Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

3D ANALYSIS OF S-WAVE PROPAGATION IN SOFT DEPOSITS

By Masahiro Iida1

ABSTRACT: During the 1985 Michoacan earthquake (surface-wave magnitude = 8.1), large-amplitude seis-
mograms of extremely long duration were recorded in the lake-bed zone of Mexico City. Large amplification
in soft surficial deposits has not been well understood. Ground subsidence of the soft clay deposits might heavily
influence S-wave propagation. In the present study, we attempt to explain both S-wave amplification and prop-
agation time between surface and borehole recordings under the action of gravity on the basis of a 3D nonlinear
finite-element (FE) technique where a bilinear model with a Mohr-Coulomb criterion is used. We identify body-
wave (S-wave) time sections in the recordings with a cross-correlation analysis. The FE simulation successfully
predicts the S-wave amplification and propagation time, which cannot be explained by a purely 1D linear model.
Although subsidence changes S-wave linear amplification, S-wave propagation time decreases little.
Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

INTRODUCTION a 1D analysis of borehole seismograms of the May 31, 1990,


The September 19, 1985 Michoacan earthquake (Ms = 8.1) earthquake (Ms = 5.8), main parts of the seismograms behave
caused very severe damage in Mexico City inside the Valley similar to body waves at lake-bed stations. Iida et al. (1994)
of Mexico, which is approximately 400 km from the epicenter analyzed borehole data of long duration at the Tlacotal (lake-
in the Pacific Ocean. The Valley of Mexico is divided into bed) station from the October 24, 1993, earthquake (Ms = 6.7)
three geotechnical zones: the hill, transition, and lake-bed and indicated that only short-period waves of <2.0 s might be
zones (Fig. 1). While the hill zone occupies the southwestern interpreted as body waves. Bodin et al. (1997) analyzed bore-
areas, the lake-bed zone extends to the central and northeastern hole seismograms at the Roma-C station in the lake-bed zone
areas. The transition zone is located between these two zones. to estimate dynamic deformations of shallow sediments. Iida
Although the earthquake produced very low accelerations at and Kinoshita (unpublished paper, 1999) applied a cross-cor-
epicentral stations, seismic waves were greatly amplified in- relation technique and a statistical-model technique with an
side the valley, particularly in the lake-bed zone, and durations information criterion to borehole data obtained at the same
of the lake-bed seismograms were surprisingly long. The large Roma-C station and identified body and surface waves. They
amplification and the long coda have not been explained sat- also interpreted surface waves as fundamental-mode Love
isfactorily, as explained in this section. waves excited by a deep structure. Iida (1999) suggested that
A basic model to investigate site amplification is a 1D ver- the deep structure corresponded to the Mexican volcanic belt
tical S-wave model. Seed et al. (1988) tried to explain re- with a depth of a few kilometers.
sponse spectra of seismograms observed in the lake-bed zone In the lake-bed zone of Mexico City, ground subsidence of
by a 1D model using a hill seismogram as the input motion. several tens of centimeters per year has been observed (Hiriart
Singh and Ordaz (1993) inferred that the long coda in the lake- and Marsal 1969; Marsal 1975). Ground subsidence caused by
bed zone could be simulated by applying a 1D amplification gravity might greatly affect strong motions. In the present
to a recent broadband hill seismogram of long duration. Baez study, we investigate effects of subsidence of soft surficial de-
et al. (1994) questioned their simulations, which assumed that posits on strong motions on the basis of a 3D nonlinear finite-
the input hill seismograms were affected by negligible feed- element (FE) technique to explain large amplification of S
back from the lake-bed zone, and concluded that, in general, waves in soft surficial deposits. Because surface waves have
the assumption was not justified. More importantly, their re- long wavelengths, only body waves are expected to be largely
productions exclude the possibility of surface waves. Although influenced by ground subsidence. First, we confirm this hy-
their studies implicitly assumed that hill seismograms were pothesis and then identify body-wave (S-wave) time sections.
mostly S-waves, Iida (1999) recently demonstrated that a hill Both S-wave amplification and propagation time are estimated
strong-motion record was mostly surface waves by a cross- by both the FE simulation under the action of gravity and a
correlation analysis between surface and borehole recordings. 1D linear model without gravity and are compared with ana-
Therefore, their 1D amplification simulations are not very rea- lytical results for the borehole seismograms. Borehole seis-
sonable. mograms, together with a shallow underground profile at the
On the other hand, Chavez-Garcia and Bard (1993a,b) borehole site, are really effective for these purposes.
showed that gravity waves traveling in surficial deposits pro- BOREHOLE STRONG-MOTION ACCELEROGRAMS
posed by Lomnitz (1990) were not a likely explanation for
lake-bed seismograms. Chavez-Garcia et al. (1995) explained We analyze accelerograms recorded at two strong-motion
long-period (3 to 5 s) wave trains as Love waves due to a deep borehole stations in the lake-bed zone (Fig. 1), which are op-
structure and also showed that short-period (<2 s) surface erated by the National Disaster Prevention Center of Mexico.
waves could not travel in surficial deposits because of large Each borehole station has one surface and two underground
attenuation and dispersion. instruments. Accelerograms with long coda were recorded at
Borehole seismograms are useful to study 1D vertical prop- both stations during the September 14, 1995, earthquake (Ms
agation; but no borehole seismograms were obtained during = 7.3), which occurred on the Pacific coast in the state of
the 1985 earthquake. Ordaz et al. (1992) showed that, through Guerrero. These seismograms are one of the best-quality data
sets ever obtained in Mexico after being synchronized in De-
1
Earthquake Res. Inst., Univ. of Tokyo, Tokyo 113-0032, Japan. cember 1994. Since then, initial P phases have also been re-
Note. Discussion open until February 1, 2000. To extend the closing corded by a remote-trigger equipment installed at the Iguala
date one month, a written request must be filed with the ASCE Manager station located 100 km south of the Valley of Mexico, in spite
of Journals. The manuscript for this paper was submitted for review and
possible publication on May 4, 1998. This paper is part of the Journal
of the large P-S time interval in seismograms observed in
of Geotechnical and Geoenvironmental Engineering, Vol. 125, No. 9, Mexico City.
September 1999. 䉷ASCE, ISSN 1090-0241/99/0009-0727–0740/$8.00 The shallow underground profiles at the two strong-motion
⫹ $.50 per page. Paper No. 18274. borehole stations are shown in Table 1 (Yamashita Architects
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 727

