Geophysical Journal International

Geophys. J. Int. (2010) 180, 1242–1252 doi: 10.1111/j.1365-246X.2009.04476.x

Surface wave dispersion measurements from ambient seismic noise

analysis in Italy

Hongyi Li,1,2 Fabrizio Bernardi3 and Alberto Michelini3

1 Key Laboratory of Geo-detection (China University of Geosciences, Beijing), Ministry of Education, China. E-mail: lih@cugb.edu.cn
2 School of Geophysics and Information Technology, China University of Geosciences, Beijing 100083, China
3 Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata, 605, Rome 00143, Italy

Accepted 2009 December 7. Received 2009 September 22; in original form 2008 October 25

Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022

SUMMARY
We present the surface wave dispersion results of the application of the ambient noise method
to broad-band data recorded at 114 stations from the Istituto Nazionale di Geofisica e Vul-
canologia (INGV) national broad-band network, some stations of the Mediterranean Very
Broadband Seismographic Network (MedNet) and of the Austrian Central Institute for Me-
teorology and Geodynamics (ZAMG). Vertical-component ambient noise data from 2005
October to 2007 March have been cross-correlated for station-pairs to estimate fundamental
mode Rayleigh wave Green’s functions. Cross-correlations are calculated in 1-hr segments,
stacked over periods varying between 3 months and 1.5 yr. Rayleigh wave group dispersion
curves at periods from 8 to 44 s were determined using the multiple-filter analysis technique.
The study region was divided into a 0.2◦ × 0.2◦ grid to invert for group velocity distribu-
tions. Checkerboard tests were first carried out, and the lateral resolution was estimated to be
about 0.6◦ . The resulting group velocity maps from 8 to 36 s show the significant difference
of the crustal structure and good correlations with known geological and tectonic features
in the study region. The Po Plain and the Southern Alps evidence lower group veloci-
ties due to soft alluvial deposits, and thick terrigenous sediments. Our results also clearly
showed that the Tyrrhenian Sea is characterized with much higher velocities below 8 km than
the Italian peninsula and the Adriatic Sea which indicates a thin oceanic crust beneath the
GJI Seismology

Tyrrhenian Sea.
Key words: Tomography; Surface waves and free oscillations; Crustal structure.

the inhomogeneous distribution of earthquakes; (2) strong depen-

1 I N T RO D U C T I O N
Seismic studies of the velocity structure of the Earth have been
traditionally based on observations of seismic waves emitted by
earthquakes or explosions. Based on those ballistic waves, many
results about the Earth’s structure have been obtained by measuring
the traveltimes of body waves, dispersion curves of the surface
waves and waveform modelling. In Italy and surrounding areas,
many traditional surface wave studies based on earthquake data
have been made (e.g. Panza et al. 1980; Mantovani et al. 1985;
Pontevivo & Panza 2002; Chimera et al. 2003; de Lorenzo et al.
2003; Marone et al. 2004; Raykova et al. 2006).
Surface waves at different periods are sensitive to the Earth struc-
ture at different depths. In general, short periods tend to sample the
slow layers closer to the surface, whereas the long periods are sen-
sitive to faster velocities found deeper in the Earth. Thus, surface
wave dispersion measurements have been widely used to study the
structure of the Earth’s crust and upper mantle.
Measurements made from earthquake-based surface waves, how-
ever, have several limitations: (1) inadequate path coverage due to
dency on the determination of origin time and source parameters
and (3) the difficulties in making short-to-intermediate period dis-
persion measurements from teleseismic events since high frequency
information is usually lost due to intrinsic attenuation and scattering
along ray paths.
Recently, it has been shown that a random wavefield generated
by scatters in the near-surface structure of the Earth, includes infor-
mation on the Green’s function of a wave travelling along the path
between two receivers. This information can be extracted by com-
puting the cross-correlation between records at two receivers over
sufficiently long times (Weaver & Lobkis 2001a,b, 2004; Derode
et al. 2003; Snieder 2004; Wapenaar 2004; Larose et al. 2005).
In seismology, two types of random wavefields have been used to
retrieve Green’s function by cross-correlating coda waves and suf-
ficiently long ambient noise recordings. Seismic coda waves are
produced by multiple scattering of seismic waves from small-scale
heterogeneities in the lithosphere (Sato & Fehler 1998). By stack-
ing cross-correlation functions of coda waves in Mexico, Campillo
and Paul (2003) extracted the Rayleigh and Love waves Green’s

1242 C 2010 The Authors

Journal compilation 
C 2010 RAS

Figure 1. Distribution of seismic stations in Italy and adjacent regions used in this study, denoted by triangles. The thick straight lines are the section lines of
the vertical profiles (AA , BB and CC ) shown in Fig. 10.
functions. Ambient seismic noise is excited by randomly distributed
natural and artificial sources such as sea waves, atmospheric pertur-
bations, traffic and human activities. Shapiro and Campillo (2004)
and Sabra et al. (2005a) have confirmed that surface wave Green’s
functions can be extracted by cross-correlating long time-series of
seismic ambient noise. Extensive work has been performed using
the ambient noise method in the United States and the Pacific North-
west (Sabra et al. 2005b; Shapiro et al. 2005; Bensen et al. 2008;
Liang & Langston 2008), in South Korea (Kang & Shin 2002; Cho
et al. 2007), in China (Yao et al. 2006, 2008; Li et al. 2009), in
New Zealand(Lin et al. 2007) and in Europe (Yang et al. 2007;
Stehly et al. 2009). In general, ambient noise method provides sur-
face wave dispersion measurements at shorter periods with higher
resolution than traditional methods under the conditions of greatly
densifying seismic networks and uniform noise field distribution.
The Italian peninsula features a complex tectonic setting which
results in strong lateral heterogeneities of the seismic velocity struc-
ture (Panza et al. 1980; Mantovani et al. 1985; Pontevivo & Panza
2002; Li et al. 2007). In recent years, high quality, continuous
broad-band recordings from the recently installed INGV national

