You are on page 1of 13

Bulletin of the Seismological Society of America, Vol. 98, No. 1, pp. 288–300, February 2008, doi: 10.

1785/0120070063

Effects of Love Waves on Microtremor H/V Ratio


by Sylvette Bonnefoy-Claudet, Andreas Köhler, Cécile Cornou, Marc Wathelet,
and Pierre-Yves Bard

Abstract The horizontal-to-vertical (H/V) method has the potential to significantly


contribute to site effects evaluation, in particular in urban areas. Within the European
project, site effects assessment using ambient excitations (SESAME), we investigated
the nature of ambient seismic noise in order to assess the reliability of this method.
Through 1D seismic noise modeling, we simulated ambient noise for a set of various
horizontally stratified structures by computing efficiently the displacement and stress
of dynamic Green’s functions for a viscoelastic-layered half-space. We performed ar-
ray analysis using the conventional semblance-based frequence-wavenumber method
and the three-component modified spatial autocorrelation method on both vertical and
horizontal components and estimated the contribution of different seismic waves
(body/surface waves, Rayleigh/Love waves) at the H/V peak frequency. We show that
the very common assumption that almost all the ambient noise energy would be car-
ried by fundamental-mode Rayleigh waves is not justified. The relative proportion of
different wave types depends on site conditions, and especially on the impedance con-
trast. For the 1D horizontally layered structures presented here, the H/V peak fre-
quency always provides a good estimate of the fundamental resonance frequency
whatever the H/V peak origin (Rayleigh wave ellipticity, Airy phase of Love waves,
S-wave resonance). We also infer that the relative proportion of Love waves in ambient
noise controls the amplitude of the H/V peak.

Introduction
The horizontal-to-vertical (H/V) spectral ratio (i.e., the layer, leading to H/V curves providing a consistent estimate
ratio between the Fourier amplitude spectra of the horizontal of the site amplification. This body-wave interpretation has
and the vertical component of microtremors) was first intro- been contradicted in several articles highlighting the relation-
duced by Nogoshi and Igarashi (1971), and was spread ship between the H/V and the ellipticity of fundamental
widely by Nakamura (1989, 1996, 2000). These authors Rayleigh-wave mode (Lachet and Bard, 1994; Kudo,
pointed out the correlation between the H/V peak frequency 1995; Bard, 1998; Fäh et al., 2001; Bonnefoy-Claudet,
and the fundamental resonance frequency of the site and pro- Cornou, et al., 2006) and thus has seriously questioned
posed to use the H/V technique as an indicator of the under- the existence of any simple direct correlation between H/V
ground structure features. A large number of experiments peak value and the actual site amplification. Recently,
(Lermo and Chavez-Garcia, 1993; Gitterman et al., 1996; Bonnefoy-Claudet, Cotton, et al. (2006) highlighted the lim-
Seekins et al., 1996; Fäh, 1997), supported by several theo- ited knowledge within the scientific community on the rela-
retical 1D investigations (Field and Jacob, 1993; Lachet and tive contribution of body and surface waves to the actual
Bard, 1994; Lermo and Chavez-Garcia, 1994; Wakamatsu ambient noise wave field.
and Yasui, 1996; Tokeshi and Sugimura, 1998; Bonnefoy- The goal of the present article is to investigate the rela-
Claudet, Cornou, et al., 2006), has shown that noise syn- tionship between the ambient noise wave-field nature and
thetics computed using randomly distributed near-surface composition and the effectiveness of the H/V method to es-
sources lead to H/V ratios sharply peaked around the funda- timate site response parameters (resonance frequency and
mental resonance frequency of SH waves whenever the sur- amplification factor). We simulate seismic ambient noise
face layer exhibits a sharp impedance contrast with the for a set of various realistic horizontally stratified structures,
underlying stiffer formations. However, the theoretical basis perform array analysis using the conventional semblance-
of the H/V technique is still unclear as two opposite explana- based frequency-wavenumber method (CVFK) and the
tions have been proposed. Nakamura (1989, 2000) claims three-component modified spatial autocorrelation method
that the H/V spectral ratio peak, by removing effects of sur- (3C-MSPAC) on both vertical and horizontal components,
face waves, provides the S-wave resonance in the soft surface and finally estimate the contribution of different seismic

288
Effects of Love Waves on Microtremor H/V Ratio 289

waves (body/surface waves, Rayleigh/Love waves) at the In our simulations, we do not take into account the effect
H/V peak frequency. of oceanic waves, which propagate over large distances in
the low-frequency range (i.e., ≪ 1 Hz). Except for the
Noise Data Sets deepest model (M3.3), where the resonance frequency is
0.45 Hz, the resonance frequency of each site is large enough
Ten horizontally stratified structures consisting of vari- not to be influenced by propagation of such low-
ous sedimentary layers overlying half-space, characterized
frequency waves.
mainly by their impedance contrast, have been selected. Ta-
Recently, Bonnefoy-Claudet, Cornou, et al. (2006)
ble 1 details the soil profiles used, which have been selected
showed that at a given observation site, local and near-
to be as close as possible to actual soil properties. A nonex-
surface sources dominate the cultural ambient noise wave
haustive review of soil properties performed at 136 different
field. Therefore we set up a source location to fulfill these
sites shows indeed that the soil structures considered in this
article are consistent with actual distribution of the S-wave conditions. Spatial locations of sources and receivers are ex-
velocity contrast observed at those sites (Fig. 1). Moreover, pressed in a grid step the size of which varies according to the
attenuation values have been selected to model values typi- sedimentary layer thickness. The grid step is then set to 2 m
cally observed or considered for sedimentary deposits for the shallow model (M2.4), to 4 m for intermediate thick-
(Hauksson et al., 1987; Fukushima et al., 1992; Wang et al., ness models (M2.2, M3.1, M3.2, M4.1, M4.2, M4.3, and
1994; Jongmans et al., 1998; Di Giulio et al., 2003). M4.4), to 8 m and 16 m for deep models (M2.3 and
The seismic noise recorded in urban areas (cultural M3.3, respectively). 333 sources are randomly distributed
noise) can be considered as mainly caused by human activity in space and time. The maximum distance between the cen-
(people walking, cars, factories, etc.) and propagates mainly tral receiver and sources is 361 grid step (Fig. 2a) and the
as high-frequency surface waves (> 1 Hz) that attenuate depth of the sources is 0.5 grid step. The minimum and maxi-
within several kilometers in distance and depth (McMamara mum distances between two receivers are 1 and 40 grid steps,
and Buland, 2004; Bonnefoy-Claudet, Cotton, et al., 2006). respectively (Fig. 2b).

