Vous êtes sur la page 1sur 9

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/267561864

Contribution of site effects and soil-structure


interaction to building damage in Golcuk,
Turkey, during the 1999 Kocaeli...

Conference Paper June 2005

CITATIONS READS

0 27

2 authors:

Hiroshi Arai Nelson Pulido


Building Research Institute, Japan National Research Institute for Earth Science
38 PUBLICATIONS 364 CITATIONS 54 PUBLICATIONS 412 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Application of State of the Art Technologies to Strengthen Research and Response to Seismic,
Volcanic and Tsunami Events, and Enhance Risk Management in the republic of Colombia View
project

International Seismic Network ISN-NIED View project

All content following this page was uploaded by Nelson Pulido on 09 December 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Proc. ICOSSAR 2005, G. Augusti, G. I. Schueller, M. Ciampoli (eds) / Millpress, Rotterdam, ISBN 90 5966 040 4

Contribution of site effects and soil-structure interaction to building


damage in Golcuk, Turkey, during the 1999 Kocaeli earthquake
Hiroshi Arai & Nelson Pulido
Earthquake Disaster Mitigation Research Center, National Research Institute for Earth Science and Disaster Prevention, Kobe,
Japan

Keywords: the 1999 Kocaeli earthquake, building damage, site effect, soil-structure interaction, microtremor

ABSTRACT: The effects of the S-wave velocity (VS) profiles on the R/C building damage in Golcuk, Turkey,
during the 1999 Kocaeli earthquake are examined considering soil-structure interaction (SSI). Microtremor ar-
ray measurements were conducted at seven sites, and a joint inverse analysis of observed dispersion curves
and horizontal-to-vertical (H/V) spectral ratios accurately results in VS profiles down to seismic bedrock at the
sites. With the estimated VS profiles and a source asperity model of the main shock, we performed a strong
ground motion simulation at the sites. Based on the simulated ground motions and the observed damage sta-
tistics at the sites, simplified SSI analytical models of Turkish R/C buildings groups are identified for evaluat-
ing the building damage ratios. The results of parametric studies using the identified buildings models indi-
cate that the damage ratios for 1-7 storied R/C buildings are amplified by a factor of 3-10 or more due to the
site effects, however, those for 1-4 storied buildings get smaller by a factor of 0.5-0.8 due to the SSI effects.