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
Ambient seismic noise analysis in Italy 1243
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
broad-band network, the MedNet network and the ZAMG network
(Fig. 1) have become available. These networks provide an unprece-
dented dense station coverage to perform seismological analysis in
Italy.
The main purpose of this study is to apply the ambient seismic
noise technique to Italy to obtain high-resolution Rayleigh wave,
fundamental mode, dispersion curves. Vertical-component time-
series recorded at 114 broad-band stations in Italy and adjacent
regions between 2005 October and 2007 March are cross-correlated
to yield estimated Rayleigh wave Green’s functions. The resulting
group velocity maps from 8 to 36 s can provide important indications
on the crust and the upper mantle of Italy and surrounding regions.
2 D ATA P R O C E S S I N G A N D G R O U P
VELOCITY MEASUREMENTS
We collected continuous broad-band vertical-component seismic
data recorded by 114 stations from the INGV national network
(Amato & Mele 2008), MedNet and ZAMG from 2005 October
1244 H. Li, F. Bernardi and A. Michelini
48˚
stack stack
ARSA

TUE
200701
200701
46˚

DOI 200609

44˚ 200607 200607

GROG CING

ARSA_TUE

DOI_SACS
SACS

42˚
CIGN 200601
200601
AMUR

40˚

200507

200507

38˚
ESLN

Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022

200501

36˚

-300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300

6˚ 8˚ 10˚ 12˚ 14˚ 16˚ 18˚ 20˚ LAG (s) LAG (s)

stack stack stack

200701 200701 200701

CING_GROG

CING_AMUR

ESLN_CIGN
200607 200607 200607

200601 200601 200601

-300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300
LAG (s) LAG (s) LAG (s)

Figure 2. Monthly cross-correlations computed for five station-pairs (shown on the upper left-hand panel) for frequency band 0.05-0.2 Hz: ARSA-TUE,
DOI-SACS, CING-GROG, CING-AMUR and ESLN-CIGN. The stacks over all the available months are displayed at the top of each panel.

to 2007 March. In a manner similar to the standard, noise cross- 0.05–0.2 Hz. For the station-pair ARSA-TUE, the amplitude of the
correlation processing procedure (Shapiro and Campillo 2004; positive correlation lag representing waves travelling from ARSA
Sabra et al. 2005a; Bensen et al. 2007) the data were (i) windowed to TUE is much smaller than the negative lag. One-sided cross-
in one-hour length time series, (ii) removed trend and mean value, correlations are also clearly observed for other four station pairs
(iii) high pass filtered at 0.01 Hz, (iv) decimated to 1 sample per in Fig. 2, the correlation lags showing large amplitudes represent
second while preventing aliasing and (v) whitened. waves propagating from DOI to SACS, GROG to CING, CING
The processed 1-hr-long traces were cross-correlated between all to AMUR and CIGN to ESLN, respectively. Previous studies have
station-pairs and then stacked all the available cross-correlations shown that the noise recorded in Europe originates mainly from
for each station-pair into a single time-series. Attention has been the northern Atlantic Ocean, and the Mediterranean Sea is also an
made to maintain a minimum distance of three wavelengths for important noise source region (Friedrich et al. 1998; Marzorati &
each station-pair when calculating the cross-correlation. The num- Bindi 2006; Pedersen et al. 2007; Marzorati & Bindi 2008), which
ber of cross-correlations in the stack varies between 3 months and result in one-sided asymmetric cross-correlations. We also noticed
1.5 yr depending on station availability. Theoretically, the resulting that the cross-correlations computed in winter show a higher signal-
cross-correlations should show the same velocity and dispersion to-noise ratio (SNR) than those in summer. Although the cross-
features on both positive and negative lags since waves travelling in correlations are asymmetric and display seasonal variations due to
opposite directions along the path between a station-pair sample the non-uniformal noise distribution, the cross-correlations generally
same structure. However, due to not perfectly isotropic noise field show coherent waveforms and contemporaneous arrivals of surface
distribution, asymmetric cross-correlations are often observed. waves from month to month.
In this study, we first investigated the dependence of the cross- Fig. 3 gives monthly cross-correlations over the frequency band
correlations on the noise source distribution. Fig. 2 shows monthly 0.02–0.05 Hz. Compared with the Fig. 2, on most of station-pairs
cross-correlations for five station-pairs over the frequency band except for CING-AMUR, we observed clear signals simultaneously


C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
 Ambient seismic noise analysis in Italy 1245

stack stack

200701
200701

200609
200607 200607

ARSA_TUE

DOI_SACS
200601
200601

200507

200507

Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022

200501

-300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300
LAG (s) LAG (s)

stack stack stack

200701 200701 200701

CING_GROG

CING_AMUR

ESLN_CIGN
200607 200607 200607

200601 200601 200601

-300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300 -300 -200 -100 0 100 200 300
LAG (s) LAG (s) LAG (s)

Figure 3. Monthly cross-correlations computed for five station-pairs for frequency band 0.02-0.05 Hz: ARSA-TUE, DOI-SACS, CING-GROG, CING-AMUR
and ESLN-CIGN. The stacks over all the available months are displayed at the top of each panel.

