of the ellipticity (H/V) of Rayleigh
waves and the transmission response of single layer over half-space models, Geophys. J. Int., 184, 793-800, doi:10.1111/j.1365-
246X.2010.04863.x

18
Uebayashi, H., 2003. Extrapolation of irregular subsurface structures using the horizontal-to-vertical spectral ratio of long-period
microtremors, Bull. seismol. Soc. Am., 93, 570582.

Vucetic, M. & Dobry, R., 1991. Effects of the soil plasticity on ciclyc response, J. Geotech. Eng. Div., ASCE, 117, No.1, pp. 89-107.

Wathelet, M., 2008. An improved neighborhood algorithm: parameter conditions and dynamic scaling, Geophys. Res. Lett., 35,
L09301, doi:10.1029/2008GL033256.

Wathelet, M., Jongmans, D., Ohrnberger, M. & Bonnefoy-Claudet, S. 2008. Array performances for ambient vibrations on a shallow
structure and consequences over Vs inversion, J. Seismol., 12, pp. 1-19.

19
FIGURE CAPTIONS

Figure 1. a) Geological plan of the studied area; downtown LAquila is bounded by a dashed curve. b) Stratigraphic columns for
surveys S2, S3 and S8 described in the text. The velocity models available at sites S3 and S8 are also shown. c). Contour map of high-
frequency peak (f1) plotted together with the most severe grade damage on residential buildings (modified after Del Monaco et al.
2013). d) Two geological cross-sections (AA and BB) are shown (modified after MS-AQ Working Group 2010). The horizontal lines
indicate the spatial extension of the arrays.

Figure 2. Digital elevation model (DEM) of downtown LAquila. Stations of the six 2-D arrays (AL, AS, AN, DUO, PZZ and SMP)
installed in downtown are plotted. The three geotechnical surveys (S2, S3 and S8 of Fig. 1) are plotted as circles. The linear 1-D array
of geophones (MASW experiment nearby S8 survey) is also indicated in the map. FAQ2 and MARG are two seismic stations installed
after the mainshock and recorded many aftershocks. The two stations that recorded the mainshock (AQK and AQU) are plotted as
triangles. AQV (out of map) is a strong-motion site in the Aterno river valley; it recorded the maximum PGA during the 2009
mainshock.

Figure 3. a) HVNSR curves computed for the average horizontal components for each station of the 2-D arrays (AL, AS, AN, DUO,
PZZ and SMP) installed in downtown. b) FAS for the central station of each array. Red, green and blue curves show the mean FAS for
the vertical, NS and EW component, respectively. The vertical black line shows the value of f0 evaluated by HVNSR curves. c)
Directional analysis on the HVNSR curves computed at the central station of each array. Frequency and direction (the angle is
measured clockwise from North) are plotted on the x- and y-axes, whereas the color scale is proportional to the amplitude level of
HVNSR curves.

Figure 4. a) Surface-wave dispersion curves resulting from FK analysis at different arrays. All the dispersion curves were plotted
together. b) Dispersion curves of the arrays deployed in the northern (AN, DUO, SMP and PZZ), and c) in the southern part (AS, AL
and MASW) of downtown.

Figure 5. Inversion results for the southern part of the city (AS, AL and 1-D MASW array). The black and the gray curves show the
experimental and the fitting inverted models, respectively. The comparison between observed dispersion curves and theoretical
Rayleigh dispersion curves is shown in (a), the comparison in terms of spatial autocorrelation curves is plotted in (b), and the actual
H/V noise curve is compared to the theoretical Rayleigh wave ellipiticity in (c). The resulting Vp and Vs profiles are shown in (d), the
mean Vs profile is also indicated by a black curve.

Figure 6. Inversion results for the northern part of the city (AN, DUO, SMP and PZZ). The black and the gray curves show the
experimental and the fitting inverted models, respectively. The comparison between observed dispersion curves and theoretical
dispersion curves is shown in (a), taking into account the fundamental mode and the three first higher modes of Rayleigh waves. The
comparison in terms of spatial autocorrelation curves is plotted in (b), and the actual H/V noise curve is compared to the theoretical
Rayleigh wave ellipticity in (c). The resulting Vp and Vs profiles are shown in (d), the mean Vs profile is also indicated by a black
curve.

Figure 7. a) Comparison between the best Vs model obtained in northern (black curve) and southern (dashed gray curve) part of the
city. b) The same in (a) for the uppermost 120 m.

