J. Phys.

Earth, 42, 377-397, 1994

Numerical and Theoretical Investigations on the

Possibilities and Limitations of Nakamura's
Technique

Corinne Lachetl 1,* and Pierre-Yves Bard 1'2

1 Laboratoire de G_??_ophysique Interne et Tectonophysique

,
Observatoire de Grenoble, IRIGM-BP 53X,
F-38041 Grenoble cedex, France
2 Laboratoire Central des Ponts et Chaussees
,
58 Bd Lefebvre, F-75732 Paris cedex 15,
France

Numerical simulation of noise is used to investigate the characteristics of the spectral

ratio between horizontal and vertical components (H/V ratio) and its sensitivity to
various parameters in order to better appreciate the reliability of the technique proposed
by Nakamura (1989) to estimate site amplification effects from single station noise
recordings. Noise is simulated as the signal produced at a single site by a set of superficial
sources (unidirectional forces or dipoles) disposed all around with random amplitude
and time delay. Individual signals from a single source are computed by the discrete
wave number technique.
Synthetic calculations for 15 soil profiles show that this ratio exhibits a single, clear
peak, the location of which is independent of the source excitation function, but strongly
correlated with the local geological structure: its frequency is very close to the S wave
resonance frequency. This peak appears to be mainly controlled by the polarization
curve of the fundamental Rayleigh waves, which in turn exhibits a sharp peak around
the fundamental resonance mode of the sedimentary structure. A similar result is found
for the H/V ratio obtained for incident plane SV waves.
In contrast, the amplitude of this peak exhibits a poor correlation with the ground
motion amplification of S waves at resonance frequency. It is shown to be related with
a high sensitivity on the value of the Poisson's ratio in the uppermost layer presumed
to be the noise source layer, and, though to a much lesser extent, on the mean distance
between site and noise sources.
It is concluded that Nakamura's method can clearly allow the resonance frequency
of a given sedimentary site to be measured very efficiently and very cheaply, but that
its use for deriving the amplification at resonance frequency seems still premature from
a theoretical point of view.

Received January 20, 1994; Accepted September 26, 1994

* To whom correspondence should be addressed .

377
378 C. Lachet and P.-Y. Bard

1. Introduction

Several recent studies of microtremor recordings have been made to investigate the
spectral ratio between the horizontal and vertical components (e.g., Nakamura, 1989).
In his study, Nakamura (1989) suggests that this ratio is a good technique for site effect
evaluation since it shows a peak which corresponds to the predominant frequency of
the soil. It has also been shown that the predominant frequency of microtremors is
dependent on the type of soil considered (rock, alluvium, etc.) (Bouckovalas and Krikeli,
1991; Finn, 1991). This clearly points to the fact that we could use microtremors for
microzonation studies. Moreover, knowledge of the resonance frequency of the soil
could be used in predicting the kinds of buildings which are likely to suffer the greatest
damage (Ohmachi et al., 1991). This method seems to be very suitable for site effect
evaluation in urban areas, since it requires only noise recorded by one three-component
station. Some recordings made in Mexico City, Oaxaca and Acapulco (Lermo and
Chavez-Garcia, 1993) and in France (Duval, 1992) provided encouraging results.
However, no reliable theoretical basis has yet been proposed to support this method.
The aim of this numerical study is to check the influence of different parameters on the
horizontal to vertical spectral ratio (henceforth referred to as the H/V ratio). For
example, source effects are likely to have a major impact on the response of a site. It
is then necessary to check if the H/V ratio obtained from microtremor recordings tells
us more about the response characteristics of the site than about the excitation function
of the source. Moreover, unlike earthquakes, the waves associated with microtremors
are generated by surface sources such as wind, road traffic, trains, machine vibrations,
etc. These waves consist of a mixture of surface and body waves (P and S), which may
not have traveled through the whole sedimentary structure. Furthermore, the stresses
involved in microtremors are very low compared with those that produce earthquakes.
One of the main goals of this study is also to investigate the sensitivity of the H/V ratio
to the local geological conditions.
We use a method of numerical simulation of urban noise to better understand the
meaning of the H/V ratio and the influence of different parameters. For this, a multiple
source model is used (Lachet, 1992) to numerically simulate urban noise. This work
can be divided into two main parts. We first looked at the position of the H/V peak
and its relation to the resonance frequency for different source types and for varying
geological structures. We also compare the results obtained with different types of
incident waves. The second point of the study was to investigate the amplitude of the
H/V peak. We try here to find which parameters have a control on this amplitude to
see if the amplitude of the H/V peak is reliable enough to be used in amplification studies.

