Comparison of Seismic Ground Response

Analyses Under Large Earthquakes

Nozomu Yoshida

Indian Geotechnical Journal

ISSN 0971-9555
Indian Geotech J
DOI 10.1007/s40098-014-0104-8

1 23

Your article is protected by copyright and

all rights are held exclusively by Indian
Geotechnical Society. This e-offprint is for
personal use only and shall not be selfarchived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com.

1 23

Indian Geotech J $SULO-XQH

DOI 10.1007/s40098-014-0104-8

ORIGINAL PAPER

Comparison of Seismic Ground Response Analyses Under Large

Earthquakes
Nozomu Yoshida

Received: 5 August 2013 / Accepted: 24 January 2014

Indian Geotechnical Society 2014

Abstract Equivalent linear, truly nonlinear, and effective

stress analyses are compared at 268 sites under 11 large
earthquake motions in order to grasp general feature of the
applicability of the equivalent linear method and total
stress analysis. Equivalent linear analysis shows larger
peak acceleration and smaller peak displacement than total
stress nonlinear analysis, but other indices such as instrumental seismic intensity, peak velocity, and spectral
intensity are nearly same to each other. Therefore, it is
applicable when truly nonlinear analysis is applicable. On
the other hand, all earthquake motion indices are much
smaller in the effective stress analysis compared with the
total stress analysis at the liquefiable sites, except two
indices peak ground displacement and maximum strain that
are much larger. Therefore, it is recommended to use
effective stress analysis at these sites. However, as indices
are overestimated, use of the total stress can be said to
result in conservative design when inertia force is interested at.
Keywords Total stress analysis 
Equivalent linear analysis  Effective stress analysis 
Earthquake motion index

Introduction
Static analysis based on the seismic coefficient has been
made long time. Recently, however, seismic response
analysis becomes a powerful and important tool for the
N. Yoshida (&)
Department of Civil and Environmental Engineering, Tohoku
Gakuin University, Chuo 1-13-1, Tagajo, Miyagi, Japan
e-mail: yoshidan@tjcc.tohoku-gakuin.ac.jp

design of structures. There are several reasons that the

seismic response analysis becomes important. One of the
big reasons is that input earthquake motions becomes large,
under which design based on the static analysis frequently
results in too conservative design. In order to make the
design based on the seismic response analysis, seismic
response analysis of ground also becomes important as a
tool to obtain the input earthquake motions to the structure.
Input earthquake motion was increased significantly in
Japan, especially after the 1995 Kobe earthquake and more
after the 2007 Niigataken-chuetsu-oki earthquake. Magnitude of the design earthquake motion, for example, exceeds
1 G in the design of nuclear power plants. In Japan,
maximum recorded earthquake motion in the past is usually used as design earthquake motion, as level 2 ground
motion. Design input earthquake motions with maximum
acceleration more than 0.5 m/s2 is not an extraordinary
case, but frequently appears in the practical design of civil
and building structures. In these situations, maximum shear
strain frequently reaches a few percent or more.
Various methods and computer programs have been
proposed for the seismic response analysis of ground.
One of the frequently used computer programs is
SHAKE [1], which is based on an equivalent linear
method. The name SHAKE is now used as if it is a common noun, and it is also used as common noun in this
paper. Unlike the name equivalent, however, it is an
approximate method. Therefore, it is believed that it cannot
be applied at large strains. For example, Ishihara [2] suggested to use different constitutive models and methods of
the seismic response analysis depending on different shear
strain ranges and soil properties as shown in Fig. 1. If
maximum shear strain yields several percent, therefore, it is
recommended to use the truly nonlinear method because
the equivalent linear method is not applicable. However,

123

Indian Geotech J

engineers have a tendency to use SHAKE partly because of

its easy handling compared with the truly nonlinear
method, and partly because of the huge accumulation of
experience to use the equivalent linear analysis.
Next problem is soil liquefaction. Considering that the
definition of the soil liquefaction by means of shear strain
is usually 5 % double amplitude axial strain under the triaxial test condition, which corresponds to 3.75 % shear
strain, effective stress analysis considering the excess
porewater pressure generation is necessary if maximum
strain yields several percent. However, use of the effective
stress analysis is more difficult than that of total stress
analysis. Therefore, total stress analysis is frequently carried out even when shear strain becomes very large.
There were not a few reports that compare various
analytical methods. Some reported that, however, SHAKE
is applicable, some are not. There seem at least two reasons
to make this confliction. The one is that comparison is
made on very few numbers of case studies, usually one or
two, in which case obtained conclusion can be accidental
except that the mechanism is clearly explained such as Ref.
[3] which shows two shortages of SHAKE and improvement to overcome it. The second reason is that there is no
common method to evaluate accuracy of the result of the
seismic response analysis.
Looking at past researches to show applicability of the
seismic response analysis, agreement of the maximum
acceleration, similarity of waveforms, etc. have been used
as indices to discuss applicability. It is well known that,
however, maximum acceleration is not a good ground
motion index for long period structures.
For example, Yamahara [4] described that damage to a
structure is controlled by the energy absorbed in the