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1. Location Map of Valley of Mexico Showing Hill, Transition, and Lake-Bed Zones. Two Strong-Motion Borehole Stations in
Lake-Bed Zone Are Indicated by Solid Circles

TABLE 1. Shallow Underground Profiles at Two Strong-Mo- by very low S-wave velocities and extremely low densities.
tion Borehole Stations The Roma-C station is located on a thin soft deposit, whereas
P-wave S-wave Depth of the Zaragoza station is located on a thick deposit. Large non-
Depth velocity velocity Density instrumentation linear behavior of shallow deposits is not expected, because
(m) (m/s) (m/s) (g/cm3) (m) the main material is high-strength clay. Small nonlinear be-
(1) (2) (3) (4) (5) havior was recognized at only one station during the 1985
(a) Roma-C station earthquake (Singh et al. 1988). A constant damping coefficient
3–5 1,430 90 1.2 0
of h = 0.05 (a quality factor of Q = 10) is used for the shallow
5–12 1,430 30 1.1 profiles. The h value will be discussed in a later section.
12–25 1,430 55 1.1 Figs. 2 and 3 show the two horizontal components of the
25–33 1,430 80 1.2 30 accelerograms and their Fourier spectra at the two borehole
33–36 1,430 200 1.4 stations for the 1995 event, respectively. Vertical components
36–44 1,430 130 1.4
are not used in this study. Durations of the original recordings
44–55 1,780 400 1.5
55–65 1,580 250 1.5 are extraordinarily long (>5 min). We use a 150-s time section
65–102 1,750 430 1.7 102 with large amplitude (Fig. 2). The sampling time interval is
(b) Zaragoza station 0.01 s. A frequency range of 0.125–5.0 Hz (0.2–8.0 s) is
studied. Spectra in the frequency domain are smoothed by the
3–5 1,420 40 1.1 0
5–23 1,420 28 1.1
Parzen’s spectral window with a bandwidth of 0.2 Hz. The
23–38 1,420 42 1.1 30 maximum amplitude of the surface seismograms is about 30
38–40 1,420 180 1.2 gal. (cm/s2). We observe a very long coda at the two stations.
40–52 1,420 76 1.1 The difference between the two stations is the frequency con-
52–57 1,420 135 1.4 tents of surface seismograms. Accelerograms recorded at the
57–67 1,700 460 1.4
67–80 1,450 140 1.5
underground instruments do not have the dominant frequencies
80–83 1,750 450 1.6 83 seen in the surface seismograms and do not include high-fre-
quency components of more than 1.0 Hz (<1.0 s).
For reference, only surface seismograms were recorded
& Engineers Inc. and Oyo Corp. 1990). The shallow profiles at six stations in Mexico City during the 1985 earthquake
were well determined by suspension PS logging in 1988. The (Anderson et al. 1986). A maximum amplitude of approxi-
lake-bed zone consists of a 10 to more than 100 m deposit of mately 190 gal. was obtained at the SCT (Secretaria de Co-
highly compressible, high water content clay, underlain by municaciones y Transportes) station in the lake-bed zone, and
sands. Hence, the soft clay deposit is primarily characterized record duration reached 3 min at two lake-bed stations.
728 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