Table 1
Soil Profiles Considered for the 1D Horizontally Layered Models
Model N H, (m) α (m=sec) β (m=sec) ρ (g=cm3 ) Qp Qs Zc Zv Fo Fell Fa

M2.2 2 25 1350 200 1.9 50 25 6.6 5 1.95 1.93 2.1


— 2000 1000 2.5 100 50
M2.3 2 83 1350 667 1.9 50 25 2.0 1.5 1.94 - 3.3
— 2000 1000 2.5 100 50
M2.4 2 10 1350 200 1.9 20 10 6.6 5 4.83 4.83 5.2
— 2000 1000 2.5 100 50
M3.1 3 18 1350 250 1.9 50 25 4.6 3.5 2.11 2.15 2.3
18 1350 330 1.9 50 25
— 2000 1000 2.5 100 50
M3.2 3 18 1350 250 1.9 50 25 5.5 4.2 2.94 2.76 3.1
18 1350 625 1.9 50 25
— 2000 1500 2.5 100 50
M3.3 3 31.25 500 250 1.9 50 25 4.0 3.0 0.45 0.73 -
375 1800 750 2.1 100 50
— 3500 2000 2.5 200 100
M4.1 2 36 1350 250  2z 1.9 50 25 4.7 3.6 2 2 2.3
— 2000 1000 2.5 100 50
M4.2 2 36 1350 250  5z 1.9 50 25 4.1 3.1 2.45 2.6 2.8
— 2000 1000 2.5 100 50
M4.3 2 36 1350 250  9z 1.9 50 25 3.5 2.7 3 3.4 3.6
— 2000 1000 2.5 100 50
M4.4 2 36 1350 250  15z 1.9 50 25 3.0 2.3 3.94 4.9 —
— 2000 1000 2.5 100 50

Number of layers, N; thickness, H; P-wave velocity, α; S-wave velocity, β; quality factors for P and S waves, Qp
and Qs , respectively; S-wave impedance contrast, Zc ; S-wave velocity contrast, Zv ; theoretical fundamental resonance
frequency, Fo ; ellipticity peak frequency of Rayleigh waves, Fell ; Airy phase frequency of Love waves, Fa . For the
three-layered models (M3) and the gradient models (M4), the impedance and velocity contrasts are defined by
considering the average S-wave velocity (by means of mean time propagation) within the sediments. z
represents depth.
290 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

(a) 1D models (b) Real sites


20 %
38 %
60 %
32 %

20 %

30 %

Figure 1. (a) Pie chart showing the distribution of the S-wave velocity contrasts for the 1D models presented in Table 1. (b) Distribution of
the actual S-wave velocity contrasts observed (or estimated) at 136 sites, including a broad range of sedimentary structures (shallow and deep
basin, for instance). Details of the soil properties have been found in the scientific literature dealing with site effects. The white color shows
the relative proportion of low velocity contrast (Zv < 2:5); the gray color shows the relative proportion of moderate velocity contrast
(2:5 ≤ Zv ≤ 3); the black color shows the relative proportion of high-velocity contrast (Zv > 3). The numbers indicate the value of the
relative proportion in percent.

(a) Receiver - source geometry (b) Receivers geometry


150
20
Y direction (grid step)

75 10

0 0

−75 −10

−20
−150

−150 −75 0 75 150 −20 −10 0 10 20


X direction (grid step) X direction (grid step)

Figure 2. (a) Spatial distribution of the 333 sources (crosses) and 42 receivers (dots). (b) Zoom on the array layout. See text for values of
grid steps used for each model.

The ambient noise wave field is computed at each re- In this study, Green functions are computed up to
ceiver of all soil models by using the wavenumber-based 14.28 Hz using a number of frequencies equal to 1024.
code developed by Hisada (1994, 1995), which computes The response of the structure is convolved with the large
efficiently the displacement and stress of dynamic Green’s number of sources approximated by surface forces with
functions for a viscoelastic-layered half-space. This code, random direction and amplitude (Moczo and Kristek,
which computes Green functions due to point sources for 2002). The source time function is a deltalike signal
viscoelastic horizontally stratified media, allows sources band-pass filtered (for numerical considerations) between
and receivers at very close depths, and significantly reduces 0.5 and 14.28 Hz for all models except the deepest model
the range of wavenumber integration, especially for the case (M3.3). For this latter, which exhibits a resonance frequency
in which sources and receivers are close to the free surface. at 0.45 Hz, the source time function is filtered between 0.1
Effects of Love Waves on Microtremor H/V Ratio 291

and 14.28 Hz to allow sufficient low-frequency energy to ex- logarithmic scale. A fraction of the central frequency (fc )
cite the sedimentary structure. However, as the ambient noise defines the frequency bandwidth (0.9–1.1 fc ). We select
simulation does not involve effects of very far sources such the time-window length as being 50 times the central period
as oceanic—or coastal—waves, when combined with our (1=fc ). The CVFK method is applied independently on both
Earth model (which is an infinite homogeneous half-space), vertical and horizontal components. We do not rotate hori-
the low-frequency energy is spread out in the half-space. zontal components into radial and transverse components.
The time series obtained at each receiver location is the We do this simplification because sources are randomly dis-
sum of time histories due to all sources’ contributions. The tributed in space (X, Y plane) (Bonnefoy-Claudet, 2004).
final duration of seismograms is 405 sec. One can refer to Figure 3 shows the comparison of phase velocity estimates
Bonnefoy-Claudet et al. (2004) and Bonnefoy-Claudet, computing for both north–south and east–west components.
Cornou, et al. (2006) for successful application of Hisada’s Although slight differences can be observed on phase velo-
code to simulate an ambient noise wave field in 1D media. city estimates, the main trend is the same on both compo-
nents. Therefore, for clarity we show the next results for
Methods the north–south component only.