arriving on both positive and negative correlation lags which indi- resolution depending on interstation distance (Levshin et al. 1989).
cated that the directionality of the noise source is weak and the We tested the α values for various distance ranges using synthetic
noise field is close to diffusive in this frequency, similar noise be- seismograms. We found α = 6.25, 12.5 and 25 for distances between
haviours have been also reported by Pedersen et al. (2007) in the 0–100, 100–250 and 250–1000 km, respectively.
Baltic shield. Fig. 5 shows group velocity dispersion curves measured from
In this study, in order to simplify data analysis and enhance the six station-pairs along paths through distinct geological regions. As
SNR, we averaged the positive and negative correlation lags to create seen in Fig. 5, dispersion curves measured from station-pairs TRI-
a symmetric cross-correlation, and more than 3000 station-pairs BOB generally showed lower values than those from DOI-SACS,
have been discarded by selecting only cross-correlations with SNR MCEL-MAON and SERS-OSKI.
larger than six to ensure correctly identified arrival-time of surface In general, Rayleigh waves sample down to depths approximately
waves. Fig. 4 gives an example of a cross-correlation record section one-third their wavelength. Low group velocities at short periods
with respect to station CING. This figure shows that a variety of (<15 s) usually are associated with sedimentary layers in the shal-
azimuths produce clear signals with physically reasonable moveouts low part of the crust. Note that the path TRI-BOB is through the
(∼2.7 km s−1 ). Po Plain, a region considered to have very thick terrigenous sedi-
The group velocity dispersion measurements were made using the ments (Scarascia & Cassinis 1997; Waldhauser et al. 1998, 2002;
multiple-filter technique (Dziewonski et al. 1969; Herrmann 1973). Kummerow et al. 2004; Tesauro et al. 2008). At longer periods,
The waveforms were narrow bandpass filtered with the operator exp Rayleigh waves sample predominantly the upper mantle. The dis-
[− α(ω − ω0 )2 /ω20 ], where ω0 is the center frequency. Since there persion curve between stations SERS and OSKI shows well the fast
is a trade-off between resolution in the time and frequency domains group velocities occurring at periods between 15 and 25 s when
with such filtering, the tunable parameter α is used to best balance compared to the other paths.


C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
1246 H. Li, F. Bernardi and A. Michelini
48˚
46˚
44˚
CING
42˚
40˚
38˚
36˚
8˚ 10˚ 12˚ 14˚ 16˚ 18˚
Figure 4. A cross-correlation record section centred at the station CING which is bandpass filtered from 8 to 50 s. Traveltime–distance plot of cross-correlated
traces are shown on the right-hand panel with the corresponding paths delineated by black lines on the left-hand panel. The grey line indicates the approximate
arrival times for Rayleigh waves in this band.
3 I N V E R S I O N F O R G RO U P
VELOCITY MAPS
By applying the data selection criteria mentioned in previous sec-
tion, the Rayleigh wave dispersion measurements from more than
2800 paths were used to invert for group velocity maps at periods
from 8 to 36 s. The study region was divided into a 0.2◦ × 0.2◦ grid,
and velocities between grid nodes were calculated with bilinear in-
terpolation. The Occam’s inversion scheme was adopted to invert
for group velocity distributions. The Occam’s inversion method pro-
posed by Constable et al. (1987) and deGroot-Hedlin & Constable
(1990) seeks a smooth model that satisfies the observations while
allowing for a fine grid for model discretization. Here, we just give
a brief description of the method. The velocity model is represented
as m with N unknowns and the observation data of M paths as d.
The observational errors are stored in a diagonal matrix w, and
the smoothness of the model is denoted by roughness R defined
as
R = Rx + R y = ||δx m||2 + ||δ y m||2 ,
where δ x m and δ y m are vectors composed of first-order spatial par-
tial derivatives in x and y directions, respectively. In our inversion,
the objective is to minimize the following function:
||wd − w F(m)||2 + μ(||δx m||2 + ||δ y m||2 ).
The first term describes the fitting to observational data, the second
term represents the smoothness constraint, and μ is a smoothing
factor. The selection of the smooth parameter μ is, however, some-
what subjective. As the value of μ is increased, the resulting model
becomes smoother because a larger amount of model conditioning
is applied. We performed a series of tests with different μ to deter-
mine its optimal value by considering the trade-off curve between
data misfit and model roughness as shown in Fig. 6.
800
600
Distance (km)
400
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
200
0
0 100 200 300 400
20˚ t (s)
Optimal results were obtained when the decrease of residual
started to slow down at decreasing μ. Thus the final model is the
result of a compromise between achieving more details and main-
taining a reasonable smoothness for the velocity model. From Fig. 6,
we found that the misfit began to decrease gently as μ is smaller
than 2, and the decrease of roughness became to slow down when μ
is larger than 0.5. The value of μ between 0.5 and 2 could provide
reasonable model resolution, and we selected μ = 1 in the present
inversion. A more detailed discussion of this method is given by
Constable et al. (1987) and deGroot-Hedlin & Constable (1990),
and a recent application can be found in Huang et al. (2003) and Li
et al. (2009).
The resolution of surface wave tomography depends primarily
on path coverage and the azimuthal distribution of paths. Fig. 7
shows the number of dispersion measurements at different periods
and the path coverage at 15 and 30 s. The path coverage is best for
periods 10–26 s with over 2000 measurements. Coverages at shorter
or longer periods become sparse, but the spatial pattern of coverage
remains similar. As it can be seen in Fig. 7, the overall coverage
(1)
pattern is reasonably good, however, the path distribution is not
even in the study region. Owing to the few available stations on the
eastern side of the Adriatic Sea, the azimuthal distribution of paths
is generally poor east of peninsular Italy. In general, path density
is highest within the Italian peninsula and it decreases gradually
(2)
towards the perimeter of the study area.
Checkerboard tests were performed to estimate resolution.
Fig. 8(a) shows a theoretical velocity model that is discretized into
a 0.2◦ × 0.2◦ grid as the real model used in our inversion. The size
of alternatively high- and low-velocity cells is 0.6◦ . Each cell has
a constant velocity 5 per cent above or below an average velocity
of 2.7 km s−1 . Synthetic data of group velocities were calculated
according to the actual paths at 15 and 30 s period. Finally, the
data were used to reconstruct the velocity model. The resulting

