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

net/publication/234404071

Evidence of soil-building resonance, building frequency shift and (maybe) site-

city interaction during the 2002 Molise seismic sequence

Article · March 2003

CITATIONS READS

2 277

5 authors, including:

Maria Rosaria Gallipoli Angelo Masi

Italian National Research Council Università degli Studi della Basilicata
130 PUBLICATIONS   2,005 CITATIONS    145 PUBLICATIONS   2,016 CITATIONS   

SEE PROFILE SEE PROFILE

Marco Mucciarelli Felice Carlo Ponzo

National Institute of Oceanography and Applied Geophysics - OGS Università degli Studi della Basilicata
275 PUBLICATIONS   4,064 CITATIONS    140 PUBLICATIONS   1,208 CITATIONS   

SEE PROFILE SEE PROFILE

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

Progetto Reluis - Coordinamento Progetto speciale Osservatorio Sismico delle Strutture e Monitoraggio View project

PRESTo@CE3RN View project

All content following this page was uploaded by Marco Vona on 29 October 2017.

The user has requested enhancement of the downloaded file.

 Bulletin of the Seismological Society of America, Vol. 94, No. 5, pp. 1943–1953, October 2004

Analysis of RC Building Dynamic Response and Soil-Building Resonance

Based on Data Recorded during a Damaging Earthquake
(Molise, Italy, 2002)
by Marco Mucciarelli, Angelo Masi, Maria Rosaria Gallipoli, Paolo Harabaglia,
Marco Vona, Felice Ponzo, and Mauro Dolce

Abstract During the 2002 seismic sequence in Molise (Italy), the town of Bonefro
suffered moderate damage (IMCS ⳱ VII) except for two reinforced concrete (RC)
buildings. These buildings are located on soft sediments, close to each other and
very similar in design and construction. The main difference is the height: the most
damaged one (European Macroseismic Scale damage 4) has four stories, whereas
the less damaged (EMS damage 2) has three stories. The M 5.4 shock on 31 October
damaged both of them. The second shock on 1 November (M 5.3) increased the
damage on the four-story building substantially, just while a 5-min. seismic recording
was taken. We analyzed the recorded data by four different techniques: short-time
fourier transform (STFT), wavelet transform (WT), horizontal-to vertical spectral
ratio (HVSR), and horizontal-to-vertical moving window ratio (HVMWR). All the
results agree upon the estimate of the main building frequency before the second
shock and upon the shift of frequency due to damage. All the fundamental frequencies
(pre-, during, and postdamage) are in the range 2.5–1.25 Hz. The fundamental fre-
quency of the less damaged building was estimated at about 4 Hz.
To test if the soil–building resonance effect could have increased the damage, we
also evaluated the soil fundamental frequency by three different techniques: noise
HVSR, strong motion HVSR of seven aftershocks, and 1D modeling based on a
velocity profile derived from noise analysis of surface waves (NASW) measurements.
The results are again in good agreement, showing that resonance frequencies of the
soil and of the more damaged building are very close.

Introduction
During recent years, many studies paid attention to the
relation between damage and the amplitude, duration, and
frequency content of the motion at a site (Trifunac and To-
dorovska, 2000), the soil–building resonance (Lomnitz,
1999; Takewaki, 2001) and the dynamic soil–structure in-
teraction (Ganev et al., 1995; Trifunac and Todorovska,
2000; Wolf and Song, 2002), the evolution of dynamic char-
acteristics of buildings during and after damaging earth-
quakes (Trifunac et al., 2001a; Trifunac et al., 2001b; Gal-
lipoli et al., 2003a), and the energy transferred back to the
soil from a vibrating building (Wirgin and Bard, 1996; Gue-
guen et al., 2000; Gueguen et al., 2002; Mucciarelli et al.,
2003a; Tsogka and Wirgin, 2003).
The aim of this work is to analyze the contribution to
damage by soil–building resonance and the shift of building
fundamental frequency due to the reduction of structural
stiffness caused by damage, taking advantage of the peculiar
damage that occurred in the town of Bonefro (Italy) during
the Molise, 2002, seismic sequence.
On 31 October and 1 November 2002, two earthquakes,
M 5.4 and M 5.3, hit the area at the border between Molise
and Puglia in southern Italy. The distribution of the observed
intensities for the 31 October shock, M 5.4, is shown in
Fig. 1, according to QUEST (2002). Figure 1 also reports the
epicenter location and focal mechanism of the main shock,
following the data published by the National Institute of
Geophysics and Volcanology (INGV-Harvard, 2002). The
damage pattern qualified the quake as intensity VII on the
Mercalli-Cancani-Sieberg scale (MCS), with one notable ex-
ception: the village of San Giuliano di Puglia (IX MCS). The
quake there caused a tragedy deeply felt all over Italy: the
primary school collapsed, killing 27 pupils and 1 teacher.
Following the M 5.4 shock on 31 October 2002, our
group participated in many survey and study activities in the

