Soil Dynamics and Earthquake Engineering 113 (2018) 406–414

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Variation of high frequency spectral attenuation (Kappa) in vertical arrays T

⁎
Gülüm Tanırcan , S.Ümit Dikmen
Department of Earthquake Engineering, Kandilli Observatory and Earthquake Research Institute, Boğaziçi University, Istanbul, Turkey

A R T I C LE I N FO A B S T R A C T

Keywords: Near surface attenuation parameter kappa (κ ) of S-waves is calculated from 52 earthquake acceleration re-
High frequency spectral attenuation cordings at surface and base level (Vs > 760 m/s, namely engineering bedrock) of three downhole arrays and at
Kappa (κ) two outcrop stations in Istanbul, Turkey. Path dependent (κR ) and site dependent (κ 0 ) components of κ are
Vertical seismic arrays calculated from epicentral distance (Repi)- κ correlations. Estimated κ 0 values are for the outcrop stations are in
Istanbul
the range of 26–30 ms, while for the surface and base levels of the arrays are 40–53 ms and 18–23 ms, respec-
tively. A strong correlation is observed between relative amplification factors of downholes and κ 0 differences
between surface-base levels. On the other hand, κ 0 not approaching to zero at base stations suggests that several
other factors still contribute to other than path and site effects. Given the earthquake scenario and stochastic
simulation approach, 10 ms decrease in κ 0 is found to increase response spectral acceleration up to 0.2 g in
0.1–0.2 s structural periods.

1. Introduction stochastic simulations of ground motion [7]. In this respect, κ char-

The concept of the high frequency attenuation parameter kappa (κ )
was first introduced by Anderson and Hough [1] observing an ap-
proximate linear decay of the acceleration spectrum at frequencies
higher than a specific frequency (fe). Subsequently, they proposed the
following mathematical form for the high frequency part of the accel-
eration spectrum, A(f) of the waveform containing source and path
effects,
A (f ) = A0 exp(−πκf ) for f > fe
where A0 is the spectral amplitude and f is the frequency. If accel-
eration spectrum is available, κ can be readily estimated from the slope
of the spectrum over a range of frequencies (fe to fe+Δf) as
κ = −λ /π where λ = [ln (A fe ) − ln (A fe +∆ ) ]/Δf
where A is the spectral acceleration at selected frequencies. They also
postulated that κ has site dependent, κ 0 and distance dependent, κR
components and a linear relationship with epicentral distance, Repi ex-
ists as;
κ = κ 0 + κR. R epi
Hence, κ 0 is the intercept of the linear relationship between κ and
Repi. Since then, this concept has attracted considerable attention in the
seismological community, especially by the researchers working on site
amplifications [2], ground motion modeling equations [3–6] and
acterizing the high frequency decay of spectra is a valuable parameter
in site studies. Consequently it is used (1) in the source related studies
in order to study self-similarity of the source spectrum (2) in the gen-
eration of synthetic ground motion using point-source or finite-fault
stochastic or hybrid simulation approaches; even in physics based si-
mulations using theoretical Green’s functions; (3) in the calibration of
ground-motion prediction equations (GMPEs) based on stochastic si-
mulations and; (4) in the engineering seismology community in prob-
abilistic seismic hazard assessment (PSHA) for critical facilities, where
(1)
it is common to perform site specific response analyses.
Observing the rather empirical nature of this parameter, researchers
continued their endeavors to broaden their understanding of the con-
cept. Thus, over the years a number of complementary techniques have
been proposed to estimate the κ 0 [8–10]. Particular effort was devoted
(2) to find a plausible correlation between κ 0 and site shear wave velocities
[3,11-12].
In this respect, a range of κ 0 values calculated at the engineering
bedrock layer with Vs > 760 m/s, underlying layers with lower Vs
(referred to as the base level of the downhole arrays in this study) and/
or at the outcrop stations can help realistic calculation of site specific
(3)
strong ground motion simulations. In this study, the parameters of Eq.
(3) are investigated utilizing the data compiled from the three seismic
downhole arrays and two outcrop stations in Istanbul operated by
Kandilli Observatory and Earthquake Research Institute (KOERI). The
arrays have different subsoil and topographical conditions, as well as
urban fabric around them. The motivation of the study has risen from

⁎
Corresponding author.
E-mail addresses: birgore@boun.edu.tr (G. Tanırcan), umit.dikmen@boun.edu.tr (S.Ü. Dikmen).

https://doi.org/10.1016/j.soildyn.2018.06.016

0267-7261/ © 2018 Elsevier Ltd. All rights reserved.

