1142

Retaining walls damaged in the Chi-Chi

earthquake
Y.S. Fang, Y.C. Yang, and T.J. Chen
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

Abstract: This paper investigates the failure of three gravity walls due to the 1999 Chi-Chi earthquake. Characteristics
of the damaged walls were carefully recorded and backfill materials behind the damaged walls were collected and
tested in the laboratory. Both the simplified analysis based on the Mononobe–Okabe method and the simplified
dynamic analysis based on the Richards–Elms method were adopted. For the first case, the sliding of concrete wall
blocks along the construction joint was observed. It was found that, during the earthquake, the frictional resistance at
the untreated construction joint was not sufficient to resist the dynamic lateral thrust. For the second case, the retaining
wall settled significantly and tilted about its toe. Seismic analysis of the wall indicated that, under the same horizontal
acceleration, the factor of safety against bearing capacity failure was lower than that against overturning and sliding. A
stability check against bearing capacity failure for the retaining wall should never be neglected. For the third case, the
retaining wall built on top of the Che-Lung-Pu fault was severely damaged by the fault rupture. During the earthquake,
the vertical displacement of the hanging wall uplifted the backfill, causing the wall to overturn. Horizontal displace-
ment of the hanging wall caused the wall to slide and the soil in front of the toe to heave.
Key words: analysis, bearing capacity, case study, earthquake, failure, retaining wall.
Résumé : Cet article étudie la rupture de trois murs poids causées par le tremblement de terre de Chi-Chi en 1999.
Les caractéristiques des murs endommagés ont été notées soigneusement et les matériaux de remblai à l’arrière des
For personal use only.

murs endommagés ont été prélevés et soumis à des essais en laboratoire. On a adopté l’analyse simplifiée basée sur la
méthode de Mononobe–Okabe de même que l’analyse dynamique simplifiée basée sur la méthode Richards–Elms. Dans
le premier cas, on a observé le glissement de blocs du mur de béton le long du joint de construction. On a trouvé que,
durant le tremblement de terre, la résistance au frottement le long du joint de construction non traité n’était pas suffi-
sante pour résister à la poussée dynamique latérale. Dans le deuxième cas, le mur de soutènement s’est affaissé de
façon significative et s’est incliné autour de son pied. Une analyse séismique du mur a indiqué que, sous la même
accélération horizontale, le coefficient de sécurité contre la rupture en capacité portante de la fondation était plus faible
que contre le renversement et le glissement. Une vérification de la stabilité contre la rupture en capacité portante d’un
mur de soutènement ne devrait jamais être négligée. Dans le troisième cas, le mur de soutènement construit sur la
faille de Che-Lung-Pu a été endommagé considérablement par la rupture de la faille. Durant le tremblement de terre, le
déplacement vertical du mur suspendu a soulevé le remblai, causant le renversement du mur. Le déplacement horizontal
du mur suspendu a produit un glissement du mur, et le sol en avant du pied s’est soulevé.
Mots clés : analyse, capacité portante, étude de cas, tremblement de terre, rupture, mur de soutènement.

[Traduit par la Rédaction] Fang et al. 1153

Introduction translated as Ji-Ji or Gi-Gi), as indicated in Fig. 1. The focal

depth was about 8.0 km. The rupture of ground along the
On 21 September 1999 at 1:47 a.m., a disastrous earth- Che-Lung-Pu fault was about 105 km long. The frequency
quake with a magnitude of 7.3 on the Richter scale struck of shaking ranged from 1.0 to 4.0 Hz. The horizontal peak
Taiwan. The Central Weather Bureau located the epicenter at ground acceleration (PGA) measured at strong ground mo-
23.85°N, 120.82°E near the town of Chi-Chi (sometimes tion station TCU-065 was as high as 774 Gal (0.79g) as
listed in Table 1. In this incident, 49 542 dwellings were to-
Received 2 April 2002. Accepted 31 July 2003. Published on tally damaged and 42 746 dwellings were partially damaged.
the NRC Research Press Web site at http://cgj.nrc.ca on Most unfortunately, 2432 people were killed, and 46 people
31 October 2003.
are still missing. Immediately after the earthquake, the Na-
Y.S. Fang1 and Y.C. Yang. Department of Civil tional Science Council of Taiwan mobilized more than 1200
Engineering, National Chiao Tung University, 1001 Ta Hsueh scientists and engineers to assess the various kinds of dam-
Road, Hsinchu 30010, Taiwan, Republic of China. age due to the earthquake. The authors joined the investiga-
T.J. Chen. Sinotech Engineering Consultants Ltd., tion project and studied the failure of retaining structures
171 Nanking East Road, Sec. 5, Taipei 105, Taiwan, due to the earthquake. For more information regarding the
Republic of China.
strong ground motion characteristics of the Chi-Chi earth-
1
Corresponding author (e-mail: ysfang@cc.nctu.edu.tw). quake, the reader is referred to Tsai and Huang (2000) and

Can. Geotech. J. 40: 1142–1153 (2003) doi: 10.1139/T03-055 © 2003 NRC Canada
 Fang et al. 1143

Fig. 1. Location of strong ground motion stations (TCU) and sites investigated.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14
For personal use only.