1943
1944 M. Mucciarelli, A. Masi, M. R. Gallipoli, P. Harabaglia, M. Vona, F. Ponzo, and M. Dolce
Figure 1. CMT for the 31 October 2002 shock
(INGV-Harvard, 2002), with distribution of the ob-
served intensities following the QUEST (2002) field
survey.
epicenter area, namely, cooperation in the macroseismic sur-
vey (QUEST, 2002), a preliminary study of the causes of
damage enhancement in San Giuliano (Mucciarelli et al.,
2003b), horizontal-to-vertical spectral ratio (HVSR) micro-
tremor measurements in all the damaged municipalities
(Gallipoli et al., 2003b), and a systematic survey and anal-
ysis of the seismic vulnerability and damage characteristics
of the building stock of San Giuliano (Dolce et al., 2003).
Figure 2. View of buildings (east side): the most damaged is in the foreground.
During the field trip, a striking damage pattern appeared
in the town of Bonefro. A VII degree on the MCS intensity
scale was estimated for the whole town, with light damage
in the historical center. However, two reinforced concrete
buildings suffered heavy damage (Fig. 2). The two buildings
are located on soft sediments, close to each others (10 m)
and very similar in design and construction, with the excep-
tion of their height (four stories vs. three stories). The M 5.4
shock on 31 October 2002 damaged both buildings: slight
damage (European Macroseismic Scale [EMS] damage 1) in
the three story building, and substantial damage European
(EMS damage 3) in the four story building. The second shock
on 1 November (M 5.3) increased the damage level of the
two buildings to EMS damage 2 and damage 4, respectively.
To investigate a possible cause for the damage difference,
during the survey, it was decided to evaluate the frequency
of the two buildings by recording ambient noise. The mea-
surement was under way just when the M 5.3 shock took
place. As sometimes occurs in science, serendipity helps. By
pure chance, it was possible to get the first strong-motion
measurement inside an European building while it was suf-
fering heavy damage from an earthquake.
The present paper deals with the following analyses:
(1) investigation of the seismic recording on the top floor of
the most damaged building using four different techniques,
namely, short-time Fourier transform (STFT), wavelet trans-
form (WT), horizontal-to-vertical moving window ratio
(HVMWR), and HVSR; (2) an estimate of the fundamental
frequency of the less damaged building; (3) investigation of
the soil resonance frequency using three different tech-
niques: noise HVSR, strong motion HVSR of seven after-
shocks and 1D modeling based on a velocity profile derived
from Noise Analysis of Surface Waves (NASW) measure-
ments (Louie, 2001).
Analysis of RC Building Dynamic Response and Soil-Building Resonance of Data Recorded during a Damaging Earthquake
Building Characteristics and Damage Analysis
The Institute for Public Housing (IACP) of the Province
of Campobasso constructed the most damaged building at
the beginning of the 1980s. It has a reinforced concrete (RC)
framed structure with four stories and two (about 5 m long)
by six (about 3 m long) bays. The ground story is partially
infilled, having one-sided masonry infilled frames as trans-
verse exterior frames. Internal beams span only in the lon-
gitudinal direction, except for two transverse frames having
shallow beams like the internal longitudinal beams. In the
other direction, the stiffness of the resisting elements is sym-
metrically distributed with the stiff stair-structure placed in
a central position (Fig. 3). The interstory height is about 3 m.
At the ground story, the columns have 0.35 ⳯ 0.35 m di-
mensions. Infills are made of two layers of hollow brick
masonry having a total thickness of about 200 mm and
showing poor mechanical characteristics.
As already said, it is particularly interesting to compare
the seismic behavior of the building under examination to
the nearby three-story building (also constructed by IACP
in the 1980s with an almost identical design; Fig. 2). While
the former one suffered heavy structural and nonstructural
damage (damage level 4 according to the EMS98 scale), the
latter suffered only slight damage to nonstructural elements.
The damage to the four-story building was heavy and exten-
sive at ground level, whereas there was substantially no
damage on the upper stories. Structural and nonstructural
damage on the ground story also shows a nonuniform dis-
tribution in the symmetrical transverse direction. The col-
umns suffered the most important structural damage (weak
column–strong beam mechanism). In addition, masonry in-
fills show wide cracking and collapse at ground level. The
heavy damage suffered by the ground-floor columns (par-
ticularly at the top of the central column of the east side
exterior frame) is due to the combined action of cyclic shear
and, mainly, of cyclic flexure under axial compression. The
latter type of damage is greatly facilitated when hoops are
widely spaced and not suitably closed around longitudinal
bars, as in the structure under examination (the survey shows
Figure 3. Building plan (first floor).
1945
that hoops were bent at a 90⬚ angle), thus providing an un-
satisfactory confinement action.
The availability of pictures before and after the 1 No-
vember event allows us to follow-up the damage evolution
in the most damaged column (Fig. 4). According to the sup-
posed type of damage, crushing of the compression zone on
both faces of column with spalling of the concrete cover
occurred first (after the 31 October event). Subsequently,
crushing of the concrete core was accompanied by buckling
of longitudinal bars and by hoop failure (after the 1 Novem-
ber event). It is worth noting that, as a consequence of this
damage mechanism, there is not only a substantial stiffness
degradation, but columns can lose also their ability to carry
vertical loads, thus leading to a potential collapse hazard.
Further, this type of damage is very common, as shown by
the damage distribution observed in RC buildings after de-
structive earthquakes (Penelis and Kappos, 1997). An eval-
uation of the dynamic characteristics of the most damaged
building without accounting for the damage effects is re-
ported by the National Seismic Service (SSN) (2002). The
authors devised an elastic model using the SAP2000 code,
