Engineering Geology 166 (2013) 237244

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Sliding displacements due to subduction-zone earthquakes

Alfredo Urza a, John T. Christian b,
a
b

Prototype Engineering Inc., 57 Westland Avenue, Winchester, MA 01890, USA

Prototype Engineering, Inc., 36E Seven Springs Lane, Burlington, MA 01803, USA

a r t i c l e

i n f o

Article history:
Received 7 May 2013
Received in revised form 9 August 2013
Accepted 11 August 2013
Available online 30 August 2013
Keywords:
Acceleration
Arias intensity
Displacements
Earthquakes
Slope stability
Subduction zone

a b s t r a c t
Empirical studies of earthquake ground motions have developed relations between sliding displacement and
acceleration ratio and other parameters such as Arias intensity. Computations using strong motion records
from the Maule 2010 Chile M = 8.8 earthquake indicate that the published relations do not conform well to
the computed displacements, and some tend to be unconservative. Extensions to the empirical equations
incorporating Arias intensity are, if anything, less accurate. These results suggest that these empirical relations
may not apply to subduction zone events and indicate that further study using records from other recent
subduction zone events is appropriate. Examining the analytical solutions for sliding displacements induced by sinusoidal shaking and the denition of Arias intensity leads to an improved normalization for sliding displacements.
When this improved normalization is applied to the records from three different earthquakes in different parts of
Chile, the results are nearly identical, and the results for the Chi Chi and Northridge earthquakes are very close to
those from the Chilean events. Suggestions for practical use of the new normalization relations are provided.
2013 Elsevier B.V. All rights reserved.

1. Introduction
Estimating the displacements that occur when a potentially unstable
mass of soil or rock is shaken by earthquake stress waves can be a
difcult and daunting task. The amplitude, frequency content, and
duration of the incoming signal are often not well known, and even
probabilistic descriptions involve large uncertainties. The properties of
the soils and rocks are also known imperfectly. Furthermore, the entire
process of deformation and amplication is strongly non-linear. In such
a situation it is not surprising that engineers and seismologists have
resorted to simplied analytical methods, of which one of the most
widely used is the simplied sliding-block model (Newmark, 1965).
The sliding-block analysis assumes that the sliding mass rests on a
plane and that the shear strength of the interface between block and
plane is known. The strength can be frictional or non-frictional, and it
can be allowed to decrease as a function of the sliding displacement,
although most applications assume the strength is constant. The input
earthquake motion is an acceleration timehistory that describes the
motion of the plane. If at any time the block is at rest with respect to
the plane (that is, it is moving along with the plane) and the acceleration of the plane is less than the acceleration that can be transmitted
by the interface, the block continues to move along with the plane.
Once the acceleration of the plane exceeds a critical value, the block
begins to slide along the plane. It continues to slide until (a) the acceleration of the plane falls below the critical value ac and (b) frictional forces
Corresponding author. Tel.: +1 781 272 3196.
E-mail addresses: alfredo.urzua@bc.edu, prototypeengineering@comcast.net
(A. Urza), jtchrist36@comcast.net (J.T. Christian).
0013-7952/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.enggeo.2013.08.005

between block and plane have slowed the relative sliding so that the
block and plane have the same absolute velocity. In effect, the motion
consists of a series of cumulative sliding pulses.
Although the sliding-block method is easily implemented on a
computer, the calculations can become tedious when the analysis is to
be used for zoning purposes. Jibson (2007), using a suite of accelerograms
from several earthquakes and referring to a series of computational studies described in his earlier papers and reports, proposed empirical relationships between the computed displacement, DR, and the ratio (ac/
amax) between the critical acceleration, ac, and the maximum acceleration
in the timehistory, amax. One possible shortcoming of these relations is
that they may not include information on many seismological parameters, such as magnitude, local intensity, duration, and dominant period.
Jibson proposed several modications to his basic empirical relation, in
particular incorporating the Arias (1970, 1973, 1993) intensity, dened as
Ia

2g

t
0

a d

in which g is the acceleration of gravity, a is the ground acceleration, and

is the variable of integration equal to time. It is customary to integrate
over the entire duration of the strong motion record (t = length of record), but t can be any time after the start of the record so that Ia is actually
a function of time. If there is no other indication, Ia is taken as the value
integrated over the entire record. The Arias intensity has units of velocity
and is usually expressed in meters per second (m/s) or centimeters per
second (cm/s). Jibson also proposed a relation incorporating magnitude.
It should be noted that Jibson's displacements, and the displacements
discussed further on in this paper, are not observed values but are the