IDENTIFICATION OF S WAVES was longer than 1,400 m (=700 m/s ⫻ 2 s). Therefore, the
Iida and Kinoshita (unpublished paper, 1999) investigated effects of ground subsidence on surface waves, which have
basic characteristics of surface-wave propagation in strong- long wavelength relative to the thickness (several tens of me-
motion seismograms at the Roma-C station. Based upon the- ters) of surficial soft deposits, are not examined in this study.
oretical calculations of amplification factors for shallow and
Cross-correlation of Borehole Recordings
deep structures, they suggested that surface waves with periods
of more than about 2 s could be interpreted as fundamental- To identify S waves, we conduct a running (temporal) cross-
mode Love waves excited in a deep structure with depths of correlation analysis between the surface and borehole record-
a few kilometers (the Valley of Mexico has a depth of several ings at the two stations (Fig. 4). Judging from the very low
hundred meters). The theoretical phase velocity for the Love S-wave velocities of surficial layers with thickness of several
waves was estimated to be >700 m/s, so that the wavelength tens of meters at the two borehole stations (Table 1), nearly
Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 2. Strong-Motion Accelerograms (Two Horizontal Components) Recorded at Two Borehole Stations during September 14, 1995
Earthquake

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 729

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 2. (Continued )

vertical incidence of S waves is expected. Table 2 shows the we expect a peak at a lag time of 0.0 s. Thus, we judge the
propagation times (lag times) of vertically incident, direct S early time sections between 30 and 70 s as S waves at both
waves traveling between the lower borehole instrument and stations. The late time sections after 70 s are not judged as S
the ground surface and primary and secondary predominant waves. Most importantly, the lag time of the cross-correlation
frequencies (periods) of S waves in shallow deposits at the function is always smaller than that estimated from the S-wave
two borehole stations using the S-wave shallow velocity pro- profile. This will be interpreted in a later section.
file. Therefore, a peak of the cross-correlation function should To examine frequency dependence of wave types, we fur-
appear at the lag time if seismic motions are mostly S waves. ther conduct a similar analysis with bandpass filters. Two con-
For example, in the case of the Roma-C station, a peak should trasting examples are displayed in Fig. 5. Regarding the Roma-
appear at a lag time of 0.75 s. For any kind of surface waves, C recordings, although high-frequency seismic waves in the
730 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Fourier Spectra of Strong-Motion Accelerograms (Two Horizontal Components) Recorded at Two Borehole Stations

FIG. 4. Cross-Correlation Coefficients for Different Time Intervals in East-West Direction between Surface and Lower Borehole Re-
cordings at Two Borehole Stations (t Is Time Interval; Period Range Is 0.2–8.0 s)

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 731

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


TABLE 2. Propagation Times of Vertically Incident Direct S Transfer Functions
Waves and Primary and Secondary Predominant Frequencies
of S Waves in Shallow Depositsa To understand actual wave amplification more clearly, Fig.
Primary Secondary 6 shows transfer functions for early (S-wave) recordings
Propagation predominant predominant (shown by bold lines) for the three pairs of vertical-array in-
time frequency (period) frequency (period) struments at the two borehole stations. These transfer functions
Station (s) [Hz (s)] [Hz (s)] are calculated from Fourier spectral ratios. In the transfer func-
(1) (2) (3) (4) tion for the pair of 0 and 102 m at the Roma-C station, a peak
Roma-C 0.747 0.40 (2.47) 0.93 (1.08) at 0.4 Hz (2.5 s) and another peak at 1.0 Hz (1.0 s) can be
Zaragoza 1.453 0.22 (4.58) 0.57 (1.74) assessed. At the Zaragoza station, we notice a smooth peak at
a
Calculated using shallow underground profiles in Table 1.
0.2 Hz (5.0 s) and a sharp peak at 0.7 Hz (1.4 s) for the pair
of 0 and 83 m.
1D linear transfer functions for vertically incident S waves
are calculated for the three pairs of vertical-array instruments
frequency range of >0.7 Hz (<1.5 s) might be interpreted as and are compared with transfer functions for the early (S-
mostly body waves, low-frequency waves of <0.3 Hz (>3.0 s) wave) recordings (Fig. 6). The linear transfer functions are
Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