1 INTRODUCTION Recent studies, for example, have shown that (1)


surface (Rayleigh and Love) waves dominate in mi-
The Kocaeli earthquake of August 17, 1999 de- crotremors, (2) the frequency-wave number (F-k)
stroyed over 60,000 houses and buildings in the spectral analysis (Capon, 1969) of microtremor ver-
northwest area of Turkey. In Golcuk, in particular, a tical motions measured with arrays of sensors at a
large number of low- and medium-rise buildings
sustained either partial or complete collapse typi-
cally of a soft first story. Fig. 1 shows spatial distri-
bution of collapse ratios for the low- and medium-
rise reinforced-concrete (R/C) buildings in Golcuk,
which is based on the results of reconnaissance sur-
vey performed by Architectural Institute of Japan
(AIJ) (AIJ et al., 2001). The damage to R/C build-
ings was concentrated at several areas in the north of
Ataturk street, which is the main street running in
the east-west (EW) direction. The concentration of
building damage could be due to the effects of un-
known underground structures on seismic motions,
i.e., site effects. In fact, most of the northern area of
the main street is located on a plain while the south
is on a hill, where the building damage was slight.
To evaluate site effects quantitatively, S-wave ve-
locity (VS) profiles of sedimentary deposits should
be properly determined down to seismic bedrock. It
is, however, too expensive and time-consuming to
estimate VS profile using conventional geophysical
or geotechnical methods with boreholes. As an eco-
Figure 1. Map showing microtremor array observation sites
nomical and practical substitute, microtremor meas- (Sites A-G) and distribution of collapse ratios for low- and me-
urements that can be readily performed on the dium-rise R/C buildings in Golcuk during the 1999 Kocaeli
ground surface have been used. earthquake (AIJ et al., 2001).
- 111 -
site can yield dispersion characteristics of Rayleigh The H/V spectral ratios are also derived for mi-
waves, and (3) the inverse analysis of dispersion crotremors observed at all the sensors in the arrays
data results in a sedimentary VS structure at the site used. The definition of microtremor H/V spectral ra-
(e.g., Horike, 1985; Matsushima and Okada, 1990; tio, (H/V)m, used in this study is
Tokimatsu et al., 1992). It was also revealed that the
horizontal-to-vertical (H/V) spectral ratios (Naka- PNS + PEW
mura, 1989) of microtremors measured with one ( H / V )m = (1)
PUD
three-component sensor at a site correspond to those
of surface waves and reflect a bedrock VS structure where PUD is the Fourier power spectrum of mi-
at the site (e.g., Tokimatsu and Miyadera, 1992; To- crotremor vertical motion, and PNS and PEW are those
kimatsu, 1997; Arai and Tokimatsu, 2004). These of two orthogonal horizontal motions (e.g., Arai and
findings indicate that the joint inverse analysis of Tokimatsu, 2004). With the H/V spectral ratios at
both microtremor dispersion curve and H/V spec- the sensors, their average spectrum and standard de-
trum can be promising to estimate the VS profile viations are determined for each Site A-G. The re-
down to bedrock at a site. sulting H/V spectra at Sites B and E, for example,
The objectives of this article are to introduce the are shown in Fig. 3 as open circles and thin lines.
use of the microtremor methods for estimating VS Using the observed dispersion and H/V data at
structures and to examine the effects of the VS pro- Sites A-G, the joint inverse analyses are conducted
files on the strong ground motions and building with the theoretical formulas of Rayleigh wave dis-
damage ratios in Golcuk during the 1999 earthquake, persion curve and surface wave H/V spectrum con-
considering soil-structure interaction (SSI). sidering the effects of fundamental and higher
modes (Tokimatsu et al., 1992; Tokimatsu, 1997;
2 MICROTREMOR MEASUREMENTS Arai and Tokimatsu, 2004). In the joint inverse ana-
lysis, the generalized (non-linear) least-squares
Microtremor measurements using two-dimensional method using the adaptive biweight estimation
horizontal arrays of sensors were conducted at seven (Tukey, 1974) and the modified Marquardts tech-
sites, subsequently called Sites A, B, C, D, E, F, and nique combined with the singular value decomposi-
G. As shown in Fig. 1, Sites A-F are located in the tion (e.g., Wiggins, 1972; Yuan and Nazarian, 1993)
heavily damaged areas on the plain and their damage is employed. The soil structure at each site is as-
ranks vary drastically depending on the site geologi- sumed to be horizontally stratified and consists of 6-
cal conditions, while Site G is on the hill where the 8 layers in the inversion. Further details of the in-
building damage was rarely seen. verse analysis used in this study can be found else-
The measurement system used consists of ampli- where (Arai and Tokimatsu, 2004).
fiers, lowpass-filters, 24-bit A/D converters, and a
note-type computer, all built in a portable case; and
three-component velocity sensors with a natural pe- 1000
(a) Site B (b) Site E
riod of 1 s. Six sensors were placed on the ground 800

surface to form a circular array with a sensor in the 600

center. The minimum array radius used was 2.5 m at 400


each site, and the maximum ones were 41, 103, 108,
200
31, 39, 42, 10 m at Sites A-G, respectively. With Observed
Inverted
Observed
Inverted
each array, microtremors were measured simultane- 0
0.02 0.1
Period (s)
1 5 0.02 0.1
Period (s)
1 5

ously and digitized at an equal sampling rate. The


sampling rate varied from 100 to 500 Hz, depending Figure 2. Dispersion curves of microtremor vertical motions
compared with those of Rayleigh waves theoretically computed
on the site conditions and array radius used. About for the inverted soil layer models at Sites B and E.
20-40 sets of data segments with 2048 or 4096
points each were selected from the digitized motions,
and used for the following spectral analyses. 10
(a) Site B (b) Site E
Observed
Std. Deviation
Inverted
3 VS PROFILES FROM MICROTREMOR F-K 1

AND H/V ANALYSES


Observed
Std. Deviation
The high-resolution F-k spectral analysis (Capon, 0.1
Inverted
0.02 0.1 1 5 0.02 0.1 1 5
1969) is used to determine dispersion curves of mi- Period (s) Period (s)
crotremor vertical motions recorded at Sites A-G. Figure 3. H/V spectra of microtremors compared with those of
The resulting dispersion curves at Sites B and E, for surface waves theoretically computed for the inverted soil layer
example, are shown in Fig. 2 as open circles. models at Sites B and E.