where the infills were also considered by modeling them as
equivalent struts. The panels either not well connected to the
frame structure or having large openings were not taken into
Figure 4. Damage evolution in the most damaged
column before and after 1 November event.
1946 M. Mucciarelli, A. Masi, M. R. Gallipoli, P. Harabaglia, M. Vona, F. Ponzo, and M. Dolce
account; therefore, only infills in the external transverse
frames were considered. Even though the designers consid-
ered horizontal forces, beams span in just one direction, the
longitudinal (east–west), as usual for building structures
without a proper seismic design. When beams are not pres-
ent, a strip of the tile lintel RC floor models the connection,
among columns. The results of the free vibration analysis
show that the first two mode shapes are prevailingly lateral
(modal mass [MM] equal to about 80%) having a frequency
equal to 1.9 and 2.2 Hz, the first in the longitudinal east–
west direction. The third mode is prevailingly torsional hav-
ing frequency equal to 5 Hz; the fourth and fifth modes are
again lateral (MM equal to about 10%) with frequencies
equal to 5.9 and 7.1 Hz, respectively.
Seismic Recordings Inside the Most
Damaged Building
For the first time in Europe, we recorded the nonline-
arity of building dynamic response during a real quake. The
main frequency of the taller building decreased because of
the reduction of structural stiffness due to damage. The
measurements were taken after the damage caused by the
31 October 2002 M 5.4 shock, and before, during, and after
the 1 November 2002 M 5.3 shock. The recording consists
of 1 min of ambient noise, 3 min of strong motion, and
finally 1 min of ambient noise. All the measurements were
taken on the stairs on the top floor of the taller building with
a 24 bit PRAX A/D converter attached to a Lennartz 3d-Lite
seismometer, 1 Hz, 400 V/m/sec gain with a sampling ratio
of 125 Hz. The north–south and east–west components cor-
respond to the transversal and longitudinal components, re-
spectively.
Data Analysis to Detect the Fundamental Building
Frequency and the Shift Due to Damage
To recognize the building behavior in static and dy-
namic conditions and the frequency variation due to damage,
we carried out different analyses of the time histories re-
corded on the taller building.
Short-Time Fourier Transform
To evaluate the effects of the nonlinear behavior of the
building under strong input motion, we processed the data
using the STFT (Ewins, 1994; Gabor, 1946; Spina et al.,
1996). STFT produces a signal representation in both the
frequency and the time domain to observe the variation in
time of the spectral properties of the recorded accelerations.
The method uses the Fourier transform for small sections of
signal using the “windowing” technique. To achieve well-
fit results, a large number of points are necessary to pick the
time-dependent frequency variations. At the same time, a
short temporal window is needed to get a good description
of the variations of the dynamic characteristics as a function
of time. The right compromise was found to be a 4.096-sec
window (512 points), which slides 1.0 sec ahead at each step.
Since the most important pieces of information are concen-
trated at the beginning of the portion of signal considered, a
temporal filter was used to damp the points after the first
0.75 sec, so that they quickly achieve negligible values. Fig-
ure 5a,b shows the north–south and east–west components
of the amplitude of the STFT, normalized to better evaluate
the trend of frequency. The amplitude trend of the north–
south normalized component clearly shows the main fre-
quency before the M 5.3 shock to be at about 2.5 Hz. During
strong motion, it decreases to about 1.3 Hz and stabilizes
after the shock at around 1.9 Hz. The amplitude trend of the
east–west normalized component is less clear, probably due
to the interference between two modes very close in fre-
quency. Once again, it is possible to read a frequency value
before the shock at about 3.2 Hz, a minimum during the
shock at about 1.8 Hz, and a final value after the shock at
about 2 Hz.
The strong reduction of the fundamental frequency dur-
ing the maximum seismic input is a consequence of the non-
linear behavior of frame members. High values of seismic
forces produce, in correspondence of beams and columns
ends, some phenomena, such as plastic mechanisms, crack
opening, and slip bonds that increase structural damping and
decelerate the oscillation of the building. When the seismic
excitation decreases, the nonlinear effects also decrease, and
thus the frequency increases again. The final values of fre-
quency are smaller than the initial ones because of the ac-
cumulation of damage. Generally, the difference in the shift
of the fundamental frequency observed for north–south and
east–west components can be ascribed to a different damage
level suffered by the building in the two directions. A pro-
nounced directional effect of the seismic shock and a dif-
ferent resistance of building in the two directions can pro-
duce different damage levels and collapse modality, as a
function of the input angle of the earthquake. In this case,
the STFT showed a similar damage level for both north–
south and east–west directions, contained in 48% and 43%
of this first frequency shift, respectively.
Wavelet Transform
The second analysis performed on the time histories was
WT. We used the WT provided by the Aoyama Gaukin
University–Vallen Wavelet software, developed by Vallen-
Systeme GmbH, Germany, and the Aoyama Gaukin Uni-
versity of Tokyo, Japan.
Figure 6a shows the WT of the north–south component:
before the seismic event, the main frequency is at about 1.5
Hz, during the M 5.3 shock; the frequency trend decreases;
and finally, the main frequency becomes 1.3 Hz. Another
frequency appears at 4 Hz probably due to a second mode
of building vibration. Figure 6b shows the WT of the east–
west component: it is possible to read the frequency before
the seismic event at about 2 Hz, and after the M 5.3 shock,
the main frequency is at about 1.3 Hz. The main difference
is actually at the beginning of the recordings. The signal-to-
Analysis of RC Building Dynamic Response and Soil-Building Resonance of Data Recorded during a Damaging Earthquake 1947