Shear
strain

10-6

10-5

Small
strain

10-4

10-3

Medium strain

10-2

Large
strain

10-1
Failure
strain

Elastic

Failure

Linear
elastic
model
Linear
method

Viscoelastic
model
Equivalent
linear
method

Load history
tracing type
model
Step-by-step
integration
method

Fig. 1 Change in soil properties with shear strain and corresponding

modelling principle and method of response analysis [2]

123

Figure 2 shows soil profiles and SPT-N value of the

grounds used in this paper. These 268 sites had been used
to evaluate applicability of the conventional method to
evaluate onset of liquefaction during the 1995 Kobe
earthquake [7].
Soils are classified into 6 categories, i.e., sand, silt, clay,
gravel, humus, and others. Here, others include concrete
and cobble, etc. In the modeling in this paper, humus soils
are treated same as clay, and others are neglected by
replacing them by the soil below it. The SPT-N values are
averaged in the same soil layer. If there is no N value data
in a layer, it is set same with the one below the layer. These
modeling may not be good if actual behavior of the ground
is investigated, but are not because the purpose of this
paper is to investigate different ground model.
Unit weight of each layer was included in the original
data. The shear wave velocity Vs is evaluated based on the
specifications for highway bridges [8], i.e.,
Vs 100N 1=3
Vs 80N 1=3

clayey soil)
sandy soil, other soils)

sf 19N kPa

Effect of
load-repetition
Effect of
loading rate

Method of
response
analysis

Analyzed sites and modeling of mechanical properties

Shear strength of the clayey soil is evaluated as an

average value suggested by JGS [9]

Elasto-plastic

Model

structure and the peak acceleration is not a good index.

Nagahashi and Kobayashi [5] showed that peak acceleration is a good index for short period range (0.10.3 s), but
peak velocity and peak displacement are good indices for
periods 0.31.0 s and more than 1.0 s, respectively. Tong
et al. [6] showed that peak acceleration does not agree with
damage rate of houses.
Detailed investigation is difficult only from the similarity of the waveform. Applicability must be evaluated by
using objective and effective ground motion indices.
In this paper, accuracy of the total stress analysis by
equivalent linear and truly nonlinear methods is evaluated
by comparing the effective stress analysis for more than
250 sites and under various types of input earthquake
motions.

On the other hand, the internal friction angle / is

evaluated based on Hatanaka and Uchida [10] as
p
/ 20 20N1 degree
3
where N1 is corrected N value considering the effective
over burden stress, and is evaluated as [6]
N1

170N
100

r0v

r0v in kPa

Indian Geotech J
0

10

15

20

10

15

20

Sandy soil
Clayey soil
Silty soil
Gravelly soil
Humic soil
Other

10

N value
0 30 60

15

20

Fig. 2 Soil profiles

The hyperbolic model

s

G0 c
1 sf c=G0

is used for shear stress sshear strain c relationships, and

the Masings rule is used to obtain damping characteristics,
where G0 is an elastic shear modulus.
Liquefaction strength of the sandy soils is evaluated
based on the design specification of road bridges [6]. Shear
stress ratio RL is calculated as

p
0:0882 Na =1:7
Na \14
RL
p
4:5
0:0882 Na =1:7 1:6  106 Na  14
Na  14

6
where Na is corrected SPT-N value considering fines
contents Fc as
Na C1 N1 C2

where

123

Indian Geotech J

C1

8
>
<
>
:

60%  Fc  100%
0%  Fc \10%

Fc =20  1
0


C2

0%  Fc \10%
10%  Fc \60% ;

1
Fc 40=50

Fc  10=18

10%  Fc  100%

and N1 is corrected SPT-N value considering the effective

overburden stress obtained from Eq. (4).
This shear stress ratio (liquefaction strength) corresponds to 20 cycles of loading. Shear stress ratio at other
cycles of loading is evaluated based on Seed et al. [11].

Earthquake Motions
Input earthquake motion must have various and typical
natures in order to obtain the general feature of the
response. Eleven recorded earthquake motions which have
different characteristics are employed so as to satisfy this
requirement. They are shown in Table 1 with several
earthquake motion indices which are explained in Earthquake Motion Indices Section. Duration of the earthquake
T in the table is computed from the time difference when
powers are 5 and 95 %.
The input earthquake motions are classified into two
categories, inland earthquakes and ocean trench type
earthquake.
Waveforms of the input motion are shown in Figure 3.
Generally speaking, duration of the earthquake is shorter in
the inland earthquakes, but intensity is larger. Figure 4
shows response spectra of the input motions under 5 %
damping.
These waves are applied as the earthquake motion of the
outcropped base. Layers with Vs = 300 m/s is added at the
bottom of the ground in Fig. 2 because SHAKE requires

semi-infinite region beneath the subsoil. As, the shear wave

velocity of the base layer is confirmed not to affect the
behavior in the subsoil [12], it can be any value; a layer
with Vs = 300 m/s is considered to be an engineering
seismic base layer (bedrock) in the Japanese design specification for port facilities [13].