- 112 -
S-Wave Velocity (m/s) S-Wave Velocity (m/s) S-Wave Velocity (m/s) Table 1. Asperity parameters of source fault model in the 1999
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
0 Kocaeli earthquake (after Pulido, 2004). The number in paren-
(a) Site A (b) Site B (c) Site C
theses is variation coefficient (uncertainty) of the parameter.
50
3.5 km/s West (+/-50%)
100 Rupture Velocity
4.8 km/s East (+/-50%)
150 Rise Time 2.5-4.5 s (+/-50%)
Stress Drops of
200 140, 148, 271 bar (+/-50%)
Asperities 1-3
250 Stress Drops of
22, 92, 128 bar (+/-50%)
Segments 1-3
300
0 f max 5 Hz (+/-20%)
(d) Site D (e) Site E (f) Site F
a b a = 7.5 (+/-30%)
50 1 / [1+( f / f max) ]
b = 0.47 (+/-20%)
100 = 113 (+/-50%)
Q=f
= 1.2 (+/-30%)
150
Segment 4 is not used because of its small contribution.
200

250

300
used are listed in Table 1, which were inferred to
minimize the misfits between the observed and
Figure 4. S-wave velocity profiles estimated from microtremor simulated waveforms at several observation stations.
array observations at Sites A-F. At Site G, it is estimated that a Further details of the simulation procedure used can
base-rock with VS over 1.3 km/s is outcropping.
be found elsewhere (Pulido et al., 2004).
Using the simulated bedrock motions at Sites A-
G, the one-dimensional (1-D) effective stress analy-
ses are then conducted for the estimated VS profiles
at the sites. In the analysis, the modified Ramberg-
Osgood model is used for the stress-strain relations
of soils. The cyclic stress ratios causing liquefaction
are inferred from the geological information of Gol-
cuk (AIJ et al., 2001), using the equations proposed
by Tokimatsu and Yoshimi (1983). Further details of
Figure 5. Multi-asperity source fault model in the 1999 Kocaeli the effective stress analysis employed in this study
earthquake (after Pulido, 2004; Sekiguchi and Iwata, 2002).
can be found elsewhere (Shamoto et al., 1992).
Fig. 6 shows the computed ground velocity mo-
tions in the north-south (NS) and EW directions at
Fig. 4 shows the inverted VS profiles at Sites A-F. Sites A-G. Figs. 7(a)-(c) show the peak acceleration,
At Site G, it is estimated that a base-rock with VS velocity, and shear strain profiles, respectively, for
over 1.3 km/s is outcropping. The Rayleigh wave the larger NS motions obtained from the effective
dispersion curves and surface wave H/V spectra for stress analyses at the sites. The analyses results indi-
the inverted soil profiles at Sites B and E are shown cate that little part of subsurface soils liquefied at the
in Figs. 2 and 3 as solid lines, respectively. The sites. However, at the sedimentary sites (Sites A-F),
theoretical dispersion curves and H/V spectra show the peak shear strain values of subsurface soils are
fairly good agreement with the observed ones at the over 1%, and the peak velocity values on the ground
sites, indicating that the inverse analyses have been are amplified by a factor of 1.5-2 as compared with
performed with a reasonable degree of accuracy and those at a depth of 50m. This reveals that the ground
that the estimated VS profiles could be reliable. motions at Sites A-F are affected significantly by the
non-linear soil behavior during the earthquake.

4 GROUND MOTION CHARACTERISTICS IN 4.2 Comparison with Building Damage


THE 1999 KOCAELI EARTHQUAKE Recent studies have indicated that the damage rank
of R/C buildings during earthquake is consistent
4.1 Strong Ground Motion Simulation with the acceleration response spectral value, SAE, of
To evaluate strong ground motions in Golcuk during strong ground motions at the equivalent-damage pe-
the main shock, unknown bedrock outcrop motions riods of the buildings (Sakai et al., 2001). They have
at each site are first determined by a broadband fre- also revealed that the equivalent-damage period of
quency seismic motion simulation using a multi- R/C buildings is about three times longer than the
asperity source fault model shown in Fig. 5 (Pulido, fundamental (elastic) one. Fig. 8 shows the variation
2004), which is based on an inverted slip model (Se- of the R/C building existence ratios in Golcuk (AIJ
kiguchi and Iwata, 2002). The asperity parameters et al., 2001) with its fundamental period, Tb, which
- 113 -
150 0.4
(a) NS Site A
Site C
Site B
Site D T = 0.07N
b
0
0.3
Site E Site F 5
Site G
-150
150 4
(b) EW 0.2