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS

Figure 5. Top panel: Interstation paths for group velocity measurements:
SERS-OSKI, DOI-SACS, MCEL-MAON, CAFE-SBPO, AMUR-MABI
and TRI-BOB. Bottom panel: group velocity curves measured from the
paths shown on the top.
velocity models are shown in Figs 8(b) and (c) where we observe
that the resolving power is generally good in the Italian territory
and Tyrrhenian Sea but it deteriorates towards the periphery. Path
coverages at 30 s period become worse, but the spatial coverage
patterns generally remain similar.
Fig. 9 presents the results of group velocity tomography at period
8, 15, 24 and 30 s. At 8 s period, the lowermost wave velocities
are found beneath the Po Plain. At 15 s period, the Po Plain, the
northeastern side of the Apennines and Sicily are characterized by
low wave velocities. Conversely, the fastest velocities occur beneath
the Tyrrhenian Sea. At 24 and 30 s period, the distribution of low
velocities beneath the Po Plain widens to the Central and Eastern
Alps and in the Apennines. Also, much higher velocities occur

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
Ambient seismic noise analysis in Italy 1247
5.0
μ=500
μ=100
4.5 μ=50
μ=20
μ=10
misfit
μ=5
4.0 μ=2
μ=1
μ=0.5
μ=0.2
μ=0.1
μ=0.05
3.5
μ=0.01
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
0.0 0.2 0.4 0.6 0.8 1.0
roughness
Figure 6. The rms traveltime misfit and model roughness presented versus
different smoothing factor μ.
beneath the Tyrrhenian Sea whereas in Sicily the pattern of high
and low velocities reverses with higher velocities resolved.
4 DISCUSSION
4.1 Interpretation of the group velocity maps
Surface waves at different periods are sensitive to the Earth struc-
ture over different depth ranges. It is generally true that surface
waves of short period tend to sample materials close to the sur-
face, and long period surface waves are more affected by deep
structures in the Earth, although short period surface waves have a
better depth sensitivity in the shallow crust and long periods have
a relatively poor depth resolution at deep depth since surface wave
energies tend to concentrate near the surface. Thus, at short periods
(8–20 s), the group velocities are much influenced by the upper part
of the crust. Short-period low-velocity anomalies are usually a good
indicator of sedimentary basins since sediments feature low seismic
velocities.
As seen in Fig. 9, low group velocities at short periods are ob-
served beneath the Po Plain and the Apennines. These low velocities
likely reflect the thick alluvial basin of the Po river and the sedi-
mentary deposits of the Apenninic and Alpine foredeeps and the
extensional basins distributed along the external and internal fronts
of the Apennines (Waldhauser et al. 1998, 2002; Kummerow et al.
2004; Di Stefano 2005). In the period range between 8 and 20 s the
Rayleigh wave energy is mostly concentrated in the crust in the land
areas, but it likely samples the higher velocities of the upper mantle
in oceanic areas where the crust is relatively thin. This is most likely
the reason why high group velocities occur in the Tyrrhenian Sea
area where seismic sounding data have long revealed a very thin
oceanic crust (Duschenes et al. 1986; Geiss 1987; de Voogd et al.
1992; Pepe et al. 2000).
The overall pattern of velocity distribution beneath the Apennines
and southern Alps is compatible with the previous group velocity
studies that used earthquake data (Pontevivo & Panza 2002). The
northern and central Apennines and southern Alps are clearly char-
acterized with low velocities. P-wave tomographic results from (Di
Stefano et al. 1999) also revealed low-velocity anomalies beneath
the northern Apennines and high-velocity anomalies beneath the
1248 H. Li, F. Bernardi and A. Michelini
48˚ 48˚

46˚ 46˚

44˚ 44˚

2500
Number of measurements

42˚ 42˚

2000

40˚ 40˚
1500

38˚ 38˚
1000

36˚ T=15.0 s 36˚ T=30.0 s

500

Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022

8˚ 10˚ 12˚ 14˚ 16˚ 18˚ 20˚ 8˚ 10˚ 12˚ 14˚ 16˚ 18˚ 20˚
0
8 12 16 20 24 28 32 36 40 44 0 10 50 100 150 200 250 300 400 500 0 10 50 100 150 200 250 300 400 500
Period (s) path density path density

Figure 7. Left-hand panel: distribution of dispersion measurements at different periods. Middle panel: path density at 15 s. Path density is defined as the
number of rays intersecting a 0.2◦ × 0.2◦ cell. Right-hand panel: path density at 30 s.