2. Methods Used to Investigate the H/V Ratio

2.1 Simulation of urban noise

The noise recorded in towns can be considered to be caused by a set of surface
sources, randomly arranged and with varying amplitude. These sources are of different
origin, for example wind, cars, trains, road works and factories, etc. It is therefore very
difficult to simulate urban noise in a deterministic way.

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 379

Fig.1. Disposition of the multiple-source model use d for noise simulation.

Fig. 2. Source and receiver dispositions, As these two dispositions are equivalent,
we can directly use the results from the program Axitra (for one source and
several receivers), to calculate synthetic seismograms for the multiple source
model.

We choose to model the noise as a set of uniformly disposed surface sources with

random amplitude placed around a central receiver, according to the model displayed

in Fig. 1. In this study we employed 240 or 480 sources, dividing the disk into 24 slices,

with a radius varying from 100 m to 1 km. This model allows us to use a simple

arrangement of one single point source and a line of receivers, instead of a single station

and several sources, as shown in Fig. 2. We can then use the synthetic seismograms

obtained from the numerical simulation program described later. The method used

consists of a summation for each of the three components to obtain the signal recorded

by the central receiver. Four parameters must be taken into account in the calculation

to simulate the random nature of urban noise:

Different types of sources (k) are used,

•E such as explosion, unidirectional forces,

with different source functions, shown in Fig. 3.

Vol. 42 , No. 5, 1994

 380 C. Lachet and P.-Y. Bard

Fig. 3. Shape of the source functions used; both in time and frequency domains.

•E We introduce a random difference in phase, ƒÓijk, between zero and the total

duration of the seismogram (tl), for each source;

The differences in amplitude are simulated

•E by a random factor, Rijk, between 0

and 1;

•EA coefficient Ai is used to take into account the probability of having a source

inside each sector. This coefficient is proportional to the area of each sector. Clearly,

the outer sectors of the disk are bigger than the inner ones, therefore there is a higher

probability of a source in one of the outer sectors. We define Ai = (area of the i

sector)/(average area of the sectors).

Then, considering Utki,j, Urki,j, and Uvki,j as the tangential, radial, and vertical

components from a source in any i,j sector, respectively, we can write the three
components of the central receiver as the sums:

with nr being the number of sources in one slice; that is to say the number of rings of

the disk (10 or 20 in our case).

This procedure is repeated ten times, by assigning random values to the factors

ijk and Rijk to obtain several synthetic recordings. ƒÓ

2.2 Numerical simulation technique

For the numerical calculations, we use the program Axitra (Coutant, 1989) which
generates synthetic seismograms, using the discrete wave number method (Bouchon,
1981). This program calculates the response of a horizontally stratified medium, for
several receivers and one point source, situated respectively at arbitrary locations. The

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 381

only condition on the positions is that the source and receivers must not be at the same
depth. The response of the structure is then convolved with a chosen source function.
The source is defined by both its type (explosion, crack, dislocation, unidirectional force,
etc.) and its shape (step, Ricker, Dirac, triangle, etc.). For each calculation, it is necessary
to specify the duration of the seismogram (tl) and the number of frequencies calculated
(nfreq);thus the maximum frequency is given by fmax= nfreq/tl.
For this study, the source is generally situated at a depth of 2 m to simulate the
noise generated by human and industrial activity in towns.