can be interpreted as surface waves. In the intermediate fre- calculated using the shallow profiles (Table 1). First, it is found
quency range between 0.3 and 0.7 Hz (1.5 and 3.0 s), wave that the Zaragoza surface recordings contain two dominant fre-
types are uncertain probably because both body and surface quencies (Fig. 3), which correspond to the primary and sec-
waves have the same dominant frequency of 0.4 Hz (2.5 s). ondary predominant frequencies calculated from the shallow
Regarding the Zaragoza recordings, high-frequency seismic profile. Therefore, the surficial deposits, rather than the inci-
waves of >0.3 Hz (<3.0 s) are interpreted as mostly body dent wavefield, control the frequency contents of the surface
waves. Wave types in the low frequency range of <0.3 Hz seismograms recorded in the lake-bed zone.
(>3.0 s) appear very complex, and body- and surface-wave Second, we note that the linear transfer functions overesti-
phases might be mixed. mate the transfer functions calculated from the observed

FIG. 5. Cross-Correlation Coefficients for Different Time Intervals for Three Frequency Ranges in East-West Direction between Sur-
face and Lower Borehole Recordings at Two Borehole Stations (t Is Time Interval; T Is Period Range)

732 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 6. Comparisons of Transfer Functions for Early Recordings (Bold Lines) with 1D Linear Transfer Functions for S Waves (Thin
Lines) in Two Horizontal Components for Three Pairs of Vertical-Array Instruments at Two Borehole Stations. Early Recordings Are 40-
s Time Section of 30–70 s, which Are Interpreted as S Waves

records in the low frequency range at both stations. In partic- be well explained. The lag time of the cross-correlation func-
ular, we see that peaks at about 0.4 and 0.25 Hz (2.5 and 4.0 tion might be well explained as well.
s) in the linear transfer functions are much larger than those
in the observed transfer functions for the uppermost 30 m of S-WAVE PROPAGATION IN NONLINEAR SOFT SOIL
the surficial layers at the Roma-C and Zaragoza stations, re- DEPOSITS UNDER ACTION OF GRAVITY
spectively. The uppermost 30 m deposits are characterized by
very soft soils (Table 1), so that large subsidence of ground is Numerical simulations of S-wave propagation in nonlinear
expected to occur in the soft deposits. If subsidence is taken soft soil deposits under the action of gravity should be en-
into account, peaks in the observed transfer functions might couraged for the following three reasons:
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 733

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


1. In reality, soil deposits are under the action of gravity, {␴t⫹dt} = ␣ ⭈ {␴t} ⫹ {␴0} (2)
and are compressed by the surrounding soil deposits,
even if strong subsidence has not been observed. where {␴0} = initial stress vector; dt = time step; and ␣ =
2. Ground subsidence of several tens of centimeters per unknown constant. The constant ␣ is determined from the con-
year has been actually observed in the lake-bed zone of straint that the stress vector remains on the yield curved sur-
Mexico City (Hiriart and Marsal 1969; Marsal 1975). face.
Subsidence is basically caused by gravity (the weight of The following nonlinear equations of motion are solved by
soil itself). The shallow profiles (Table 1) at the two a Newton-Raphson method:
borehole stations were measured in 1988, and the strong-
motion accelerograms (Fig. 2) were recorded in 1995. [G] ⭈ {␦¨ t⫹dt} = {Qt⫹dt} (3a)
Therefore, considerable subsidence accumulated during
the 7-year period, and the soil properties in 1995 are [G] = [M] ⫹ 0.5 ⭈ dt ⭈ [C] ⫹ ␤ ⭈ dt 2 ⭈ [K] (3b)
probably different from those in Table 1.
3. The stress of a soil sample extracted from soil deposits
could be substantially relieved during sampling, whereas {Qt⫹dt} = {Pt⫹dt} ⫺ [C] ⭈ {␦˙ t ⫹ 0.5 ⭈ dt ⭈ ␦¨ t}
the soil sample was compressed by the surrounding soil
Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