Because this study is aimed at inferring the composition 3C-MSPAC Method. The spatial autocorrelation method
of the seismic noise wave field at the H/V ratio peak fre- (SPAC) (Aki, 1957) allows us to estimate the frequency-
quency, we present here the basis of the H/V method (so- dependent phase velocity from the azimuthal-averaged auto-
called Nakamura’s technique) and the array-processing correlation curve (function of frequency) without requiring
methods used (CVFK and 3C-MSPAC). knowledge about the directions of propagation and number
of incoming waves. Furthermore, the method offers a decom-
H/V position of the horizontal component records into the radial
and transversal polarized wave fields. Thus, assuming only
The H/V technique originally proposed by Nogoshi and
single-mode Rayleigh and Love waves, SPAC allows the de-
Igarashi (1971) and spread widely by Nakamura (1989,
termination of the dispersion curves and the partitioning be-
2000) consists of estimating the ratio between the Fourier
tween both types of waves in the horizontal component wave
amplitude spectra of the horizontal and the vertical compo-
field. In particular the relative proportion of Rayleigh waves
nents of the microtremor recorded at the surface. In this
α  R=R  L is estimated, where R and L indicate the
study, the H/V ratio is calculated using 30-sec time windows.
power of Rayleigh and Love waves on the horizontal com-
Fourier amplitude spectra are smoothed following Konno
ponents.
and Ohmachi (1998), with a b parameter set to 40. The quad-
One of several modifications or extensions of SPAC is
ratic mean of the horizontal amplitude spectra is used here.
the MSPAC method, which admits arbitrary array layouts
The final H/V ratio is obtained by averaging the H/V ratios
(Bettig et al., 2001). Here, we employ the extension of
from all the windows. The standard deviations on H/V ratio
MSPAC on three components (3C-MSPAC) (Köhler et al.,
curves are estimated considering the arithmetic average for
2007). The processing of data includes a sliding-window
all individual logarithms of H/V ratio computed at each time
analysis (see the section Frequency-Wavenumber Method).
window. See Atakan, Bard, et al. (2004), and Bard (2005) for
For each time window and component (vertical, radial,
further details.
and tangential) the spatial averaged autocorrelation curves
are computed. Afterward, the mean curve and its variance
2D Arrays are calculated over all windows. We estimate the best-fitting
Frequency-Wavenumber Method. The frequency-wave- combination of phase velocities and proportion of Rayleigh
number (f-k) based methods are often used for deriving waves and the corresponding uncertainties from the time-
the phase velocity dispersion curves from ambient vibration window-averaged autocorrelation curves using a simple grid
array measurements. In this study, we have used the CVFK search. For further details, see Köhler et al. (2007). Care
(Kvaerna and Ringdahl, 1986) implemented in the CAP soft- must be taken when interpreting the results obtained for
ware developed within the framework of the site effects as- α. A correct interpretation is ensured only if both Rayleigh-
sessment using the ambient excitations (SESAME) project and Love-wave velocities show reasonable results. Further-
(Ohrnberger, 2004; Ohrnberger, Schissele, Cornou, Bonne- more, α  0 only indicates absence of Rayleigh waves on
foy-Claudet, et al., 2004; Ohrnberger, Schissele, Cornou, the horizontal components. Purely vertical polarized Ray-
Wathelet, et al., 2004). Operating within sliding time win- leigh waves, as well as the total absence of Rayleigh waves
dows and narrow frequency bands, this method provides in the wave field, may explain this observation.
the wave-propagation parameters (azimuth and slowness
as a function of frequency) of the most coherent plane-wave Application to Simulated Seismic Noise. Figure 3 displays
arrivals. In this study, we use a wavenumber grid layout phase velocities estimated for the M2.2 model using the
sampled equidistantly in slowness and azimuth. The central CVFK and the 3C-MSPAC methods. Both methods success-
frequency of each band is selected to be equally spaced in a fully identify the fundamental Rayleigh mode (using the ver-
292 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

(a) SPAC - Vertical (b) SPAC - Horizontal


1200 1200
(m/s)

800
800

400
400

0
2 10 0
2 10

(c) FK - Vertical (d) FK - North-South


1200 1200
(m/s)

800 800

400 400

0
2 10 0
Frequency (Hz) 2 10

(e) FK - East-West
Figure 3. Phase velocity estimates (thin black lines) and stan-
dard deviation obtained by using the three-components modified-
1200 spatial autocorrelation method 3C-MSPAC for (a) vertical and
(b) horizontal components. Normalized histogram distributions of
phase velocities derived from the ensemble of the wave-propagation
(m/s)

800
estimates obtained for each individual time-frequency cell using the
CVFK method (gray scale: white and black colors indicate normal-
400 ized values of 0 and 1, respectively), for (c) vertical, (d) north–
south, and (e) east–west components. The theoretical dispersion
curves for the fundamental, first and second Rayleigh-wave modes
0 (black lines) and the theoretical dispersion curves for the fundamen-
2 10
Frequency (Hz) tal, first, and second Love-wave modes (dashed black lines) are also
displayed.

tical components) and the fundamental Love mode (using the Origin of the H/V Peak
horizontal components) in the simulated seismic noise wave
field. At high frequencies (i.e., above 6 Hz), the CVFK meth- The H/V curves and apparent phase velocities computed
od also detects higher Rayleigh modes, whereas the 3C- from noise synthetics for each soil model are shown in Fig-
MSPAC method failed in identifying higher modes. These ob- ures 4 to 13. Models have been classified into three groups
servations agree with Okada (2003) and Asten (2006), who according to the impedance contrast between sediment and
show the limitation of standard SPAC technique (Aki, 1957) bedrock (see Table 1):
to identify higher modes in ambient noise wave field. Note 1. High-impedance contrast models (i.e., strictly above 4):
that recent improvements of modified SPAC processing tech- models M2.2, M2.4, M3.1, M3.2, M4.1, and M4.2
nique have been done to better identify and account for high- (Figs. 4 to 9).
er modes (Asten, 2006). However, the comparison between 2. Moderate-impedance contrast models (i.e., between 3
array-noise-based methods is not the aim of the present ar- and 4): models M3.3, and M4.3 (Figs. 10 and 11).
ticle, and one may refer to Okada (2003) for such a compar- 3. Low-impedance contrast models (i.e., below 3): models
ison. Thus, the phase velocity estimates shown in the M2.3 and M4.4 (Figs. 12 and 13).
following are computed using the CVFK method; the relative
proportion of Rayleigh waves (α) is computed using the 3C- Whatever the soil model and the impedance contrast, the
MSPAC method. H/V curves always exhibit a peak in the vicinity of the fun-
Effects of Love Waves on Microtremor H/V Ratio 293