2.3 Computation of the horizontal to vertical spectral ratio

The technique developed by Nakamura (1989) is based on an estimation of the
transfer function using microtremors. Usually, the transfer function of surface layers
is given by TF = SHS/SHB,with SHSbeing the horizontal microtremor spectrum at the
surface and SHBthe horizontal microtremor spectrum on the substratum. Considering
that artificial noise is not only propagated as body waves, but comprises an important
part of Rayleigh waves, it is necessary to make a correction to remove the effect of
surface waves. Nakamura assumes that the effect of Rayleigh waves is included in the
vertical spectrum at the surface (SVS)and not at the base ground (SVB), then it could
be defined as ES = SVS/SVB.Assuming also that the effect of Rayleigh waves is equal
for the vertical and horizontal components, Nakamura gives the new transfer function
as TFF= (SHB/SVS)/(SHB/SVB). Three recordings gave results where the ratio SHB/SVBis
nearly 1 for a relatively wide frequency range. Thus, Nakamura considers the horizontal
to vertical spectral ratio for microtremors as an estimation of the transfer function for
body waves.
Figure 4 displays schematically how we calculate here the H/V ratio from synthetic
seismograms. The example shown is generated with the multiple source model described
before, to obtain a three component synthetic noise recording.
We calculate the fast Fourier transform of the three components and smooth the
spectra obtained. We have used two types of smoothing function. The first one was
applied on a linear frequency scale. For different calculations with different geological
structures, the maximum energy is situated at different positions in the frequency domain,
so we have to take different smoothing widths for each case. The second smoothing
function is a triangle window applied on a logarithmic frequency scale. The advantage
is that the smoothing width is a function of the frequency, so we can keep it constant
(here equal to 0.1) for all cases. Then we take one of the two smoothed horizontal
spectra (or the average of the two, given by Sh- =(St + Sr)/2) and the vertical one,
and calculate their ratio.

3. Position of the H/V Ratio in the Frequency Domain

3.1 Variation of the source characteristics

The first part of this work consists of investigating the spectral shape of the H/V
ratio. Nakamura (1989) suggests that this ratio peaks at the S waves resonance frequency
of the layered structure (fs), and that the location of this peak (fn) does not depend on
the source characteristics.

Vol. 42, No. 5, 1994

382 C. Lachet and P.-Y. Bard

Fig.4. Procedure used to calculate the H/V ratio from synthetic seismograms.

We consider in this section a simple structure consisting of one sedimentary

horizontal layer over a homogeneous half-space. The physical characteristics of this
structure, S1, are given in Table 1. The multiple source model described above is used
to calculate the average H/V spectra for ten synthetic noise recordings. As described
above, the simulation program allows one to choose different source types and functions.
The same procedure is repeated for different source characteristics, so as to compare
the H/V ratios obtained. The first source type used was an explosion (represented by
a triple dipole), with two different functions, step and Ricker (Fig. 3). The spectra
obtained are displayed in Fig. 5. However, another source type, a unidirectional force,
given by its three directions (x, y, z), is probably more representative of the urban noise
than sources such as double-couples or explosions. The H/V ratios are thus calculated
for the same simple geological structure, with a source defined by a contribution of
unidirectional forces (in the three directions), and three types of functions: step, Ricker
and pseudo-Dirac. The results are displayed in Fig. 6.
The H/V ratios obtained for different source characteristics, all clearly exhibit a
peak whose position is constant regardless of the source type and source function. In
this particular simple case, it is situated at a frequency of approximately 10Hz. Based
on these results, for randomly distributed surface sources the H/V peak position is
independent of the source characteristics. Moreover, calculations made on different
geological structures with various sources (Lachet, 1993) confirm this result.

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 383

Table 1. Geological structure for the simple theoretical models.

3.2 Influence of the geological structure

We now investigate the influence of the geological structure on the H/V peak
position. Synthetic seismograms are thus computed with the multiple source model for
a number of different soil structures. A constant source type is chosen for this set of
calculations, characterized by unidirectional forces (radial, tangential, and vertical) and
a pseudo-Dirac function. Two types of structures are considered.
In the first step, six simple theoretical models are studied with varying velocity
contrast and thickness of the layers. The physical parameters for these models are given
in Table 1, and the computational parameters in Table 2. Figure 7 shows the average
of the spectra obtained from ten synthetic noise recordings for each of the six geological
structures described above.
In the second step, a similar procedure is applied for a few (more complex) real
geological structures from three different regions: Ashigara Valley (Japan), Turkey Flat
(California) and coastal sites located in SE France. Geotechnical surveys carried out

Vol. 42, No. 5, 1994

384 C. Lachet and P.-Y. Bard

explosion - step
explosion - Ricker

Fig. 5. Spectra for the simple structure S1, with an explosion source type and
step (left) and Ricker (right) source functions. Smoothing window=1 Hz.

force - step force - Ricker force - Dirac

Fig. 6. Spectra for the simple structure S1, with a unidirectional forces source
type, and step, Ricker and Dirac source functions (respectively from left to
right). Smoothing on a logarithmic scale, window =0.1. The dotted line gives
the standard deviation.