deposits at the original underground position. Therefore, ⫺ [K] ⭈ {␦t ⫹ dt ⭈ ␦˙ t ⫹ (0.5 ⫺ ␤) ⭈ dt 2 ⭈ ␦¨ t} (3c)
the real soil deposits are expected to be harder than the
measurement values. where [M], [C], [K], {␦}, and {P} = mass matrix, damping
matrix, stiffness matrix, displacement vector, and external
Among the above three factors, factor 1 can be almost per- force vector, respectively; and ␤ = constant of a Newmark’s
fectly taken into account by the simulations, whereas factor 2 ␤ method. Owing to nonlinearity of the soil, (3) cannot be
can be considered only partly, as explained in the following.
Factor 3 might not be directly related to gravity and is not
expected to be considered adequately. In our case, this factor
is probably negligible because the shallow profiles were ac-
curately measured (see Section 2). To summarize, the simu-
lations are able to represent real soft deposits considerably
well.
It would be reasonable to use the same soil model between
dynamic and static analyses. Because the surface material is
clay and liquefaction has never occurred during shaking, we
use a bilinear model with a Mohr-Coulomb criterion to rep-
resent nonlinearity of the clay. The weak point of this model
is that the effects of ground-water withdrawal or gradual dis-
sipation of excess pore pressures in soft, saturated clay cannot
be considered. Such consideration is beyond the scope of this
study, although the effects might be rather significant.
An FE technique is a powerful tool for estimating static
stress and deformation in various types of soils [e.g., Bergado
et al. (1995) and Alfaro et al. (1997)]. The technique is also
powerful when dynamic wave propagation is calculated for
soils compressed by a static analysis. We apply a nonlinear FE
numerical simulation to the lake-bed clay to estimate the am- FIG. 7. Model Composed of Upper 3D Nonlinear FE Volume
plification and propagation time of S waves under the action and Lower 1D Linear Volume. Depths of Two Horizontal Bound-
of gravity. We use a 3D model because the effects of a building aries (D1 and D2) Are Shown in Table 3
or horizontal heterogeneity in an underground structure can be
considered, and because the two horizontal components of
seismic motions can be simultaneously excited. Consequently, TABLE 3. Soil Properties Used in 3D FE Technique
a complete 3D Mohr-Coulomb criterion is adopted. In this
Reduction ratio Frictional
study, however, we basically investigate 1D nonlinear behavior Depth in second Cohesion angle
of soils under the action of gravity, because no building and (m) Material rigidity (g/cm2) (degrees)
no horizontal heterogeneity exist in an underground structure. (1) (2) (3) (4) (5)
As a basic model for comparison, a 1D linear model without
(a) Roma-C stationa
gravity is used.
0–32 Clay 0.5 250 5
Method 32–36 Clay 0.5 350 5
36–44 Clay 0.5 350 5
A bilinear model with a 3D Mohr-Coulomb failure criterion 44–52 Sand 0.5 500 30
is assumed for the soil. The soil is in the yield range if one (b) Zaragoza stationb
of the following inequalities is satisfied: 0–36 Clay 0.5 250 5
␴1 ⫺ ␴2 > (␴1 ⫹ ␴2) ⭈ sin ␾s ⫹ 2cs ⭈ cos ␾s (1a) 36–40 Clay 0.5 350 5
40–52 Clay 0.5 250 5
␴1 ⫺ ␴3 > (␴1 ⫹ ␴3) ⭈ sin ␾s ⫹ 2cs ⭈ cos ␾s (1b) 52–56 Clay 0.5 350 5
56–64 Clay 0.5 350 5
␴2 ⫺ ␴3 > (␴2 ⫹ ␴3) ⭈ sin ␾s ⫹ 2cs ⭈ cos ␾s (1c) a
Depth of bottom plane of 3D volume, D1 = 52 m; depth where lower
where ␴1, ␴2, and ␴3 (␴1 > ␴2 > ␴3) = principal stresses; cs = borehole instrument is installed, D2 = 102 m.
b
Depth of bottom plane of 3D volume, D1 = 64 m; depth where lower
cohesion; and ␾s = frictional angle. If the soil is in the yield borehole instrument is installed, D2 = 83 m.
range, the stress vector {␴t⫹dt} is expressed by
734 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 8. Comparisons of Surface Accelerograms (Two Horizontal Components) at Two Borehole Stations. Each Set of Diagrams
Shows Observed Seismograms for 40 s (30–70 s) (Upper), 1D Linear Seismograms (Middle), and FE Simulated Seismograms (Lower).
Lower Borehole Recordings Are Used as Input Motions

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 735

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 8. (Continued )

solved explicitly. The equations are solved by the following Once accelerations are determined, velocities and displace-
three-step procedure: ments are evaluated by the following:

1. By assuming that the soil is elastic, stresses and external {␦˙ t⫹dt} = {␦˙ t} ⫹ 0.5 ⭈ dt ⭈ {␦¨ t ⫹ ␦¨ t⫹dt} (4a)
force are approximately calculated. {␦t⫹dt} = {␦t} ⫹ dt ⭈ {␦˙ t} ⫹ 0.5 ⭈ dt 2 ⭈ {␦¨ t} ⫹ ␤ ⭈ dt 2 ⭈ {␦¨ t⫹dt ⫺ ␦¨ t}
2. If the soil is in the yield range, stresses are adjusted
(4b)
according to the aforementioned way of (2).
3. Calculations of external force are repeated until it is con- These calculations are performed using a computer code de-
verged. veloped by Ishihara and Miura (1993).
736 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 9. Comparisons of Fourier Spectra of Surface Accelerograms (Two Horizontal Components) Shown in Fig. 8