G. Tanırcan, S.Ü. Dikmen
the fact that, to authors’ knowledge, no study is available estimating the
spectral attenuation parameters at the downhole arrays in Turkey with
a reasonably large strong motion data set. Moreover, κ estimation for
this part of Istanbul, namely the western side of Bosporus, is not
available.
In brief, the objectives of this study are threefold encompassing the
calculation of κ at three downhole arrays (both surface and base levels)
and outcrop stations, differentiation of site and path components and
finally the investigation of possible correlation of the κ 0 with site re-
lated parameters to complement existing rock κ 0 empirical equations.
2. Study area, downhole arrays and outcrop stations
Istanbul is in close proximity to active fault lines [13]. A number of
large earthquakes, at least eight times since the 15th Century, have hit
the region in the past [14]. Parsons [15] in his study published in 2004
estimated the probability of having a ML = 7 + earthquake in the
Marmara Sea Faults, northern extension of the North Anatolian Fault
(NAF), as 70% in 30 years. A recent study, by Murru et al [16] proposed
that the time dependent combined probability of having a M > 7.0
earthquake in the Marmara Region is as high as 47%. Hence, the city,
with a population of 15 + million, is under major earthquake risk.
Subsequently, countless collaborative efforts of public bodies and
universities were made and are still ongoing to decrease the earthquake
risk of the city, including ground motion simulations based on scenario
earthquakes [17–20] and several geotechnical and geophysical in-
vestigation projects at city scale. For instance, part of the investigations
in the past included Vs30 mapping of the city by shallow borings [21]. In
smaller scale, deep borings were done particularly on the European side
where sedimentary formations of Tertiary-Quaternary ages exist
[22,23]. Microtremor measurements were carried out on sedimentary
formations such as Avcılar, Zeytinburnu and Ataköy [24–27].
In local scale, three downhole arrays were deployed at the west side
of Istanbul; namely at Ataköy (ATK), Fatih (FTH) and Zeytinburnu
(ZYT) to investigate the dynamic behavior of soils. The arrays are si-
tuated at distances about 5.0–9.0 km, from each other (Fig. 1). All three
have different subsoil profiles and urban environment, as noted earlier.
Strong motion accelerometers are placed at various depths of the arrays
including one at the surface and one at about 20 m deep into the en-
gineering bedrock, with the objective to avoid the fractured zone. The
latter is acknowledged as the base layer in this study. The base layers
sensors (deepest sensors) are at 140 m, 120 m and 288 m depths for
ATK, FTH and ZYT arrays, respectively. All instruments are force
Fig. 1. Locations of the downhole arrays (ATK, ZYT, FTH) and outcrop stations (MECLI and OKCIO) used in the study together with simplified geology map of the
Istanbul (Adapted from [22].) (left) and their shear wave velocity profiles (right) (ATK-K [29], FTH and ZYT per [28,29]ATK-P [30,31]).
407
Soil Dynamics and Earthquake Engineering 113 (2018) 406–414
balance type triaxial digital recording accelerometers set to 200 Hz
sampling frequency. Further detailed information about the subsoil
formation at these arrays, as well as the arrays themselves, can be found
in relevant studies published earlier [28–33]. Note that there are two
profiles available for the ATK array; one obtained from the study by
Parolai et al [30,31] using waveform deconvolution method (ATK-P in
Fig. 1) and another one reported by Kurtuluş [29] using the subsurface
soil data (ATK-K in Fig. 1) [34].
In addition to the downhole arrays, KOERI operates the Istanbul
Rapid Response System (IRRS) in the metropolitan area of Istanbul
[35]. The system currently has almost 100 strong motion recording
stations at various locations in the city. However, at the Western side
(i.e. the European side) of the city, only the two of these stations,
namely MECLI and OKCIO, are situated on engineering bedrock [34].
The distance between the downhole arrays and these stations are in the
range of 3.6–15.3 km (Fig. 1). The instruments installed at these sta-
tions are three-component accelerometers with 18-bit analog to digital
converter. The recordings are made at 100 Hz sampling frequency. The
Vs30 values at these stations are determined to be above 800 m/s during
Istanbul microzonation studies [21,34]. However, detailed velocity
profiles are not available for these stations. A summary of the shear
wave velocity averages and geological units of the downhole arrays and
outcrop stations are given in Table 1.
3. Earthquake data
An earthquake data set consisting of 52 earthquakes that occurred
between 2004 and 2016 are compiled for the purposes of this study.
Strong motion recordings at surface and base sensors of downhole ar-
rays and recordings at outcrop stations are considered. Local magni-
tudes of these earthquakes are in the range of ML= 3.0–4.8. Epicentral
distances are within 200 km and hypocentral depths are at less than
20 km. The epicentral distribution of earthquakes is shown in Fig. 2.
Yet, due to various reasons, mostly instrument malfunction, not all the
earthquakes were recorded at all the downhole arrays and the selected
outcrop locations. The earthquakes, which data are available, are in-
dicated in Table 2.
4. Data analysis
To compute the κ values standard pre-processing procedures has
been followed. First the three component waveforms are baseline cor-
rected which followed by the visual selection of P-wave and S-wave
G. Tanırcan, S.Ü. Dikmen Soil Dynamics and Earthquake Engineering 113 (2018) 406–414

Table 1
Information about the stations used in the study.
Station Name Latitude / Longitude Geological Unit Vs30 (m/s) Vs (at base) (m/s) Sensor Depths (m) # of Eqs. Recorded # of Eqs. Used for regression

ATK 40.988N/28.849E Pliocene-Miocene 340 700 0 18 10

140 11
FTH 41.019N/28.951E Pliocene-Miocene 350 1100 0 43 24
136
ZYT 40.986N/28.980E Pliocene-Miocene 200 900 0 37 21
288 27
OKCIO 41.0496N/28.966E Paleozoic 800 + n/a 0 27 19
MECLI 41.071N/28.996E Paleozoic 800 + n/a 0 23 19

onsets. Depending on the magnitude and distance of the earthquakes,
5–15 s portion of the S and noise window is selected for spectral ana-
lysis. For some of the records noise window is chosen from post-event
waveform, if there is no existing pre-event time. Afterwards, the se-
lected portion of the waveforms is 10% tapered and Fourier spectra are
calculated. Signal to Noise Ratio (SNR) of each recording is calculated
and the recordings satisfying SNR > 3 condition are kept in the ana-
lysis. The Nyquist frequency of the data set is 100 Hz for the downhole
arrays and 50 Hz for the outcrop stations. Instrument response does not
exist in the wide frequency band of interest. Corner frequencies (fc) of
the events are estimated from displacement amplitude spectra plotted
in log-log scale.
Acceleration spectra are smoothed using Konno & Ohmachi’s [36]
scheme with the smoothing coefficient b fixed to 20. Both smoothed and
unsmoothed spectra are taken into consideration while manually
picking the frequencies where the spectral decay starts (fe) and ends (fx)
in lin-log scale. Exemplarily, three different earthquake recordings of
ATK base station are used to demonstrate the process in Fig. 3. Selection
of noise and signal windows and their corresponding spectra are illu-
strated in the same graphs. Record and spectrum processing constraints
suggested by Ktenidou et al. [8] are applied for estimating fe and fx.
According to these constraints, only data satisfying fe > fc and fe-
fx≥ 10 Hz conditions are utilized. Resulting fe and fx values vary be-
tween 4 and 20 Hz and 15–48 Hz, respectively. Exponential fitting is
performed with a nonlinear least square model of selected high
frequency data as defined by Anderson and Hough [1].
In an earlier study on site response and amplification, authors re-
ported that the resonance frequencies at FTH, ATK and ZYT as 1.9, 0.9,
0.5 Hz determined from horizontal to vertical spectral ratio (HVSR) and
1.8, 0.9–1.0 and 0.7 Hz determined using Standard Spectral Ratio (SSR)
approach, respectively [34]. Hence, first resonance frequencies of the
sites are low enough to not to hinder the kappa estimation in the higher
frequency range. In the same study, resonance frequencies are reported
to be in the 3.5–4.0 Hz band for both outcrop stations. This range being
outside of the frequency band used to estimate κ , confirms that the
calculations will not be affected by the rock amplification. Moreover as
several studies showed that the wider spectral windows range applied
to estimate the κ value (here is 10 Hz or above) would decrease the
effect of local peaks, (due to site effect and downgoing wave con-
tamination) that would bias the measured results [34].
Analysis is performed for NS and EW components of acceleration
spectra separately. Due to observed high noise level, vertical compo-
nent is kept outside the analysis. The high noise level, especially above
8.0 Hz is attributed to the urban noise, particularly to the vehicular
traffic, since all three downhole arrays are situated within the city. In
general linear spectral decay is immediately noticed at base level sta-
tion recordings. For surface recordings, on the other hand, decay is
relatively difficult to choose possibly due to the seismic existing noise,
namely cultural noise, in the high frequency component. For instance, a
recent study indicates the existence of traffic induced noise at

Fig. 2. Epicenters of earthquakes in the data set (red stars). Black squares are the location of outcrop stations and downhole arrays. Black lines indicate the active
faults in the region [1]. PIF: Prince Islands Fault.