in several sites in these areas give precise physical characteristics of the geological
structures (Bard, 1993; Cramer and Real, 1992; Samano and Seo, 1992; Sawada, 1992;
Seo, 1992 for Ashigara Valley and Turkey Flat; Duval, 1992 for SE France). These
geotechnical data are given in Tables 3-5 and computational parameters in Table 2.
The H/V ratios obtained for each of these sites are shown in Fig. 8.
A simple look at the general shape of the H/V ratios obtained in these various
cases indicates that they present a clear peak situated at a varying frequency fn. A

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique385

Table 2. Computational parameters.

comparison of the results obtained for the different sites clearly shows the influence of
the geological structure on the H/V peak position.
The next step is to compare fn with the resonance frequency fs, corresponding to
the vertical S wave resonance. These frequencies fs were computed for each of the
geological structures with a simple 1-D program based on the reflectivity method
(Kennett and Kerry, 1979). Figure 9 provides a summary comparison between fs and
fn for all the geological structures presented before. It shows overall good agreement
between the peak positions obtained for noise and those obtained for vertical S waves,
from which we conclude that the fn frequency derived from H/V spectra is probably a
reliable indication of the fundamental resonance frequency fs of a horizontally layered
structure.

4. Interpretation

4.1 Polarization curves of Rayleigh waves

As described above, urban noise is believed to be generated by surface sources and
should therefore mainly consist of surface waves. In layered media like those considered
in this study, two main kinds of surface waves may exist, Love waves and Rayleigh
waves. Love waves are polarized only in the horizontal direction and therefore lead to
infinite H/V ratios. Rayleigh waves however, are polarized in both horizontal (radial)
and vertical directions, so that the peak observed in the H/V ratio may be related to
the polarization curve of Rayleigh waves. The polarization of Rayleigh waves was
therefore investigated for each of the models considered using a set of programs

Vol. 42, No. 5, 1994

386 C. Lachet and P.-Y. Bard

Site S1 Site S2 Site S3

Site S4 Site S5
Site S6

Fig.7. Spectra for the six theoretical structures. Smoothing on a logarithmic

scale, window =0.1. The dotted line gives the standard deviation.

(Herrmann, 1985) which calculate the dispersion values and eigenfunctions for a given
velocity model for Love or Rayleigh waves. The polarization curves obtained for three
simple structures (one layer over a half-space, models S1 to S3) are shown in Fig. 10.
The results for some of the more complex real sites (a few examples in each region:
Ashigara Valley, Turkey Flat, and SE France) are shown in Fig. 11. These polarization
curves exhibit several peaks corresponding to a vanishing of the vertical component.
It is interesting to compare the polarization curves obtained for each of the sites
considered with the H/V ratios calculated from noise simulation, for these same sites
(Figs. 7 and 8). Figure 12 displays a comparison between fn and the frequency of the
first peak on the polarization curve, fp: the H/V peak position in noise simulation
generally corresponds to the first peak in the polarization curves. In other words, the
shape of the H/V ratio is widely controlled by fundamental Rayleigh waves which, in
turn, are closely related with the resonance phenomena.

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 387

Table 3. Geological structure for the sites at Ashigara Valley.

Site KS1Site KS2

Site KR1

Table 4. Geological structure for

Site
the sites
V1 at Turkey Flat. Site V2

4.2 SV waves with varying incidence angle

Since for near sources (as on an urban scale) noise also comprises an important

contribution of body waves, it is also interesting to investigate the H/V ratio for body

waves. A 1-D program based on the reflectivity technique was used to compute the

surface response to SV waves with an incidence angle varying from 5•‹ to 89•‹, with a

step of 5•‹. The results are shown in Fig. 13 for the simple structure S1 and the site

KS2 at Ashigara Valley, whose geological structures are described in Tables 1 and 3,

respectively. These ratios exhibit rather stable peaks, the first of which corresponds to
the fundamental resonance frequency. This position is almost independent of the

incidence angle (except in the immediate vicinity of the critical angle).

It is also interesting that, for body waves, the H/V ratio also shows several peaks

Vol. 42, No. 5, 1994

388 C. Lachet and P.-Y. Bard

Table 5. Geological structure for the sites at

SE FranceSite fv1 Site sbm1

Site sbm2

Site fv2

Site larv1

Site fv3

Site larv2

for each site situated at higher frequencies. A comparison with the polarization curves
of Rayleigh waves obtained for these two sites (see Figs. 11 and 12) demonstrated that
these peaks generally correspond to one another, and to the higher resonance modes
of the structure. These higher frequency peaks are not observed on the H/V ratios
derived from noise simulation. This may perhaps be explained by the fact that noise is
composed not only of Rayleigh and SV waves but also of Love and SH waves.