Figure 5. Normalized short-time Fourier transform of strong motion recording atop

the more damaged building for (a) north–south component, (b) east–west component.

noise ratio in this window is much worse than in the fol-
lowing.
Horizontal-to-Vertical Spectral Ratio
Recent work shows how HVSR is able to detect the fun-
damental mode of buildings (Irie and Nakamura, 2000; Cas-
tro et al., 1998, 2000; Mucciarelli and Monachesi 1999;
Mucciarelli et al., 1999; Chávez-Garcı́a and Cardenas-Soto,
2002; Volant et al., 2002; Gallipoli et al., 2003a). The theo-
retical justifications for the use of the HVSR with noise mea-
surements inside buildings are still unclear, and we believe
that more convincing data are needed. A possible explana-
tion of the observed good agreement between HVSR and
other system identification techniques might be the follow-
ing: excluding rocking modes, in the linear domain, without
significant soil–structure interaction and considering the
building as a 1D damped oscillator, the vertical component
of motion is transferred through the building considered as
a filter with very high rigidity. On the contrary, the horizon-
tal components undergo significant amplification. Thus, the
ratio between vertical and horizontal components has a
meaning similar to the HVSR for soils: we divide an ampli-
fied component by another one that we assume is mainly
unchanged by the propagation across the structure.
We estimated a separate HVSR for each 1 min of re-
cording, as an average of the two horizontal components.
The data were band-pass filtered between 0.1 and 25 Hz,
detrended, baseline corrected, and tapered. Then we per-
formed FFT on 50 log-equispaced bands per decade, apply-
ing the processing methodology proposed by Castro et al.
1948 M. Mucciarelli, A. Masi, M. R. Gallipoli, P. Harabaglia, M. Vona, F. Ponzo, and M. Dolce

Figure 6. Wavelet transform of strong motion recording atop the more damaged
building for (a) north–south component, (b) east–west component