Method of Analysis and its Characteristics

Two different total stress analyses, equivalent linear
method and truly nonlinear method, and one effective
stress analysis are used. A computer program DYNEQ [14]
is used for the equivalent linear analysis. DYNEQ is an
open source program that has various functions based on
complex moduli method, among which technique same
with SHAKE is used in the calculation, and is referred as
SHAKE. A computer program YUSAYUSA-2 [15], which
is also an open source program, is used for the truly nonlinear and effective stress analyses. It will be referred as
YUSAYUSA in the followings.
These two programs are examined under the same condition, i.e., the same ground model and the same stressstrain
relationships. The only difference is that velocity proportional damping is considered in YUSAYUSA for the stability
of numerical integration. Since the Rayleigh damping is
employed, viscous coefficient matric [c] is calculated from
the mass matrix [m] and stiffness matrix [k] as
c am bk

where a and b are constants and are set 0.090 and 0.00091,
respectively, so that average damping ratio between 0.5 Hz
and 6 Hz becomes 1.43 % (Q = 35).
Method of each analysis is briefly explained in the
followings.

Table 1 Earthquake motions

Record name

Inland

Port Island GL-33 m, NS 1995 Kobe

543.5

JMA Kobe, NS

1995 Kobe

Hakuta, NS

1999 Tottoriken-seibu

JMA Kawaguchi, EW

2004 Niigataken-chuetsu

Kashiwa-Kariwa Service
hole SG1, NS

2007 Niigataken-chuetsu-oki

Ichinoseki-nishi, EW

2008 Iwate-Miyagi-nairiku

66.5

29.0

5.5

300.6

67.1

7.8

818.0

45.9

5.1

5.8

464.9

50.7

4.1

719.2

834.2

7.7

5.7

307.2

41.4

7.3

1675.8

146.6

38.3

6.5

789.0

155.3

9.6

347.2

45.2

5.3

219.4

42.3

9.25

6.31

1434.6

65.1

22.2

6.0

569.4

73.8

1978 Miyagiken-oki

286.8

31.5

8.1

4.9

147.4

22.6

16.3

Akita port

1983 Nihonkai-chubu

235.3

29.4

14.3

5.1

147.8

33.7

46.6

Hachinohe office, EW

1994 Sanriku-haruka

675.5

45.2

13.3

5.8

436.8

70.9

12.9

Hanasaki, EW
Taiki, EW

1994 Hokkaido-toho-oki
2003 Tokachi-oki

350.7
499.8

24.8
77.2

6.9
51.2

5.3
5.9

205.0
378.8

31.7
67.3

26.5
61.1

Ocean trench Kaihoku bridge EW

123

Earthquake

amax cm/s2 vmax cm/s dmax cm IJMA a0 cm/s2 SI cm/s T s

Type

10.8

Acceleration (cm/s )

Indian Geotech J

400
200
0
-200
-400
-600

Port Island, GL-33 m, NS

10

15

20
2

Acceleration (cm/s )

Acceleration (cm/s )

Time (s)
500

JMA Kobe, NS

0
-500
-1000

10

15

Kaihoku

200
0
-200
0

10

15

20

0
-500
0

10

15

20

40

Akita Port

0
-200
0

10

15

20

25

30

35

40

45

50

55

60

Acceleration (cm/s )

Time (s)

Acceleration (cm/s )

35

200

1500
1000
500
0
-500
-1000

Kawaguchi

10

15

20

500
Hachinohe

0
-500
0

10

15

20

25

30

35

40

Time (s)

Acceleration (cm/s )

Time (s)

Acceleration (cm/s )

30

Acceleration (cm/s )

Hakuta

500

Time (s)

400
Kashiwazaki

200
0
-200
-400

10

15

20

Hanasaki

10

15

20

25

30

35

40

45

50

55

60

Time (s)

Acceleration (cm/s )

Time (s)

Acceleration (cm/s )

25

Time (s)

Acceleration (cm/s )

Time (s)

1500
1000
500
0
-500
-1000

Ichinoseki-nishi

10

15

20

500
Taiki

0
-500

10 15 20 25 30 35 40 45 50 55 60 65 70 75 80

Time (s)

Time (s)