408
G. Tanırcan, S.Ü. Dikmen Soil Dynamics and Earthquake Engineering 113 (2018) 406–414

Table 2
Earthquake data set used in the study (n/a: not available, √: kappa values are not utilized since they do not comply with constraints applied for selection procedure
√√: average horizontal kappa values used for regression, base sensor is considered at downhole arrays).
# Date Time (UTC) Latitude (N) Longitude (E) Depth (km) ML ATK FTH ZYT MECLI OKCIO

1 May 16, 2004 03:30:48 40.70 29.33 10 4.3 n/a n/a n/a n/a √√
2 Sep.29,2004 15:42:07 40.79 29.02 13 4 n/a n/a n/a n/a √√
3 Mar.12,2008 18:53:31 40.62 29.01 10 4.8 n/a n/a n/a n/a √
4 Oct.03,2010 17:49:26 40.84 28.12 11.8 4.3 n/a n/a n/a n/a √√
5 Oct. 20,2012 23:42:20 40.81 28.28 11.0 3.3 n/a √√ √√ n/a n/a
6 Oct.21,2012 0:48:09 40.62 29.00 10.6 3.3 n/a √√ √√ n/a n/a
7 Oct.26,2012 3:37:36 40.43 28.72 8.3 4.0 n/a √√ √√ n/a n/a
8 Oct.31,2012 9:55:51 40.44 28.72 7.1 3.5 n/a n/a √√ n/a n/a
9 Mar.19,2013 12:44:39 42.13 29.58 12.0 4.5 n/a √√ √√ √√ √√
10 Apr.9,2013 11:42:24 40.54 28.13 5.0 3.4 n/a √√ √√ n/a n/a
11 Jul.7,2013 7:55:38 40.72 28.98 6.8 3.1 n/a √ √ n/a n/a
12 Aug.6,2013 19:18:12 39.84 29.75 7.5 4.0 n/a √√ √√ n/a n/a
13 Aug.7,2013 5:20:32 39.20 29.45 9.5 4.4 n/a √ √√ n/a n/a
14 Aug. 17,2013 18:16:27 40.41 29.12 5.5 4.3 n/a √√ √√ √√ √√
15 Aug. 29,2013 06:20:45 40.35 27.45 14.8 4.1 n/a √√ √√ √√ √
16 Sep. 25,2013 13:39:44 40.77 27.42 7.5 3.5 n/a √√ √ √ n/a
17 Oct. 03,2013 10:25:56 40.09 28.70 13.8 3.7 n/a √√ √ √ n/a
18 Nov. 10, 2013 02:09:23 40.76 30.27 13.8 3.5 √√ √√ n/a n/a n/a
19 Nov. 27, 2013 04:13:38 40.85 27.92 10.8 4.7 √√ √√ √√ √ √√
20 Nov. 27, 2013 04:21:30 40.83 27.92 12.5 3.8 √√ √√ √√ √√ √√
21 Jan.30,2014 2:54:33 40.67 29.27 10.1 3.1 √ √√ √√ n/a n/a
22 Feb. 5, 2014 01:56:44 41.37 28.62 14.6 3.8 √√ √√ n/a √√ n/a
23 Feb.17,2014 23:59:21 40.61 27.69 7.1 3.0 n/a n/a √√ n/a n/a
24 May 11, 2014 5:07:00 40.67 27.04 11.0 3.6 n/a √√ √√ n/a n/a
25 Jul. 3, 2014 05:04:46 40.21 27.93 11.8 4.5 √√ √√ √√ n/a n/a
26 Aug. 03,2014 10:39:34 40.61 29.16 3.4 3.6 n/a √√ √√ √√ √
27 Aug. 03,2014 22:22:26 40.61 29.17 11.5 4.1 n/a √√ √√ √√ n/a
28 Oct,8,2014 3:08:49 40.76 27.48 19.2 3.3 n/a √√ √ n/a n/a
29 Oct. 22,2014 17:10:16 40.41 30.11 7.5 4.5 n/a √√ √√ √√ √√
30 Nov.28,2014 2:30:07 39.35 29.02 3.9 4.6 n/a n/a √√ n/a n/a
31 Dec. 16,2014 09:02:25 40.15 27.08 12.9 4.4 n/a √√ √ n/a √
32 Dec.24,2014 0:21:19 40.35 27.75 12.8 3.2 n/a √√ √√ n/a n/a
33 Jan. 17,2015 00:42:40 39.88 30.39 7.0 4.2 n/a √√ √ √ n/a
34 Jan. 23,2015 10:19:20 40.07 28.59 5.0 4.5 n/a √√ √ √√ n/a
35 Feb.1,2015 10:46:32 40.70 27.50 5.6 3.5 n/a √√ √√ n/a n/a
36 Apr. 29, 2015 04:40:53 42.08 29.30 15.2 4 n/a √√ n/a √√ √√
37 May 11, 2015 04:16:28 40.42 29.13 8.1 3.9 n/a √ √ √√ √√
38 Aug.13,2015 01:01:28 40.70 29.28 9.4 3.8 n/a n/a n/a n/a √√
39 Oct. 16,2015 04:35:06 40.45 29.17 5.4 3.5 √ √√ √ n/a √
40 Oct.28,2015 16:20:02 40.82 27.76 14.3 4.5 √√ √√ √√ √√ √√
41 Nov.16,2015 15:45:43 40.83 28.75 12.6 4.2 √ n/a √√ √√ √
42 Dec. 05,2015 20:53:51 40.44 29.07 13.2 3.7 √√ √√ √√ √√ √√
43 Dec. 15,2015 01:13:38 42.24 29.71 18.7 4 √ √√ √ √√ √√
44 Jan. 30,2016 09:03:13 40.79 28.06 9.8 3.5 √√ √√ √√ √√ √√
45 Mar. 28,2016 17:23:46 40.73 27.54 15.5 3.7 √√ √√ n/a n/a n/a
46 Mar. 27,2016 05:03:54 40.82 27.90 14.7 3.7 √√ √√ n/a n/a √
47 Jun. 7,2016 04:09:45 40.26 29.14 11.5 4.6 √√ √√ n/a √√ √√
48 Jun. 7,2016 08:02:15 40.27 29.15 6.5 3.5 n/a √√ n/a n/a n/a
49 Jun. 25,2016 05:40:11 40.71 29.19 9.3 4.5 √ √√ n/a √√ √√
50 Jul. 9,2016 14:20:51 40.71 29.19 9.5 3.6 √ √√ n/a √√ √
51 Jul.17,2016 08:55:41 40.71 29.18 9.4 4 √√ √√ n/a n/a √√
52 Sep.30,2016 14:09:37 40.67 29.18 5.3 3.6 n/a n/a n/a n/a √√

frequencies higher than 8 Hz at ATK array especially during daytime ⎡ ⎤

when traffic activity is high [33]. Difference of κ between the two
horizontal directions is within 25% at all base stations. However, this
difference increases to 40% at the surface stations Earthquake record-
ings, which the kappa values of the two horizontal components exhibit
a difference greater than 25% with respect to each other are excluded
from the analysis as recommended in by Ktenidou et al. [8]. Subse-
quently horizontal kappa (κave ) values, for the earthquakes meeting this
criterion, are calculated as the arithmetic mean of horizontal κ values.
The remaining list of earthquakes following this screening process is
given in Table 2.
Fourier acceleration spectrum of an acceleration record can be de-
scribed by Brune [37,38] considering point source model with a single
corner frequency.
(2πf )2 ⎥ −πfκ
Y (f ) = CM0 G (R) ⎢
⎢ 2⎥
e A (f )
⎣
() f
⎢ 1+ f ⎥
c ⎦ (4)
In Eq. (4) M0 is the seismic moment in units of Nm and fc is the
corner frequency in Hz. fc is given by 0.4906β(Δσ/ M0 )1/3 [37,39]
where Δσ is stress drop in MPa and β is the average shear wave velocity
in the vicinity of earthquake source. Term in the squared parenthesis is
the source effect. A(f) accounts for the site effects arisen from im-
pedance contrast in shallow medium. C is a constant to account for
average source radiation pattern for S waves, near surface amplifica-
tion, the average crustal density and partitioning of the total shear wave
energy into the horizontal components. C controls the low frequency
spectral amplitudes. G (R) is the geometrical spreading function. Among
them stress parameter (Δσ) and anelastic attenuation operator; κ de-
termine the behavior of high frequency spectral amplitudes. κ