(1990). The result is shown in Figure 7. The HVSR of the
first minute of ambient noise (before the shock) shows the
main peak at 2.45 Hz. The HVSR comprising the M 5.3 shock
(second minute) estimates the peak at 1.5 Hz. The HVSR of
the third minute contains a smaller quake and shows again
the fundamental frequency at 1.5 Hz. The HVSR of the fourth
minute, composed of the earthquake coda, estimates the fre-
quency at 1.92 Hz. Finally, the last HVSR of the fifth minute
(ambient noise again) yields a 1.92 Hz frequency. The HVSR
technique on separate pieces of recording identifies the main
vibrational mode of the building and the shift of frequency
due to damage with values close to those provided by other
techniques.
To complete the HVSR analysis on the two RC build-
ings, we recorded 5 min of ambient noise on the less dam-
aged three-story building (top floor). The noise recordings
were taken after the M 5.3 of the November 1 shock, thus
estimating the frequency after the second damaging event.
Figure 8 shows the HVSR functions of the two components:
the HVSR of the north–south component estimates peak of
the main frequency at 4 Hz, and the HVSR of the east–west
component shows two main frequencies at 2.5 and 5 Hz.
Analysis of RC Building Dynamic Response and Soil-Building Resonance of Data Recorded during a Damaging Earthquake
The three story building roughly respects the rule of thumb
estimate of frequency as a function of height: Period ⳱ (0.10
⬃ 0.05) ⳯ number of stories.
Horizontal-to-Vertical Moving Window Ratio
To better exploit the capability of the HVSR technique
in identifying the frequency shift, a new approach was de-
veloped: the HVMWR (horizontal-to-vertical moving win-
dow ratio). This technique is a combination of a spectrogram
(or evolutionary spectral density, as introduced by Liu 1970)
and HVSR. The spectral ratio is represented in the time–
frequency space by overlapping moving windows. The
frequency–domain transform is a fast Fourier one. It is cal-
culated on a subset of 512 samples with an increment of 12
samples (500 samples overlap) for 100 equispaced fre-
quency values in the range of 0.1–10 Hz.
Figure 9a,b shows the results for both components.
Clockwise from bottom left, each figure represents the rele-
vant horizontal component of acceleration, the HVMWR, the
average HVMWR (that is the sum over time divided into
the number of overlapping windows), and an enlargement
of the former at selected frequencies.
Figure 9a shows the results for the north–south com-
ponent. The HVMWR shows that the main frequency before
1949
Figure 7. HVSR analysis for five different
time windows of strong-motion recording atop
the more damaged building for the average of
the north–south and east–west components.
The first minute includes ambient noise before
the quake, minutes 2 and 3 comprise the strong
motion part, minute 4 is the coda of the earth-
quake, and minute 5 is again ambient noise.
Figure 8. HVSR analysis for ambient noise
recordings atop the less damaged building.
the shock is about 2.5 Hz. The seismic event energizes
higher modes, and the fundamental one decreases at around
1.7 Hz. After the strong motion, the main frequency is at 1.9
Hz. The enlargement of the average is centered on the fun-
damental frequency: the two peaks at 2.5 Hz and 1.9 Hz are
evident. The behavior is similar for the ratio between east–
west and vertical components (Fig. 9b): the HVMWR shows
that the main frequency before the shock is about 3 Hz.
During the seismic event, higher frequencies are energized,
and the fundamental one decreases below 2 Hz. After the
strong motion, the main frequency is slightly above 2.0 Hz.
The enlargement of the average shows the frequency range
from 4 to 9 Hz: besides the higher modes from 7.5 to 8.5
Hz, it is interesting to note a peak just above 4 Hz.
In Table 1, the frequencies estimated by the four tech-
niques for each component before (Fbeg), during (Fmin), and
after (Fend) the quake are reported. The results estimated by
STFT, HVSR, and HVMWR are in good agreement. On the
contrary, the frequencies evaluated by WT are slightly un-
derestimated. The agreement among all the techniques ap-
pears more clearly by considering the percentage frequency
shift, computed as follows:
DFmax ⳱ (Fmin ⳮ Fbeg)/Fbeg
1950 M. Mucciarelli, A. Masi, M. R. Gallipoli, P. Harabaglia, M. Vona, F. Ponzo, and M. Dolce

Figure 9. HVMWR analysis of the strong-motion recording atop the more damaged
building for (a) north–south component, (b) east–west component (see text for details).