J. Phys. Earth
 Possibilities and Limitations of N akamura's Technique 389

a
Site KS1 Site KS2

Site KR2

b
Site V1 Site V2

Site l arv1 Site larv2

Site sbm1 Site sbm2

Site fv1 Site fv2

Site fv3

Fig. 8. H/V ratios for the sites at: (a) Ashigara Valley (Japan), (b) Turkey Flat
(California), (c) SE France. Smoothing on a logarithmic scale, window =0.1.
The dotted line gives the standard deviation.

Vol. 42, No. 5, 1994

c
390 C. Lachet and P.-Y. Bard

Fig. 9. Plot of the H/V peak frequency for each site; fn: H/V peak frequency for
noise simulation, fs: H/V peak frequency for vertically incident S waves. The
dashed line is the fs=fn line.

Fig. 10. Polarization curves of Rayleigh waves for the one layer structures. The
vertical axis gives the amplitude for the radial component normalized with
respect to the vertical one (Ur/Uz).

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 391

Fig. 11. Polarization curves of Rayleigh waves for the sites at Ashigara Valley,
Turkey Flat, and SE France. The vertical axis gives the amplitude for the
radial component normalized with respect to the vertical one (Ur/Uz).

Fig. 12. Plot of the H/V peak frequency for each site; fn: H/V peak frequency
for noise simulation, fp: frequency of the fundamental peak in the polarization
curves of Rayleigh waves. The dashed line is the fp =fn line.

Vol. 42, No. 5, 1994

392 C. Lachet and P.-Y. Bard

Fig. 13. H/V ratios obtained for SV waves with varying incidence angle, for one
simple structure (left) and one site at Ashigara Valley (right).

Fig. 14. Plot of the H/V peak amplitude for each site. With An being the H/V
peak amplitude for noise simulation and As the amplitude of the horizontal
component for vertically incident S waves.

5. Amplitude of the H/V Peak

5.1 Comparison of the amplitude for noise and vertical S waves

The second objective of this numerical study is to investigate the amplitude of the
H/V peak (An). Nakamura (1989) suggests that we can estimate the amplification of
seismic motion due to resonance of surface layers simply from the maximum spectral
amplitude of the H/V ratio. A first raw test of this strong assumption is provided
by comparing the amplitude (An) of the H/V peaks obtained from noise simulation with
the maximum amplitude of the transfer function for vertically incident S waves (As).
This is done in Fig. 14, where An and As are plotted for each of the geological structures
described above. Obviously, there is poor agreement between the H/V peak amplitude
and that for vertical S waves. For some cases (KS2, sbm1, sbm2, larv1, larv2, fv1), the
H/V peak amplitudes are much larger than the vertical S wave amplification. As these
sites correspond to large Poisson's ratios at the surface (see Tables 3-5), we investigated

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Technique 393

a b

Fig. 15. H/V ratios for the simple structure S2, with: (a) Varying Poisson's ratio
in the sedimentary layer. The S wave velocity is constant and equal to 300 m/s
whereas the P wave velocity, VP (sed), is variable. The H/V peak amplitude
varies between 4.75 and 17.5. (For a constant source depth of 2 m, and a
maximum source-receiver distance of 300 m), (b) Varying distance d, between
the farthest source and the central receiver. The H/V peak amplitude varies
between 4 and 8.7. (For a constant source depth of 2 m, and VP(sed) = 600 m/s),
(c) Varying source depth. The H/V peak amplitude remains between 7.5 and
9.5. (For a constant maximum source-receiver distance of 300 m, and VP
(sed) = 600 m/s.) The dotted line gives the standard deviation.

the variations of Poisson's ratio and source location to see if they influence the H/V
peak amplitude derived from noise simulation.

5.2 Influence of Poisson's ratio

The influence of Poisson's ratio in the sedimentary layer is first studied for the
simple structure S2 by varying the P wave velocity while fixing the S wave velocity to
300 m/s. The results obtained are displayed in Fig. 15(a), and show significant variations

Vol. 42, No. 5, 1994

c
 394 C. Lachet and P.-Y. Bard

a site fv1 b

Fig. 16. Spectra for the site fv1 (SE France);

(a) Original geological structure of the site.
(b) Introduction of a superficial 4-m-thick
layer with a Poisson's ratio of 0.25. The ampli-
tude shows a large decrease (from 35 to 4.5).
The dotted line gives the standard deviation.
(c)Plot of the H/V peak amplitude for noise simu-
lation, A,,, (for the sites KS2, fv1, sbm1, sbm2,
larv1, larv2; with their original geological struc-
ture in one case, and with a surface Poisson's
ratio of 0.25 in the other case) against the am-
plitude, As, for vertically incident S waves.