409
G. Tanırcan, S.Ü. Dikmen
Fig. 3. (Left) Examples of recorded earthquakes at ATK base station. Acceleration recordings are normalized for a better visualization. Boxes in the recordings show
pre-event or post-event noise window, as well as windows taken from S waves onset. (right) Corresponding Fourier amplitude spectra (FAS) of the noise (grey line)
and signal (black line) windows. Thick red line shows the smoothed FAS of the signal.
calculation with Anderson and Hough method [1] considers a fre-
quency independent anelastic equation parameter, Q in the range be-
tween fe and fx. Hence, Eq. (3) can be simplified and rewritten in terms
of region specific Q as;
κ = κ 0 + R epi /(βQ)
It should be noted that Eq. (5) with constant Q results in smaller κ
values when compared to those calculated with frequency-dependent
Q. Moreover, Q dominates the κ value in large distances, hence un-
certainty in κ increases with increasing distances.
5. Results of this study
Once κave is determined for the downhole arrays (at surface and base
levels) and the outcrop stations (surface level) for each event recorded,
their variation with respect to epicentral distance, Repi is determined
per Eq. (3). A linear fit of data yields the intercept, κ 0 and distance
dependent, κ R components. Discussion on the use of Repi as the distance
measure for κ is given in detail by Ktenidou et al. [8] and Edwards et al.
[6].
In Fig. 4, κave variation with Repi is presented for the outcrop sta-
tions. Both OKCIO and MECLI stations, as would be anticipated, have
Fig. 4. Distribution of average horizontal kappa (κ ave ) with epicentral (Repi)
distance for outcrop stations OKCIO and MECLI. The solid and dashed black
lines indicate the regression lines determined by least square method.
Soil Dynamics and Earthquake Engineering 113 (2018) 406–414
the same κR value, 0.14 ms/km. However, κ 0 exhibit slightly different
values as 30 ms and 26 ms at OKCIO and MECLI, respectively. Con-
sidering the short distance between these stations, a common κ 0 value is
determined again using the kappa estimates of both stations yielding a
mean value of 28 ms.
(5)
Fig. 5 compares κave variation with Repi at base and surface level of
downhole arrays. In general, base records show less κave variability than
that of the surface records particularly for the first 100 km epicentral
distance. In the same figure, regression lines at the base level of three
downhole arrays and at outcrop stations are compared as a separate
graph (Fig. 5d). Linear regression lines of κave at three base sensors have
similar slopes, κ R yielding 0.22 ms/km at ATK and ZYT and 0.19 ms/km
at FTH, reflecting the similar regional attenuation characteristics. κ 0
values at the base levels are lower than what is calculated at outcrop
stations, as would be anticipated due to the downgoing waves reflected
from the surface.
κ 0 values at all surface stations are larger than those at base stations.
Overall κ 0 values range from 40 to 54 ms at the surface and 18–23 ms at
the base of downhole arrays. This indicates that the response on the
surface is affected by the filtering and amplification of the waves pro-
pagating through the soft shallow layers. The waveforms travelling
through the softer layers result in the amplification of accelerations
especially at low frequencies. However, since the amplification at
higher frequencies is relatively lower, the resulting high frequency
slope of the spectra will be higher. On the other hand, at higher fre-
quencies (> 8.0 Hz) the signals may also receive additional energy from
cultural noise, thereby influencing the shape of the spectrum. Also, κ 0 at
all base stations do not approach to zero. This suggests that several
other factors including; site effects originated at deeper site, possibly
contribute to the formation of κ 0 . Furthermore, the relatively large κ 0
value at the base level of ATK can be attributed to the availability of
limited amount of earthquakes at that station.
Even though κ 0 at surface should be treated with care due to cultural
noise, it is still worthwhile to compare it with the κ 0 values obtained at
base in order to see a general trend. The difference between κ 0 at sur-
face and κ 0 at base, namely Δκ 0 is the largest (33 ms) at FTH array. This
410
G. Tanırcan, S.Ü. Dikmen Soil Dynamics and Earthquake Engineering 113 (2018) 406–414

Fig. 5. Distribution of average horizontal kappa (κ ave ) with epicentral (Repi) distance for surface and base stations for ATK (a), FTH (b) and ZYT (c) downhole arrays.
The solid black and red lines indicate the regression lines determined by least square method for base and surface stations, respectively. In (d) regression lines at the
base of three downhole arrays and surface of OKCIO and MECLI outcrop stations are compared.

Table 3
κ 0 and κR estimations at surface and base sensors of downhole arrays and outcrop stations (*Biased due to limited data points).

Station Name Sensor Depth (m) κ 0 (ms) κR (ms/km) R2 Relative Site Amplif. Resonant Freq. (Hz) κ 0h prediction eq. (ms)
Dikmen & Tanircan [34] Dikmen & Tanircan [34] Aşkan et al.[44] Bora et al. [41]

ATK 0 40 0.23 0.22 7.5 0.9–1.0 45.5 43.3

140 23 0.22 0.39 – –
FTH 0 54 0.13 0.07 16 1.8–1.9 45.5 43.3
136 21 0.19 0.20 – –
ZYT 0 42 0.19 0.24 8.5 0.5–0.7 – 39.5
288 18 0.22 0.38 –
OKCIO 0 30 0.14 0.17 – 3.9 – 49.5 *
MECLI 0 26 0.14 0.18 – 3.8 – 49.5 *