DFmax is similar for both components and equal to about
35%. Similar values of the frequency decrease consequent
to heavy damage were also evaluated by other authors during
postearthquake survey and experimental investigations (e.g.,
respectively, Trifunac et al., 2001a; Dolce et al., 2001).
It is worth noting that the frequency values found by all
techniques (with the exception of the WT technique) for the
building after damage are higher than those ones computed
in SSN (2003) on the building model before damage. More
refined models are under scrutiny (see conclusions), and they
will probably be able to take into account more structural
details and thus provide a better frequency estimate for the
undamaged building. Furthermore, the longitudinal compo-
nent (east–west) of the damaged building shows a funda-
Analysis of RC Building Dynamic Response and Soil-Building Resonance of Data Recorded during a Damaging Earthquake 1951

mental mode of vibration at a frequency higher than the
transversal (north–south) one, whereas the contrary happens
in the (undamaged) building model. Again, a more accurate
modeling of the building before and after damage is strongly
needed.
Data Analysis to Detect the Resonance between
Soil and Buildings
In the analysis of earthquake damage, two important
aspects should be considered: (1) if the earthquake carries
significant energy near the period of the structure and (2) if
the foundation’s soft soil has its predominant period near the
structure’s period. The presence of soft soil (clays) and such
peculiar damage has led as to suspect a possible soil–struc-
ture resonance. The stratigraphy of the soil involves a layer
of clays overlying a very thick marl formation. The depth of
the interface has been estimated at about 30 m.
We investigated the fundamental frequency of the soil
with three different techniques: noise HVSR, strong motion
HVSR of seven aftershocks, and 1D modeling based on a
velocity profile derived from NASW measurements. We per-
formed the noise HVSR near the most damaged building and
in the historical center using the same instrumentation and
processing technique described before. The resulting HVSR

Table 1
Estimate of Frequency Variation of the Most Damaged Building, Obtained with Different Techniques
N–S Frequency (Hz) E–W Frequency (Hz)
FBEG FEND FMIN (DF)MAX FBEG FEND FMIN (DF)MAX

STFT 2.5 1.9 1.3 ⳮ48% 3.2 2 1.8 ⳮ43%

WT 1.5 1.3 1.0 ⳮ33% 2 1.3 1.0 ⳮ50%
HVMWR 2.5 2.1 1.8 ⳮ28% 2.7 2 1.5 ⳮ44%
HVSR 2.45 1.9 1.5 ⳮ38% 2.8 1.9 1.8 ⳮ35%