The entire analysis is composed of the following two main


steps (Fig. 7):

1. Static analysis: We fix the bottom plane of the 3D non-


linear FE volume at a depth of D1 (m) and perform static
nonlinear soil subsidence under the action of gravity. The
depth of D1 (m) is chosen considering that very soft de-
posits are concentrated in the upper 25–40 m at both
stations (Table 1).
2. Dynamic analysis: Early S-wave time sections of 30 to
70 s in the two horizontal components of the observed
seismograms are applied as excitation to a depth of D2
(m) where the lower borehole instrument is installed. Al-
though a 1D linear propagation of vertically incident S
waves is performed between D2 and D1 (m), a 3D non-
linear propagation is conducted between D1 and 0 (m).
We calculate surface motions for the compressed soil de-
posits. The depths of the two horizontal boundaries (D1
and D2) are shown in Table 3.

The 3D volume with a horizontal extension of 40 ⫻ 40 m


is modeled by rectangular prism elements with a base of 5 ⫻
5 m and a height of 4 m. The calculation time interval is 0.01
s. Horizontal and vertical extensions and a damping coefficient
of h = 0.05 in the 3D volume can attenuate artificial wave
reflections from the bottom and side planes. The constant el-
FIG. 10. FE Simulated Transfer Functions for 40 s (30–70 s) ement sizes are determined on the basis of the minimum seis-
(Two Horizontal Components) at Two Borehole Stations. Pair of mic wavelength L to be analyzed. The wavelength L is the
Surface and Lower Borehole Seismograms Is Used minimum velocity V of the shallow profiles (Table 1) multi-
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 737

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 11. Comparisons of Cross-Correlation Coefficients for Different Time Intervals (East-West Component) at Two Borehole Sta-
tions (t Is Time Interval; Period Range Is 0.2–8.0 s). Each Set of Diagrams Shows Observed Seismograms for 40 s (30–70 s) (Upper),
1D Linear Seismograms (Middle), and FE Simulated Seismograms (Lower). Lower Borehole Recordings Are Used as Input Motions

plied by the minimum period T to be analyzed. The wave- FE simulated transfer functions at both stations are shown
length L is estimated by in Fig. 10. They might be compared with two kinds of transfer
functions—the observed and 1D linear transfer functions in
L= V ⭈ T = 30 (m/s) ⫻ 0.8 (s) = 24 (m)
Fig. 6. Effects of subsidence due to gravity are easily recog-
Hence, the element sizes are valid for dominant periods of nized by comparing FE simulated transfer functions with 1D
seismic motions. linear ones (thin lines in Fig. 6). As the peak frequencies are
The shallow soil profiles are shown in Table 1. Soil nonlin- identical between both functions, we can confirm that soil non-
ear parameters are roughly evaluated on the basis of a soil linearity, if any, is very minor. The FE simulation satisfactorily
study (Jaime 1987). The yield strength assumed for the soils explains the observed transfer functions (bold lines in Fig. 6)
is given in Table 3. at both stations.
Fig. 11 shows comparisons of running cross-correlation
Results functions at both stations. Although the 1D linear model can-
Figs. 8 and 9 show comparisons of surface seismograms and not satisfactorily explain lag times between the borehole and
their Fourier spectra at both stations, respectively. Although surface recordings, the FE simulation successfully explains
the 1D linear model overestimates wave amplitudes, ampli- them, particularly at the Roma-C station. Although lag times
tudes of FE simulated seismograms are comparable to those may be partly explained by oblique incidence of S waves in
of observed seismograms. Soil behavior in the soft surficial the 1D linear model, the FE simulation is able to make a better
deposits is within linear range in the FE simulation. At the explanation than the oblique incidence of S waves.
Roma-C station, frequency contents are similar between FE Here, we interpret the effects of ground subsidence. Judging
simulated and observed seismograms, although 1D linear seis- from transfer functions, S-wave amplification was considera-
mograms include more high-frequency components than the bly changed due to subsidence, although S-wave propagation
other two seismograms. Regarding Zaragoza seismograms, FE was nearly linear. On the basis of lag times of cross-correlation
simulated seismograms have lower frequencies than observed functions, S-wave propagation time decreased a little owing to
seismograms. subsidence. Thus, subsidence changes the soil properties, so
738 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

FIG. 12. Depth Distributions in Shear Strain and Stress in East-West Direction for Uppermost 30 m of Surficial Layers at Roma-C and
Zaragoza Stations. Maximum Dynamic Values and Initial Static Values Are Shown