(a) Inland earthquakes

(b) Ocean trench earthquakes

Fig. 3 Input earthquake motions

1000

100

10
0.1

PI
Ichinoseki
Kobe
Kawaguchi
Hakuta
Kashiwazaki

Period (s)

10

5% Damping

Displacement (cm)

1000

Akita
Hanasaki
Kaihoku
Kaiki
Hachinohe

Velocity (cm/s)

Acceleration (cm/s )

5% Damping

100

10

1
0.1

Akita
Hanasaki
Kaihoku
Kaiki
Hachinohe

Period (s)

PI
Ichinoseki
Kobe
Kawaguchi
Hakuta
Kashiwazaki

10

100

5% Damping

PI
Ichinoseki
Kobe
Kawaguchi
Hakuta
Kashiwazaki

10
Akita
Hanasaki
Kaihoku
Kaiki
Hachinohe

1
0.1

10

Period (s)

Fig. 4 Response spectra of input motion

123

Indian Geotech J

SHAKE
Amplification factor

GL-1.5 m

SHAKE [1] is based on the multiple reflection theory;

equation of motion is solved in the frequency domain.
Equivalent linear method is used in order to consider nonlinear behavior, in which stiffness and damping ratio at the
effective stress ceff are used in whole analysis. The effective
stress is evaluated from the maximum strain cmax as

0.1

Downhole
Suspension
Observed

0.01
0.1

10

where a is an adjusting parameter and 0.65 is usually used.

Two shortages are known in SHAKE [1] in addition to
the fact that it cannot express the accurate stressstrain
behavior because it is produced by the complex moduli.
The first shortage of SHAKE is overestimation of the
maximum acceleration. The mechanism is shown in Fig. 5.
Since stressstrain relationships is a linear line passing the
point A (ceff, s = Gceff), maximum stress at c = cmax (s1 at
point C) is larger than the actual stress (s2 at point B),
which overestimation of the maximum shear stress result in
overestimation of the maximum acceleration (See Total
Stress vs. Effective Stress section on the mechanism).
The second shortage is typically seen in Fig. 6. Amplification is significantly underestimated at high frequency.
Maximum acceleration is underestimated by this mechanism.
These two factors work in the opposite way. The first
mechanism overestimates maximum acceleration and the
second mechanism underestimates maximum acceleration.
Degree of effects depends on the degree of nonlinearity. This
may be one of the reasons why evaluations of SHAKE scatter.

10

Frequency (Hz)

Fig. 6 Underestimation of amplification at high frequencies [4]

Fa
con ilure
fin lin
ing e a
str t low
e ss
Ph
as
et F
ra ailu
ns re
fo
rm line
lin
e

ceff acmax

10

= Bu

v

v

v v
0

'v
'v

v' = m
v

Bp 2

m tan = 1

YUSAYUSA

Fig. 7 Stress-path model in YUSAYUSA

YUSAYUSA [12] is a truly nonlinear seismic response

analysis program. It uses hyperbolic model, Eq. (5), for
backbone curve and the Masings rule is used to develop
hysteresis curve. Both total and effective stress analyses
can be made. Excess porewater pressure generation is
evaluated in the effective stress analysis by specifying the
stress paths in the shear stress-effective overburden stress

plane as shown in Fig. 7, where Bp and Bu are parameters

to control excess porewater pressure generation. They are
evaluated so that R20 and R5 coincide with the liquefaction
strength defined in Analyzed sites and modeling of
mechanical properties section.

Earthquake Motion Indices

Relationship between
maximum strain and
maximum stress

C
Overestimation of
maximum stress

2
A

B
Stress-strain curve

Stress-strain curve
used in the analysis
O

eff

max

Fig. 5 Mechanism of overestimation of maximum shear stress [4]

123

Many earthquake motion indices have been proposed,

which indicates difficulty to express the intensity of the
ground motion in one index. In this paper, several frequently used indices are used.
(1) Peak acceleration Peak acceleration is the most frequently used index because it directly related to
inertia force of the structure. However, it is sometimes too sensitive.
(2) Peak velocity Applicability of the peak acceleration as
an earthquake motion index becomes less for longer
period structure. Instead of it, peak velocity becomes a

Indian Geotech J

IJMA 2 log a0 0:94

11

where a0 is the acceleration in cm/s2; sum of the

duration where acceleration exceeds a0 is just 0.3 s in
the filtered acceleration-time history. It is defined by
using three directional components, but only onedirectional component is used in this paper. The
instrumental seismic intensity is considered to be
average of PGA and PGV [17].
(5) Spectral Intensity Spectral intensity SI is first proposed
by Housner [18] by integrating the response spectrum
from 0.1 s to 2.5 s. It is modified to divide 2.4 s in order
to make the dimension of the seismic intensity to be
same with velocity in Japan. Then it is calculated as
1
SI
2:4

Z2:5
Sv;0:2 dt

12

0:1

where Sv,0.2 is a velocity response spectrum under 20 %

damping. Equation (12) indicates that SI is an average
of the velocity spectrum between 0.1 and 2.5 s. Pseudo
velocity spectrum is sometimes used instead of velocity
spectrum, but velocity spectrum is used in this paper.