Figure 8. Checkerboard resolution tests with a grid spacing of 0.6◦ . (a) theoretical model for T = 15 s with 5 per cent velocity disturbance; (b) inversion
result at T = 15 s path density and (c) inversion result at T = 30 s path density.

southern Apennines and the Calabrian arc; at 22 km depth, a con-
tinuous low-velocity zone is observed beneath the whole Apenninic
belt, and the width of the low-velocity zone gets narrower in the
southern Apennines than in the northern Apennines. Studies from
Mele et al. (1996, 1997) also observed evidence for a broad low-
velocity and high-attenuation anomaly of Pn and Sn phases beneath
the Apenninic area. However, some inconsistencies also exist. Low-
velocity anomalies from P-wave tomography are imaged beneath
the western Sicily and the Aeolian volcanoes between 8 and 22 km,
which are interpreted to be crustal magmatic bodies, probably con-
nected with thermal anomalies at depth; high-velocity anomalies
are observed beneath the eastern Sicily (the Mt Etna volcano) be-
tween 8 and 22 km, and they suggested that is probably associated
with the presence of intrusive bodies and the absence of a wide
magma chamber in the crust (Di Stefano et al. 1999). Moreover,
local seismic tomography studies underneath the Mt Etna volcano
from Villaseñor et al. (1998) found that a high-velocity anomaly is
located in the depth range 2–12 km beneath the southeastern flank
of the volcano and a low-velocity-zone extends from beneath the
Crater region to a depth of 10 km, however, significantly different
tomographic results are also proposed for the same area, for exam-
ple, Aloisi et al. (2002) revealed that a major high-velocity-body
is recognized in the upper 10 km beneath the Crater area and the
southeastern flank of the volcano, and low velocities are found to
the west and the east of the high-velocity-body and also the south-
western flank of the volcano. In our study, most of the Sicily stands
out as having a low velocity at 15 s. Since the accuracy of earth-
quake location and arrival-time picking, as well as data coverage, all
influence the final tomographic results based on earthquake data, es-
pecially at shallow depths, in order to address above controversies,
further studies with a finer parametrization and higher resolution
are needed in the future. At this period, there is also a new feature


C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS

Figure 9. Estimated Rayleigh wave group velocity maps at periods of 8, 15, 24 and 30 s. Period is also indicated in the lower right-hand corner of each map.
recovered from our ambient noise tomography is—the velocity on
the Tyrrhenian side of the northern Apennines which is much higher
than that on the Adriatic side. This relevant contrast is well visible
at periods 24 and 30 s.
In the period range between 20 and 35 s, Rayleigh waves are
primarily sensitive to the thickness of crust and the shear velocities
in the lower crust and uppermost mantle. The group velocities can
reflect the Moho depth variation which correlates inversely with

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
Ambient seismic noise analysis in Italy 1249
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
group velocities—high velocities in regions with a thin crust and
low velocities in regions with a thick crust.
In Fig. 9, the velocity maps at 24 and 30 s still exhibit low ve-
locities associated with the Alps and northern-central Apennines.
The low velocities beneath the central Alps are probably related to
the thick crust in this region, where the crust has been estimated to
be about 50 km (Waldhauser et al. 2002; Kummerow et al. 2004;
Li et al. 2007; Tesauro et al. 2008). The low-velocity anomalies
1250 H. Li, F. Bernardi and A. Michelini

beneath the northern-central Apennines are also reported by pre-

vious work using earthquake data (e.g. Mele et al. 1996, 1997; Di
Stefano et al. 1999; Pontevivo & Panza 2002). Deep crustal roots,
beneath mountain regions resulting from isostatic compensation,
can explain the low velocities in the northern-central Apennines.
The high velocities are observed underneath Sicily which is differ-
ent from Di Stefano et al. (1999) and Lucente et al. (1999)’s results
featured with low-velocity anomalies at 38 km depth. However,
attenuation studies from Mele et al. (1996, 1997) did not show a
high-attenuation anomaly of Pn and Sn phases beneath the Sicily,
which is usually associated with a low-velocity anomaly. Since the
path density and the resolution of the Sicily island become much
poorer at longer period as shown in Figs 7 and 8, it is hard to make
further analysis based on our results.
We also note that the low-velocity feature beneath the Apennines

Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022

extends as far south as 41◦ latitude approximately. In the southern
part of the Apennines relatively higher velocities occur. The veloc-
ity contrast between the northern and southern Apennines has been
also observed by other authors (Di Stefano et al. 1999; Pontevivo &
Panza 2002). This transition corresponds to a well-known bound-
ary between the northern and southern Apenninic arcs (Patacca &
Scandone 1989).

4.2 Shear wave velocity structures

From the group velocities obtained in the previous section, we
extracted the pure path dispersion curves for each grid node. Al-
though the sensitivity kernel of group velocity has a complicated
relation with the structure, in general, longer period Rayleigh waves
are more affected by deeper structures, therefore, these data may
provide valuable insights into the lateral variation of shear wave
velocities in different depth ranges.
Inversions for the shear wave velocity structure at each node of
the 0.2◦ × 0.2◦ grid was carried out in this study with a program
developed by Herrmann & Ammon (2004) using linear steps. The
S-wave velocity in each layer is taken as the inversion parameter, and
P-wave velocity and density are calculated from S-wave velocity
with empirical formulas. The same initial model is used for all
nodes in our inversion, then the inversion was repeated 15 times
with 15 different initial models by referring to recent tomography
results (Di Stefano 2005; Li et al. 2007), and the final S-wave
velocity structure was averaged from the 15 inversion solutions.
As we know, shear wave structure inversion based on surface wave
dispersion data suffers from poor resolution and non-uniqueness
due to strong non-linearity, therefore, we did not put much effort
into seeking the true model’s possible ranges in this study. Our
goal is to obtain a preliminary 3-D structure, which could satisfy
our group velocity dispersion data and show the essential crustal
and upper-mantle features of different tectonic units in the study
area.
Figure 10. Vertical cross-sections of shear wave velocity along the lines
Fig. 10 shows shear wave velocity structure along three verti- AA , BB and CC in Fig. 1. Topography is plotted above each profile (black
cal profiles from the surface down to 40 km depth. As seen in the area).

profile AA , there are obvious differences in the structures along

the profile. It is remarkable that shear wave velocities beneath the Sea. In the profile BB crossing Sicily to the southern Apennines,
northern Tyrrhenian Sea clearly display much higher values than the lateral variation of velocities is still evident. The Sicily island
those beneath the northern Apennines and the northern Adriatic shows lower velocities at depth 10–15 km than the southern Apen-
Sea at depth deeper than 8 km, which are consistent with the shal- nines, however, at depth larger than 20 km, the pattern is reversed
low Moho depth beneath the Tyrrhenian Sea area where seismic and the Sicily island is characterized with high velocities. Although
sounding data have revealed a very thin oceanic crust (Duschenes some geophysical studies have suggested that the local low-velocity
et al. 1986; Geiss 1987; de Voogd et al. 1992; Pepe et al. 2000), and anomaly in this region is associated with crustal magmatic bodies,
normal continental crust beneath the Apennines and the Adriatic more detailed studies are needed to be performed in the future.