array also produces the largest relative site amplification among the
three downhole arrays. Whereas, Δκ 0 at ZYT and ATK downhole arrays
are 24 ms and 17 ms and their relative site amplifications are 8.5 and
7.5, respectively. The Δκ 0 value determined for ATK is at the upper
bound of the range, 14 ± 3 ms, determined by Dikmen et al. [33]
analyzing traffic induced seismic noise data. Furthermore, regression
line of ATK is drawn based on results from only 10–11 events. Hence
the limited availability of data might be the reason of such difference.
Nonetheless, the results clearly indicate that κ 0 has a strong correlation
with local site amplification which accounts for the properties of soil
from base to surface. In Table 3, site amplification at downhole arrays,
as reported by Dikmen and Tanırcan [34] and Δκ 0 values calculated in
this study are compared.
On the other hand, a linear correlation of κ 0 with the resonant
frequencies of sites or with the depth to engineering bedrock (Vs > 760)
is not observed (Table 3).
6. Comparison with earlier studies
Ktenidou et al. [8] have made a comprehensive review of the
worldwide studies on kappa. In the studies reported, average rock site
κ 0 at the base level of downhole sites ranges between 17 and 26 ms. A
recent study published by Lai et al. [40] have shown even a larger
scatter in κ 0 values. The researchers estimated κ 0 values between 12 and
78 ms at base levels of downhole sites in Taiwan. They attributed such a
wide difference to the geological environment. Edwards et al. [6] re-
ported that very hard rock site (Vs30 = 1000–3000 m/s) surface κ 0
ranges between 10.6 and 21.0 ms, obtained using different methods.
Disregarding the extreme high values in the Lai et al study, κ 0 at the
base of ATK, FTH and ZYT downholes are comparable with the findings
of the above mentioned studies.
Regarding the surface values, Bora et al. [41] investigated κ 0 values
using strong motion database [42] from Turkey and Italy and reported
site class specific κ 0 as 39.5 ms for very soft soil site (NEHRP E), 43.3 ms
for soft soil site (NEHRP D), 41.6 ms for stiff soil site (NEHRP C) and
49.5 ms for rock site (NEHRP B) in Turkey. Their rock κ0 value, as noted
by them, is biased since their estimation is derived with limited data
points. κ 0 values calculated at the base level of ATK, FTH and ZYT
stations are very similar to those calculated at rock stations in Italy
(21.2 ms) by Bora et al. [41]. Previously, Akinci et al [43] found the
average value of distance and site-independent κ 0 as 55 ms in the
Marmara Region. Furthermore Askan et al. [44] proposed a κ 0 value for
Northwestern Turkey as 45.5 ms for soft soil site and 37.7 ms for stiff
soil site. Per NEHRP the FTH and ATK sites are classified as a D site,
while ZYT is classified as E site [34]. κ0 at FTH is higher than average κ 0
reported both by Bora et al. [41] and Askan et al. [44] for NEHRP D
sites but matches the average value suggested by Akinci et al [43]. On
the other hand, κ 0 of the ATK is slightly lower than the above