Comparison and Discussion

Equivalent Linear vs. Truly Nonlinear
Results under 11 earthquake motions are shown in one
figure in the followings, in which result of each earthquake
motion is distinguished by the symbols shown in Fig. 8.

Inland eq.
Port Island
JMA Kobe
Hakuta
Kawaguchi
Kashiwazaki
Ichikan-nishi
Ocean trench eq.
Akita Port
Hachinohe
Hanasaki
Kaihoku Br.
Taiki

15

1.5:1
1:1

(4)

Fig. 8 Legend to distinguish

input earthquake

PGAEQ (m/s )

(3)

good index. Here, this velocity is absolute velocity, but

not relative velocity from the base although relative
velocity is frequently used as outputs because it is
directly obtained by solving the equation of motion.
Peak displacement Displacement is a very important
index for the design of the underground structures.
Comparison of the relative displacements with
recorded one is very difficult because vertical array
records are necessary and because it depends on the
method of integration as will be seen in in Fig. 11.
However, it can be done between analyses.
Instrumental seismic intensity Seismic intensity has
been used to express the degree of damage or ground
shaking. Because of its nature, it takes time to
evaluate the seismic intensity. After the 1995 Kobe
earthquake, Japan Meteorological Agency made an
seismic intensity which can be calculated from the
acceleration time history and which is consistent with
traditional JMA seismic intensity [16]. Evaluation of
the seismic intensity can be done very quickly.
Instrumental seismic intensity IJMA is evaluated as

10

10

15

PGANL (m/s )

Fig. 9 Comparison of peak ground accelerations

Figure 9 compares PGAs by all analyses, where subscript EQ indicates equivalent linear method (SHAKE) and
NL indicates truly nonlinear method (YUSAYUSA). Two
lines are drawn in the figure whose gradients are 1:1 and
1.5:1. PGA by SHAKE is always larger than that by
YUSAYUSA except only several cases. PGA is overestimated more than 1.5 times in SHAKE in many cases which
agrees with the discussion in Ref. [4] and Fig. 5.
Peak ground velocity, PGV, is compared in Fig. 10.
Both PGVs are nearly the same although that by the
equivalent linear method seems to be a little larger than
that by the nonlinear analysis. There are, however, two
exceptions. PGV by nonlinear method shows much larger
value than that of equivalent linear method under Kawaguchi ( ) and Kaihoku ( ). This is not caused from the
difference of the analytical methods but other reason.
Input earthquake motions are integrated to obtain the
absolute velocity by the Fourier transform in SHAKE and
Newmarks b method (b = 0.25) in YUSAYUSA, respectively. Zero-th order or constant term is neglected in SHAKE,

123

Indian Geotech J
0.30

2.0

0.25

PGD EQ (m)

PGVEQ (m/s)

1.5

1.0

0.20
0.15
0.10

0.5
0.05

0.0

0.00

0.5

1.0

1.5

2.0

0.05

0.10

0.20

0.25

0.30

6.5

7.0

Fig. 12 Comparison of PGD

Fig. 10 Comparison of peak ground (absolute) velocity

4

7.0

Kawaguchi

6.5

Ichikan-nishi

6.0

Kaihoku

Hakuta

Kashiwazaki

Taiki

Port Island

IJMA,EQ

Velocity (Newmark ) (m/s)

0.15

PGDNL (m)

PGVNL (m/s)

5.5
5.0

JMA Kobe

4.5

Akita port

Hachinohe
Hanasaki
0
0

4.0
4.0

4.5

5.0

5.5

6.0

IJMA,NL

Velocity (Fourier) (m/s)

Fig. 13 Comparison of JMA seismic intensity

Fig. 11 Maximum velocity of input earthquake motions

123

200

150

SIEQ (cm/s)

but is also integrated in YUSAYUSA. Figure 11 compares

maximum velocity integrated by the Fourier transform and by
the Newmarks b method. Almost all earthquake motion
shows the same maximum velocity in two methods, but
maximum velocity by the Newmarks b method is much
larger than that by Fourier transform for Kaihoku and JMA
Kawaguchi, which is the reason why PGVs by YUSAYUSA
are much larger than those by SHAKE in Fig. 10. The base
line correction is made on all earthquake motions except two
earthquakes. It indicates that the base line may be important
when absolute velocity is evaluated.
Considering these it can be concluded that PGVs are
nearly the same for both SHAKE and YUSAYUSA
although those by SHAKE is a little larger than those by
YUSAYUSA.
Figure 12 compares displacements at the ground surface
relative to base, PGD. PGD by SHAKE is in general
smaller than that by YUSAYUSA. This behavior can be
explained by the mechanism in Fig. 5, too. Since stiffness