0 2
3 6
0.1
-150
0 5 10 15 20 25 30 35 40 45 50 7
N=1 89
Time (s) 0
>9
0 0.2 0.4 0.6 0.8 1 1.2
Figure 6. Ground velocity motions in NS and EW directions Fundamental Period, T (s)
b
obtained from strong ground motion simulation for Sites A-G.
Figure 8. Variation of R/C building existence ratios in Golcuk
(AIJ et al., 2001) with its fundamental period.
2
Peak Acc. (cm/s ) Peak Vel. (cm/s) Peak Shear Strain
-4 -3 -2 -1
0 200 400 600 0 50 100 150 10 10 10 10
0 (a) 0 (b) 0
2000
50 50 50 Site A Site D
Site B Site E
Site C Site F
100 100 100 1500 Site G

150 150 150 (c)


Site A 1000
200 200 200 Site B
Site C
Site D
250 250 250 Site E
Site F 500
Site G
300 300 300

Figure 7. Peak acceleration, velocity, and shear strain profiles 0


0 0.2 0.4 0.6 0.8 1 1.2
for NS direction motions computed from 1-D effective stress Period (s)
analyses at Sites A-G. Figure 9. Acceleration response spectra (h=5%) of NS ground
motions estimated at Sites A-G.

is based on the following empirical formula (e.g., 2000


Kobayashi et al., 1996): Simulated (Deterministic)
with Src.+Site Errors
(Std. Deviation)

Tb = 0.07 N (2) 1500 E

C
where N is the number of building story. In the fig- 1000 A
ure, the most R/C buildings in Golcuk are 2-6 sto- F D
B
ried and their fundamental periods are about 0.2-0.4 500

s. Thus, the equivalent-damage periods of the R/C G


with Src.+Site Errors
(Max. Dispersion)
buildings in Golcuk range in about 0.6-1.2 s. 0
1 2 3 4 5
Fig. 9 shows the acceleration response spectra Damage Rank
(h=5%) of the larger NS ground motions estimated Figure 10. Acceleration response spectral values at periods of
at Sites A-G. Based on the response spectra and the 0.6-1.2 s of ground motions estimated at Sites A-G and their
equivalent-damage periods, the SAE values at periods relationships with site damage ranks in Fig. 1 (AIJ et al., 2001).
of 0.6-1.2 s are evaluated for the sites and their rela-
tionships with the site damage ranks (Fig. 1, AIJ et
al., 2001) are illustrated in Fig. 10 as open circles. damage ratios near the sites are determined for two
The evaluated SAE values for Sites A-G are fairly damage criteria, slightly and heavily ones, which are
consistent with the observed damage ranks at the defined as the cases that the R/C building damage
sites, indicating that the estimated VS profiles and grades judged using the EMS-98 (European Seis-
strong ground motions could be reasonably reliable. mological Commission, 1998) are over G1 and G3,
respectively. In Fig. 11, the R/C building damage ra-
tios for the slightly and heavily criteria near Sites A-
5 DAMAGE RATIO SIMULATION OF R/C G are shown as open and solid circles, respectively.
BUILDINGS IN GOLCUK The building damage ratios for both criteria are dis-
persive at any of the 1-7 story numbers.
5.1 Building Damage Statistics
5.2 Buildings Group Models with Soil-Structure
In order to examine a possibility for estimating R/C
Interaction for Damage Ratio Evaluation
building damage ratios during earthquake, the results
of damage investigation by AIJ (AIJ et al., 2001) are Using the observed damage ratios in Fig. 11 and the
first rearranged for Sites A-G. In this study, the estimated ground motions at Sites A-G, simplified
- 114 -
analytical models of Turkish R/C buildings groups 1
(29)(104)(110) (97) (148) (76) (12) (0) (0) (0)