411
G. Tanırcan, S.Ü. Dikmen
Fig. 6. κ0 plotted against Vs30 values. Empty squares show the surface κ0, while
solid filled squares show κ0 vs Vs values at the base. Κ0 - Vs30 information of
station #3401 is adapted from Askan et al. [44]. Edwards and Fah [1] and Van
Houtte et al. [3] empirical equations are also given for comparison purpose.
mentioned estimations. Whereas for ZYT, the κ 0 value is comparable
with the findings of Bora et al. [41] for NEHRP E sites.
Pursuing the availability of a possible Vs30 - κ 0 correlation has been
the subject of many researchers [3,11,45]. Yet there seems to be a
considerable scatter in such correlations. In Fig. 6, Vs30 - κ 0 couples
obtained from this study are shown graphically together with two
empirical equations of Van Houtte et al. [3] and Edwards & Fah [12].
Equation proposed by Van Houtte et al. [3] is valid for stations where
Vs30 is greater than 500 m/s. Edwards and Fah [12], on the other hand,
utilized a dataset encompassing all soils classes with an upper bound of
1200 m/s for Vs30.
The Vs30 - κ 0 information at station #3401 is from [44]. Station
#3401 is also located in Istanbul, approximately 4.0 km east of the
study area in the center of the ancient peninsula. κ 0 values of OKCIO,
MECLI and station #3401 fall into the ± 1σ of the equation proposed
by Van Houtte et al. [3]. However, the equation of Edwards and Fah
proposed for all types of soil seem to overestimate κ 0 of ZYT, while
underestimating the κ 0 of FTH. Such deviations can be attributed to the
characteristics and the content of data, as well as the urban fabric. For
instance the existence of high frequency environmental noise in built
environment, site related issues, site specific amplification which
cannot be explained with 1-D linear approaches, topographic effects or
backward reflection of energy from buildings nearby to site [46,47].
Moreover, it was suggested in a number of publications that the Vs30 is
not a sufficient proxy in classifying soil deposits [34]. Hence, a major
part of the difficulty in identifying a plausible Vs30 - κ 0 relationship can
be attributed to this point too.
In Eq. (5), assuming shear wave velocity at crust, β is 3.5 km/s, the
slope corresponds to an equivalent Q value of 1298 at the base of ATK
and ZYT and 1504 at the base of FTH. Numbers imply that crustal at-
tenuation at FTH is lower than those at ATK and ZYT.
Q reaches to 2040 at outcrop surface stations, OKCIO and MECLI.
The largest Q (the lowest attenuation) is estimated at FTH surface
station as 2200. Previous studies with Turkish data show significant
variation of Q as well. Q is estimated by Askan et al. [44] as 2164 for
soft soil, and overall in Turkey by Bora et al. [41] as 1462. It is ne-
cessary to acknowledge several other frequency dependent attenuation
studies with Turkish data [43,48]. Those studies are not concerned with
frequencies higher than 10–15 Hz, as expected. Hence, in order to make
the comparison between frequency independent and dependent Q va-
lues, Q at 10 Hz is set from frequency dependent results. Moreover, in
seismic hazard analysis, attenuation characteristics of strong ground
motion are mostly seek in the frequency range of 0.1–10 Hz. Among
above mentioned studies, Q is estimated in Aegean region, Turkey by
Kurtulmus and Akyol [49] as 1507 and 332 (f=10 Hz) in short
(10–70 km) and in large distance (120–200 km) ranges, respectively.
Additionally, Q estimation of an earlier study by Akinci et al [43] is 507
(f = 10 Hz) for Marmara Region. Considering the simplifications on the
equation, Q values determined in this study are in agreement with those
412
Soil Dynamics and Earthquake Engineering 113 (2018) 406–414
reported in previous studies
7. Discussion
As utilized also in this study, the use of vertical seismic networks,
namely the seismic downhole arrays, has contributed greatly to the
body of knowledge about the dynamic behavior of soils and wave
propagation as well as achieving further insight on κ. Through these
studies, including some reported above, authors suggest that κ 0 at a
given site can be more precisely defined as a combination of two
components rather than one as given in Eq. (1). Thus the new form of
Eq. (3) will read as,
κ = (Δκ 0 + κ 0base ) + κR. R epi (5)
where κ 0base
is the intercept, believed to be emerging from various ef-
fects including source, and reflections in the travel path. κR is the path
effect between the source and the bedrock at site and finally ∆κ 0 is
purely due to the near surface geology.
Furthermore, since κR represents the contribution of the path, it can
be hypothesized that this value should be constant for a region of
reasonable size. This obviously requires assuming that the physical
characteristics of the propagation path between the source and any
point in the region do not change drastically in the frequency range
between fe and fx. With the same token, one can assume that the κ 0base
value will remain constant within a region of reasonable size. The κ 0base
values of ATK and ZYT arrays reported above also reflect this result.
Consequently, the ∆κ 0 value remains as the only site specific compo-
nent of a particular point within the region. Yet, on the other hand, it is
well known that the site response at a site within the urban environ-
ment is a function of not only the near surface soil profile, but topo-
graphy and the urban fabric as well [34,46,50,51] .
Recent studies on ground motion prediction equations indicate that
a slight increase in κ 0 , reduces the Fourier acceleration amplitude in
high frequency range as well as spectral values at corresponding
structural periods [4,5]. Thus in ground motion simulation studies,
station-dependent κ 0 is recommended. A set of ground motion simula-
tion is performed to see the differences in 5% damped response spectral
acceleration, Sa at short structural periods, when site specific κ 0 and
empirically obtained κ 0 are utilized as simulation inputs.
A moderate earthquake (Mw=6.0) and the characteristic earth-
quake (Mw = 7.2) on the Marmara Sea eastern fault segment (Princes’
Island Fault, PIF), [13] which is the most active fault in the Marmara
Sea, [52,53] are simulated following a response spectral ground motion
prediction equation (GMPE) from empirical source and duration which
is proposed by Bora et al. [5,54]. In their approach point source ap-
proximation has been used to make a reasonable approximation for
large events in high frequencies. Further information can be sought in
the related reference. Assuming the earthquake epicenter is in the
middle of the segment, FTH and ATK arrays are approximately at equal
distance from the earthquake (Repi = 28 km). Mean Sa of 100 random
simulations at ATK and FTH are given in Fig. 7. Simulations were
performed with constant stress drop values, 35 MPa (maximum bound)
and 10 MPa (average value). When the characteristic earthquake is
considered, a 11 ms increase in κ 0 causes a considerable decrease in Sa
in the range of 0.1–0.2 s structural periods. The amount of reduction is
approximately 1.0 m/s2 at 0.1 s and 1.8 m/s2 at 0.2 s at FTH, when
stress drop is set as 35 MPa. Whereas, when the moderate earthquake is
considered these values get halved, resulting a 0.5 m/s2 increase at 0.1 s
and 0.2 s. Simulations with lower stress drop value, on the other hand
decrease the overall spectral amplitudes as expected (Fig. 7). However
effect of different κ 0 on spectral values is still visible and not less than
0.4 m/s2, particularly on short period range. Since the Vs30 values of
FTH and ATK are almost the same, the difference in Sa, would not be
seen if the Vs30 -κ 0 equations are used in the simulation. It is also ne-
cessary to note that, in point source stochastic simulation approach,
G. Tanırcan, S.Ü. Dikmen
Fig. 7. 5% damped acceleration response spectra plotted for different κ0 values for an earthquake of (a) Mw= 7.2, Repi= 28 km, Vs30= 340 m/s (b) Mw= 6.0,