(a) (a)

Amplitude
Amplitude

10 10

1 1

1 1

Alpha
Alpha

(b) (b)
0.5 0.5

1200 (c) 1200 (c)

V (m/s)
V (m/s)

800 800

400 400

0 0
1200 (d) 1200 (d)

NS (m/s)
NS (m/s)

800 800

400 400

0 0
2 10 4 10
Frequency (Hz) Frequency (Hz)

Figure 4. H/V ratio and array analysis results for the high-im- Figure 5. H/V ratio and array analysis results for the high-im-
pedance contrast model M2.2. (a) H/V ratio curve (thick black line) pedance contrast model M2.4. See Figure 4 for legend.
and standard deviation (grayed area), theoretical ellipticity curves of
the fundamental, first, and second Rayleigh-wave modes (thin black
lines), and theoretical 1D transfer function computed for vertically
incident SH waves (dashed black lines). (b) Relative proportion of
Rayleigh waves (Alpha) and associated standard deviation esti-
mated by using the 3C-MSPAC method. (c) Normalized histogram (a)
distributions of phase velocities for vertical component derived
Amplitude

from the ensemble of the wave-propagation estimates obtained 10


for each individual time-frequency cell (gray scale): white and black
colors indicate normalized values of 0 and 1, respectively; theo- 1
retical dispersion curves for the fundamental, first, and second
Rayleigh-wave modes, black lines. (d) Normalized histogram dis-
tributions of phase velocities for north–south component derived 1
Alpha

from the ensemble of the wave-propagation estimates obtained (b)


0.5
for each individual time-frequency cell (gray scale): white and black
colors indicate normalized values of 0 and 1, respectively; theore-
tical dispersion curves for the fundamental, first, and second Ray- 1200 (c)
leigh-wave modes, black lines; theoretical dispersion curves for the
V (m/s)

fundamental, first and second Love-wave modes, dashed black 800


lines; theoretical group velocity for the fundamental Love mode,
dash-dotted black lines. Vertical black lines display plus and minus 400
standard deviation of the H/V ratio peak frequency calculated at the
central receiver. 0
1200 (d)
NS (m/s)

damental resonance. In the case of high-impedance contrast 800


(Zc > 4), there is a good agreement between the H/V ratio
400
peak frequency (called Fh=v hereafter), the fundamental re-
sonance frequency (called Fo hereafter), and the ellipticity 0
peak frequency of the fundamental Rayleigh wave mode 1 10
Frequency (Hz)
(called Fell hereafter) (parts a in Figs. 4 to 9). Although
the CVFK method suffers huge limitations close to Fh=v
(due to the vanishing of the energy on the vertical compo- Figure 6. H/V ratio and array analysis results for the high-im-
nent; see Scherbaum et al., 2003), we can systematically pedance contrast model M3.1. See Figure 4 for legend.
294 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

(a) ticed for the M3.3 model at 1.8 Hz (Fig. 10a). The origin of
this peak cannot be deduced from our array analysis; it could
Amplitude

10
be due either to a higher Rayleigh-wave mode or an S-wave
resonance within a surficial layer of the soil profile (Fig. 10c,
1
d). Once again the presence of the fundamental Love mode
can be identified from the phase velocity estimates on hor-
1 izontal component (Figs. 10d and 11d), and the relative
Alpha

(b) proportion of Rayleigh waves (α) exhibit values between


0.5
0.1–0.2 at the H/V ratio peak frequency (Figs. 10b and 11b).
In case of low-impedance contrast (Zc ≤ 3), we can dis-
(c)
tinguish two different situations. In the first case, model
V (m/s)

1200
M2.3, there is no ellipticity peak, but the H/V curve exhibits
800 one peak whose frequency (Fh=v ) agrees with Fo (both in
400 frequency and amplitude) (Fig. 12a). The phase velocity es-
0 timates could correspond to those of fundamental Rayleigh-
(d) wave mode, fundamental Love-wave mode and/or S waves
within the bedrock (Fig. 12d). It is thus quite difficult to iden-
NS (m/s)

1200
tify clearly which waves are responsible for the H/V peak in
800
such a situation. However, because the H/V peak amplitude
400 matches the theoretical amplification factor, it strongly sug-
0 gests that the H/V ratio is due to an S-wave resonance. In the
1 10
Frequency (Hz) second case, model M4.4, the fundamental Rayleigh-wave
mode exhibits an ellipticity peak but there is a large discre-
pancy between Fh=v and Fell , while there is a good agreement
Figure 7. H/V ratio and array analysis results for the high-im- between Fh=v and Fo (deviation less than 20%) (Fig. 13a).
pedance contrast model M3.2. See Figure 4 for legend. Again, at the H/V ratio peak frequency, the Rayleigh-wave
phase velocity estimates do not fit the theoretical phase ve-
locities of the fundamental Rayleigh-wave mode (Fig. 13c),
while on the horizontal component, phase velocities agree to
identify the presence of the fundamental Rayleigh mode in the theoretical dispersion curve of fundamental Love-wave
the vertical noise wave field from Fh=v up to the maximum mode (Fig. 13d). The relative proportion of Rayleigh waves
computed frequency (parts c in Figs. 4 to 9). At higher fre- (α) exhibits values between 0.1 and 0.2 (Fig. 13b) at the H/V
quency, we can also note the presence of higher modes, peak frequency.
which have, however, no influence on the fundamental
H/V peak. Such higher modes may have an influence in case
of multiple H/V peaks, this is a topic out of the scope of the
Discussion
present article. Note also that phase-velocity estimates ob- The two initial hypotheses on the origin of the H/V ratio
tained by using the horizontal component fit the fundamental peak discussed in the beginning of the present article were
Love mode rather well (Figs. 4d to 9d). This is explained by (1) the S-wave resonance within the sedimentary fill (Naka-
the relative proportion of Rayleigh waves (α) exhibiting val- mura, 1989, 2000), and (2) the horizontal polarization of the
ues of about 0.2 at the H/V ratio peak frequency (Figs. 4b to Rayleigh waves (ellipticity peak) (Lachet and Bard, 1994;
9b), strongly indicating that noise synthetics are dominated Kudo, 1995; Bard, 1998; Bonnefoy-Claudet, Cornou, et. al.,
by Love waves in the horizontal plane. This might be a bias 2006). The in-depth investigation of noise wave-field struc-
related to the simulation method, but such a domi- ture at the H/V peak frequency shows that the second hypoth-
nance has also been actually observed at several sites in esis (Rayleigh wave ellipticity) is fulfilled in the case of high-
Europe by Köhler et al. (2006) impedance contrast, but is not always satisfied in the case of
In case of moderate-impedance contrast (3 < Zc ≤ 4), low or moderate contrast. The CVFK analysis performed on
there is a discrepancy between Fh=v and Fell , suggesting that horizontal components, and the low values of the relative
the H/V ratio peak is not due to the fundamental Rayleigh proportion of Rayleigh waves α (from 0.05 to 0.2), reveal
wave mode (Figs. 10a and 11a). Moreover, at low frequency, the systematic presence of fundamental Love mode in the
the phase velocity estimates do not fit the theoretical phase simulated ambient noise wave field (up to 85%).
velocities of the fundamental Rayleigh-wave mode, suggest- In the case of moderate- and high-impedance contrasts,
ing that the fundamental Rayleigh mode is not responsible the Airy phase frequency of fundamental Love mode
for the H/V ratio peak frequency. Nevertheless, there is a matches the H/V ratio peak frequency as shown by frequency
good agreement between Fh=v and Fo (deviation less than dependent group velocity of Love waves (Figs. 4 to 11).
20%). A second, small-amplitude H/V ratio peak can be no- Because the Airy-phase frequency of Love waves corre-
Effects of Love Waves on Microtremor H/V Ratio 295