are identified to simulate the observed damage ratios 0.9


>G1 (G2+G3+G4+G5)
>G3 (G4+G5)
at the sites, based on the earthquake response analy- 0.8

sis considering soil-structure interaction (SSI). The 0.7


0.6
procedure of SSI buildings group modeling and iden- 0.5
tification has been basically proposed by Nagato and 0.4
Kawase (2001), and the outline is as follows: 0.3
0.2
1) The multi-degrees-of-freedom (MDOF) build- 0.1

ing systems with degrading tri-linear force- 0


1 2 3 4 5 6 7 8 9 10
displacement relations and sway-rocking soil springs Number of Story, N
(e.g., Parmelee, 1970) are inferred to represent stan- Figure 11. Variation of R/C building damage ratios near Sites
dard 1-7 storied R/C buildings with SSI (Fig. 12(a)). A-G with story number (AIJ et al., 2001). The number in pa-
In the figure, building yield strength (base shear co- rentheses on the top of the figure is that of total building sam-
efficient), cy, is not determined. The existing prob- ples for each story.
ability density distribution for the yield strengths of
buildings group is then assumed to be the log-
normal one having a reference value of cy0 (Fig. (a)
12(b)), which is based on the results of R/C building
investigations by Shibata (1980). This leaves only
unknown reference yield strength of buildings group,
cy0, to be identified.
2) The strength demands of the 1-7 storied
MDOF building systems with SSI, cyD, are computed
for the simulated ground motions at Sites A-G. In
the analyses, the required displacement ductility de-
mands for the slightly and heavily damage criteria
are assumed to be equal to 1.5 and 3.5, respectively
(e.g., Sakai et al., 2001).
3) Using the cyD value for each damage criterion
and the inferred yield strength distribution, p(cy/cy0),
the building damage radio, RD-SSI, is presented as:
(b) 1
0.9
c yD
RD SSI =
0.8
p ( x c y 0 ) dx (3) 0.7
0 0.6
0.5
for each story number. The reference yield strength 0.4

value of buildings group, cy0, is then sought using 0.3


0.2
the iteration analysis to minimize the misfit between 0.1

the computed damage ratio RD-SSI and the observed 0


0 0.5 1 1.5 2 2.5 3 3.5 4
Normarized Base Shear Coef.
one. Repeating the analyses for all the damage ratios
data of the 1-7 storied R/C buildings shown in Fig. Figure 12. R/C buildings group model with SSI. (a) Multi-
11 and averaging the analyses results, thus, the stan- degrees-of-freedom building system with sway-rocking soil
dard models of the 1-7 storied Turkish R/C buildings springs. (b) Existing probability density distribution for nor-
groups with SSI are finally identified. malized yield strength of R/C buildings (after Shibata, 1980).

Open and solid circles in Fig. 13 show the identi- 1


fied reference yield strengths of the 1-7 storied R/C 0.9
buildings in Golcuk, cy0, for the slightly and heavily 0.8

damage criteria, respectively. For both damage crite- 0.7


0.6
ria, the estimated cy0 values are concentrative and
0.5
show a distinguishable trend in which they decrease 0.4
with increasing the number of building story, N. 0.3 Estimated from
Based on the results, regression curves for standard 0.2 >G3 (G4+G5)
>G1 (G2+G3+G4+G5)
cy0-N relationships of the 1-7 storied R/C buildings 0.1 0.39
Regression (C =0.78/N )
y0

in Golcuk are then determined as 0


1 2 3 4 5 6 7 8 9 10
Number of Story, N
c y 0 = 0.78 N 0.39 0.05 (standard deviation) (4) Figure 13. Reference yield strengths of R/C buildings with SSI
in Golcuk and their regression curves identified using observed
and shown in Fig. 13 as solid and broken lines. damage statistics and simulated ground motions at Sites A-G.
- 115 -
Thick solid lines in Fig. 14 show the R/C building 1
(b) Site A (>G3)
damage ratios for the slightly and heavily damage 0.9
0.8
criteria computed from Eq. (3) using the 1-7 storied 0.7
SSI buildings group models with the identified cy0-N 0.6
relations in Eq. (4), compared with the observed 0.5
damage ratios for the criteria at Sites A, B, and G. 0.4
For each site and damage criterion, the computed 0.3

building damage rations are fairly consistent with 0.2 Observed (>G1)
0.1 (a) Site A (>G1)
the observed ones, indicating that the identified SSI 0
buildings group models are promising to estimate 1
(c) Site B (>G3) (d) Site G (>G3)
0.9
Turkish R/C building damage ratios during earth- 0.8
Observed (>G3)

quake, and also confirming again that the estimated 0.7


Simulated
(Deterministic)
VS profiles and the simulated strong ground motions 0.6
with Bldg. Error
with Src.+Site+Bldg. Errors
at the sites are reasonably reliable. 0.5 (Std. Deviation)
with Src.+Site+Bldg. Errors
0.4 (Max. Dispersion)