238

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

values computed by applying the sliding-block model to recorded

accelerograms.
Other researchers (Ambraseys and Menu, 1988; Crespellani et al.,
1998; Bray and Travasarou, 2007; Rathje and Saygili, 2008; Saygili and
Rathje, 2008; Rathje and Saygili, 2009; Hsieh and Lee, 2011; Saragoni,
2012) have also developed or used empirical relationships incorporating the acceleration ratio, Arias intensity, and other parameters such
as the destructiveness potential factor PD (Araya and Saragoni, 1984).
These authors went to some lengths to eliminate bias in the input to
the regression models, but, inevitably, the number of records made
during large subduction zone earthquakes is limited by the rarity of
such events. The recordings from three earthquakes generated off the
coast of Chile make it possible to compare the sliding displacements
computed by applying the sliding-block method to actual recordings
from subduction-zone events with the earlier empirical predictions.
This paper reports on those comparisons and proposes improvements.
2. Displacements computed from the Maule earthquake
The rst set of records used in these analyses were made in the
Maule region of Chile during the strong motion of the 2010 (M = 8.8)
earthquake (the www.renadic.cl website contains further details). The
analysis used 18 records. Each recording included both horizontal
components, and each recording was run through the analysis with
the original and reversed sense of motion, so there were 72 timehistories in total. The amplitude, frequency, and other parameters of
the accelerograms are obviously inuenced by propagation and by
local site conditions, but no further effects of local conditions were
considered. The analysis for each record proceeded as follows:
1. A value of ac/amax was selected.
2. One of the 72 time-histories was chosen.
3. The sliding-block analysis was performed for that combination of
time-history and ac/amax.
4. Steps 1 through 3 were repeated for the other values of ac/amax and
for all time-histories.
Fig. 1 displays the results. For each value of ac/amax the computed
displacements are plotted as asterisks. The dashed lines are drawn
through the values of the means and the means plus one standard

deviation at each value of ac/amax. The solid lines are drawn through
the anti-logarithms of the means and means plus one standard
deviation of the logarithms of the results. Thus, the lower solid line is
the plot of the median displacements. The lines representing the
means plus one standard deviation of both the displacements and the
logarithms of the displacements are so close that they plot on top of
each other.
3. Existing empirical relationships
The empirical relationships used in this study are, in chronological
order, as follows:
3.1 Ambraseys and Menu (1988) proposed a simple relation using
only the acceleration ratio:

2:53 


a
ac 1:09
log DR 0:90 log 1 c
 0:30
a max
a max

Logarithms are to base 10.

3.2 Crespellani et al. (1998), in the notation of this paper, proposed
0:977 1:338

DR 0:011P D kc
log DR 0:977P D 1:388kc 1:9586

PD is the destructiveness potential factor dened by Araya and

Saragoni (1984) and Saragoni and Hart (1974) as
2

P D I a =0

where 0 is the zero crossing intensity per second. The zero

crossing intensity can be calculated by counting the crossings
that occur between the times that the Arias intensity reaches
5% of its nal value and 95% of its nal value. This procedure eliminates the values from the beginning and the tail of the record to
concentrate attention on the strongest portion of the motion.
Since the zero crossing intensity has units of the inverse of
time, PD has units of meterseconds or centimeterseconds.
The parameter kc is the critical acceleration expressed in units
of g's (i.e., kc = ac/g).
3.3 Bray and Travasarou (2007) proposed relations that incorporated the exibility of the sliding mass, but they also proposed the
following relation for the case of a rigid sliding mass
2

ln DR 0:222:83 ln ac 0:333 ln ac

0:566 ln ac ln a max 3:04 ln a max

2
0:244 ln a max 0:278M7 

The notation has been changed to agree with that of the present
paper. M is the moment magnitude of the earthquake, and is a
normally distributed error term with zero mean. In Eq. (5) and
subsequent equations ln () is the natural logarithm
3.4 Based on the work of Ambraseys and Menu (1988), Jibson (2007)
proposed this relation using only the acceleration ratio

2:341 


a
ac 1:438
logDR 0:215 log 1 c
 0:5106
a max
a max

Fig. 1. Results from sliding block analysis of Maule strong motions recordings together
with observed and proposed empirical relations. Asterisks are the computed results.
Dashed lines are the mean and mean plus of standard deviation of the results, and solid
lines are the mean and mean plus one standard deviation of the logarithms of the results.