(a)

Amplitude
10

Alpha 1
(b)
0.5

1200 (c)
V (m/s)

800

400

0
1200 (d)
NS (m/s)

800

400

0
2 10
Frequency (Hz)

Figure 8. H/V ratio and array analysis results for the high-impedance contrast model M4.1. See Figure 4 for legend.

sponds to a bump of energy on both horizontal components, is a key parameter controlling the origin of the H/V ratio
the H/V ratio of the Fourier amplitude spectra may thus be peak. This origin is then not unique and involves Ray-
enhanced by this strong energy on the horizontal component. leigh-wave ellipticity, Love-wave Airy phases, and/or the
In other words, it would mean that the Airy phase of Love S-wave resonance. Nevertheless, whatever its origin, the
waves may even be responsible for the H/V ratio. H/V ratio peak frequency provides a good estimation of
To check this hypothesis, we compare, for every soil the resonance frequency, at least for the 1D media shown
model considered in this study, the frequencies of the H/V in this study.
ratio peak (Fh=v ) with the ones of the theoretical fundamental The estimation of the amplification using ambient noise-
resonance Fo (Fig. 14a), the fundamental Rayleigh mode el- based method is still a challenge in the earthquake engineer-
lipticity Fell (Fig. 14b) and the Airy phase of fundamental ing community. Nakamura (1989) claims that H/V peak
Love mode FAiry (Fig. 14c). The largest deviation observed amplitude provides a consistent estimate of the site amplifi-
between the H/V ratio peak frequency and the 1D theoretical cation function, while others (Lachet and Bard, 1994; Kudo,
resonance frequency does not exceed 20%. The Rayleigh- 1995; Bard, 1998; Fäh et al., 2001; Haghshenas, 2005; Bon-
wave ellipticity explains the H/V ratio peak only in case nefoy-Claudet, Cornou, et al., 2006) claim the opposite. To
of high-impedance contrast and only in some moderate- evaluate the robustness of the H/V method to estimate the site
impedance contrast cases (Fig. 14b). In case of low- amplification factor, we compare, for each soil model, the
impedance contrast, the ellipticity of Rayleigh waves cannot H/V ratio peak amplitude and the theoretical site amplifica-
explain the H/V peak ratio. Figure 14c indicates that the Airy tion (Fig. 15a). Except for the low-impedance cases, the H/V
phase of the Love-wave hypothesis is fulfilled only in the ratio peak amplitude observed from seismic noise synthetics
case of high or for some moderate contrasts, while it is always overestimates the site amplification factor. This
not satisfied in the case of a low-impedance contrast. These observation contradicts previous field results showing that
results strongly suggest that the S-wave impedance contrast actual H/V ratio peaks amplitude underestimates the ampli-
296 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

(a) (a)

Amplitude
Amplitude

10 10

1 1

1 1
Alpha

Alpha
(b) (b)
0.5 0.5

1200 (c) 1200 (c)


V (m/s)

V (m/s)
800 800

400 400

0 0
1200 (d) 1200 (d)
NS (m/s)

NS (m/s)
800 800
400 400
0 0
2 10 2 10
Frequency (Hz) Frequency (Hz)

Figure 9. H/V ratio and array analysis results for the high-im- Figure 11. H/V ratio and array analysis results for the moderate-
pedance contrast model M4.2. See Figure 4 for legend. impedance contrast model M4.3. See Figure 4 for legend.

(a) (a)
Amplitude
Amplitude

10 10

1 1

1 1
Alpha
Alpha

(b) (b)
0.5 0.5

(c) 1200 (c)


1600
V (m/s)
V (m/s)

800
800
400

0 0
(d) 1200 (d)
NS (m/s)
NS (m/s)

1600
800
800
400

0 0
1 8 2 10
Frequency (Hz) Frequency (Hz)

Figure 10. H/V ratio and array analysis results for the moderate- Figure 12. H/V ratio and array analysis results for the low-im-
impedance contrast model M3.3. See Figure 4 for legend. pedance contrast model M2.3. See Figure 4 for legend.
Effects of Love Waves on Microtremor H/V Ratio 297

(a) ambient noise synthetics at the central receiver of each soil


model, keeping the same source geometrical layout as shown
Amplitude

10
in the section Noise Data Sets (see Fig. 2), and using differ-
ent sources characteristics—with purely vertical point forces
1
(generation of P, SV, and Rayleigh waves) (Fig. 15b), and
purely radial point forces (generation of SV, P and Rayleigh
1 waves) (Fig. 15c). We observed lower H/V ratio peak ampli-
Alpha