Figure 8. Velocity profiles (a) used for forward computation (b) of Rayleigh dispersion curves (assuming fundamental, first and
second higher mode of Rayleigh waves) for the north sector. The experimental surface wave dispersion curves (with associated
standard deviation) inferred at SMP and PZZ arrays are also shown in (b).

Figure 9. Comparison between the 1-D response to vertically incident SH waves (black curves) and SSR results obtained from
aftershock data analysis (gray curves showing the geometrical average 1 standard deviation). SSR were computed for FAQ2 (a) and

20
MARG (b) stations (Milana et al. 2011), located in the southern and northern part of the city, respectively. The velocities models for
the computation of SH response have been provided by the best 300 models derived from inversion inversion (Figs. 5d and 6d). Q
values are set equal to the velocities values divided by 10. The panel (c) shows the H/V computed both on noise and earthquake data at
the station AQ12 used as reference site in the SSR method.

Figure 10. Normalized shear modulus and damping ratio curves used respectively for deconvolution at AQK (panel a) and
convolution analysis at AQV (panel b).

Figure 11. Deconvolution results at AQK. The different models of STRATA are shown as gray curves in terms of Vs profiles (a), shear
strain profiles (b), acceleration response spectra (c) and raw Fourier amplitude spectra (d). The mean (and mean +/- 1 standard
deviation) are plotted as thick black curve. The observed data at AQK is indicated by a red curve. Panel e shows the comparison
between deconvolved and recorded seismograms.

Figure 12. Soil column, Vs profile from cross-hole data and HVNSR curve at AQV site. The black line in the Vs profile indicates the
simplified profile used in the convolution analysis.

Figure 13. Convolution results at AQV. The different models of STRATA are shown as gray curves in terms of Vs profiles (a), shear
strain profiles (b), acceleration response spectra (c) and raw Fourier amplitude spectra (d). The mean (and mean +/- 1 standard
deviation) are plotted as thick black curve. The observed data at AQV is indicated by a red curve. Panel e shows the comparison
between deconvolved and recorded seismograms (the circle symbols on time-series indicate the PGA). The synthetic at AQV of Ameri
et al. 2012 is also plotted.

21
Table 1. List of the 2-D arrays deployed in the city. Column no. 3 shows the length of synchronous signal collected by each
array. Column no. 5 shows the wavenumber limits after the computation of theoretical array transfer function (Wathelet et al.,
2008).

Maximum Kmin-Kmax
Number of Time-period of
Array Name aperture (m)
stations recording (rad/m)

AS 15 20:00 -21:20 1197 0.0082-0.11

(29 November
2010)

AN 14 14:20 -16:00 940 0.0095-0.13

(30 November
2010)

AL 12 13:40 -15:10 170 0.054-0.22

(2 December
2010)

SMP 12 11:50 -13:00 193 0.039-0.24

(17 May 2011)

PZZ 12 09:30-11:30 85 0.132-0.65

(18 May 2011)

DUO 10 15:30-17:00 230 0.025-0.11

(9 August 2011)

22
Figure 1. a) Geological plan of the studied area; downtown LAquila is bounded by a dashed curve. b) Stratigraphic columns for
surveys S2, S3 and S8 described in the text. The velocity models available at sites S3 and S8 are also shown. c). Contour map of high-
frequency peak (f1) plotted together with the most severe grade damage on residential buildings (modified after Del Monaco et al.
2013). d) Two geological cross-sections (AA and BB) are shown (modified after MS-AQ Working Group 2010). The horizontal lines
indicate the spatial extension of the arrays.

23
Figure 2. Digital elevation model (DEM) of downtown LAquila. Stations of the six 2-D arrays (AL, AS, AN, DUO, PZZ and SMP)
installed in downtown are plotted. The three geotechnical surveys (S2, S3 and S8 of Fig. 1) are plotted as circles. The linear 1-D array
of geophones (MASW experiment nearby S8 survey) is also indicated in the map. FAQ2 and MARG are two seismic stations installed
after the mainshock and recorded many aftershocks. The two stations that recorded the mainshock (AQK and AQU) are plotted as
triangles. AQV (out of map) is a strong-motion site in the Aterno river valley; it recorded the maximum PGA during the 2009
mainshock.