Table 1. Records of peak ground acceleration In this paper, the authors introduce the behavior of retain-
(cm/s2) (Central Weather Bureau 1999). ing walls under seismic excitation. Three cases of gravity
wall failure are reported. For the first case, sliding of con-
Horizontal
crete wall blocks along construction joints was observed. For
Station Vertical East–west North–south the second case, the retaining wall settled significantly and
TCU-065 258 563 774 tilted about its toe. For the third case, the gravity wall built
TCU-072 275 371 465 on top of the Che-Lung-Pu fault was severely damaged. The
TCU-078 171 302 440 research team carefully measured and recorded the charac-
TCU-079 384 417 580 teristics of the damaged walls. Backfill materials behind the
damaged walls were collected and taken to the soil mechan-
ics laboratory. Experiments were conducted to determine the
physical properties and strength parameters of the backfill.
Idriss and Abrahamson (2000). A dedicated issue containing Using these parameters, an analysis was done to evaluate the
42 articles regarding the seismic aspects of the Chi-Chi seismic adequacy of the damaged walls. In this paper, the
earthquake was published as a Bulletin of the Seismological simplified analysis based on the Mononobe–Okabe (M–O)
Society of America in October 2001. method and the simplified dynamic analysis based on the
Ueng et al. (2001) reported the geotechnical hazards ob- Richards–Elms (R–E) method were adopted. The intention
served after the earthquake, including landslides, soil lique- of the paper is to document several case histories of dam-
faction, foundation failures, and ground movements. Fang aged walls such that lessons can be learned from this failure.
and Chen (1999), Huang (2000), and Fang et al. (2001) re-
corded the failure of quay walls, masonry walls, gravity Seismic behavior of gravity walls
walls, modular-block retaining walls, and reinforced earth-
retaining structures. This paper analyzes the seismic ade- Seed and Whitman (1970) reported outward movements
quacy of three gravity retaining walls under earthquake exci- of retaining walls and wing walls for the 1960 Chilean and
tation. Mononobe (1924) and Okabe (1924) proposed an 1964 Niigata earthquakes. The behavior of a gravity retain-
analytical method (the Mononobe-Okabe method) to esti- ing wall under horizontal acceleration ah can be explained
mate the dynamic earth pressure against the wall under with the help of Fig. 2. As the ground was shaken from left
a seismic condition. Seed and Whitman (1970) described to right, the dynamic earth pressure and inertia force of the
how to estimate the dynamic earth pressure based on the wall would act from right to left. If the resistance at the base
Mononobe–Okabe theory. Richards and Elms (1979) pro- of the wall was not sufficient to defy the dynamic thrust, ac-
posed a procedure (the Richards–Elms method) for the tive wall movement would occur. If the direction of ground
design of gravity retaining walls based on the limit- shaking was reversed, however, the wall would be thrown
displacement concept. from left to right. Due to the existence of the backfill, pas-

© 2003 NRC Canada

 1144 Can. Geotech. J. Vol. 40, 2003

Fig. 2. Seismic behavior of gravity wall. Fig. 3. Profile of the damaged retaining wall at site 1.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

sive earth pressure would act between the wall and the soil,
and only partial active wall movement could be recovered. If
the wall moved outward a small amount due to each effec- Fig. 4. Outward movement of the gravity wall along construction
tive cycle of ground shaking, with increasing effective cy- joint.
cles, the wall would gradually move away from the backfill.
For personal use only.

As indicated in Fig. 2, with the outward movement of the

wall, the active soil wedge would slide down along the fail-
ure plane, causing settlement and tension cracks in the back-
fill.

Site 1: sliding along construction joints

The damaged retaining wall was located along Taiwan
Provincial Highway 14 and is labeled as site 1 in Fig. 1. The
gravity wall was built to retain a steep excavation for high-
way construction. The wall was 4.7 m high, 2.0 m thick at
the base, and 0.6 m thick at the top as illustrated in Fig. 3.
The retaining structure was constructed by placing concrete
in five pours. It should be mentioned that no special treat-
ment had been provided at the construction joints of the
wall. During the Chi-Chi earthquake, the upper parts of the
wall gradually moved away from the backfill along the con-
struction joints. Eventually, the top two blocks fell down to
the side ditch of Highway 14 as indicated in Fig. 4. The
damaged zone was approximately 20 m long. Based on field
observations, it was suspected that the frictional resistance at
the flat construction joint was not sufficient to resist the dy-
namic earth pressure and the inertial force of the upper part
of the wall. It was clear that a more detailed analysis would
be needed

Soil testing
After the earthquake, the authors joined the damage re-
connaissance team organized by the National Science Coun-
cil of Taiwan. Backfill materials behind the damaged walls
were collected at the site and tested in the soil mechanics
laboratory of the National Chiao Tung University. Physical
property tests and standard Proctor compaction tests (ASTM
2000) were conducted. Physical properties for backfill ob-
tained from sites 1, 2, and 3 are summarized in Table 2, and

© 2003 NRC Canada

 Fang et al. 1145

Table 2. Properties of backfill materials.