Figure 10. (a) Shear wave velocity profile estimated with NASW technique. (b) Soil
amplification in Bonefro estimated with different techniques: free-field noise HVSR,
1D modeling, and HVSR of seven accelerometric recordings of aftershocks.
1952 M. Mucciarelli, A. Masi, M. R. Gallipoli, P. Harabaglia, M. Vona, F. Ponzo, and M. Dolce
shows a main resonance peak at 2.5 Hz, and the noise HVSR
performed in the historical center of Bonefro (on a limestone
outcrop) shows a flat amplification function in the range of
the frequency of interest for buildings (Fig. 10b). Then we
installed a Kinemetrics S2 accelerometer, recording seven
aftershocks in the magnitude range 4.2–3.2, with PGA in the
0.1–0.005 g range. The HVSR analysis of these records es-
timates the resonance peak at 1.8 Hz. These estimates have
been compared with that obtained from the 1D model (Fig.
10b). The Vs profile (Fig. 10a) was obtained with a 100-m-
long array of vertical geophones. The array recorded 10 min
of ambient noise, then we optimized the inversion from the
surface wave dispersion curve in the slowness-frequency
transform domain (Louie, 2001). Finally, we modeled the
soil response using the computer code EERA (Bardet et al.,
2000). The model evaluated the predominant frequency at
2.4 Hz, in good agreement with the estimates obtained with
noise and strong motion HVSR (possibly lower for nonlin-
earity of soil behavior) (Fig. 10b). The different measure-
ments and estimates lead to the conclusion that the resonance
soil frequency is in the same range as the fundamental mode
of the most damaged building before the damage, thus jus-
tifying important resonance effects between soil and struc-
ture that occurred during the earthquake. The least damaged
building, on the contrary, has a fundamental frequency that
is almost the double that of the soil.
Conclusions
Due to a fortuitous event, we recorded strong motion
inside an RC building while it was suffering damage from
an M 5.3 event. It was possible, with four different tech-
niques, to appreciate the lowering of the fundamental mode
and its subsequent migration to a value lower than that re-
corded before damage. This observation lends support to
estimates derived from numerical and analogical models.
The frequency shift did not move the fundamental mode far
enough from the soil resonance frequency, thus giving rise
to a soil–building interaction. A nearby similar building, but
with a different fundamental frequency, showed less dam-
age. The next step will be the preparation of a more detailed
structural model. For this purpose, the Italian Seismic Sur-
vey has organized an open test, providing detailed structural
information to the engineers willing to test their model on
this case history (BOB-Code Project http://adic.deseg.unige.
it/anidis2004/Bob-code_invito_a_partecipare.pdf). While
this first report is being prepared, about 10 groups from Italy,
France, Portugal, and the United States have already agreed
to participate. In the near future, it will be possible to know
if nonlinear dynamic 3D analyses will explain the results
described herein.
Acknowledgments
We thank Riccardo Caputo of the University of Basilicata, who defied
common sense and security standards and rescued the instrumentation atop
the crumbling building. Many thanks are due to Dario Albarello and Matteo
Picozzi of the University of Siena for the NASW velocity profile inversion.
Thanks to F. J. Chávez-Garcı́a and two anonymous reviewers, whose com-
ments helped improve the manuscript.
References
Bardet, J. P., K. Ichii, and C. H. Lin (2000). EERA: a computer program
for equivalent-linear earthquake site response analyses of layered soil
deposits, Ph.D. Thesis, University of Southern California, Department
of Civil Engineering.
Castro, R. R, J. G. Anderson, and S. K. Singh (1990). Site response, atten-
uation and source spectra of S waves along the Guerrero, Mexico,
subduction zone, Bull. Seism. Soc. Am. 80, 1481–1503.
Castro, R. R., M. Mucciarelli, F. Pacor, P. Federici, and A. Zaninetti (1998).
Determination of the characteristic frequency of two dams located in
the region of Calabria, Italy, Bull. Seism. Soc. Am. 88, no. 2, 503–
511.
Castro, R. R., E. Ruiz, A. Uribe, and C. J. Rebollar (2000). Site response
of the dam El Infiernillo, Guerrero-Michoacan, Mexico, Bull. Seism.
Soc. Am. 90, no. 6, 1446–1453.
Chávez-Garcı́a, F. J., and M. Cardenas-Soto (2002). The contribution of
the built environment to the ‘free-field’ ground motion in Mexico
City, Soil Dyn. Earthquake Eng. 22, 773–780.
Dolce, M., A. Masi, G. Zuccaro, F. Cacace, L. Samela, G. Santarsiero, and
M. Vona (2004). Survey and analysis of damage in San Giuliano di
Puglia, GNDT Technical Report (in Italian).
Dolce, M., D. Cardone, and F. Ponzo (2001). Retrofitting of RC framed
structures through SMA-based energy dissipating and re-centering
braces, in Proc. of the 7th International Seminar on Seismic Isolation,
Passive Energy Dissipation and Active Control of Vibration of Struc-
tures, Assisi, Italy.
Ewins, D. J. (1994). Modal Testing: Theory and Practice, Research Studies
Press, 219 pp.
Gabor, D. (1946). Theory of communication, IEE J. 93, 429–457.
