Geoscience Reference
In-Depth Information
of parameters that need to be known for non-linear response evaluation imposes a heavy
burdenonthereconnaissancestage.Inaddition,theproblemisagainthelackofgeneral-
ityoftheresults;duetothein-depthdetailsofthemechanicalbehaviourofthemediathat
arerequired,non-linearstudiescanonlybemadeforspecificsites.Themorespecificthe
results are to a given site, the more unlikely is their generalisation. Finally, the author's
experience comes mostly from working in a developing country, where the means to
makeaproperreconnaissancearelackingyetevaluationsofsiteeffectsareneededbadly.
For these reasons, this paper deals only with linear site effects, even if recognising that
it is not always straightforward to guess the modifications that non-linear behaviour will
introduce in the ground motion for the large earthquakes, those that are our real concern
in seismicrisk.
2. Estimation of site effects
Siteeffectshavebeenestimatedusingearthquakedataandambientnoiserecords.Clearly,
earthquake data are more reliable. However, in regions of moderate seismicity, installing
a temporal network that would record enough events to compute a reliable estimate of
local amplification may be too expensive. Recording ambient noise, in contrast, is easy,
fast and very cheap anywhere. For this reason, methods based on ambient noise records
have been very much used recently. In the following lines, we will review briefly the
different techniques that have been used with those two kinds of data; earthquake and
noise records.
The analysis of earthquake data to estimate site effects has made recourse to several
techniques. It could be argued, for example, that attenuation relations that obtain inde-
pendent regressions of PGA with distance and magnitude for rock and soft soil sites
provide an estimate of local amplification. This approach has the advantage of its gen-
erality. However, the disadvantages are large; site classification is extremely imprecise,
local amplification is evaluated using a single number, and a single estimate is obtained
forallsoftsoilsites.Inaddition,PGAisaverypoormeasureoflocalamplification(see,
e.g., Ch´avez-Garc´ıa and Faccioli, 2000).
Starting from the 1970s (Borcherdt, 1970), ratios of Fourier amplitude spectra of earth-
quakesrecordedonsoftsoilrelativetoareferencestationhavebeenextensivelyused.The
hypotheses on which this method is based can be clearly stated, which allows appraising
the validity of its application. This is a significant advantage of this technique. Spectral
ratios are a good estimate of local amplification provided that the reference site is effec-
tively free of site effects and is located near to the soft soil site. The proximity of the
reference station must, of course, be measured in terms of dominant wavelength. The
choice of a reference station may pose large difficulties though (see, e.g., Steidl et al.,
1996), and sometimes it is an unsolvable problem. A helpful analysis of the possible
complications one may run into was presented by Steidl et al. (1996). Their conclusion
is that surface rock sites are inevitably affected by amplification at frequencies as low
as 4 to 5Hz because of the thin weathered layer that is almost always present on rock