100

50

50

100

150

200

SINL (cm/s)

Fig. 14 Comparison of SI value

is estimated large in SHAKE, resulting strain becomes

small. It is also seen that degree of difference depends on
the earthquakes as the same symbol shows similar

7.0

0.14
6.5

0.12
6.0

0.10

IJMA,EF

Maximum strain (Equivalent linear analysis)

Indian Geotech J

0.08
0.06

5.5
5.0

0.04

4.5

0.02

4.0
3.5
3.5

0.00
0

0.04

0.08

0.12

4.0

4.5

6.0

6.5

7.0

Fig. 18 Comparison of IJMA

10

140
120

SIEF (cm/s)

PGAEF (m/s )

5.5

IJMA,NL

Maximum strain (Total stress analysis)

Fig. 15 Comparison of maximum strains

5.0

100
80
60
40

20
0

10

20

40

60

80

100 120 140

SINL (cm/s)

PGANL (m/s )

Fig. 19 Comparison of SI value

Fig. 16 Comparison of PGA

2.0

ru

PGV EF (m/s)

1.5

dz r u
1.0

t +dt
Fig. 20 Equilibrium of soil column

0.5

0.0

0.5

1.0

1.5

2.0

PGVNL (m/s)
Fig. 17 Comparison of PGV

tendency. Both analyses, for example, show similar PDG

under Kaihoku. PDGs under Hachinohe also seem similar
although PGD by SHAKE becomes smaller when PDG

becomes larger than 0.1 m. In addition, agreement is better

under the ocean type earthquake than under the inland
earthquake, except the Taiki earthquake motion.
Figures 13 and 14 compare JMA seismic intensity IJMA
and spectral intensity SI, respectively. Same with PGV,
both indices are nearly the same although those by SHAKE
are a little larger than that by YUSAYUSA. It is also
pointed out that data under the same earthquake motion
show similar characteristics.

123

Indian Geotech J

Figure 15 shows comparison of the maximum stresses

by both methods. Here, the largest strain in each ground is
used as the maximum strain. Therefore, it may not be the
same layer. If the mechanism shown in Fig. 5 works,
the shear strain by SHAKE is smaller than that by
YUSAYUSA. Validity of this mechanism was confirmed
through Figs. 5 and 12 in general. However, there are many
data that strains by SHAKE are larger than those by
YUSAYUSA. It may indicate that vibration modes are
different under each analysis.

1.4

PGD EF (m)

1.2
1.0
0.8
0.6
0.4
0.2
0.0

0.1

0.2

0.3

0.4

0.5

Total Stress vs. Effective Stress

PGDNL (m)

Maximum strain (Effective stress analysis)

Fig. 21 Comparison of PGD

1.0

0.5

0.0

0.05

0.10

0.15

Maximum strain (Total stress analysis)

Fig. 22 Comparison of maximum shear strain. a Equivalent/nonlinear b Effective stress/nonlinear

Effective stress analysis and total stress truly nonlinear

analysis are compared here. Both calculations are made by
YUSAYUSA; the only difference is consideration of
excess porewater pressure generation or not. The subscript
EF and NL indicate effective stress and total stress analyses, respectively.
The peak ground acceleration PGA is compared in
Fig. 16. Many data lie on the line PGAEF = PGANL. They
are the case where there is no liquefiable layer, in which
case results become identical. Beside these data, PGA by
effective stress analysis is smaller than that by total stress
analysis. Sometimes it reaches 1/3 or less.
The peak ground velocity PGV is compared in Fig. 17.
Again magnitude by the effective stress analysis is much
smaller than that by the total stress analysis. This is quite
different feature compared with the comparison between
equivalent linear and truly nonlinear analyses in Fig. 10, in
which PGVs are similar.
PGA

PGA
2
SI

`PGV

1
PGV

SI

Port Island
JMA Kobe
Hakuta
Kawaguchi
Kashiwazaki
Ichinoseki-nishi
Akita
Hachinohe
Hanasaki
Kaihoku
Taiki

1
1
2
1

PGD

IJMA
1

max

2
PGD

IJMA
2

max

(a) Equivalent / Nonlinear

(b) Effective stress / Nonlinear

Fig. 23 Ratio of earthquake motion indices by different seismic response analyses. a Equivalent/nonlinear. b Effective stress/nonlinear

123

Indian Geotech J
PGA

PGA
2
SI

`PGV

PGV

SI

1
1
2

PGD

IJMA
1

2
PGD

IJMA
2

max

max

(a) Equivalent / Nonlinear

(b) Effective stress / Nonlinear

Fig. 24 Average of ratio of earthquake motion indices by different seismic response analyses