Gallipoli, M. R., M. Mucciarelli, R. R. Castro, G. Monachesi, and P. Contri
(2003a). Structure, soil-structure response and effects of damage
based on observations of horizontal-to-vertical spectral ratio of mi-
crotremors, Soil Dyn. Earthquake Eng. 24, 487–495.
Gallipoli, M. R., M. Mucciarelli, S. Gallicchio, M. Tropeano, and C. Lizza
(2003b). HVSR measurements in the area damaged by the 2002 Mo-
lise, Italy, earthquake Earthquake Spectra, 20, 581–593.
Ganev, T., F. Yamazaki, and T. Katayama (1995). Observation and nu-
merical analysis of soil–structure interaction of a reinforced concrete
tower, Earthquake Eng. Structural Dyn. 24, no. 4, 491–503.
Guéguen, P., P-Y. Bard, and C. S. Oliveira (2000). Experimental and nu-
merical analysis of soil motion caused by free vibration of a building
model, Bull. Seism. Soc. Am. 90, 1464–1479.
Guéguen, P., P-Y. Bard, and F. J. Chávez-Garcı́a (2002). Site-city seismic
interaction in Mexico City-like environments: an analytical study,
Bull. Seism. Soc. Am. 92, no. 2, 794–811.
INGV-Harvard (2002). European-Mediterranean Regional Centroid-
Moment Tensors Project, www.ingv.it/seismoglo/RCMT (last ac-
cessed December 2002).
Irie, Y., and K. Nakamura (2000). Dynamic characteristics of a r/c building
of five stories based on microtremor measurements and earthquake
observations, in Proc. of 12th World Conf. Earthquake Engineering,
Wellington, (New Zealand), 500–508.
Liu, S. C. (1970). Evolutionary power spectral density of strong-motion
earthquakes, Bull. Seism. Soc. Am. 60, 891–900.
Lomnitz, C. (1999). The end of earthquake hazard, Seism. Res. Lett. 70,
387–388.
Louie, J. N. (2001). Faster, better: shear wave velocity to 100 m depth from
refraction microtremor arrays, Bull. Seism. Soc. Am. 91, no. 2, 347–
364.
Mucciarelli, M., and G. Monachesi (1999). The Bovec (Slovenia) earth-
 Analysis of RC Building Dynamic Response and Soil-Building Resonance of Data Recorded during a Damaging Earthquake
quake, April 1998: a preliminary correlation among damage, ground
motion amplification and building frequencies, J. Earthquake Eng. 3,
317–327.
Mucciarelli, M., G. Monachesi, and M. R. Gallipoli (1999). In situ mea-
surements of site effects and building dynamic behaviour related to
damage observed during the 9/9/1998 earthquake in Southern Italy,
in Proc. of ERES99 Conf., Catania, Italy, 253–265.
Mucciarelli, M., M. R. Gallipoli, F. Ponzo, and M. Dolce (2003a). Seismic
waves generated by oscillating buildings: analysis of a release test,
Soil Dyn. Earthquake Eng., 23, 255–262.
Mucciarelli, M., A. Masi, M. Vona, M. R. Gallipoli, P. Harabaglia, R.
Caputo, S. Piscitelli, E. Rizzo, M. Picozzi, D. Albarello, and C. Lizza
(2003b). Quick survey of the possible causes of damage enhancement
observed in San Giuliano after the 2002 Molise, Italy, seismic se-
quence, J. Earthquake Eng. 7, no. 4, 599–614.
National Seismic Service (SSN) (2003). Sintesi del danno e delle tipologie
edilizie prevalenti, http://ssn.protezionecivile.it (last accessed June
2004).
Penelis, G. G., and A. J. Kappos (1997). Earthquake Resistant Concrete
Structures, E & F Spon, London.
QUEST (Quick Earthquake Survey Team) (2002). Rilievo degli effetti pro-
dotti dalla sequenza sismica molisana Ottbre–Novembre 2002,
http://www.ingv.it/quest (last accessed June 2004).
Spina, D., C. Valente, and G. R. Tomlinson (1996). A new procedure for
detecting nonlinearity from transient data using the Gabor transform,
Nonlinear Dyn. 11, 235–254.
Takewaki, I. (2001). Resonance and criticality measure of ground motions
via probabilistic critical excitation method, Soil Dyn. Earthquake Eng.
21, 645–659.
Tsogka, C., and A. Wirgin (2003). Simulation of seismic response in an
idealized city, Soil Dyn. Earthquake Eng. 23, 391–402.
Trifunac, M. D., and M. I. Todorovska (2000). Long period microtremors,
microseisms and earthquake damage: Northridge, CA, earthquake of
17 January 1994, Soil Dyn. Earthquake Eng. 19, 253–267.
View publication stats
1953
Trifunac, M. D., S. S. Ivanovic, and M. I. Todorovska (2001a). Apparent
period of a building. I. Fourier analysis, J. Struct. Eng. 27, no. 5, 517–
526.
Trifunac, M. D., S. S. Ivanovic, and M. I. Todorovska (2001b). Apparent
period of a building. II. time-frequency analysis, J. Struct. Eng. 27,
no. 5, 527–537.
Volant, P. H., N. Orbovic, and F. Dunand (2002). Seismic evaluation of
existing nuclear facility using ambient vibration test to characterise
dynamic behaviour of the structure and microtremor measurements
to characterise the soil: a case study, Soil Dyn. Earthquake Eng. 22,
1159–1167.
Wirgin, A., and P-Y. Bard (1996). Effects of buildings on the duration and
amplitude of ground motion in Mexico City, Bull. Seism. Soc. Am.
86, 914–920.
Wolf, J. P., and C. Song (2002). Some cornerstones of dynamic soil–
structure interaction, Eng. Struct. 24, 13–28.
Dept. of Structures
Soil Dynamics and Engineering Geology (DiSGG)
University of Basilicata
Campus Macchia Romana
Potenza, Italy
(M.M., A.M., M.R.G., P.H., M.V., F.P., M.D.)
Institute for Environmental Analysis Methodologies—CNR
C. da S. Loja
Tito Scalo (PZ) Italy
(M.R.G.)
I.N.G.V.
Rome, Italy
(M.M.)
Manuscript received 4 September 2003.