Unified Specific Optimum Maximum dry Internal

Site classification gravity, moisture unit weight, Cohesion, friction
No. Location of backfill Gs content (%) γd,max (kN/m3) c (kN/m2) angle, φ (°)
1 Provincial Highway 14, kilometre 38 SC–SM 2.51 12.2 17.6 0.8 35.7
2 Provincial Highway 21, kilometre 71 SC 2.51 24.9 15.6 1.7 31.0
3 Taiwan Cinema Culture Town SP 2.62 11.3 18.1 1.1 38.8

gradation curves of the soil samples are illustrated in Fig. 5. Fig. 5. Grain-size distribution of backfill materials at sites 1–3.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

In the field, the contractors are generally required to com-

pact the soil to at least 90% of standard maximum dry den-
sity. To simulate the performance of the backfill in the field,
the soil sample at its optimum moisture content was com-
pacted in the laboratory with the standard Proctor method
and then tested with the direct shear apparatus. Seed and
Chan (1959) reported that samples compacted dry-of-
optimum tend to be more rigid and stronger than samples
compacted wet-of-optimum. From a practical point of view,
in this study the soils collected from the sites were com-
pacted at the optimum moisture content and then sheared to
failure. The strength parameters obtained are listed in Ta-
ble 2 and were used in the stability analyses of the retaining
walls.

Seismic pressure based analysis

For personal use only.

Kramer (1996) stated that gravity walls are customarily 1

designed by one of the following two approaches: a seismic [1] PAE = γ H 2 (1 − kv) KAE
2
pressure based approach, or a permanent displacement based
approach. Gravity walls have traditionally been designed on where the dynamic active earth pressure coefficient KAE is
the basis of seismic earth pressure. The Mononobe–Okabe given by
method is commonly used to estimate the dynamic earth
pressure due to earthquakes. The total active thrust PAE can
be expressed as

cos 2(φ − β − θ)
[2] KAE = 2
 sin(φ + δ) sin(φ − θ − i) 
cos θ cos β cos (δ + β + θ) 1 +
2

 cos (δ + β + θ) cos (i − β) 

 k  where av is the vertical acceleration). Pseudo-static accelera-

[3] θ = tan−1  h  tions are generally considered smaller than anticipated peak
 1 − kv 
accelerations. Whitman (1990) reported that values corre-
sponding to one third to one half of the peak ground surface
γ is the unit weight of soil, H is the height of the wall, φ is acceleration were commonly used with factors of safety
the internal friction angle of soil, δ is the wall friction angle, (FSs) of 1.0–1.2. Under the combined effect of static and
i is the slope of the ground surface behind the wall, β is the earthquake load, the NAVFAC DM-7.2 design manual of
slope of the back of the wall to the vertical, kh is the hori- the U.S. Department of the Navy (1982) recommends that an
zontal seismic coefficient (= ah/g, where ah and g are the FS between 1.1 and 1.2 is acceptable. It is reported that
horizontal acceleration and gravitational acceleration, re- gravity walls designed by the traditional approach have gen-
spectively), and kv is the vertical seismic coefficient (= av/g, erally performed quite well in earthquakes.

© 2003 NRC Canada

 1146 Can. Geotech. J. Vol. 40, 2003

Seed and Whitman (1970) suggested that the total active Fig. 6. Factor of safety (FS) of wall at site 1 as a function of
thrust PAE could be resolved into the static active thrust PA seismic coefficient.
and the dynamic increment ∆PAE components:
[4] PAE = PA + ∆PAE
PA acts at H/3 above the base of the wall, and ∆PAE acts at
0.6H above the base of the wall. The magnitude and the
point of application of ∆PAE suggested by Seed and Whit-
man were adopted in this study for estimation of overturning
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

moment of the wall. The static active thrust PA is calculated

using Coulomb theory with the cohesion of soil neglected.
The cross section of the collapsed wall at site 1 is shown
in Fig. 3. It is clear in Figs. 3 and 4 that sliding failure oc-
curred along the construction joint A–B. Therefore, stability
analysis was carried out only for the upper two blocks of the
wall. It is true that, due to the earthquake excitation, sliding
failure could occur along any of the construction joints. To
limit the scope of this study, however, the FSs were only de-
termined for blocks 4 and 5 of the wall which actually
failed. The coefficient of friction at joint A–B was assumed
to be 0.5. It is obvious that the actual value of the friction
coefficient depends on the roughness of the upper and lower
surfaces in contact.
The wall friction angle δ between the backfill and the wall
is commonly assumed to be between φ/2 and 2φ/3. The
study by Seed and Whitman (1970) indicated that, for δ = 0
For personal use only.