The instrumental seismic intensity IJMA is compared in

Fig. 18 and the spectral intensity SI is compared in Fig. 19.
Similar with PGA and PGV, those by the effective stress
analysis are frequently much smaller than those by the total
stress analysis. These observations can be explained by
considering the upper bound ground motion caused by the
nonlinear behavior [19].
Left side of Fig. 20 shows equilibrium of a infinitesimally small element, from which equation of motion or
equation of wave motion can be derived. On the other
hand, right side of Fig. 20 shows equilibrium of a soil
column above the weak layer. If shear stress of the weak
layer reaches shear strength, average acceleration above
this layer reaches upper bound. This mechanism can
explain the reason why PGA by the effective stress analysis
is smaller than that by the total stress analysis.
There are also upper bounds on IJMA and SI values, but
the mechanism is somewhat different from the upper bound
of PGA. IJMA uses frequency between 0.5 to 8 Hz and SI
uses period between 0.1 to 2.5 s. Period of the wave

become longer as nonlinear behavior becomes significant,

and if period becomes longer than the period that is used in
these indices, earthquake motion does not contribute these
indices. Indices by the effective stress analysis are sometimes much smaller than that by the total stress analysis,
which indicates that nonlinear behavior is significant when
liquefaction occurs.
Peak ground displacements are compared in Fig. 21 and
maximum shear strains are compared in Fig. 22. Those by
the effective stress analysis are sometimes much larger
than those by the total stress analysis, which also suggests
significant nonlinear behavior.
Summary
Figure 23 shows summary of the comparisons. Ratios of
earthquake motion indices by the equivalent linear analysis
to that by the truly nonlinear analysis are shown by means
of rader chart in Fig. 23a for each input earthquakes. Here,
cmax indicates maximum shear strain. The average values

123

Indian Geotech J

of the ratio is also shown in Fig. 24. Since accuracy of the

seismic response methods is discussed by means of several
indices, the rader chart is one of the best methods to see the
general nature of the applicability, although it has not been
used in the past researches as only one or two indices are
investigated.
The PGA is always larger in the equivalent analysis
than the nonlinear analysis regardless of earthquakes. The
average of overestimation is 1.41 although it scatters a
little. On the other hand, ratios for instrumental seismic
intensity IJMA are always very close to 1.0 and average of
them is 1.01. In the same manner, average of SI is 1.05
and ratios scatters around 1.0 depending on earthquakes.
Again scattering is small. The average of PGV is 0.94,
but in increases two exceptional data explained in Fig. 11.
It becomes 1.04 if these two data are neglected. On the
other hand, PGDs are underestimated by the equivalent
linear method; the average underestimation is 0.80. In the
same manner, maximum shear strain cmax is also underestimated with average of 0.75. Scatters of ratios by
different earthquakes are larger than those of SI, IJMA and
PGV.
The peak ground acceleration PGA is much smaller in
the effective stress method; the average ratio is 0.60. Here
it is noted that cyclic mobility behavior, by which pulsy
acceleration may appear, is considered in YUSAYUSA,
but it sometimes do not appear significantly. Therefore, this
ratio may be different if other computer programs are used.
However, these pulsy wave usually does not affect other
indices.
The peak ground absolute velocity PGV, Spectral
intensity SI, and instrumental seismic intensity IJMA are
also much smaller in the effective stress analysis than those
the total stress nonlinear analysis; the average values are
0.70, 0.51 and 0.89, respectively. Scattering of IJMA is very
small. Scattering of PGV and SI are a little larger than that
of IJMA. On the other hand, peak displacement PGD and
maximum shear strain cmax are much larger in the effective
stress analysis than in the total stress analysis. The average
values are 2.26 and 5.11 for PGD and gmax, but they may
not have practical meaning as the scatterings are very large.

(1)

(2)

(3)

(4)

The peak ground acceleration PGA is overestimated by

the equivalent linear analysis compared with the truly
nonlinear analysis; the overestimation is 1.41. It
decreases significantly in the effective stress; PGA by
effective stress is 0.6 times of PGA by nonlinear analysis.
The peak ground displacement PGD is underestimated by the equivalent linear analysis; the average
value is 0.80. It increases significantly in the effective
stress analysis with average of 2.26.
Other indices such as peak ground absolute velocity
PGV, seismic intensity IJMA, and spectral intensity SI
are nearly the same between the equivalent linear
analysis and truly nonlinear analysis; the average
ratios are 1.04, 1.01, and 1.05 respectively. However,
they are significantly small in the effective stress
analysis; the average ratios compared with the total
stress analysis are 0.70, 0.87, and 0.51, respectively.
Vibration mode may be different depending on
different analyses because maximum shear strain
and PGD sometimes shows different nature.