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999 / 739

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740


that linear wave propagation through the 1988 structure (Table gestions made by an editor, an editorial board member, and three anon-
1) is different from linear wave propagation through the real ymous reviewers greatly improved the manuscript.
(1995) structure.
Fig. 12 displays depth distributions in the shear strain and APPENDIX. REFERENCES
stress for the uppermost 30 m of surficial layers at both sta- Alfaro, M. C., Hayashi, S., Miura, N., and Bergado, D. T. (1997). ‘‘De-
tions. It seems that the maximum dynamic strain depends upon formation of reinforced soil well-embankment system on soft clay
softness of soil deposits, and the maximum dynamic stress foundation.’’ Soils and Found., 37, 33–46.
Anderson, J. G., et al. (1986). ‘‘Strong ground motion from the Michoa-
gradually increases with depth. The initial static strain and can, Mexico, earthquake.’’ Sci., 233, 1043–1049.
stress values are considerably small relative to the maximum Baez, G., Flores, J., Mateos, J. L., Mendez, R. A., Novaro, O., and Se-
dynamic ones. ligman, T. H. (1994). ‘‘Comment on the origin of long coda observed
At the present stage, there are two limitations in the FE in the lake-bed strong-motion records of Mexico City.’’ Bull. Seismo-
simulation. One is that an artificial horizontal plane for ground logical Soc. of Am., 84, 2015–2018.
subsidence was fixed at a depth of several tens of meters. This Bergado, D. T., Chai, J. C., and Miura, N. (1995). ‘‘FE analysis of grid
reinforced soil wall-embankment system on soft clay foundation.’’
should lead to excessive subsidence (overloading). The other Soils and Found., 17, 447–471.
is that the element size (5 ⫻ 5 ⫻ 4 m) is too large for high- Bodin, P., Gomberg, J., Singh, S. K., and Santoyo, M. (1997). ‘‘Dynamic
frequency components. The minimum effective wavelength is deformations of shallow sediments in the Valley of Mexico, Part I:
Downloaded from ascelibrary.org by Marco Ferro on 03/16/23. Copyright ASCE. For personal use only; all rights reserved.