and δ = φ/2, the seismic active coefficients KAE determined

using the Mononobe–Okabe theory were nearly identical. It
was concluded that values of KAE computed for δ = φ/2 The variation of FS of the wall (blocks 4 and 5) as a func-
could be considered satisfactory for most conditions. In this tion of the horizontal seismic coefficient kh is shown in
study, it was assumed that δ = φ/2. Assuming the backfill Fig. 6. Since for most earthquakes the horizontal accelera-
had been compacted to a relative compaction of 90% before tion components are considerably greater than the vertical
the earthquake, the unit weight of the compacted backfill components, Seed and Whitman (1970) concluded that kv
would be 17.8 kN/m3. For the gravity wall, the unit weight can be neglected for practical purposes. In this study, the
of concrete was assumed to be 23.6 kN/m3. vertical ground acceleration was neglected. Under the static
To evaluate the stability of a retaining wall, the NAVFAC condition (kh = kv = 0), the FS against sliding obtained using
DM-7.2 design manual (U.S. Department of the Navy 1982) the Mononobe–Okabe method was found to be 2.18. Das
requires checks for (i) resistance against overturning, (ii) re- (1999) suggested that, under a static condition, the FS
sistance against sliding, (iii) allowable pressures on the base, against sliding, overturning, and bearing capacity failure
(iv) settlement, and (v) overall stability of the wall. For the should be at least 1.5, 2.0, and 3.0, respectively. It was obvi-
wall damaged at site 1, it is obvious that the resistance ous that the gravity wall was safe from sliding without the
against sliding would be the point of emphasis. The factor of earthquake excitation.
safety against sliding (FSsliding) can be expressed as Figure 6 shows that the FS against sliding decreased with
∑ FR increasing horizontal ground acceleration. As the horizontal
[5] FSsliding = acceleration exceeds 0.24g, the FS against sliding would be
∑ FD
less than 1.0. This means the frictional resistance at the
where Σ FR is the sum of the horizontal resisting forces, and construction joint A–B would be insufficient to resist the
Σ FD is the sum of the horizontal driving forces. The factor dynamic soil thrust. In Table 1, the peak horizontal accelera-
of safety against overturning (FSoverturning) about the toe can tions in east–west and north–south directions measured at
be expressed as station TCU-072 during the Chi-Chi earthquake were 0.38g
and 0.47g, respectively. These values are apparently greater
∑ MR than the yield acceleration of 0.24g. It is possible that, as
[6] FSoverturning =
∑ MO mentioned in the second section of this paper, the wall
moved a small amount due to each effective cycle of ground
where Σ MR is the sum of the resisting moments about the shaking. After several effective cycles, the wall gradually
toe, and Σ MO is the sum of the overturning moments about moved away from the backfill and eventually fell into the
the toe. The computation of FS against sliding and overturn- side ditch. In Fig. 6, sliding failure is shown to be more
ing under earthquake excitation is introduced in most soil likely to occur than overturning, since the FS against sliding
dynamics textbooks, such as those by Das (1993) and is significantly lower than the FS against overturning for the
Kramer (1996). same kh value.

© 2003 NRC Canada

 Fang et al. 1147

Permanent displacement-based analysis Fig. 7. Variation of FS as a function of soil strength.

This approach allows the designer to take into consider-
ation the wall inertia effect and leads to the conclusion that
there might be some lateral movement of a wall even for
mild earthquakes. Design procedures based on Richards and
Elms (1979) and Whitman and Liao (1985) procedures for
estimation of permanent displacement are available. In the
Richards and Elms analysis, the weight of the wall Ww re-
quired to prevent wall motion is expressed as
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

1 
[7] Ww =  γ H 2(1 − kv) KAE  C IE
2 
The coefficient KAE can be determined using eq. [2], and
the wall inertia factor CIE is given by
cos (β + δ) − sin(β + δ) tan φ b
[8] C IE =
(1 − kv) (tan φ b − tan θ)

where tan φ b is the friction coefficient at the base of the wall.

The FS to take into account the effects of dynamic soil pres-
sure and wall inertia is defined as
W
[9] FS =
Ww

where W is the weight of the wall for equilibrium in the potential movement of the steep slope. The installation of a
For personal use only.