The last term is the standard deviation of the regression. Incorporating Arias intensity in his regression, Jibson arrived at
log DR 0:561 log Ia 3:833 logac =a max 1:474  0:616 7

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

239

3.5 Saygili and Rathje (2008) and Rathje and Saygili (2008) proposed
six relations. The rst and simplest is






ac
ac 2
ac 3
20:39
42:61
ln DR 5:524:43
a max
a max
a max


ac 4
28:74
0:72 ln a max ln D
a max
8
In this equation is a standard normal variate with zero mean and
unit standard deviation, and lnD is the standard deviation of the
natural logarithm of the displacement. In the subsequent parts
of this paper, this is identied as Saygili and Rathje (2008) #1.
Their sixth relation incorporates both the period and the Arias
intensity:






ac
ac 2
ac 3
ln DR 4:274:62
46:53
21:49
a max
a max
a max


9
ac 4
31:66
0:57 ln a max 1:14 ln T
a max
0:86 ln Ia ln D
It is identied as Saygili and Rathje (2008) #6.
3.6 Rathje and Saygili (2009) proposed a relation incorporating the
magnitude of the earthquake






ac
ac 2
ac 3
42:49
19:64
ln DR 4:894:85
a max
a max
a max


ac 4
29:06
0:72 ln a max 0:89M6
a max
10
3.7 By incorporating data from the 1999 Chi Chi Taiwan earthquake
(M = 7.3) Hsieh and Lee (2011) proposed relations incorporating
Arias intensity for soil and rock sites. Their equation for rock sites is
log DR 0:788 log Ia 10:166 ac 5:95ac log I a 1:779
 0:294

11

4. Results from empirical equations

Fig. 2 compares the predictions of the various empirical equations
with the median curves in Fig. 1 for the displacements calculated from
the observed data for the Maule set of records from the 2010 earthquake. The data points for the individual time histories are not plotted
to avoid cluttering the gure. Ambraseys and Menu (1988) and Jibson
(2007) equations use the acceleration ratios alone (Eqs. 2 and 6), and
the corresponding curves were developed by inserting the appropriate
acceleration ratios into the empirical equations. The error terms at the
end of some of the equations were not included in the calculations.
Crespellani et al. (1998), Jibson (2007), Bray and Travasarou, 2007;
Rathje and Saygili, 2008; Saygili and Rathje, 2008; Rathje and Saygili,
2009, and Hsieh and Lee (2011) proposed equations incorporating
other parameters, including Arias (1970, 1973, 1993) intensity. To
investigate the applicability of this type of formulation, the analysis
proceeded as follows:
1. For each record the Arias intensity and the other seismic parameters
were computed. Because the acceleration is squared for Ia, the sense
of the motion is irrelevant.
2. For each record and for each of the proposed equations that incorporate parameters in addition to ac/amax and for each value of ac/amax,
the parameters were inserted in the proposed equation to obtain a
predicted value of DR. This gave, for each record, for each equation,
and for each value of ac/amax, a set of predictions of DR.

Fig. 2. Results from empirical relations compared to anti-logarithm of mean of the logarithms of the computed results from Maule 2010. The data points from Fig. 1 have been
removed for clarity.