C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS


Another obvious feature in the profile BB is that at depth deeper
than 15 km, the southern Tyrrhenian sea region features with high

velocities which reflect its thin oceanic crust. Along the profile CC ,
the lowest velocities at shallow depth occurred beneath the Po Plain
which is associated with thick cover of alluvial deposits in this
area.
5 C O N C LU S I O N S
In this study, we have obtained group velocity maps of Rayleigh
waves between 8 and 36 s, using ambient noise data recorded at
114 broad-band seismic stations from INGV, MedNet and ZAMG.
Cross-correlations are computed in one-hour segments with a total
duration ranging between 3 months and 1.5 yr. Group velocity
dispersion curves at periods from 8 to 44 s were determined using
the multiple-filter analysis technique. The lateral resolution in the
Italian peninsula and Tyrrhenian Sea is estimated to be ∼0.6◦ , but
a poor resolution is found on the northeastern side of the Italian
peninsula because of the few available stations.
Our group velocity maps show that at increasing period, the
velocity distribution pattern changes systematically and correlates
well with the known geological and tectonic features—sedimentary
basins and lateral variation of crustal thickness. Compared with
previous surface wave studies which have relied on traditional
earthquake-based measurements, this study has a much denser data
coverage and higher resolution. In general, previously found fea-
tures are confirmed and, in addition, our results showed that the
Tyrrhenian Sea is clearly featured with much higher velocities be-
low 8 km due to its thin oceanic crust.
AC K N OW L E D G M E N T S
We are grateful to the Centro Nazionale Terremoti (CNT) of INGV,
MedNet and ZAMG data center for providing the data. We thank
Zhongxian Huang for offering the inversion code. We benefited from
discussions with R. B. Herrmann and Lupei Zhu. This research was
supported in part by the 2004–2006 DPC-S4, 2007–2009 DPC-S3
projects and the open fund (No. GDL0708).
REFERENCES
Aloisi, M., Cocina, O., Neri, G., Orecchio, B. & Privitera, E., 2002. Seismic
tomography of the crust underneath the Etna volcano, Sicily, Phys. Earth
planet. Inter., 134, 139–155.
Amato, A. & Mele, F.M., 2008. Performance of the INGV National Seismic
Network from 1997 to 2007, Ann. Geophys., 51(2/3), 417–431.
Bensen, G.D., Ritzwoller, M.H., Barmin, M.P., Levshin, A.L., Lin, F.,
Moschetti, M.P., Shapiro, N.M. & Yang, Y., 2007. Processing seismic
ambient noise data to obtain reliable broad-band surface wave dispersion
measurements, Geophys. J. Int., 169, 1239–1260.
Bensen, G.D., Ritzwoller, M.H. & Shapiro, N.M., 2008. Broadband ambient
noise surface wave tomography across the United States, J. geophys. Res.,
113, doi:10.1029/2007JB005248.
Campillo, M. & Paul, A., 2003. Long-range correlations in the diffuse
seismic coda, Science, 299, 547–549.
Chimera, C., Aoudia, A., Sarao, A. & Panza, G.F., 2003. Active tectonics
in Central Italy: constraints from surface wave tomography and source
moment tensor inversion, Phys. Earth. planet. Inter., 138, 241–262.
Cho, K.H., Herrmann, R.B., Ammon, C.J. & Lee, K., 2007. Imaging the
crust of the Korean peninsula by surface wave tomography, Bull. seism.
Soc. Am., 97, 198–207.