static condition. For this method, an FS of 1.5 is generally shearing key would be another efficient method to prevent
required. The variation of FS against sliding based on the sliding failure at the construction joint.
method proposed by Richards and Elms (1979) is also
shown in Fig. 6. It is clear that as the horizontal ground ac- Site 2: bearing capacity failure under the
celeration exceeds 0.15g, the FS would be less than 1.0, toe
which means that the frictional resistance at the construction
joint will not be able to resist the horizontal soil thrust and The damaged retaining wall was built near Sun-Moon
wall inertia force. When the ground shaking is equal to or Lake and is labeled as site 2 in Fig. 1. The gravity wall was
greater than the yield acceleration, permanent wall displace- built to retain the steep slope for the construction of Provin-
ment will occur. cial Highway 21. After the Chi-Chi earthquake, the retaining
Figure 7 shows how the variation of soil property affects wall settled significantly and tilted about its toe as shown in
the results of analysis. Based on the laboratory test results, Fig. 9. Little lateral movement of the wall was observed.
the internal friction of the backfill is assumed to be 35.7°. The active soil wedge behind the wall collapsed, however a
Under the horizontal shaking of 0.15g, the calculated FS large portion of the wall body remained undamaged. The
against sliding (R–E method) is 1.06. If the φ angle of soil beetle-nut plantation on the uphill slope remained stable dur-
is reduced to 34.0°, however, the FS against sliding (R–E ing the earthquake.
method) would drop to only 0.95. The analytical results are The wall was 2.5 m high, 1.25 m thick at the base, and
relatively sensitive to the variation of soil strength. It is pos- 0.5 m thick at the top, as shown in Fig. 10. Weep holes were
sible that locally poor compaction or the presence of weaker fabricated in the wall 0.25 m apart horizontally. The dam-
soil triggered the 20 m long wall failure at site 1, not failure aged zone was about 40 m long. Figure 1 shows that site 2 is
for the whole length of wall. quite close to the epicenter of the Chi-Chi earthquake. It is
The Foundation Design Code of Taiwan adopted the possible that the unexpected strong ground motion is the
Mononobe–Okabe method to determine the dynamic earth main cause of the damage. Adequacy of the gravity wall un-
pressure under seismic loading. Central Taiwan, the most se- der seismic excitation is discussed in the following section.
verely damaged area, was included in zone 2 of the seismic
zoning chart of Taiwan. The design horizontal acceleration Seismic adequacy
recommended for zone 2 is only 0.23g. On 21 September, The variation of FS for the gravity wall at site 2 as a func-
however, the horizontal PGA measured at stations TCU-072 tion of the horizontal seismic coefficient kh is shown in
and TCU-065 was 0.47g and 0.79g, respectively. Under such Fig. 11. It was assumed that the foundation soils had been
a strong vibration, the retaining walls designed with a hori- compacted and had the same properties as the backfill. The
zontal acceleration of 0.23g probably could not remain sta- resisting force against sliding at the base of the wall can be
ble. About 2 months after the earthquake, the damaged wall expressed as
at site 1 had been repaired. Shearing reinforcements were
 2φ   2c 
fabricated at the construction joint as indicated in Fig. 8. [10] ∑ FR = ∑ V tan   + B  
During construction, steel H-piles were driven to resist the  3 3

© 2003 NRC Canada

 1148 Can. Geotech. J. Vol. 40, 2003

Fig. 8. Shearing reinforcement at construction joint.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14
For personal use only.

Fig. 9. Overturning of gravity wall at site 2.

where Σ V is the sum of the vertical forces, B is the width at and should exhibit good resistance against overturning. In
the base of the wall, and c is the cohesion. Fig. 11, at any kh value, the FS against overturning is appar-
Under a static condition, the FS against sliding is only ently higher than that against sliding. This analysis indicates
1.15. This value is unacceptable for design, since it is less that the wall should slide instead of overturn under seismic
than the required FS of 1.5. In Fig. 11, the FS against sliding shaking. This finding is contrary to the wall behavior ob-
decreases with an increase in the horizontal acceleration. served in the field.
With the Mononobe–Okabe method, the FS against sliding Tatsuoka et al. (1998) reported that, under a seismic load,
would be less than 1.0 for a horizontal acceleration greater gravity and semigravity walls could display bearing capacity
than 0.05g. With the Richards–Elms method, the wall would failure under the toe. As indicated in Fig. 12, the incremen-
start to slide as the horizontal acceleration exceeds 0.025g. It tal seismic forces ∆PAE and khW would increase the contact
is clear in Fig. 10 that the wall is relatively short and thick pressure at the toe, qmax. If qmax exceeds the ultimate bearing

© 2003 NRC Canada

 Fang et al. 1149

Fig. 10. Profile of damaged wall at site 2.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14
For personal use only.

Fig. 11. Factor of safety of wall at site 2 as a function of seis- Fig. 12. Assumed contact pressure at the base of the wall.
mic coefficient.

The ultimate bearing capacity of the foundation soil was

capacity, qult, of the underlying soils, settlement would occur
determined with the following equation proposed by
at the toe, causing the retaining wall to tilt. The FS against
Meyerhof (1963):
bearing capacity failure is defined as
qult [12] qult = cNcFcsFcdFci + qNqFqsFqdFqi
[11] FS =
q max + 0.5γBNγ Fγ sFγ dFγ i