approximately 24 m. As the lowest S-wave velocity is about Three-dimensional strains and rotations recorded on a seismic array.’’
30 m/s, the effective frequency (period) range is less than Bull. Seismological Soc. of Am., 87, 528–539.
about 1.25 Hz (more than about 0.8 s). This condition might Chavez-Garcia, F. J., and Bard, P.-Y. (1993a). ‘‘Gravity waves in Mexico
City? I. Gravity perturbed waves in an elastic solid.’’ Bull. Seismolog-
be severe for the Zaragoza site, because the shallow profile ical Soc. of Am., 83, 1637–1655.
has thick low-velocity layers. In the effective frequency (pe- Chavez-Garcia, F. J., and Bard, P.-Y. (1993b). ‘‘Gravity waves in Mexico
riod) range, our results are quite satisfactory (Fig. 9). These City? II. Coupling between an anelastic solid and a fluid layer.’’ Bull.
parameters were determined considering memory and CPU Seismological Soc. of Am., 83, 1656–1675.
time of a normal workstation. The limitations are expected to Chavez-Garcia, F. J., Ramos-Martinez, J., and Romero-Jimenez, E.
be overcome considerably. (1995). ‘‘Surface-wave dispersion analysis in Mexico City.’’ Bull. Seis-
mological Soc. of Am., 85, 1116–1126.
Fukushima, Y., Kinoshita, S., and Sato, H. (1992). ‘‘Measurement of Q-
DISCUSSION AND CONCLUSIONS 1 for S waves in mudstone at Chikura, Japan: Comparison of incident
and reflected phases in borehole seismograms.’’ Bull. Seismological
A constant damping coefficient of h = 0.05 (a quality factor Soc. of Am., 82, 148–163.
of Q = 10) was used for the shallow profiles. Damping coef- Hiriart, F., and Marsal, R. J. (1969). ‘‘Nabor carrillo, el hundimiento de
ficient values are uncertain in Mexico City in the low-fre- la ciudad de Mexico y proyecto Texcoco, secretaria de hacienda y
quency range. In previous studies [e.g., Sanchez-Sesma et al. credito publico.’’ Mexico City (in Spanish).
(1988) and Ordaz et al. (1992)], h values of 0.02–0.07 (Q Iida, M. (1999). ‘‘Excitation of high-frequency surface waves with long
values of 7–25) were usually used. Our h value was deter- duration in the Valley of Mexico.’’ J. Geophys. Res., Solid Earth, 104,
7329–7345.
mined with moderate confidence by the 1D linear and FE sim- Iida, M., Ordaz, M., Taniguchi, H., Gutierrez, C., and Santoyo, M. (1994).
ulations using the borehole seismograms. Frequency-depen- ‘‘Interpretation of wave field inside the Mexico Valley on the basis of
dent h (Q) values might be more reasonable (Fukushima et al. borehole data.’’ Proc., 9th Japan Earthquake Engrg. Symp., Vol. 3,
1992; Satoh et al. 1995). If we assume frequency-dependent E121–E126.
h values (larger h for lower frequency), 1D linear transfer Ishihara, T., and Miura, F. (1993). ‘‘Nonlinear seismic response analysis
functions (thin lines in Fig. 6) become smaller in the lower method for 3-D soil-structure interaction systems.’’ Proc., JSCE, To-
kyo, 465, 145–154 (in Japanese).
frequency range, compared with the results shown here, and Jaime, A. (1987). ‘‘Foundation engineering in Mexico City: General as-
will probably satisfactorily match transfer functions for re- pects and subsoil conditions.’’ Proc., Int. Symp. Geotech. Engrg. Soft
cordings (bold lines in Fig. 6). Soils, Vol. 2, Mexican Society of Soil Mechanics, Mexico City, 225–
However, h values cannot change lag times in cross-corre- 243.
lation functions. Judging from the fairly good agreement in Lomnitz, C. (1990). ‘‘Mexico 1985: The case of gravity waves.’’ Geo-
lag times between the FE simulation and the observation (Fig. phys. J. Int., 102, 569–572.
Marsal, R. J. (1975). The lacustrine clays of the Valley of Mexico.’’ Rep.
11), surface recordings cannot be interpreted without consid- 07/75, Universidada Nacional Autonoma de Mexico, Mexico City.
ering the effects of ground subsidence. Although we are un- Ordaz, M., Santoyo, M. A., Singh, S. K., and Quaas, R. (1992). ‘‘Analysis
certain about real h values, we obtain reasonable amplitude of the borehole recordings obtained in Mexico City during the May
matches with a constant h of 0.05. Therefore, subsidence can 31, 1990 earthquake.’’ Proc., Int. Symp. Effects Surface Geol. Seismic
make a more reasonable interpretation than damping. Motion, Vol. 1, Association for Earthquake Disaster Prevention, Tokyo,
In the present study, we assumed that ground subsidence 155–160.
Sanchez-Sesma, F., Chavez-Perez, S., Suarez, M., Bravo, M. A., and
influenced only body waves, because surface waves were ex- Perez-Rocha, L. E. (1988). ‘‘The Mexico earthquake of September 19,
cited in a deep structure relative to the thickness of soft sur- 1985—on the seismic response of the Valley of Mexico.’’ Earthquake
ficial layers. Thus we identified body-wave (S-wave) phases Spectra, 4, 569–589.
in the lake-bed seismograms. Both S-wave amplification and Satoh, T., Kawase, H., and Sato, T. (1995). ‘‘Evaluation of local site
propagation time were estimated, using a 3D nonlinear FE effects and their removal from borehole records observed in the Sendai
simulation under the action of gravity and a 1D linear model Region, Japan.’’ Bull. Seismological Soc. of Am., 85, 1770–1789.
Seed, H. B., Romo, M. P., Sun, J. I., Jaime, A., and Lysmer, J. (1988).
without gravity. The FE simulation successfully predicted both ‘‘Relationships between soil conditions and earthquake ground mo-
the S-wave amplification and propagation time, which could tions.’’ Earthquake Spectra, 4, 687–729.
not be well explained by the purely 1D linear model. Singh, S. K., et al. (1988). ‘‘The Mexico earthquake of September 19,
1985—A study of amplification of seismic waves in the Valley of Mex-
ACKNOWLEDGMENTS ico with repect to a hill zone site.’’ Earthquake Spectra, 4, 653–673.
Singh, S. K., and Ordaz, M. (1993). ‘‘On the origin of long coda observed
The writer thanks Roberto Quaas of the National Disaster Prevention in the lake-bed strong-motion records of Mexico City.’’ Bull. Seismo-
Center of Mexico for supplying the strong-motion accelerograms used in logical Soc. of Am., 83, 1298–1306.
this study. The writer appreciates the support of the Japan International Yamashita Architects & Engineers Inc. and Oyo Corp. (1990). ‘‘Estudios
Cooperation Agency (JICA) and the JICA Mexico project committee. de subsuelo para el proyecto de el centro de prevencion de desastres
Research funds were provided by the Ministry of Education, Japan. Sug- sismicos en los estados unidos mexicanos.’’ Mexico City (in Spanish).

740 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / SEPTEMBER 1999

J. Geotech. Geoenviron. Eng., 1999, 125(9): 727-740

You might also like