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS
Ambient seismic noise analysis in Italy 1251
Constable, S.C., Parker, R.L. & Constable, C.G., 1987. Occam s inversion:
A practical algorithm for generating smooth models from electromagnetic
sounding data, Geophysics, 52, 289–300.
deGroot-Hedlin, C. & Constable, S., 1990. Occam’s inversion to generate
smooth, two-dimensional models from magnetotelluric data, Geophysics,
55, 1613–1624.
de Lorenzo, S., Iannaccone, G. & Zollo, A., 2003. Modeling of Rayleigh
waves dispersion in the Sannio region (Southern Italy) from seismic active
refraction data, J. Seism., 7, 49–64.
de Voogd, B., Truert, C., Chamot-Rooke, N., Huchon, P., Lallemant, S. &
Le Pichon, X., 1992. Two-ships deep seismic soundings in the basin of
the eastern Mediterranean Sea (Pasiphae cruise), Geophys. J. Int., 109,
536–552.
Derode, A., Larose, E., Tanter, M., de Rosny, J., Tourim, A., Campillo,
M. & Fink, M., 2003. Recovering the Green’s function from field-field
correlations in an open scattering medium, J. acoust. Soc. Am., 113,
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
2973–2976.
Di Stefano, R., Chiarabba, C., Lucente, F. & Amato, A., 1999. Crustal and
uppermost mantle structure in Italy from the inversion of P-wave arrival
times: geodynamic implications, Geophys. J. Int., 139, 483–498.
Di Stefano, R., 2005. Subduction-collision structure beneath Italy: high res-
olution images of the Adriatic-European-Tyrrhenian lithospheric system,
PhD. thesis, Naturwissenschaften, Eidgenssische Technische Hochschule
ETH Zrich.
Duschenes, J., Sinha, M.C. & Louden, K.E., 1986. A seismic refraction
experiment in the Tyrrhenian Sea, Geophys. J. R. astr. Soc., 85, 139–
160.
Dziewonski, A., Bloch, S. & Landisman, M., 1969. A technique for the
analysis of transient seismic signals, Bull. seism. Soc. Am., 59, 427–
444.
Friedrich, A., Krüger, F. & Klinge, K., 1998. Ocean-generated microseismic
noise located with the Gräfenberg array, J. Seism., 2, 47–64.
Geiss, E., 1987. A new compilation of crustal thickness in the Mediterranean
area, Ann. Geophys., 5B(6), 623–630.
Herrmann, R.B., 1973. Some aspects of band-pass filtering of surface waves,
Bull. seism. Soc. Am., 63, 663–671.
Herrmann, R.B. & Ammon, C.J., 2004. Surface waves, receiver func-
tions and crustal structure, in Computer Programs in Seismology,
Version 3.30, Saint Louis University, http://www.eas.slu.edu/People/
RBHerrmann/CPS330.html.
Huang, Z., Su, W., Peng, Y., Zheng, Y. & Li, H., 2003. Rayleigh wave
tomography of China and adjacent regions, J. geophys. Res., 108(B2),
2073, doi:10.1029/2001JB001696.
Kang, T.S. & Shin, J.S., 2002. Surface-wave tomography from ambient
seismic noise of accelerograph networks in southern Korea, Geophys.
Res. Lett., 33, 1–5, doi:10.1029/2006GL027044.
Kummerow, J., Kind, R., Oncken, O., Giese, P., Ryberg, T., Wylegalla,
K., Scherbaum, F. & TRANSALP Working Group, 2004. A natural and
controlled source seismic profile through the Eastern Alps: TRANSALP,
Earth planet. Sci. Lett., 225, 115–129.
Larose, E., Derode, A., Corennec, M., Margerin, L. & Campillo, M., 2005.
Passive retrieval of Rayleigh waves in disordered elastic media, Phys. Rev.
E, 72, 046607, doi:10.1103/PhysRevE.72.046607.
Levshin, A.L., Yanovskaya, T.B., Lander, A.V., Buckchin, B.G., Barmin,
M.P., Ratnikova, L.I. & Its, E.N., 1989. Seismic Surface Waves in a Later-
ally Inhomogeneous Earth, ed. Keilis-Borok, V.I., Kluwer, Norwell, MA.
Li, H., Michelini, A., Zhu, L., Bernardi, F. & Spada, M., 2007. Crustal
velocity structure in Italy from analysis of regional seismic waveforms,
Bull. seism. Soc. Am., 97, 2024–2039.
Li, H., Su, W., Wang, C.-Y. & Huang, Z., 2009. Ambient noise Rayleigh
wave tomography in western Sichuan and eastern Tibet, Earth planet. Sci.
Lett., 282, 201–211.
Liang, C. & Langston, C.A., 2008. Ambient seismic noise tomogra-
phy and structure of eastern North America, J. geophys. Res., 113,
doi:10.1029/2007JB005350.
Lin, F., Ritzwoller, M.H., Townend, J., Bannister, S. & Savage, M.K., 2007.
Ambient noise Rayleigh wave tomography of New Zealand, Geophys. J.
Int., 170, 649–666, doi:10.1111/j.1365-246X.2007.0341.x.
1252 H. Li, F. Bernardi and A. Michelini
Lucente, F.P., Chiarabba, C., Cimini, G.B. & Giardini, D., 1999. Tomo-
graphic constraints on the geodynamic evolution of the Italian region,
J. geophys. Res., 104, 20 307–20 327.
Mantovani, E., Nolet, G. & Panza, G.F., 1985. Lateral heterogeneity in
the crust of the Italian region from regionalized Rayleigh wave group
velocities, Ann. Geophys., 3, 519–530.
Marone, F., Van der Lee, S. & Giardini, D., 2004. three-dimensional upper-
mantle S-velocity model for the Eurasia-Africa plate boundary region,
Geophys. J. Int., 158, 109–130.
Marzorati, S. & Bindi, D., 2006. Ambient noise levels in north central Italy,
Geochem. Geophys. Geosyst., 7, doi:10.1029/2006GC001256.
Marzorati, S. & Bindi, D., 2008. Characteristics of ambient noise cross
correlations in northern Italy within the frequency range of 0.1-0.6 Hz,
Bull. seism. Soc. Am., 98, 1389–1398.
Mele, G., Rovelli, A., Seber, D. & Barazangi, M., 1996. Lateral variations
of Pn propagation in Italy: evidence for a high-attenuation zone beneath
