Geoscience Reference
In-Depth Information
Langston (1979). It was applied to earthquake data by Lermo and Ch´avez-Garc´ıa (1993)
and in many recent papers. H/V spectral ratios of earthquake records have a weaker
theoretical basis than spectral ratios relative to a reference station. However, they have
been successful to estimate site effects. The reasons are the same behind the success of
receiver functions; the decrease of shear-wave velocity with decreasing depth brings the
incoming rays closer to vertical. As a consequence, vertical motion includes mostly P
waves (little affected by near-surface impedance contrasts) while horizontal components
consist mostly of shear waves. The difference between horizontal and vertical motions,
as measured by spectral ratios, would cancel common information related to source and
regional path propagation and enhance the differences, mainly due to the near surface
amplification of horizontal motion. This technique has been successful and has provided
nice estimates of local amplification. It offers an alternative when there is no reference
station, or when the reference station failed to record the same events as the soft soil sta-
tions.Itworksbetterwhensiteeffectsareimportant,i.e.,whenlargeimpedancecontrasts
significantly redress theincoming rays.
Very often, however, earthquake data are not available to estimate site amplification. For
this reason, and following the work of Japanese seismologists, microtremors or ambi-
ent noise records have been used to characterise seismic response. The use of ambient
noise records to estimate local amplification was proposed a long time ago (see Bard,
1999, for a thorough review). The analysis of Fourier spectra of microtremors and the
computation of spectral ratios of microtremors relative to a reference site have both
been used. However, in recent years, we have seen the thriving of the microtremors
H/V spectral ratios (referred to simply as H/V in the following) on a very large scale.
This paper will not go through a detailed review which can be consulted, for example,
in Bard (1999) or in the many documents produced by the European project SESAME
(http://sesame-fp5.obs.ujf-grenoble.fr/index.htm).Onlyafewexperiencebasedconsider-
ations will bediscussed.
Most authors agree that H/V gives the dominant period of the soft soil layers. Of course
the term dominant period may loose its meaning in many cases where site effects are
stillimportant.Inthosecases,H/Visnotveryuseful.Inadditiontodominantperiod,the
possibility of using H/V to estimate the level of amplification has been the subject of a
large discussion. In the author's opinion, noise can be used to determine amplification
level in some cases. Horike et al. (2001) have shown among others that amplification
computed from H/V of noise is reliable when local amplification is caused by a marked
impedance contrast at a single interface and the amplification level is important. This
means that, if H/V ratios are large, the results are usually dependable. However, when
their amplitude is small it could indicate either that amplification is not significant (and
then we need not worry about site effects) or that amplification is significant but not
due to a single impedance contrast but resulting from a complex situation (e.g., velocity
gradient). This isa large problem and one that has no general solution at present.
Probably the more reliable results are those that compare different techniques among
them. It is likely that real cases cannot depend on a single technique to estimate site