5.3 Effects of VS Profiles and Soil-Structure 0.3


0.2
Interaction on Building Damage Ratios 0.1
Imaging the cases that the sedimentary VS profiles 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
and/or SSI at Sites A-G are missing, the correspond- Number of Story, N Number of Story, N
ing 1-7 storied R/C building damage ratios, RBD-SSI, Figure 14. Simulated R/C building damage ratios computed
RD-FIX, and RBD-FIX, are similarly computed for the from Eq. (3) using 1-7 storied SSI buildings group models with
slightly and heavily damage criteria, where the su- identified cy0-N relations in Eq. (4), compared with observed
perscript B and the subscript FIX indicate the condi- damage ratios at Sites A, B, and G.
tions that the VS profile and SSI at the site are ig-
nored, respectively. 1000
Fig. 15 shows the variations of the RD-FIX / RBD-FIX (a) >G1 (G2+G3+G4+G5) (b) >G3 (G4+G5)
values simulated for the slightly and heavily damage Site A
Site B
Site E
Site F
100
criteria at the sites with number of building story, Site C
Site D
Site G

thus, presenting the site effects on the 1-7 storied


10
R/C building damage ratios in Golcuk. In the figure,
for the slightly and heavily criteria, the R/C building
damage ratios at the sedimentary sites (Sites A-F) 1

are amplified by factors of 1-10 and 3-30 or more,


respectively, as compared with those at the rock- 0.1
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
outcropping site (Site G). Number of Story, N Number of Story, N
Fig. 16 shows the variations of the RD-SSI / RD-FIX Figure 15. Variations of RD-FIX / RBD-FIX values simulated at
values computed for the both damage criteria at the Sites A-G with number of building story, thus, site effects on
sites with number of building story, thus, meaning 1-7 storied R/C building damage ratios in Golcuk. The damage
the SSI effects on the R/C building damage ratios in ratios at the sedimentary sites (Sites A-F) are amplified by a
Golcuk. In the figure, for both criteria, the damage factor of 3-10 or more due to the site effects, as compared with
ratios for 1-4 storied (low-rise) R/C buildings get those at the rock-outcropping site (Site G).
smaller by a factor of 0.5-0.8 due to the SSI effects
at the sedimentary sites (Sites A-F), however, those 2
for 5-7 storied buildings do not. (a) >G1 (G2+G3+G4+G5) Site A
Site B
Site E
Site F
Site C Site G
1.5 Site D
5.4 Effects of Source, Site, and Structural Un-
certainties on Building Damage Ratios
1
To investigate the effects of the uncertainties of seis-
mic source fault, sedimentary soil profile (site), and 0.5
structural performance on the R/C building damage
(b) >G3 (G4+G5)
ratio evaluation in this study, the Monte Carlo simu- 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
lation is performed. In the simulation, variable coef- Number of Story, N Number of Story, N
ficient values of the source asperity parameters used,
based on seismological considerations, are listed in Figure 16. Variations of RD-SSI / RD-FIX values simulated at Sites
Table 1 as the numbers in parentheses. Those of the A-G with number of building story, thus, SSI effects on R/C
building damage ratios in Golcuk. The damage ratios for 1-4
soil layer (site) parameters (thickness, density, initial storied (low-rise) R/C buildings get smaller by a factor of 0.5-
S-wave velocity, shear strength, reference shear 0.8 due to the SSI effects at the sedimentary sites (Sites A-F),
strain, and maximum damping ratio) and the build- however, those for 5-7 storied buildings do not.