(b) tudes in the cases of vertical and radial sources than in cases
0.5
of random sources (Fig. 15c). Because no Love waves are
excited by using such radial and vertical forces, these results
1200 (c) suggest a large influence of Love waves on the H/V peak am-
V (m/s)

800 plitude, at least in our simulation models. Moreover, recent


studies show that most of the energy in the actual noise wave
400 field is propagated by Love waves: at least 50% in Bonnefoy-
0 Claudet, Cotton, et al. (2006), and more than 70% in Köhler
et al. (2006), who measured the relative proportion of Ray-
1200 (d)
leigh waves (α) at different European sites. Therefore, the
NS (m/s)

800 relative proportion of Love waves in ambient noise wave


field may control the amplitude of the H/V peak, as already
400
discussed in Arai and Tokimatsu (2000), who also show the
0 influence of Love waves on the shape of the H/V ratio curves.
2 10
Frequency (Hz)
Conclusions
Figure 13. H/V ratio and array analysis results for the low-im- The H/V and the conventional f-k and 3C-MSPAC meth-
pedance contrast model M4.4. See Figure 4 for legend.
ods have been applied to 1D seismic noise synthetics com-
puted at the surface of various horizontally layered media
using Hisada (1994, 1995) codes, which efficiently compute
the displacement and stress of dynamic Green’s functions for
fication given by the site-to-reference earthquake method viscoelastic-layered half-space. At the H/V ratio peak fre-
(e.g., Haghshenas [2005]). Note that in the case of the quency, the relative contribution of different types of seismic
site-to-reference earthquake method, 1D and 2D/3D site ef- waves (body and surface waves) to the simulated ambient
fects are included in the amplification estimate, while this noise wave field strongly depends on the source character-
study focuses on 1D site effects only. Moreover, in our simu- istics (especially on the direction of source force amplitude)
lation, the direction of point source forces is randomly dis- and on the impedance contrast between sediment and bed-
tributed, and such randomness could be different from actual rock. (1) For high-impedance contrast (Zc > 4), the simu-
field ambient vibration sources. Therefore, we simulate new lated seismic noise wave field is mainly composed by

5
(a) (b) (c)
4
3
Fhv (Hz)

0.5
0.5 1 2 3 4 5 0.5 1 2 3 4 5 0.5 1 2 3 4 5
Fo (Hz) Fell (Hz) Fairy (Hz)

Figure 14. H/V ratio peak frequency and standard deviation plotted against (a) the 1D theoretical resonance frequency, (b) the ellipticity
peak frequency of the fundamental Rayleigh mode, and (c) the Airy phase frequency of fundamental Love mode. Results are shown for each
model according to the values of the impedance contrast: high, circles; moderate, diamonds; low, squares.
298 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

(a) Random sources (b) Vertical sources (c) Radial sources


24

18
Ahv
12

0
0 2 4 6 0 2 4 6 0 2 4 6
Ao Ao Ao

Figure 15. H/V ratio peak amplitude and standard deviation plotted against the site amplification given by the 1D theoretical transfer
function. H/V curves are calculated for different sets of noise synthetics computed with various directions of point source forces: (a) randomly
generated in 3D space, (b) vertical, and (c) radial. Results are shown for each model according to the values of the impedance contrast: high,
circles; moderate, diamonds; low, squares.

both fundamental Rayleigh- and Love-wave modes. (2) For case, this simulation work indicates that the H/V method
moderate-impedance contrast (3 < Zc ≤ 4), the fundamental is not reliable for providing a correct estimation of the
Rayleigh-wave mode exists, but is not predominant, while site amplification, at least when ambient vibrations are re-
the fundamental Love mode strongly dominates the simu- lated with local surfaces sources (i.e., for microtre-
lated wave field. (3) For low-impedance contrast (Zc ≤ 3), mors [f > 1 Hz]).
the contribution of the fundamental Rayleigh mode to the It is worth concluding by emphasizing the possibilities
seismic noise wave field is rather small, and the simulated and limitations of the H/V method for horizontally stratified
wave field is mainly composed by fundamental Love-wave media that the method always provides a good estimate of
mode and S waves. the resonance frequency while, in most cases (when surface
Although the origin of the H/V ratio peak is not unique waves dominate the ambient noise wave field), the method
and could be explained by the Rayleigh wave ellipticity and/ fails to provide the correct amplification factor. To go ahead
or the Airy phase of Love waves and/or the S-wave reso- and prolongate this work, it is now necessary to perform
nance, we show the robustness of the H/V peak frequency similar analysis not only on more complex media such as
to provide a good estimate of the 1D theoretical resonance 2D/3D structures (e.g., sedimentary basin, rough or sloping
frequency, whatever the impedance contrast (at least for interfaces between sediment and bedrock; see Cornou,
the 1D soil structures presented here). 2005; Guillier et al., 2006; Roten et al., 2006), but also
On the other hand, this study points out the influence of on real ambient noise recordings (Köhler et al., 2006).
the direction of point source force on the H/V peak ampli-
tude. When sources are vertical or radial, the simulated
H/V peak amplitude is significantly smaller than when con- Acknowledgments
sidering random sources. This clearly demonstrates a large We thank the participants of the SESAME project, especially F.
influence of the relative proportion of Love waves in the si- Cotton, for helpful discussions, comments, and suggestions. We thank
mulated ambient noise wave field on the H/V peak amplitude, M. Ohrnberger for providing the source code for array software. This work
was partly supported by the European Union research program Energy,
and simultaneously indicates that the simulation method
Environment, and Sustainable Development (European Community Con-
should be carefully worked out to include a realistic propor- tract Number EVG1-CT-2000-00026). Most of the computations were per-
tion of vertical, radial, and tangential forces to reproduce the formed at the Service Commum de Calcul Intensif de l’Observatoire de
reality. It may well be that the present assumption of random Grenoble, France.
sources with random angles is overemphasizing the role of
Love waves, but answering this question requires a much
larger number of carefully measured sites with 3C sensors References
to reliably derive the actual Rayleigh-to-Love proportion Aki, K. (1957). Space and time spectra of stationary stochastic waves, with
(which may also be different for long-period, natural micro- special reference to microtremors, Bull. Earthquake Res. Inst. 35, 415–
seisms and short-period, man-made microtremors). In any 457.
Effects of Love Waves on Microtremor H/V Ratio 299

