Site characterization by seismic noise in Istanbul, Turkey


Technical Report, pp.50-59, 2008

  • Publication Type: Other Publication / Technical Report
  • Publication Date: 2008
  • Page Numbers: pp.50-59


Istanbul is a megacity of 12 million inhabitants, who are exposed to a significant earthquake
hazard. Moreover, the considerable rate of urbanization combined with uncontrolled land use
makes such hazard even higher (Erdik et al., 2003). The main factor controlling the
earthquake hazard for Istanbul is undoubtedly the proximity of the North Anatolian Fault,
which in the Marmara Sea region forms a complex fault system (Erdik et al., 2003). From an
analysis of the available earthquake records performed by Ambraseys and Finkel (1991), the
city is estimated to be affected by a medium intensity (epicentral intensity of VII-VIII)
earthquake with an average return period of 50 years. During the Izmit and Düzce 1999
earthquakes, this scenario was worsened by the recognition of site amplifications that were
observed to locally modify the ground motion inside the metropolitan area. In particular, as
shown by several studies (Tezcan et al., 2002; Ozel et al., 2004; Beyen and Erdik, 2004;
Ansal et al., 2004), the Avcilar district in the western part of Istanbul suffered significant
damage, largely due to the amplification of the earthquake ground motion. In fact, for this
area, an intensity of VII (MSK) was assigned, while in the other districts of the metropolitan
area an intensity of VI (MSK) was generally observed (Gruenthal, 1998; Ozmen, 2004) The
anomalous amplification of the ground motion observed in the western part of the city is
considered to be mainly related to the presence soft sediments overlaying a competent seismic
bedrock. Therefore, recent studies (Ergin et al., 2004; Sørensen et al., 2006) have focused on
estimating possible site effects in the metropolitan area of Istanbul, and in particular in its
western part. Often, information is obtained through invasive techniques, such as drilling,
down/cross-hole measurements, etc. However, due to their expensive nature, the widespread
application of such techniques is only able to be performed after coming to a compromise
with respect to cost, resulting in a limited exploration depth. In fact, microzonation works are
frequently based on the use of the average shear-wave velocity in the uppermost 30m (Vs30),
which is adopted by the National Earthquake Hazard Reduction Program (NEHRP)
classification in the USA. However, several works (Wald and Mori, 2000; Stewart et al.,
2003; Park and Hashash, 2004; Di Giacomo et al., 2005; Mucciarelli and Gallipoli, 2006)
have showed that in a number of geological–geotechnical and morphological contexts, the
Vs30 classification is not always a suitable tool for site-effect estimation. Also for this reason,
non-invasive and cost effective passive seismic techniques have recently become an attractive
option for seismic site-effect studies. Especially in the last decade, environmental noise
recordings performed by single station methods to estimate horizontal-to-vertical (H/V)
spectral ratio curves (Fah et al., 2001; Arai and Tokimatsu, 2004) and by 2D micro-array
techniques to estimate surface wave dispersion curves (Scherbaum and et al., 2003) have
provided very promising results. Parolai et al. (2001) and Parolai et al. (2004) showed that the
seismic noise H/V curves exhibit a good agreement with the H/V from earthquake recordings,
especially with regard to the value of the fundamental resonance frequency of the sedimentary
cover. Therefore, performing a large number of noise measurements over a region of interest
allows a map of the fundamental frequencies to be obtained, which provides an overview of
the distribution of both the sedimentary cover thickness and, most importantly, of those areas
where the amplification of the seismic motion in the frequency band of interest for buildings
behaviour is expected.
Concerning 2D arrays, it has been shown (Tokimatsu et al., 1992; Ohori et al., 2002; Parolai
et al., 2006;) that by using Rayleigh wave dispersion curves, the characterization of the local
S-wave velocity profile can be obtained with a good accuracy, especially when a priori
information about the total sedimentary cover thickness is available in advance. Further
improvements are obtained by applying the joint inversion of phase velocity and H/V ratio
curves (Scherbaum et al., 2003; Parolai et al., 2005; Arai and Tokimatsu, 2005), which allows
the trade-off problem between the model parameters that hampers the separate inversion of
these curves to be overcome. The application of this inversion scheme has also had success in
estimating the S-wave velocity profile for sedimentary covers hundreds of meters thick
(Parolai et al., 2005). Although for the engineering-geotechnical community, the lack of high
resolution in the S-wave velocity profile can be considered a drawback, from the site-effect
point of view, passive techniques, especially in the case of sedimentary cover thicker than 30–
50 m, provide estimates of the local transfer function that are in very good agreement with
both the empirical ones (Picozzi and Albarello, 2007; Parolai et al., 2007) and those obtained
by 1D techniques (e.g., SASW and MASW) that allow the reconstruction of the shallower
part of the S-wave velocity profile with a higher resolution (Richwalski et al., 2007).
In this work, as a preliminary activity for the microzonation characterization of Istanbul, 192
single station measurements (Fig. 1a) for the estimation of the H/V curves were carried out in
the western part of the metropolitan area, in order to estimate the fundamental resonance
frequency of the sedimentary cover. In particular, 42 of these noise measurements were
performed at sites where accelerometers belonging to the permanent and temporary networks
operated by the Kandilli Observatory and Earthquake Research Institute (KOERI) are located.
Thus, for 29 sites of the considered accelerometric stations, the H/V curves from noise
recording were compared with those calculated using weak motion recordings. This
comparison of the results provided by the different methods was in fact directed towards
performing a calibration of the passive seismic techniques in the area investigated, allowing a
preliminary validation for the reconstructed sediment–bedrock interface geometry and the site
effects estimates.
In addition, in order to provide additional useful information for site effects and
microzonation studies, a series of eight 2D micro-array measurements utilizing short-period
sensors and high dynamic digitizers were carried out in selected sites of the study area. The
extended spatial autocorrelation technique (ESAC; Ohori et al., 2002; Aki, 1957; Okada,
2003), and frequency–wavenumber analysis (maximum likelihood method; Capon, 1969;
Horike, 1985) were used for the estimation of the Rayleigh wave dispersion curves and
wavefield analysis. For the first time, the resulting dispersion curves were used together with
the H/V curve in a joint inversion scheme for the estimation of the S-wave velocity profiles in
a megacity. Finally, theoretical site responses were calculated from the S-wave velocity
profiles obtained from the micro-array data using the propagator matrix method for a 1Dlayered medium (Wang, 1999).