- 116 -
ing yield strengths used in the simulation are 20% at the sites. With the estimated VS profiles and a
and 10%, respectively, which are based on the stan- source asperity model of the 1999 Kocaeli earth-
dard error values of the inverted VS profiles at Sites quake, we performed a strong ground motion simu-
A-G and those of the identified Turkish R/C build- lation at the sites. The acceleration response spectral
ing yield strengths shown in Fig. 13. In the simula- values of the simulated ground motions are fairly
tion, 10 asperity models, 20 soil profiles at each site, consistent with the observed damage ranks at the
and 10 SSI building models for each story number sites, indicating that the estimated VS profiles and
are generated randomly using the uniform probabil- ground motions could be reasonably reliable.
ity density distribution. Thus, 200 different strong Based on the simulated ground motions and the
ground motions are calculated at each site and 2,000 observed damage statistics at the sites, simplified
case building damage ratios are computed for each SSI analytical models of 1-7 storied Turkish R/C
story number at each site. buildings groups are identified for evaluating the
Chained lines and vertical error bars shown in Fig. building damage ratios. With the identified buildings
10 are respectively the standard deviations and models, the effects of the VS profiles and SSI on the
maximum dispersions of the SAE values for the R/C building damage ratios are examined parametri-
Monte Carlo simulated ground motions at Sites A-G. cally. The examination results indicate that the dam-
At the rock site (Site G), probable SAE error ratios of age ratios for 1-7 storied R/C buildings at the sedi-
the simulated ground motions are less than 20%. mentary sites are amplified by a factor of 3-10 or
This is due only to the asperity uncertainties. At the more due to the site effects, however, those for 1-4
sedimentary sites (Sites A-F), however, the SAE error storied buildings get smaller by a factor of 0.5-0.8
ratios enlarge because of the site uncertainties, and due to the SSI effects. In addition, the effects of the
their standard deviations and maximum dispersions source, site, and structural uncertainties on the build-
are about 20% and 50%, respectively. ing damage ratio evaluation in this study are investi-
Broken lines in Fig. 14 show the Monte Carlo gated using the Monte Carlo simulation. The simula-
simulated R/C building damage ratios at Sites A, B, tion results indicate that the reduction of the
and G, considering the structural uncertainties only. uncertainties of the source and site parameters is one
Chained and thin lines in the figure are respectively of the key components for predicting ground mo-
the standard deviations and maximum dispersions of tions and building damage in a future earthquake.
the simulated R/C building damage ratios at the sites,
with all the physical uncertainties considered. In the
figure, the similar trends indicated in Fig. 10 can be ACKNOWLEDGEMENTS
again confirmed. Thus, the damage estimation error
ratios due to the building uncertainties are less than The authors would like to express their sincere
about 10-20%, however, those adding the source and thanks and appreciation to Prof. Tetsuo Kubo, Uni-
site uncertainties are further amplified at a maximum versity of Tokyo, Japan, Assoc. Prof. Ryosuke Uzu-
of 50-80%. These results indicate that the uncertain- oka, Tohoku University, Japan, and Drs. Yalikun
ties of seismic source and, in particular, sedimentary Yusufu, Khosrow T. Shabestari, and Kangning Li
soil profile have significant effects on the R/C build- (formally Earthquake Disaster Mitigation Research
ing damage ratio evaluation in this study. Center, National Research Institute for Earth Science
The deterministic estimation of VS profiles and and Disaster Prevention, Japan), for their valuable
ground motions are fairly consistent with the ob- assistance in the microtremor measurements.
served damage ranks at the study sites. However, for
a prediction analysis in a future earthquake, the large
uncertainties in the source asperity location and size REFERENCES
might drastically change the input seismic bedrock Arai, H. & Tokimatsu, K. 2004. S-wave velocity profiling by
waveforms (e.g., Pulido et al., 2004), and therefore, inversion of microtremor H/V spectrum. Bull. Seism. Soc.
the overall damage assessment prediction accuracy. Am. 94(1): 53-63.
This suggests that the reduction of the uncertainties Architectural Institute of Japan, Japan Society of Civil Engi-
of the source and site parameters is one of the key neers & The Japanese Geotechnical Society. 2001. Report
components for predicting ground motions and on the damage investigation of the 1999 Kocaeli earth-
quake in Turkey. 465 pages.
building damage in a future earthquake. Capon, J. 1969. High-resolution Frequency-wavenumber spec-
trum analysis. Geophysics. 34(1): 21-38.
European Seismological Commission (Grunthal, G., Editor).
6 CONCLUSIONS 1998. European Macroseismic Scale 1998 (EMS-98). Lux-
embourg.
Microtremor array measurements were conducted at Horike, M. 1985. Inversion of phase velocity of long-period
seven sites in Golcuk, Turkey, and a joint inverse microtremors to the S-wave-velocity structure down to the
basement in urbanized area. J. Phys. Earth. 33: 59-96.
analysis of observed dispersion and H/V data accu- Kobayashi, H. et al. 1996. Evaluation of dynamic behavior of
rately results in VS profiles down to seismic bedrock building structures with microtremors for seismic micro-