(between4.75and 17.5),as a function of the VP/VSratio (related with Poisson's ratio).

In the second step, the sites with a large Poisson's ratio (KS2, sbm1, sbm2, larv1,
larv2, fv1) were modifiedto set their Poisson's ratio to 0.25 at the surface. For this, a
4-m-thick layer, with a 0.25 Poisson's ratio, was introduced at the surface of each site.
Figure 16(a), (b) displays the results obtained for site fv1 (SE France); it is clear that
the amplitude of the H/V peak shows a large decrease(from 35 to 4.5) when Poisson's
ratio in the sourcelayer is reducedfrom 0.49to 0.25. The same phenomenonis observed
for the other sites as shown in Fig. 16(c),where the H/V peak amplitude for noise
simulation (A„)is plotted against that for vertically incident S waves. In the case of a
0.25surfacePoisson's ratio, the amplitude is much smaller,indicating the large influence
of Poisson's ratio in the surface layer on the H/V peak amplitude; although there is no
clear relation between these two quantities.

5.3 Influence of the source location

The results may also depend on the relative location of the source and receivers,

. Phys. Earth

Jc
 Possibilities and Limitations of Nakamura's Technique 395

namely on the ratio between the structure characteristic thickness, the maximum
source-receiver distance, and on the source depth which may influence the excitation
of the various surface waves (especially the higher modes). The effects of these parameters
are also investigated on the simple structure S2. First we change the maximum
source-receiver distance, that is to say the radius of the multiple-source disk. As we
have seen before, the H/V peak generally corresponds to the fundamental peak of the
polarization curves of Rayleigh waves. It is therefore likely that the greater the
source-receiver distance, the bigger the Rayleigh wave contribution, and thus the higher
the H/V peak amplitude. Figure 15(b) shows that the source-receiver distance has some
effect on the H/V peak amplitude (which varies between 4 and 9), but with no particular
correlation between the two parameters.
Finally, the H/V ratios obtained for varying source depths, displayed in Fig. 15(c),
show that the H/V peak amplitude remains nearly constant for different source depths.
These results indicate that the H/V peak amplitude can not be used in a
straightforward way for amplification studies since it undergoes considerable varia-
tion with respect to parameters such as Poisson's ratio in the sedimentary structure
and (to a lesser extent) the source-receiver distance.

6. Conclusions

In this numerical study, some aspects of the properties of the horizontal to vertical

spectral ratio applied to noise simulation were investigated. The main conclusions

inferred from these synthetic calculations may be summarized as follows:

The H/V ratios obtained from noise simulation show a peak whose position is

independent of the source excitation function.

H/V ratios obtained for obliquely •E incident SV waves show several peaks whose

position is independent of the incidence angle, and whose fundamental position

corresponds to the H/V peak for noise simulation.

The shape of the H/V ratio is largely

•E controlled by the polarization curve of

fundamental Rayleigh waves;

There is a good •Eagreement between the H/V peak positions derived from noise

simulation and those obtained for vertical S waves. This shows that the H/V ratio gives
a reliable indication of the resonance frequency of a horizontally layered structure.

In contrast, the amplitude of the H/V peak was shown to be very sensitive not

only to the velocity contrast, but also to parameters such as Poisson's ratio in the
sedimentary structure and the source-receiver distance. Its use for determining the

amplification of horizontal motion for incident S waves, therefore, still seems premature

from a strictly theoretical point of view.

This numerical study allows a better understanding of the H/V ratio and shows

this technique to be of great interest for site effect and microzonation studies, in an

urban context. This method represents a relatively cheap way of determining the

resonance frequency of a site, that is the frequency of the motion that buildings are

likely to undergo in the site considered. However, it is important to consider the limits

of the application of the H/V method in the case of very low resonance frequencies

corresponding to thick sedimentary structures (f0 < 0.5 Hz). Are surface sources, like

Vol. 42, No. 5, 1994

396 C. Lachet and P.-Y. Bard

noise, able to excite the structure at such a low frequency?