Arai, H., and K. Tokimatsu (2000). Effects of Rayleigh and Love waves on Haghshenas, E. (2005). Conditions géotechniques et aléa sismique local à
microtremor H/V spectra, in 12th World Conf. on Earthquake Engi- Téhéran, Ph.D. Thesis, University Joseph Fourier, Grenoble, France,
neering, Auckland, New Zealand, paper 2232. 288 pp. (in French with English abstract).
Asten, M. (2006). Site shear velocity profile interpretation from microtremor Hauksson, K., T. Teng, and T. L. Henyey (1987). Results from a 1500 m
array data by direct fitting of SPAC curves, in 3rd Int. Symposium on deep, three-level downhole seismometer array: site response, low Q
the Effects of Surface Geology on Seismic Motion, Grenoble, France, value, and fmax, Bull. Seismol. Soc. Am. 77, 1883–1904.
paper 99. Hisada, Y. (1994). An efficient method for computing Green’s functions for
Atakan, K., P.-Y. Bard, F. Kind, B. Moreno, P. Roquette, and A. Tento, and a layered half-space with sources and receivers at close depths, Bull.
the , and SESAME-Team (2004). J-SESAME: a standardized software Seismol. Soc. Am. 84, no. 5, 1456–1472.
solution for the H/V spectral ration technique, in Proc. of the 13th Hisada, Y. (1995). An efficient method for computing Green’s functions for
World Conf. on Earthquake Engineering, paper 2270, Vancouver, a layered half-space with sources and receivers at close depths (part 2),
Canada, August 2004. Bull. Seismol. Soc. Am. 85, no. 4, 1080–1093.
Bard, P.-Y. (1998). Microtremor measurements: a tool for site effect estima- Jongmans, D., K. Pitilakis, D. Demanet, D. Raptakis, J. Riepl, C. Horrent, G.
tion?, in 2nd Int. Symp. on the Effects of Surface Geology on Seismic Tsokas, K. Lontzetidis, and P.-Y. Bard (1998). EURO-SEISTEST:
Motion, Yokohama, Japan, 3, 1251–1279. determination of the geological structure of the Volvi basin and vali-
Bard, P.-Y., and the , and SESAME-Team (2005). Report D23.12, Guide- dation of the basin response, Bull. Seismol. Soc. Am. 88, no. 2, 473–
lines for the implementation of the H/V spectral ratio technique on 487.
ambient vibrations measurements, processing and interpretation in Köhler, A., M. Ohrnberger, and F. Scherbaum (2006). The relative fraction
European Commission: Research General Directorate, Project No. of Rayleigh and Love waves in ambient vibration wave fields at dif-
EVG1-CT-2000-00026, SESAME, 62 pp.: available online at http:// ferent European sites, in 3rd Int. Symposium on the Effects of Surface
sesame‑fp5.obs.ujf‑grenoble.fr (last accessed January 2008). Geology on Seismic Motion, Grenoble, France, paper 83.
Bettig, B., P. Y. Bard, F. Scherbaum, J. Riepl, F. Cotton, C. Cornou, and D. Köhler, A., M. Ohrnberger, F. Scherbaum, M. Wathelet, and C. Cornou
Hatzfeld (2001). Analysis of dense array noise measurements using the (2007). Assessing the reliability of the modified three-component
modified spatial auto-correlation method (SPAC): application to the spatial autocorrelation technique, Geophys. J. Int. 168, no. 2,
Grenoble area, Boll. Geof. Teorica Appl. 42, nos. 3–4, 281–304. 779–796.
Bonnefoy-Claudet, S. (2004). Nature du bruit de fond sismique: implications Konno, K., and T. Ohmachi (1998). Ground-motion characteristics estimated
pour les études des effets de site, Ph.D. Thesis, University Joseph from spectral ratio between horizontal and vertical components of
Fourier, Grenoble, France, 241 pp. (in French with English abstract). microtremor, Bull. Seismol. Soc. Am. 88, no. 1, 228–241.
Bonnefoy-Claudet, S., C. Cornou, P.-Y. Bard, F. Cotton, P. Moczo, J. Kris- Kudo, K. (1995). Practical estimates of site response: state-of-art report, in
tek, and D. Fäh (2006). H/V ratio: a tool for site effects evaluation: 5th Int. Conf. on Seismic Zonation, Nice, France.
results from 1D noise simulations, Geophys. J. 167, no. 2, 827–837. Kvaerna, T., and F. Ringdahl (1986). NORSAR Scientific Report 1-86/87,
Bonnefoy-Claudet, S., C. Cornou, J. Kristek, M. Ohrnberger, M. Wathelet, Stability of various f-k estimation techniques, in Semiannu. Tech.
P.-Y. Bard, P. Moczo, D. Fäh, and F. Cotton (2004). Simulation of seis- Summ., 1 October 1985–31 March 1986, Kjeller, Norway, 29–40.
mic ambient noise: I. results of H/V and array techniques on canonical Lachet, C., and P.-Y. Bard (1994). Numerical and theoretical investigations
models, in 13th World Conf. on Earthquake Engineering, Vancouver, on the possibilities and limitations of Nakamura’s technique, J. Phys.
Canada, paper 1120. Earth 42, no. 4, 377–397.
Bonnefoy-Claudet, S., F. Cotton, and P.-Y. Bard (2006). The nature of the Lermo, J., and F. J. Chavez-Garcia (1993). Site effects evaluation using spec-
seismic noise wave field and its implication for site effects studies: a tral ratios with only one station, Bull. Seismol. Soc. Am. 83, no. 5,
literature review, Earth Sci. Rev. 79, nos. 3–4, 205–227. 1574–1594.
Cornou, C. (2005). Reports D11.10 and D17.10, Set of noise synthetics for Lermo, J., and F. J. Chavez-Garcia (1994). Are microtremors useful in site
H/V and array studies from simulation of real sites and comparison for response evaluation?, Bull. Seismol. Soc. Am. 84, no. 5, 1350–1364.
test sites in European Commission: Research General Directorate, Pro- McMamara, D., and R. Buland (2004). Ambient noise levels in the
ject No. EVG1-CT-2000-00026, SESAME, 62 pp.: available online at continental United States, Bull. Seismol. Soc. Am. 94, no. 4, 1517–
http://sesame‑fp5.obs.ujf‑grenoble.fr (last accessed January 2008). 1527.
Di Giulio, G., A. Rovelli, F. Cara, R. M. Azzara, F. Marra, R. Basili, and A. Moczo, P., and J. Kristek (2002). Report D02.09, FD code to generate noise
Caserta (2003). Long-duration asynchronous ground motions in the synthetics, in European Commission: Research General Directorate,
Colfiorito plain, central Italy, observed on a two-dimensional dense Project No. EVG1-CT-2000-00026, SESAME, 31 pp.: available online
array, J. Geophys. Res. 108, 2486–2500. at http://sesame‑fp5.obs.ujf‑grenoble.fr (last accessed January 2008).
Fäh, D. (1997). Microzonation of the city of Basel, J. Seism. 1, no. 1, 87– Nakamura, Y. (1989). A method for dynamic characteristics estimation of
102. subsurface using microtremor on the ground surface, Q. Rept. Railw.
Fäh, D., F. Kind, and D. Giardini (2001). A theoretical investigation of aver- Tech. Res. Inst. 30, no. 1, 25–30.
age H/V ratios, Geophys. J. 145, no. 2, 535–549. Nakamura, Y. (1996). Real-time information systems for hazards mitigation,
Field, E., and K. Jacob (1993). The theoretical response of sedimentary in 11th World Conf. on Earthquake Engineering, Acapulco, Mexico.
layers to ambient seismic noise, Geophys. Res. Lett. 20, no. 24, Nakamura, Y. (2000). Clear identification of fundamental idea of Naka-
2925–2928. mura’s technique and its applications, in 12th World Conf. on Earth-
Fukushima, Y., S. Kinoshita, and H. Satho (1992). Measurement of Q for quake Engineering, Auckland, New Zealand.
S-wave in mudstone at Chikura, Japan: comparison of incident and Nogoshi, M., and T. Igarashi (1971). On the amplitude characteristics of
reflected phases in boreholes seismograms, Bull. Seismol. Soc. Am. microtremor (part 2), J. Seismol. Soc. Japan 24, 26–40 (in Japanese
82, 148–163. with English abstract).
Gitterman, Y., Y. Zaslavsky, A. Shapira, and V. Shtivelman (1996). Empiri- Ohrnberger, M. (2004). Report D18.06, User manual for software package
cal site response evaluations: case studies in Israel, Soil Dyn. Earth- CAP: a continuous array processing toolkit for ambient vibration array
quake Eng. 15, no. 7, 447–463. analysis, in European Commission: Research General Directorate, Pro-
Guillier, B., C. Cornou, J. Kristek, S. Bonnefoy-Claudet, P.-Y. Bard, D. Fäh, ject No. EVG1-CT-2000-00026, SESAME, 83 pp.: available online at
and P. Moczo (2006). Simulation of seismic ambient vibrations: does http://sesame‑fp5.obs.ujf‑grenoble.fr (last accessed January 2008).
the H/V provide quantitative information on 2D–3D structure, in 3rd Ohrnberger, M., E. Schissele, C. Cornou, S. Bonnefoy-Claudet, M. Wathe-
Int. Symposium on the Effects of Surface Geology on Seismic Motion, let, A. Savvaidis, F. Scherbaum, and D. Jongmans (2004). Frequency
Grenoble, France, paper 153. wavenumber and spatial autocorrelation methods for dispersion curve
300 S. Bonnefoy-Claudet, A. Köhler, C. Cornou, M. Wathelet, and P.-Y. Bard