the Apennine, Geophys. Res. Lett., 23, 709–712.
Mele, G., Rovelli, A., Seber, D. & Barazangi, M., 1997. Shear wave atten-
uation in the lithosphere beneath Italy and surrounding regions: tectonic
implications, J. geophys. Res., 102, 11 863–11 875.
Patacca, E. & Scandone, P., 1989. Post-Tortonian mountain building in the
Apennines. The role of the passive sinking of a relic lithospheric slab,
in The Lithosphere in Italy, pp. 157–176, eds Boriani, A., Bonafede, M.,
Piccardo, G.B. & Vai, G.B., Academia Nazionale dei Lincei, Rome.
Panza, G., Mueller, S. & Calcagnile, G., 1980. The gross features of the
lithosphere-asthenosphere system in Europe from seismic surface waves
and body waves, Pure appl. Geophys., 118, 1209–1213.
Pedersen, H.A., Krüger, F. & the SVEKALAPKO Seismic Tomography
Working Group, 2007. Influence of the seismic noise characteristics
on noise correlations in the Baltic shield, Geophys. J. Int., 168, 197–
208.
Pepe, F., Bertotti, G., Cella, F. & Marsella, E., 2000. Rifted margin formation
in the south Tyrrhenian Sea: a high-resolution seismic profile across the
north Sicily passive continental margin, Tectonics, 19(2), 241–257.
Pontevivo, A. & Panza, G., 2002. Group velocity tomography and region-
alization in Italy and bordering areas, Phys. Earth planet. Int., 52, 193–
203.
Raykova, R., Morelli, A., Park, J. & Pondrelli, S., 2006. Surface wave to-
mography in Northern Apennines (Italy) derived from RETREAT records,
Geophys. Res. Abs., EGU, 8, 08223.
Sabra, K.G., Gerstoft, P., Roux, P., Kuperman, W.A. & Fehler, M.C., 2005a.
Extracting time-domain Green’s function estimates from ambient seismic
noise, Geophys. Res. Lett., 32, L03310, doi:10.1029/2004GL021862.
Sabra, K.G., Gerstoft, P., Roux, P., Kuperman, W.A. & Fehler, M.C., 2005b.
Surface wave tomography from microseism in southern California, Geo-
phys. Res. Lett., 32, L14311, doi:10.1029/2005GL023155.
Sato, H. & Fehler, M., 1998. Seismic Wave Propagation and Scattering in the
Heterogeneous Earth, Springer-Verlag and American Institute of Physics
Press, New York.
Scarascia, S. & Cassinis, R., 1997. Crustal structures in the central-eastern
Alpine sector: a revision of the available DSS data, Tectonophysics, 271,
157–188.
Shapiro, N.M. & Campillo, M., 2004. Emergence of broadband Rayleigh
waves from correlations of the ambient seismic noise, Geophys. J. Int.,
151, 88–105.
Shapiro, N.M., Campillo, M., L. Stehly & Ritzwoller, M.H., 2005. High-
resolution surface-wave tomography from ambient seismic noise, Science,
307, 1615–1618.
Snieder, R., 2004. Extracting the Green’s function from the correlation of
coda waves: a derivation based on stationary phase, Phys. Rev. E, 69,
046610, doi:10.1103/PhysRevE.69.046610.
Stehly, L., Fry, B., Campillo, M., Shapiro, N.M., Guilbert, J., Boschi, L. and
Giardini, D., 2009. Tomography of the Alpine region from observations
of seismic ambient noise, Geophys. J. Int., 178, 338–350.
Tesauro, M., Kaban, M.K. & Cloetingh, S.A.P.L., 2008. EuCRUST-07: a
Downloaded from https://academic.oup.com/gji/article/180/3/1242/555856 by guest on 09 November 2022
new reference model for the European crust, Geophys. Res. Lett., 35,
doi:10.1029/2007GL032244.
Villaseñor, A., Benz, H.M., Filippi, L., De Luca, G., Scarpa, R., Patan, G.
& Vinciguerra, S., 1998. Three-dimensional P-wave velocity structure of
Mt. Etna, Italy, Geophys. Res. Lett., 25, 1975–1978.
Waldhauser, F., Kissling, E., Ansorge, J. & Mueller, S., 1998. Three-
dimensional interface modeling with two-dimensional seismic data: the
Alpine crust-mantle boundary, Geophys. J. Int., 135, 264–278.
Waldhauser, F., Lippotsch, R., Kissling, E. & Ansorge, J., 2002. High-
resolution teleseismic tomography of upper-mantle structure using an a
priori three-dimensional crustal model, Geophys. J. Int., 150, 403–414.
Wapenaar, K., 2004. retrieving the elastodynamic Green’s function of an
arbitrary inhomogeneous medium by cross correlation, Phys. Rev. Lett.,
93, paper 254301.
Weaver, R.L. & Lobkis, O.I., 2001a. Ultrasonics without a source: thermal
fluctuation correlation at MHZ frequencies, Phys. Rev. Lett., 87, paper
134301.
Weaver, R.L. & Lobkis, O.I., 2001b. On the emergence of the Green’s
function in the correlation of a diffuse field, J. acoust. Soc. Am., 110,
3011–3017.
Weaver, R.L. & Lobkis, O.I., 2004. Diffuse fields in open system and the
emergence of the Green’s function, J. acoust. Soc. Am., 110, 2731–2734.
Yang, Y., Ritzwoller, M.H., Levshin, A.L. & Shapiro, N.M., 2007. Ambient
noise Rayleigh wave tomography across Europe, Geophys. J. Int., 168,
259–274.
Yao, H., van der Hilst, R.D. & de Hoop, M.V., 2006. Surface-wave tomogra-
phy in SE Tibet from ambient seismic noise and two-station analysis: I—
phase velocity maps, Geophys. J. Int., 166, 732–744, doi:10.1111/j.1365-
246X.2006.03028.x.
Yao, H., Beghein, C. & van der Hilst, R.D., 2008. Surface wave array
tomography in SE Tibet from ambient seismic noise and two-station
analysis: II—crustal and upper-mantle structure, Geophys. J. Int., 173,
205–219, doi:10.1111/j.1365-246X.2007.03696.x.

C 2010 The Authors, GJI, 180, 1242–1252
Journal compilation 
C 2010 RAS