This study is concerned with the use of urban noise recordings, but it is interesting
to notice that the H/V method might also give good results in the case of weak or
strong motion recordings since the ratio is also stable for body waves. This theoretical
result is supported by various experimental observations (Lermo and Chavez-Garcia,
1993; Theodulidis and Bard, 1994). Finally, it is clear that the H/V ratio method has
to be tested on several real noise recordings to check the theoretical results obtained
here, and to find its limitations.

REFERENCES

Bard, P. Y., Comparaison aveugle theorie/observations sur les effets de site: synthese des rêsultats
des experiences de Turkey Flat (Californie) et Ashigara Valley (Japon), Genie Parasismique
et Aspects Vibratoires dans le Genie Civil, 3eme Colloque National AFPS, Saint-Remy-les-
Chevreuse (France), Vol. I, ES 48-58, 1993.
Bouchon, M., A simple method to calculate Green's functions for elastic layered media, Bull.
eismol. Soc. Am., 71, 959-971, 1981. S
Bouckovalas, G. and I. Krikeli, Effect of local soil stratigraphy on microtremor measurements,
Proceedings of the Second International Conference on recent advances in geotechnical
earthquake engineering and soil dynamics, Missouri, paper No. 8.21, 1245-1251, 1991.
Coutant, 0., Programme de simulation numerique Axitra (version 1 and 2), Rapport LGIT, 1989.
Cramer, C. H. and C. R. Real, A statistical analysis of submitted site-effects predictions for the
weak motion blind prediction test conducted at Turkey Flat, USA, site-effects test area near
Parkfield, California, International symposium, The effects of surface geology on seismic
motion, ESG, Odawara, Japan, 15-20, 1992.
Duval, A. M., Evaluation des effets de site a partir de l'enregistrement du bruit de fond en site
urbain, CETE Mediterannee, rapport interne, 1992.
Finn, W. D. L., Geotechnical engineering aspects of microzonation, Fourth International Con-
ference on Seismic Zonation, Stanford, California, I, 236-253, 1991.
Herrmann, R. B., Computer Programs in Seismology, Vol. III, Surface Waves in Plane Layers,
Saint Louis University, 1985.
Kennett, B. L. N. and N. J. Kerry, Seismic waves in a stratified half space, Geophys. J. R. Astron.
Soc., 57, 557-583, 1979.
Lachet, C., Faisabilite de l'utilisation du bruit de fond. Critique theorique de la methode de
Nakamura, Rapport de stage 2eme annêe de Magistere des Sciences de la Terre, 1992.
Lachet, C., Signification du rapport spectral entre le mouvement horizontal et le mouvement
vertical. Application a l'etude des effets de site par le bruit de fond, Mêmoire de DEA,
Universite J. Fourier, Grenoble I, 1993.
Lermo, J. and F. J. Chavez-Garcia, Site effect evaluation using spectral ratios with only one
station, Bull. Seismol. Soc. Am., 83, 1574-1594, 1993.
Nakamura, Y., A method for dynamic characteristics estimations of subsurface using microtremors
on the ground surface, QR RTRI, 30, 25-33, 1989.
Ohmachi, T., Y. Nakamura, and T. Toshinawa, Ground motion characteristics of the San
Francisco Bay area detected by microtremor measurements. Proceedings of the Second
International Conference on recent advances in geotechnical earthquake engineering and soil

J. Phys. Earth
 Possibilities and Limitations of Nakamura's Tech nique 397

dynamics, Missouri, paper No. LPO8, 1643-1648, 1991.

Samano, T. and K.Seo, Prediction of ground motion at Ashigara Valley, International symposium,
The effects of surface geology on seismic motion: ESG, Odawara, Japan, 125-126, 1992.
Sawada, Y., Geotechnical data, International symposium, The effects of surface geology on
seismic motion. ESG, Odawara, Japan, 29-42, 1992.
Seo, K., A joint work for measurements of microtremors in the Ashigara Valley, International
symposium, The effects of surface geology on seismic motion. ESG, Odawara, Japan, 43-52,
1992.
Theodulidis, N. and P. Y. Bard, Horizontal to vertical spectral ratio and geological conditions:
strong motion data from Greece and Taiwan (SMART-1), Soil Dynam. Earthq. Eng. J.,
1994 (in press).

Vol. 42, No. 5, 1994