determination from ambient vibration recordings, in 13th World Conf. Wang, Z., R. Street, and E. Woolery (1994). Qs estimation for unconsoli-
on Earthquake Engineering, Vancouver, Canada, paper 946. dated sediments using first-arrival SH critical refractions, J. Geophys.
Ohrnberger, M., E. Schissele, C. Cornou, M. Wathelet, A. Savvaidis, F. Res. 99, no. B7, 13,543–13,552.
Scherbaum, D. Jongmans, and F. Kind (2004). Microtremor array
measurements for site effect investigations: comparison of analysis
methods for field data crosschecked by simulated wave fields, in
Institut de Radioprotection et de Sureté Nucléaire
13th World Conf. on Earthquake Engineering, Vancouver, Canada,
BP 17
paper 940.
92262 Fontenay-aux-Roses, Cedex, France
Okada, H. (2003). The microtremor survey, in Society of Exploration Geo- (S.B.)
physicists, American Geophysical Monograph, 13, 135 pp.
Roten, D., D. Fäh, C. Cornou, and D. Giardini (2006). Two-dimensional
resonances in Alpine valleys identified from ambient vibration wave Institute of Geosciences
fields, Geophys. J. Int. 165, no. 3, 889–905. Potsdam University
Scherbaum, F., K.-G. Hinzen, and M. Ohrnberger (2003). Determination of Karl-Liebknecht-Str. 24
shallow shear wave velocity profiles in the Cologne, Germany area Haus 27
using ambient vibrations, Geophys. J. Int. 152, no. 3, 597–612. 14476 Potsdam, Germany
(A.K.)
Seekins, L. C., L. Wennerberg, L. Marghereti, and H.-P. Liu (1996). Site
amplification at five locations in San Francisco, California: A compari-
son of S waves, codas, and microtremors, Bull. Seismol. Soc. Am. 86, Laboratoire de Géophysique Interne et de Tectonophysique, Institut de
no. 3, 627–635. Recherche et de Développement, Laboratoire Central des
Tokeshi, J. C., and Y. Sugimura (1998). On the estimation of the natural Ponts-et-Chaussées
period of the ground using simulated microtremors, in 2nd Int. Symp. Maison des Géosciences
on the Effects of Surface Geology on Seismic Motion, Yokohama, Joseph Fourier University
Japan, 2, 651–664. BP 53
38041 Grenoble Cedex 9, France
Wakamatsu, K., and Y. Yasui (1996). Possibility of estimation for amplifica-
(C.C., M.W., P.-Y.B.)
tion characteristics of soil deposits based on ratio of horizontal to ver-
tical spectra of microtremors, in 11th World Conf. on Earthquake
Engineering, Acapulco, Mexico. Manuscript received 14 March 2007

You might also like