© 2003 NRC Canada

 1150 Can. Geotech. J. Vol. 40, 2003

where Nc, Nq, and Nγ are bearing capacity factors; Fcs, Fqs, Fig. 13. Variation of FS as a function of internal friction angle, φ.
and Fγ s are shape factors; Fcd, Fqd, and Fγ d are depth factors;
and Fci, Fqi, and Fγ i are load inclination factors. It should be
noted that the resultant force acting at the base of the wall is
not vertical, and the load inclination factors would signifi-
cantly influence the estimated qult value. Unfortunately, due
to limited funding, the strength parameters c2 and φ2 of the
foundation soil were not available for this case. It was as-
sumed that the properties of the soils under the wall were
similar to those of the compacted backfill. In the analysis,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

the parameters γ 2 = 17.5 kN/m3, φ2 = 31°, and c2 = 1.7 kN/m2

were adopted for the foundation soil.
The variation of FS against bearing capacity failure is also
indicated in Fig. 11. Under the same seismic coefficient, the
FSs obtained are slightly lower than the FS against sliding.
During the Chi-Chi earthquake, the horizontal PGA recorded
at the nearby station TCU-079 was 0.59g. Under such a
strong seismic vibration, the overturning wall movement was
most probably triggered by insufficient bearing capacity un-
der the toe. Based on this case study, it could be concluded
that a stability check against bearing capacity failure for the
gravity wall should never be neglected.
It should be noted in Fig. 11 that, even under static load-
ing, the FSs against sliding and bearing failure are very low.
Figure 13 shows how the variation of soil friction angle af-
fects the analytical results prior to the earthquake. Based on
For personal use only.

the laboratory tests, the angle φ of the compacted backfill is

assumed to be 31.0°. In Fig. 13, if φ is reduced to 29°, bear- and horizontal PGA measured at station TCU-065 near site 3
ing failure would be more likely. However, the FS against were 0.26g and 0.79g, respectively. Figure 15 shows that
bearing failure increases rapidly with an increase in φ. If φ is during the earthquake the vertical displacement of the hang-
increased to 33°, the FS against bearing failure would be ing wall uplifted the backfill, causing the wall to overturn.
2.18, and sliding becomes the possible mode of failure. Horizontal displacement of the hanging wall pushed the
wall, causing it to slide, and the soil in front of the toe, caus-
ing it to heave. It may be concluded that it is of critical im-
Site 3: retaining wall damaged by fault portance for the designer to locate the active faults near the
rupture site and not to construct any structure on top or even near
active fault zones.
Figure 14 shows the failure of a gravity wall located at the
Figure 17 shows the variation of FS against sliding and
Taiwan Cinema Culture Town (site 3) near the township of
overturning for the wall as a function of the horizontal seis-
Wu-Fung. The wall was constructed right above the Che-
mic coefficient. It should be mentioned that the change of
Lung-Pu fault. During the earthquake, the fault rupture
FS in Fig. 17 is solely due to the effect of ground shaking.
caused the wall to slide and overturn. Figure 14 shows that
The procedure of analysis is the same as that for the walls at
the power pole tilted along with the retaining wall. The body
sites 1 and 2. Under a static condition, the FSs against slid-
of the wall remained unbroken, however. Ground heave in
ing and overturning are greater than the required values of
front of the toe is quite obvious in Fig. 14. The backfill be-
1.5 and 2.0, respectively. The wall should be safe from slid-
hind the wall cracked, and the buildings constructed on the
ing and overturning without earthquake excitation. Based on
backfill were severely damaged by the ground deformation
the Richards–Elms method, as the horizontal acceleration
and strong ground shaking.
exceeds 0.17g, the FS against sliding would be less than 1.0.
The failure mechanism of the damaged wall is illustrated
In Fig. 17, the FS against overturning is apparently greater
in Fig. 15. The gravity wall is 4.0 m high and 0.4 m thick at
than the FS against sliding under the same seismic coeffi-
the top. The retaining wall was built on top of the Che-
cient kh. The wall should be safe from overturning up to a
Lung-Pu fault. Figure 16 is an oblique aerial photograph
horizontal acceleration of 0.50g. It may be concluded that
showing the fault rupture cutting through the Taiwan Cinema
failure of the retaining wall was mainly caused by excessive
Culture Town. Lee et al. (2000) investigated the surface fault
fault displacement.
rupture due to the Chi-Chi earthquake. It was reported that
the vertical and horizontal displacements measured at Wu-
Fung were 2.0 and 1.3 m, respectively. The maximum verti- Conclusions
cal fault rupture reported for the Chi-Chi earthquake was as
much as 7.0 m. It is clear in Fig. 16 that the buildings This paper has investigated the failure of retaining struc-
constructed along the fault were severely damaged by the tures due to the 1999 Chi-Chi earthquake. Stability analysis
ground deformation and strong ground shaking. The vertical was performed with both the Mononobe–Okabe and the

© 2003 NRC Canada

 Fang et al. 1151

Fig. 14. Sliding and overturning of retaining wall at site 3.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14
For personal use only.