3. For each equation and at each value of ac/amax, the mean of the
logarithms of the computed values of DR among the records were
computed.
4. Curves were drawn through the points computed in step 4. These are
the curves plotted in Fig. 2.
Major observations are that Jibson's revised equation incorporating
the Arias intensity provides a poor match to the computed results for
this event and that the Ambraseys and Menu relation ts the data
best, even though it is the oldest relation and is based on the least
data. Of the other proposals, the relation proposed by Bray and
Travasarou (2007) ts the computed results well at low values of the
acceleration ratio but deviates from the computed results at higher acceleration ratios, and that proposed by Saygili and Rathje (2008) ts
best at higher values of acceleration ratio but more poorly at low values.
Rathje and Saygili's (2009) relation recovers the slope of the computed
line well but is located substantially higher in the plot.
On November 16, 2007, an M = 7.7 earthquake occurred along the
northern portion of the offshore Chilean subduction zone, and 16
recordings of strong motion were made in the Tocopilla region of northern Chile. Similarly, 22 of the recordings of the March 3, 1985, M = 8.0
offshore earthquake, which has been named after Valparaiso. These recordings were processed in the same way as the Maule recordings. Fig. 3
shows the results for the Tocopilla data, and Fig. 4 does the same for the
Valparaiso data. The histograms for seismic parameters for the three
Chilean earthquakes are in Appendix A.
Figs. 2, 3, and 4 show that there is a great deal of scatter in the predictions of the empirical equations. Relations that agree well with the
data for one event, such as the AmbraseysMenu equation in Fig. 2, do
poorly for other sets of recordings. None of the empirical relationships
is consistently good or poor.
It should be noted that all these relations have dimensional irregularities. Those expressed in terms of the dimensionless acceleration
ratio alone do not incorporate parameters describing the intensity or
duration of the strong motion. Those that do incorporate other parameters do so in ways that do not preserve dimensions such as velocity,
displacement, or time. It would be desirable to have a truly rational dimensionless expression that incorporates more strong motion parameters than the acceleration ratio.

240

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

Fig. 3. Results from empirical relations compared to anti-logarithm of mean of the logarithms of the computed results from Tocopilla 2007. The data points from Fig. 1 have
been removed for clarity.

5. Analytical derivation for idealized seismic displacements

Yegian et al. (1991) developed analytical solutions for the sliding displacements induced by single pulses of rectangular, triangular, and sinusoidal shape. If the displacement during a single pulse is DN, the total
displacement during N such pulses is DR = N DN. Yegian et al. found that
h
i
2
12
DR f ac =a max  a max NT
where f(ac/amax) is a function of the acceleration ratio and T is the period of
the pulse. The function f(ac/amax) depends on the shape of the acceleration
pulse. Fig. 5 is a plot of the function for a sinusoidal pulse. For the limiting
case that ac = 0, the value of the function is 1/2. At other values of ac/amax
the function must be calculated by iterative solution of the sliding-block

Fig. 5. Sliding displacement during single sinusoidal pulse (after Yegian et al., 1991).

equations. In most practical cases of actual sliding the value of the acceleration ratio will be between 0.05 and 0.5, a range where Fig. 5 shows the
logarithm of f(ac/amax) is nearly linear.
It should be noted that the solution for sliding-block displacement
for any prescribed shape of input accelerogram must have the functional form of Eq. (12). The specic composition of the function f(ac/amax)
depends on the details of the shape of the input motion. As mentioned
above, Yegian et al. (1991) gave solutions for the cases of sequences of
sinusoidal, rectangular, and triangular wave trains. Solutions could be
developed for a series of Dirac delta functions or any other shape of
accelerogram. The details of the function f(ac/amax) may be quite
complicated, but all solutions will be of the form of Eq. (12). The fact
leads to a suggestion for a rational way to improve the dimensionless
relations between sliding displacements and acceleration ratios. The
sinusoidal input is the obvious starting point.
For a sinusoidal acceleration pulse of period T and amplitude amax,
the Arias intensity is
Ia

2g

t
0

jaj d

2g

t
0

a max sin 2=T d

13

Integrating over a single period gives the Arias intensity for a single
pulse:
Ia

2
a
T
4g max

14

and for N pulses, with the ground acceleration expressed as Ka in

units of g (i.e., amax = Kag)
Ia

2
NgTK a
4

15

Substituting Eq. (15) into Eq. (12) and performing some algebraic
manipulation yields
 
I T
16
DR f ac =a max a
Ka

Fig. 4. Results from empirical relations compared to anti-logarithm of mean of the logarithms of the computed results from Valparaiso 1985. The data points from Fig. 1 have
been removed for clarity.

In other words, a rational way to normalize the computed or estimated sliding displacements in a dimensionless plot is to divide the
computed values by the bracketed term in Eq. (16), which has units of
displacement.