24
Figure 3. a) HVNSR curves computed for the average horizontal components for each station of the 2-D arrays (AL, AS, AN, DUO,
PZZ and SMP) installed in downtown. b) FAS for the central station of each array. Red, green and blue curves show the mean FAS for
the vertical, NS and EW component, respectively. The vertical black line shows the value of f0 evaluated by HVNSR curves. c)
Directional analysis on the HVNSR curves computed at the central station of each array. Frequency and direction (the angle is
measured clockwise from North) are plotted on the x- and y-axes, whereas the color scale is proportional to the amplitude level of
HVNSR curves.

25
Figure 4. a) Surface-wave dispersion curves resulting from FK analysis at different arrays. All the dispersion curves were plotted
together. b) Dispersion curves of the arrays deployed in the northern (AN, DUO, SMP and PZZ), and c) in the southern part (AS, AL
and MASW) of downtown.

26
Figure 5. Inversion results for the southern part of the city (AS, AL and 1-D MASW array). The black and the gray curves show the
experimental and the fitting inverted models, respectively. The comparison between observed dispersion curves and theoretical
Rayleigh dispersion curves is shown in (a), the comparison in terms of spatial autocorrelation curves is plotted in (b), and the actual
H/V noise curve is compared to the theoretical Rayleigh wave ellipticity in (c). The resulting Vp and Vs profiles are shown in (d), the
mean Vs profile is also indicated by a black curve.

27
Figure 6. Inversion results for the northern part of the city (AN, DUO, SMP and PZZ). The black and the gray curves show the
experimental and the fitting inverted models, respectively. The comparison between observed dispersion curves and theoretical
dispersion curves is shown in (a), taking into account the fundamental mode and the three first higher modes of Rayleigh waves. The
comparison in terms of spatial autocorrelation curves is plotted in (b), and the actual H/V noise curve is compared to the theoretical
Rayleigh wave ellipticity in (c). The resulting Vp and Vs profiles are shown in (d), the mean Vs profile is also indicated by a black
curve.

28
Figure 7. a) Comparison between the best Vs model obtained in northern (black curve) and southern (dashed gray curve) part of the
city. b) The same in (a) for the uppermost 120 m.

29
Figure 8. Velocity profiles (a) used for forward computation (b) of Rayleigh dispersion curves (assuming fundamental, first and
second higher mode of Rayleigh waves) for the north sector. The experimental surface wave dispersion curves (with associated
standard deviation) inferred at SMP and PZZ arrays are also shown in (b).

30
Figure 9. Comparison between the 1-D response to vertically incident SH waves (black curves) and SSR results obtained from
aftershock data analysis (gray curves showing the geometrical average 1 standard deviation). SSR were computed for FAQ2 (a) and
MARG (b) stations (Milana et al., 2011), located in the southern and northern part of the city, respectively. The velocities models for
the computation of SH response have been provided by the best 300 models derived from inversion (Figs 5d and 6d). Q values are set
equal to the velocities values divided by 10. The panel (c) shows the H/V computed both on noise and earthquakes for the station
AQ12 used as reference site in the SSR method.

31
Figure 10. Normalized shear modulus and damping ratio curves used respectively for deconvolution at AQK (panel a) and
convolution analysis at AQV (panel b).

32
Figure 11. Deconvolution results at AQK. The different models of STRATA are shown as gray curves in terms of Vs profiles (a), shear
strain profiles (b), acceleration response spectra (c) and raw Fourier amplitude spectra (d). The mean (and mean +/- 1 standard
deviation) are plotted as thick black curve. The observed data at AQK is indicated by a red curve. Panel e shows the comparison
between deconvolved and recorded seismograms (the circle symbols on time-series indicate the PGA).

33
Figure 12. Soil column, Vs profile from cross-hole data and HVNSR curve at AQV site. The black line in the Vs profile indicates the
simplified profile used in the convolution analysis.

34
Figure 13. Convolution results at AQV. The different models of STRATA are shown as gray curves in terms of Vs profiles (a), shear
strain profiles (b), acceleration response spectra (c) and raw Fourier amplitude spectra (d). The mean (and mean +/- 1 standard
deviation) are plotted as thick black curve. The observed data at AQV is indicated by a red curve. Panel e shows the comparison
between deconvolved and recorded seismograms (the circle symbols on time-series indicate the PGA). The synthetic at AQV of Ameri
et al. 2012 is also plotted.

35