Fig. 15. Schematic diagram of the wall damaged by fault rup- sured in the east–west and north–south directions was 0.38g
ture. and 0.47g, respectively. It would be reasonable to infer that
the wall moved a small amount due to each effective cycle
of ground shaking. After several effective cycles, the wall
gradually moved away from the backfill and eventually fell
into the side ditch. The installation of shearing keys or
shearing reinforcements would be effective methods to pre-
vent sliding at the construction joint.
For the second case, the retaining wall settled significantly
and tilted about its toe. Tatsuoka et al. (1998) reported that,
under a seismic load, gravity and semigravity walls could
display bearing capacity failure under the toe. If the contact
pressure under the toe, qmax, exceeds the ultimate bearing ca-
pacity, qult, of the underlying soils, settlement would occur at
the toe, causing the retaining wall to tilt. Seismic analysis of
the wall indicates that, under the same horizontal accelera-
tion, the FS against bearing capacity failure was lower than
the FS against overturning and sliding. The overturning wall
failure was most probably triggered by the insufficient bear-
ing capacity beneath the toe. It is concluded that a stability
check against bearing capacity failure for the gravity wall
Richards–Elms methods. Based on the case study for three should never be neglected.
gravity walls, the following concluding remarks are made. For the third case, the retaining wall built on top of the
For the first case, the sliding of concrete wall blocks along Che-Lung-Pu fault was severely damaged by excessive fault
the construction joints was observed. With the Mononobe– displacement. During the earthquake, vertical displacement
Okabe method, as the horizontal acceleration exceeded of the hanging wall uplifted the backfill, causing the wall
0.24g, the FS against sliding at the construction joint would to overturn. Horizontal displacement of the hanging wall
be less than 1.0. With the Richards–Elms method, the mini- caused the wall to slide and the soil in front of the toe to
mum horizontal acceleration to initiate any sliding is 0.15g. heave. The failure of the retaining wall was mainly due to
When the ground shaking is equal to or greater than the the excessive fault rupture. It is of critical importance for the
yield acceleration, permanent wall displacement will occur. engineer not to construct any structure on top of or near an
During the Chi-Chi earthquake, the horizontal PGA mea- active fault.

© 2003 NRC Canada

 1152 Can. Geotech. J. Vol. 40, 2003

Fig. 16. Che-Lung-Pu fault cut through Taiwan Cinema Culture Town (after Sino-Geotechnics Research and Development Foundation
1999).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14
For personal use only.

Fig. 17. Factor of safety of wall at site 3 as a function of seis- Acknowledgements

mic coefficient.
The writers wish to acknowledge the National Science
Council of the Republic of China (NSC 89-2921-Z-319-005-
10) for the financial assistance that made this investigation
possible.

References
ASTM. 2000. Standard test methods for laboratory compaction
characteristics of soil using standard effort (D698-91). In 2000
Annual Book of ASTM Standards. Vol. 04.08. American Soci-
ety for Testing and Materials (ASTM), Conshohocken, Pa.
pp. 78–85.
Central Weather Bureau. 1999. PGA data (ASCI) of Chi-Chi earth-
quake [online]. Available from http://www.cwb.gov.tw.
Das, B.M. 1993. Principles of soil dynamics. PWS-Kent Pub-
lishing Company, Boston, Mass.
Das, B.M. 1999. Principles of foundation engineering. 4th ed.
PWS Publishing Company, Boston, Mass.
Fang, Y.S., and Chen, T.J. 1999. Damage at Taichung harbor due to
1999 Chi-Chi earthquake. Geotechnical Engineering, SEAGS,
30(3): i–v.
Fang, Y.S., Chen, T.J., Yang, Y.C., and Tang, C.C. 2001. The be-
havior of retaining walls under 1999 Chi-Chi earthquake. In
Proceedings of the 4th International Conference on Recent
Advances in Geotechnical Earthquake Engineering and Soil
Dynamics, San Diego, Calif., 26–31 Mar. 2001. Paper 10.06.
University of Missouri-Rolla, Rolla, Mo.

© 2003 NRC Canada

 Fang et al. 1153

Huang, C.C. 2000. Investigation of the damaged soil retaining Sino-Geotechnics Research and Development Foundation. 1999.
structures during the Chi-Chi earthquake. Journal of the Chinese Taiwan Chi-Chi earthquake 9.21.1999 — Bird’s eye view of
Institute of Engineers, 23(4): 417–428. Cher-Lung-Pu fault. 1st ed. Sino-Geotechnics Research and De-
Idriss, I.M., and Abrahamson, N.A. 2000. Geotechnical aspects of velopment Foundation, Taipei, Taiwan.
the earthquake ground motions recorded during the 1999 Chi- Tatsuoka, F., Koseki, J., Tateyama, M., and Horii, K. 1998. Seismic
Chi earthquake. In Proceedings of the International Workshop stability against high seismic loads of geosynthetic-reinforced
on Annual Commemoration of Chi-Chi Earthquake, Taipei, Tai- soil retaining structures: keynote lecture. In Proceedings of the
wan, 18–20 Sept. 2000. Vol. 3. National Center for Research on 6th International Conference on Geosynthetics, Atlanta, Ga.,
Earthquake Engineering, Taipei, Taiwan. pp. 9–22. 25–29 Mar. 1998. Vol. 1. Industrial Fabrics Association Interna-
Kramer, S.L. 1996. Geotechnical earthquake engineering. Prentice- tional, Roseville Minn. pp. 103–142.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