- 117 -
zonation mapping. Proc. CD-ROM. 11th WCEE: Paper ID isting soil deposit. J. Struct. Constr. Eng. AIJ. 433: 113-119
1769. (in Japanese with English abstract).
Matsushima, T. & Okada, H. 1990. Determination of deep geo- Shibata, A. 1980. Prediction of the probability of earthquake
logical structures under urban areas. BUTSURI-TANSA. damage to reinforced concrete building groups in a city.
43(1): 21-33. Proc. 7th WCEE. 4: 395-402.
Nagato, K. & Kawase, H. 2001. Damage evaluation models of Tokimatsu, K. 1997. Geotechnical site characterization using
reinforced concrete buildings based on the building damage surface waves. Proc. 1st Intl. Conf. Earthquake Geotechni-
statistics and simulated strong motions. J. Struct. Constr. cal Engineering. 3: 1333-1368.
Eng. AIJ. 544: 31-37 (in Japanese with English abstract). Tokimatsu, K. & Yoshimi, Y. 1983. Empirical correlation of
Nakamura, Y. 1989. A method for dynamic characteristics es- soil liquefaction based on SPT N-value and fines content.
timation of subsurface using microtremor on the ground Soils and Foundations. JGS. 23(4): 56-74.
surface. Quarterly Report of the Railway Technical Re- Tokimatsu, K. & Miyadera, Y. 1992. Characteristics of
search Institute. 30(1): 25-33. Rayleigh waves in microtremors and their relation to un-
Parmelee, R. A. 1970. The influence of foundation parameters derground structures. J. Struct. Constr. Eng. AIJ. 439: 81-
on the seismic response of interaction systems. Proc. 3rd 87 (in Japanese with English abstract).
Japan Earthquake Engineering Symposium. Tokimatsu, K., Shinzawa, K., & Kuwayama, S. 1992. Use of
Pulido, N. 2004. Broadband frequency asperity parameters of short-period microtremors for VS profiling. J. Geotech. Eng.
crustal earthquakes from inversion of near-fault ground mo- ASCE. 118(10): 1544-1588.
tion. Proc. CD-ROM. 13th WCEE: Paper ID 751. Tukey, J. E. 1974. Introduction to todays data analysis. Proc.
Pulido, N. et al. 2004. Strong ground motion estimation in the the Conference on Critical Evaluation of Chemical and
sea of Marmara region (Turkey) based on a scenario earth- Physical Structural Information. D. R. Lide, Jr. and M.A.
quake. Tectonophysics. 391: 357-374. Paul (Editors), National Academy of Sciences, Washington,
Sakai, Y. et al. 2001. Investigation on indices of representing D. C.: 3-14.
destructive power of strong ground motions to estimate Wiggins. R. A. 1972. The general linear inversion problem:
damage to buildings based on the 1999 Chi-chi earthquake, Implication of surface waves and free oscillations for earth
Taiwan. J. Struct. Constr. Eng. AIJ. 549: 43-50 (in Japa- structure. Reviews of Geophysics and Space Physics. 10(1):
nese with English abstract). 251-285.
Sekiguchi, H. & Iwata, T. 2002. Rupture process of the 1999 Yuan, D. & Nazarian, S. 1993. Automated surface wave
Kocaeli, Turkey, earthquake estimated from strong-motion method: Inversion technique. J. Geotech. Eng. ASCE.
waveforms. Bull. Seism. Soc. Am. 92(1): 300-311. 119(7): 1112-1126.
Shamoto, Y., Tokimatsu, K. & Ariizumi, K. 1992. Applicabil-
ity of a one-dimensional effective stress analysis to an ex-

- 118 -
View publication stats

Vous aimerez peut-être aussi