Repi= 28 km, Vs30= 340 m/s. Solid lines are mean Sa of 100 random simulations.
Vs30 dependent site classification is utilized to constrain the site effects.
Certainly, Sa at longer structural periods would be different if site
specific amplifications are integrated into simulations. Applicability of
the equivalent point-source model for large events is not in the scope of
this research paper.
8. Conclusion
In this paper, we have presented the results of a study made to
determine the high frequency decay of the Fourier spectra of strong
ground motion data compiled from Northwestern Turkey. The data
used was recorded at three downhole arrays (both surface and base)
and outcrop stations are from 52 earthquakes with magnitudes
3.0 < ML < 4.8. Difference in kappa at base and surface stations and
its possible correlation with site resonance frequency, relative site
amplification and depth to bedrock depth are examined. The results are
also investigated to see whether they correspond to the Vs30-κ 0 equa-
tions in the literature.
The κ 0 values obtained in this study are comparable with the pre-
vious studies done for the same region except the one calculated at
surface level FTH array. κ 0 is the highest among all stations implying
low attenuation of waves in the high frequency range. FTH array also
exhibits the highest amplification among the downhole arrays. The
reason can be directly attributed to the two high impedance contrast
interfaces existing in the FTH profile [34].
One of the remarkable points derived from the analysis is the dif-
ference between κ 0 at surface and κ 0 at base. Furthermore, the differ-
ence, Δκ 0 exhibits values with similar proportions to that of the SSR’s.
However, a correlation between relative site amplification and Δκ 0 is
not attempted here. More data is needed to make a quantitative state-
ment. The lowest κ 0 value is found in the deepest array, ZYT. A cor-
relation of κ 0 with the resonant frequency of sites or with the depth to
engineering bedrock (Vs > 760 m/s) is not found. Also, a plausible re-
lationship does not seem to exist between the Vs30 and κ 0 or Δκ 0 . This is
partially attributed to the insufficiency of Vs30 as a single proxy in
seismic classification of soil deposits. Furthermore, the attenuation and
amplification of seismic waves involve a number of parameters that
cannot be packed into a single proxy as Vs30 to arrive at sound results.
Considering previous studies using downhole array data also, the
surface κ 0 values are influenced both by the soft deposits overlying the
engineering bedrock, as well as the cultural noise (built environment,
traffic etc.). Hence, the κ 0 may not reflect the true characteristics of the
high frequency decay at surface level if especially high frequency rich
noise exists in the data. This result suggests that the κ equation pro-
posed by Anderson and Hough [1], i.e. Eq. (3) needs to be extended to
reflect the local site effects as proposed by Eq. (5) in this study.
Furthermore, it should be noted that recordings of downhole arrays,
except the surface recordings, are more or less affected by reflections
from different velocity interfaces depending on the depth they are
overlain. As shown by Bonilla et al [55], if the base sensor is deployed
deep enough, recordings are not affected by the downgoing wave
413
Soil Dynamics and Earthquake Engineering 113 (2018) 406–414
effects. If this condition is not satisfied, kappa values might be over-
estimated. To reduce this effect as much as possible, selection of wider
spectral window away enough from amplification is recommended in
the literature.
Acknowledgments
Authors wish to acknowledge the financial support provided for the
deployment of the downhole arrays granted by the Scientific and
Technological Research Council of Turkey (TUBITAK) (Project no:
108M057) Istanbul Metropolitan Municipality, German Research
Center for Geosciences (GFZ) and Bogazici University. We thank Aybige
Akinci for her thorough review that significantly improved the sub-
mitted manuscript. Authors also extend their gratitude and appreciation
to D. Sinan Akkar for sharing point-source stochastic simulation code
using empirical source and duration.
References
[1]. Anderson JG, Hough SE. A model for the shape of the fourier amplitude spectrum of
acceleration at high frequencies. Bull Seismol Soc Am 1984;74:1969–93.
[2]. Campbell KW. Estimates of Shear-Wave Q and κ0 for unconsolidated and semi-
consolidated sediments in Eastern North America. Bull Seismol Soc Am
2009;99:2365–92. http://dx.doi.org/10.1785/0120080116.
[3]. van Houtte C, Drouet S, Cotton F. Analysis of the origins of κ (kappa) to compute
hard rock to rock adjustment factors for GMPEs. Bull Seismol Soc Am
2011;101:2926–41. http://dx.doi.org/10.1785/0120100345.
[4]. Laurendeau A, Cotton F, Ktenidou OJ, Bonilla LF, Hollender F. Rock and stiff-soil
site amplification: dependency on VS30 and Kappa (κ0). Bull Seismol Soc Am
2013;103:3131–48. http://dx.doi.org/10.1785/0120130020.
[5]. Bora SS, Scherbaum F, Kuehn N, Stafford P, Edwards B. Development of a response
spectral ground-motion prediction equation (GMPE) for seismic-hazard analysis
from empirical fourier spectral and duration models. Bull Seismol Soc Am
2015;105:2192–218. http://dx.doi.org/10.1785/0120140297.
[6]. Edwards B, Ktenidou OJ, Cotton F, Abrahamson N, Van Houtte C, Fäh D. Epistemic
uncertainty and limitations of the κ0 model for near-surface attenuation at hard
rock sites. Geophys J Int 2015;202:1627–45. http://dx.doi.org/10.1093/gji/
ggv222.
[7]. Boore DM. Simulation of ground motion using the stochastic method. Pure Appl
Geophys 2003;160:635–76. http://dx.doi.org/10.1007/PL00012553.
[8]. Ktenidou OJ, Gélis C, Bonilla LF. A study on the variability of Kappa (κ) in a
Borehole: implications of the computation process. Bull Seismol Soc Am
2013;103:1048–68. http://dx.doi.org/10.1785/0120120093.
[9]. Drouet S, Cotton F, Guéguen P. VS30, κ, regional attenuation and Mw from ac-
celerograms: application to magnitude 3-5 French earthquakes. Geophys J Int
2010;182:880–98. http://dx.doi.org/10.1111/j.1365-246X.2010.04626.x.
[10]. Oth A, Parolai S, Bindi D. Spectral analysis of K-NET and KiK-net data in Japan,
Part I: database compilation and peculiarities. Bull Seismol Soc Am
2011;101:652–66. http://dx.doi.org/10.1785/0120100134.
[11]. Edwards B, Fäh D, Giardini D. Attenuation of seismic shear wave energy in
Switzerland. Geophys J Int 2011;185:967–84. http://dx.doi.org/10.1111/j.1365-
246X.2011.04987.x.
[12]. Edwards B, Fäh D. A Stochastic ground-motion model for Switzerland. Bull Seismol
Soc Am 2013;103:78–98. http://dx.doi.org/10.1785/0120110331.
[13]. Omer Emre, Duman TY, Ozalp S, Saroglu F, Olgun S, Elmaci H, et al. Active fault
database of Turkey. Bull Earthq Eng 2016:1–47. http://dx.doi.org/10.1007/
s10518-016-0041-2.
[14]. Ambraseys NN, Finkel CF. Long‐term seismicity of Istanbul and of the Marmara Sea
region. Terra 1991;3:527–39. http://dx.doi.org/10.1111/j.1365-3121.1991.
tb00188.x.
[15]. Parsons T. Recalculated probability of M ≥ 7 earthquakes beneath the Sea of
G. Tanırcan, S.Ü. Dikmen
Marmara, Turkey. J Geophys Res Solid Earth 2004:109. http://dx.doi.org/10.
1029/2003JB002667.
[16]. Murru M, Akinci A, Falcone G, Pucci S, Console R, Parsons T. M ≥ 7 earthquake
rupture forecast and time-dependent probability for the sea of Marmara region,
Turkey. J Geophys Res Solid Earth 2016;121:2679–707. http://dx.doi.org/10.
1002/2015JB012595.
[17]. Ansal A, Akinci A, Cultrera G, Erdik M, Pessina V, Tönük G, et al. Loss estimation in
Istanbul based on deterministic earthquake scenarios of the Marmara Sea region
(Turkey). Soil Dyn Earthq Eng 2009;29:699–709. http://dx.doi.org/10.1016/j.
soildyn.2008.07.006.
[18]. Akinci A, Aochi H, Herrero A, Pischiutta M, Karanikas D. Physics-based broadband
ground-motion simulations for probable Mw≥7:0 earthquakes in the marmara sea
region (Turkey). Bull Seismol Soc Am 2017;107:1307–23. http://dx.doi.org/10.