Considering these, it is recommended to use effective

stress analysis at the liquefiable site in order to make
rational design, but total stress analysis is applicable
because it shows larger earthquake motion indices which
result in conservative design. Use of total stress analysis is
not recommended for the design of underground structures
because displacement and strain of the ground are underestimated such as piles and pipes.
Equivalent linear method is applicable in many fields
when there is no liquefiable layer because all the earthquake motion indices are nearly same with those by the
truly nonlinear analyses. Use of equivalent linear method is
not recommended for the case where maximum acceleration is a key factor of the design. Even in this case, however, it can be used because use of equivalent linear
analysis result in a conservative design. Use of total stress
is important when displacement of the ground is of interested such as piles and pipes.

References
Concluding Remarks
Three seismic response analyses, the equivalent linear
analysis, the total stress truly nonlinear analysis, and the
effective stress analysis, are compared by using more than
250 sites and 11 large earthquake motions. Since typical
grounds and earthquake motions are included, the conclusions obtained here shows general feature of the analysis.
The following conclusions are obtained.

123

1. Schnabel PB, Lysmer J, Seed HB (1972) SHAKE-A Computer

program for earthquake response analysis of horizontally layered
sites, Report No. EERC72-12. University of California, Berkeley
2. K. Ishihara (1982) Evaluation of soil properties for use in
earthquake response analysis, 9th International Symposium on
Numerical Models in Geomechanics, Zurich, 237259, 1982
3. Yoshida N, Kobayashi S, Suetomi I, Miura K (2002) Equivalent
linear method considering frequency dependent characteristics of
stiffness and damping. Soil Dyn Earthq Eng 22(3):205222
4. Yamahara H (1971) Seismic response of structure considering
energy balance. J Struct Eng 187:6974 (in Japanese)

Indian Geotech J
5. Nagahashi S, Kobayashi H (1987) Maximum amplitude as a
simple index to represent strength of earthquake motion. J Struct
Eng 181:1522 (in Japanese)
6. Tong H, Yamazaki F, Sasaki H, Matsumoto SA (1994) relationship between seismic ground motion severity and house
damage ratio, 9th Japan Earthquake Engineering Symposium,
22992304, 1994
7. Public Work Research Institute (1996) Report on accuracy of method
to identify liquefaction during the Hyogoken-nambu earthquake.
Public Work Research Institute, Bunkyo-ku (in Japanese)
8. Japan Road Association (2012) Specifications for highway
bridges, Part V (seismic design). Japan Road Association, Chiyoda-ku (in Japanese)
9. Japanese Geotechnical Society (2004) Japanese standards for
geotechnical and geoenvironmental investigation methods-standards and explanations. Japanese Geotechnical Society, Bunkyoku (in Japanese)
10. Hatanaka M, Uchida A (1996) Empirical correlation between
penetration resistance and internal friction angle of sandy soils.
Soils Found 36(4):19
11. Seed HB, Idriss IM, Arango I (1981) Evaluation of liquefaction
potential using field performance data. J Geotech Eng, ASCE
109(3):458482
12. Y. Usami, T. Sasaki, N. Yoshida (2009) Effect of velocity
structure under seismic base layer on earthquake response,

13.

14.

15.

16.
17.

18.
19.

Annual Meeting of Tohoku Branch of JSCE, 317318, 2009. (in

Japanese)
The Over-seas Coastal Area Development Institute of Japan
(2007) Technical standards and commentary of port and harbour
facilities in Japan. The Over-seas Coastal Area Development
Institute of Japan, chiyoda-ku (in Japanese)
N. Yoshida (1995) A computer program for dynamic response analysis of level ground by equivalent linear method, Revised in 2004
(version 3.25), Tohoku Gakuin University; http://www.civil.tohokugakuin.ac.jp/yoshida/computercodes/index.html, 1995
N. Yoshida, I. Towhata, YUSAYUSA-2 and SIMMDL-2, theory
and practice, revised in 2003 (version 2.1), Tohoku Gakuin
University and University of Tokyo; http://www.civil.tohokugakuin.ac.jp/yoshida/computercodes/index.html, 1991
Muramatsu I (1996) On JMA instrumental seismic intensity.
JSEEP NEWS 150:2736 (in Japanese)
Kinoshita S, Ohtake M (eds) (2000) Fundamentals of strong
motion, National Re-search Institute for Earthquake Science and
Disaster Prevention, Tsukuba http://www.k-net.bosai.go.jp/k-net/
gk/publication/
Housner GW (1965) Intensity of earthquake ground shaking near
the causative fault, 3WCEE, I, III-94-III-115, 1965
Yoshida N (1999) Large earthquake motion and ground-nonlinear
problem. Jishin J, Association for Development of Earthquake
Prediction 28:6674 (in Japanese)

123