Hall Inc., Englewood Cliffs, N.J. Tsai, Y.B., and Huang, M.W. 2000. Strong ground motion charac-
Lee, C.T., Kang, K.H., Cheng, C.T., and Liao, C.W. 2000. Surface teristics of the Chi-Chi, Taiwan, earthquake of September 21,
rupture and ground deformation associated with the Chi-Chi, 1999. Earthquake Engineering and Engineering Seismology,
Taiwan, earthquake. Sino-Geotechnics, 81: 5–16. [In Chinese.] 2(1): 1–21.
Meyerhof, G.G. 1963. Some recent research on the bearing capac- Ueng, T.S., Lin, M.L., and Chen, M.H. 2001. Some geotechnical
ity of foundations. Canadian Geotechnical Journal, 1: 16–26. aspects of 1999 Chi-Chi, Taiwan, earthquake: special lecture. In
Mononobe, N. 1924. Consideration into earthquake vibrations and Proceedings of the 4th International Conference on Recent
vibration theories. Journal of the Japanese Society of Civil Engi- Advances in Geotechnical Earthquake Engineering and Soil
neers, 10(5): 1063–1094. Dynamics, San Diego, Calif., 26–31 Mar. 2001. University of
Okabe, S. 1924. General theory of earth pressure and seismic sta- Missouri-Rolla, Rolla, Mo. Paper SPL-10.1.
bility of retaining walls and dam. Journal of the Japanese Soci- U.S. Department of the Navy. 1982. Foundations and earth struc-
ety of Civil Engineers, 10(5): 1277–1323. tures: design manual NAVFAC DM-7.2. U.S. Department of the
Richards, R.J., and Elms, D.G. 1979. Seismic behavior of gravity Navy, Naval Facilities Engineering Command, Washington,
retaining walls. Journal of the Geotechnical Engineering Divi- D.C.
sion, ASCE, 105(4): 449–464. Whitman, R.V. 1990. Seismic design behavior of gravity retaining
Seed, H.B., and Chan, C.K. 1959. Structures and characteristics of walls. In Proceedings of the ASCE Specialty Conference on De-
compacted clay. Journal of the Soil Mechanics and Foundations sign and Performance of Earth Retaining Structures. American
For personal use only.

Division, ASCE, 85(5): 87–128. Society of Civil Engineers, Geotechnical Special Publication 25,
Seed, H.B., and Whitman, R.V. 1970. Design of earth retaining pp. 817–842.
structures for dynamic loads. State-of-the-art paper. In Proceed- Whitman, R.V., and Liao, S. 1985. Seismic design of retaining
ings of the 1970 ASCE Specialty Conference on Lateral Stresses walls. U.S. Waterways Experiment Station, Vicksburg, Miss.
in the Ground and Design of Earth Retaining Structures, Ithaca, Miscellaneous Paper GL-85-1.
N.Y. American Society of Civil Engineers, New York. pp. 103–
147.

© 2003 NRC Canada

 This article has been cited by:

1. Riccardo Conti, Giulia M.B. Viggiani, Simone Cavallo. 2013. A two-rigid block model for sliding gravity retaining walls. Soil
Dynamics and Earthquake Engineering 55, 33-43. [CrossRef]
2. Giovanni Biondi, Ernesto Cascone, Michele Maugeri. 2013. Displacement versus pseudo-static evaluation of the seismic
performance of sliding retaining walls. Bulletin of Earthquake Engineering . [CrossRef]
3. Mahmoud Yazdani, Ali Azad, Abol hasan Farshi, Siamak Talatahari. 2013. Extended “Mononobe-Okabe” Method for Seismic
Design of Retaining Walls. Journal of Applied Mathematics 2013, 1-10. [CrossRef]
4. S. Caltabiano, E. Cascone, M. Maugeri. 2012. Static and seismic limit equilibrium analysis of sliding retaining walls under different
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by National Chiao Tung University on 04/27/14

surcharge conditions. Soil Dynamics and Earthquake Engineering 37, 38-55. [CrossRef]
5. Chulmin Jung, Antonio Bobet, Gabriel Fernández. 2010. Analytical solution for the response of a flexible retaining structure with
an elastic backfill. International Journal for Numerical and Analytical Methods in Geomechanics 34:13, 1387-1408. [CrossRef]
6. Mandar M. Dewoolkar, A. H. C. Chan, Hon-Yim Ko, Ronald Y. S. Pak. 2009. Finite element simulations of seismic effects
on retaining walls with liquefiable backfills. International Journal for Numerical and Analytical Methods in Geomechanics 33:6,
791-816. [CrossRef]
7. CHING-CHUAN HUANG. 2009. EFFECTS OF HORIZONTAL AND VERTICAL SEISMIC LOADS ON THE
BEARING CAPACITY OF A SURFACE FOOTING ADJACENT TO A SLOPE. SOILS AND FOUNDATIONS 49:2,
249-258. [CrossRef]
For personal use only.