1785/0120160096.
[19]. Spagnuolo E, Akinci A, Herrero A, Pucci S. Implementing the effect of the rupture
directivity on PSHA for the city of Istanbul, Turkey. Bull Seismol Soc Am
2016;106:2599–613. http://dx.doi.org/10.1785/0120160020.
[20]. Mert A, Fahjan Y, Pinar A, Hutchings L. Strong ground motion simulations around
Prince Islands Fault. Tek Dergi/Tech J Turk Chamb Civ Eng 2014;25:1757–83.
http://dx.doi.org/10.18400/TD.94957.
[21]. Inc OYO. J. Production of microzonation report and maps —on European Side
(south). Final Report to Istanbul Metropolitan Municipality. Istanbul.
[22]. Oktay FY, Gokasan E, Sakinc M, Yaltirak C, Imren C, Demirbag E. The effects of the
North Anatolian Fault zone on the latest connection between Black Sea and Sea of
Marmara. Mar Geol 2002;190:367–82. http://dx.doi.org/10.1016/S0025-
3227(01)00246-8.
[23]. Consultants IP, OYO Corporation. The Study on a Disaster Prevention/Mitigation
Basic Plan in İstanbul Including Seismic Microzonation. Istanbul; 2002.
[24]. Kudo K, Kanno T, Okada H, Ozel O, Erdik M, Sasatani T, et al. Site-specific issues
for strong ground motions during the Kocaeli, Turkey, earthquake of 17 August
1999, as inferred from array observations of microtremors and aftershocks. Bull
Seismol Soc Am 2002;92:448–65. http://dx.doi.org/10.1785/0120000812.
[25]. Ozel O, Sasatani T, Kudo K, Okada H, Kanno T, Tsuno S, et al. Estimation of S-wave
velocity structures in avcilar-Istanbul from array microtremor measurements. J
Hokkaido Univ Fac Sci Ser VII Geophys 2004;12:115–29.
[26]. Sørensen MB, Oprsal I, Bonnefoy-Claudet S, Atakan K, Mai PM, Pulido N, et al.
Local site effects in Ataköy, Istanbul, Turkey, due to a future large earthquake in
the Marmara Sea. Geophys J Int 2006;167:1413–24. http://dx.doi.org/10.1111/j.
1365-246X.2006.03204.x.
[27]. Karagöz Ö, Citak S , Chimoto K, Yamanaka H, Özel O , Yalcinkaya E. et al. Array
Observation of Microtremors in Tekirdag and Zeytinburnu (Turkey), for S-wave
Profiling. Second Eur. Conference Earthq. Eng. Seismol; 2014.
[28]. Ansal A, Kurtulus A, Tonuk G, Cetiner B. Monitoring and modeling of the local site
response during earthquakes by downhole arrays. IStanbul; 2011. Retrieved from
〈http://uvt.ulakbim.gov.tr/uvt/index.php?Cwid=9&vtadi=TPRJ&ano=147832_
733ea56c1c1ea038fc71d6e6fd305420〉.
[29]. Kurtuluş A. Istanbul geotechnical downhole arrays. Bull Earthq Eng
2011;9:1443–61. http://dx.doi.org/10.1007/s10518-011-9268-0.
[30]. Parolai S, Ansal A, Kurtulus A, Strollo A, Wang R, Zschau J. The Ataköy vertical
array (Turkey): insights into seismic wave propagation in the shallow-most crustal
layers by waveform deconvolution. Geophys J Int 2009;178:1649–62. http://dx.
doi.org/10.1111/j.1365-246X.2009.04257.x.
[31]. Parolai S, Bindi D, Ansal A, Kurtulus A, Strollo A, Zschau J. Determination of
shallow S-wave attenuation by down-hole waveform deconvolution: a case study in
Istanbul (Turkey). Geophys J Int 2010;181:1147–58. http://dx.doi.org/10.1111/j.
1365-246X.2010.04567.x.
[32]. Dikmen SU, Edincliler A, Pinar A. Northern Aegean Earthquake (Mw=6.9): ob-
servations at three seismic downhole arrays in Istanbul. Soil Dyn Earthq Eng
2015;77:321–36. http://dx.doi.org/10.1016/j.soildyn.2015.06.008.
[33]. Dikmen SU, Pinar A, Edincliler A. Near-surface attenuation using traffic-induced
seismic noise at a downhole array. J Seismol 2016;20:375–84. http://dx.doi.org/
10.1007/s10950-015-9533-9.
[34]. Dikmen SÜ, Tanırcan G. Site amplification and resonance frequency in the urban
environment. Soil Dyn Earthq Eng 2018;105:160–70. http://dx.doi.org/10.1016/j.
soildyn.2017.12.010.
414
Soil Dynamics and Earthquake Engineering 113 (2018) 406–414
[35]. Erdik M, Fahjan Y, Ozel O, Alcik H, Mert A, Gul M. Istanbul earthquake rapid
response and the early warning system. Bull Earthq Eng 2003;1:157–63. http://dx.
doi.org/10.1023/A:1024813612271.
[36]. Konno K, Ohmachi T. Ground-motion characteristics estimated from spectral ratio
between horizontal and vertical components of microtremor. Bull Seismol Soc Am
1998;88:228–41.
[37]. Brune JN. Tectonic stress and the spectra of seismic shear waves from earthquakes.
J Geophys Res 1970;75:4997–5009. http://dx.doi.org/10.1029/
JB075i026p04997.
[38]. Brune JN. J Geophys Res 1971;76:5002.
[39]. Eshelby JD. Determination of the elastic field of an ellipsoidal inclusion, and re-
lated problems. Proc R Soc Lond 1957:376–96. http://dx.doi.org/10.1098/rspa.
1957.0133.
[40]. Lai TS, Mittal H, Chao WA, Wu YM. A study on Kappa value in Taiwan using
borehole and surface seismic array. Bull Seismol Soc Am 2016;106:1509–17.
http://dx.doi.org/10.1785/0120160004.
[41]. Bora SS, Cotton F, Scherbaum F, Edwards B, Traversa P. Stochastic source, path
and site attenuation parameters and associated variabilities for shallow crustal
European earthquakes. Bull Earthq Eng 2017;15:4531–61. http://dx.doi.org/10.
1007/s10518-017-0167-x.
[42]. Akkar S, Sandıkkaya MA, Şenyurt M, Sisi AA, Ay B, Traversa P, et al. Reference
database for seismic ground-motion in Europe (RESORCE). Bull Earthq Eng
2014;12:311–39. http://dx.doi.org/10.1007/s10518-013-9506-8.
[43]. Akinci A, Malagnini L, Herrmann RB, Gok R, Sørensen MB. Ground motion scaling
in the Marmara region, Turkey. Geophys J Int 2006:166. http://dx.doi.org/10.
1111/j.1365-246X.2006.02971.x.
[44]. Askan A, Sisman FN, Pekcan O. A regional near-surface high frequency spectral
attenuation (kappa) model for northwestern Turkey. Soil Dyn Earthq Eng
2014;65:113–25. http://dx.doi.org/10.1016/j.soildyn.2014.06.007.
[45]. Chandler AM, Lam NTK, Tsang HH. Near-surface attenuation modelling based on
rock shear-wave velocity profile. Soil Dyn Earthq Eng 2006;26:1004–14. http://dx.
doi.org/10.1016/j.soildyn.2006.02.010.
[46]. Petrovic B, Parolai S. Joint deconvolution of building and downhole strong-motion
recordings: evidence for the seismic wavefield being radiated back into the shallow
geological layers. Bull Seismol Soc Am 2016;106:1720–32. http://dx.doi.org/10.
1785/0120150326.
[47]. Petrovic B, Parolai S, Pianese G, Dikmen SU, Moldobekov B, Orunbaev S, et al.
Joint deconvolution of building and downhole seismic recordings: an application to
three test cases. Bull Earthq Eng 2017. http://dx.doi.org/10.1007/s10518-017-
0215-6.
[48]. Akinci A, D’Amico S, Malagnini L, Mercuri A. Scaling earthquake ground motions
in western Anatolia, Turkey. Phys Chem Earth, Parts A/B/C 2013;63:124–35.
http://dx.doi.org/10.1016/j.pce.2013.04.013.
[49]. Kurtulmuş TÖ, Akyol N. Crustal attenuation characteristics in western Turkey.
Geophys J Int 2013;195:1384–94. http://dx.doi.org/10.1093/gji/ggt318.
[50]. Gueguen P, Bard P-Y, Oliveira C. Experimental and numerical analysis of soil
motions caused by free vibrations of a building model. Bull Seismol Soc Am
2000;90:1464–79. http://dx.doi.org/10.1785/0119990072.
[51]. Guéguen P, Bard PY, Chávez-García FJ. Site-city seismic interaction in Mexico City
- Like environments: an analytical study. Bull Seismol Soc Am 2002;92:794–811.
http://dx.doi.org/10.1785/0120000306.
[52]. Erdik M, Demircioglu M, Sesetyan K, Durukal E, Siyahi B. Earthquake hazard in
Marmara Region, Turkey. Soil Dyn Earthq Eng 2004;24:605–31. http://dx.doi.org/
10.1016/j.soildyn.2004.04.003.
[53]. Ergintav S, Reilinger RE, Cakmak R, Floyd M, Cakir Z, Dogan U, et al. Istanbul’s
earthquake hot spots: geodetic constraints on strain accumulation along faults in
the Marmara seismic gap. Geophys Res Lett 2014;41:5783–8. http://dx.doi.org/10.
1002/2014GL060985.
[54]. Bora SS, Scherbaum F, Kuehn N, Stafford P. Fourier spectral- and duration models
for the generation of response spectra adjustable to different source-, propagation-,
and site conditions. Bull Earthq Eng 2014;12:467–93. http://dx.doi.org/10.1007/
s10518-013-9482-z.
[55]. Bonilla LF, Steidl JH, Gariel JC, Archuleta RJ. Borehole response studies at the
Garner Valley Downhole Array, Southern California. Bull Seismol Soc Am
2002;92:3165–79. http://dx.doi.org/10.1785/0120010235.