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

241

6. Normalized relation applied to three sets of recordings

There are three populations of computed sliding displacements for a
range of values of acceleration records and for three different subduction zone earthquakes recorded in different parts of Chile (Maule in
the central south, Tocopilla in the north, and Valparaiso in the center).
Fig. 6 displays the mean values of the displacements normalized
according to Eq. (16) and the means plus and minus one standard
deviation. The results for the three events are very close, conrming
that the normalization of Eq. (16) is rational and reproducible.
Although the slightly curved lines in Fig. 6 could be employed to
estimate displacements during other earthquakes, straight lines would
be easier to use. Therefore, regressions were run to obtain linear relations between the logarithms of the normalized computed displacements and the acceleration ratios. It should be noted that these
regressions were run only to nd the best linear approximations to
the slightly curved lines in Fig. 6 and do not represent further statistical
analysis of the data. Fig. 7 shows the resulting set of straight lines for the
three sets of recordings. The equations for the normalized displacement
are all of the form:
log DN 0 1 ac =a max

17

Table 1 contains the values of the parameters for each case, as well as
the parameters for all three sets of records taken together. The results
for the three events fall so close to each other that, for engineering
purposes, Eq. (17) could be simplied to
log DN 0:14:3ac =a max

18

This procedure gives results for the median lines that are identical to
what would be obtained by regressing on all the data, but the intermediate step reveals how close the normalized median relations are to a
straight line. As Fig. 6 demonstrates, the relations are already very close
to linear, so it is not surprising that the values of R2 for the regressions
to obtain the straight lines of Fig. 7 are all either 0.997 or 0.998. When

Fig. 7. Linear regression lines for the results in Fig. 6.

linear regression is performed on all the results underlying Fig. 6, the

values of R2 are: for Maule 0.898, for Tocopilla 0.888, for Valparaiso
0.896, for all three Chilean earthquakes 0.892, and for the Chilean earthquakes plus Chi Chi and Northridge (discussed further below) 0.861.
An interesting result from these analyses is that, for any single suite
of accelerograms, the slope of the linear regressions through the mean
values of logarithms of normalized displacements are independent of
the normalizing factor. Any differences due to the choice of normalizing
factor are reected in the intercepts. These statements are not true for
the regressions through the means plus and minus one standard
deviation. Appendix B demonstrates why this is true.
7. Application to other earthquake records
This work originally dealt with sliding displacements induced by
subduction zone earthquakes with specic reference to events in
Chile. However, the relations developed ought to be applicable to earthquakes in other regions with different seismic generating mechanisms.
A full investigation of this topic is beyond the scope of the present effort,
but as a preliminary step, the authors applied the methodology to
records from the January 21, 1999, Chi Chi earthquake (M = 7.6) and
the January 17, 1994, Northridge earthquake (M = 6.7). The Chi Chi
event was a shallow thrust earthquake, and the Northridge event occurred along a blind thrust fault. In order to concentrate on the motions
that might cause sliding distress, approximately 50 records with the
largest peak ground accelerations were used for each earthquake.
Table 1
Values of parameters for Eq. (16).

Fig. 6. Normalized displacements computed from accelerograms recorded during the

three earthquakes in Chile for a range of acceleration ratios. For each acceleration ratio
and each earthquake the results are the means and standard deviations of the logarithms
of the computed displacements. The solid lines are for the Maule earthquake, the dashed
lines for the Tocopilla earthquake, and the short dashed lines for the Valparaiso records.
In each case the top line is for the mean plus one standard deviation, the middle line for
the mean, and the bottom line for the mean minus one standard deviation.

Case

Maulemean of log displacement

Maulemean plus s. d. of log displacement
Maulemean minus s. d. of log displacement
Tocopillamean of log displacement
Tocopillamean plus s. d. of log displacement
Tocopillamean minus s. d. of log displacement
Valparaisomean of log displacement
Valparaisomean plus s. d. of log displacement
Valparaisomean minus s. d. of log displacement
All threemean of log displacement
All threemean plus s. d. of log displacement
All threemean minus s. d. of log displacement

0.124509
+0.040330
0.289348
0.050656
+0.077725
0.179036
0.120225
+0.055346
0.295797
0.101725
+0.059039
0.262429

4.26177
4.12274
4.40080
4.32436
4.01717
4.63156
4.29918
4.18101
4.41734
4.29435
4.11832
4.47038

242

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

Fig. 8 is a plot of the dimensionless mean displacements computed

according to the proposed procedure for the records from the ve earthquakes (Maule, Tocopilla, Valparaiso, Chi Chi, and Northridge). The results
are similar despite the substantial differences in faulting, propagation geometry, and local conditions, which, for example, would call for different
attenuations or ground motion relations in a probabilistic seismic hazard
analysis. Therefore, it is not surprising that there are some differences
among the results for different seismic conditions, even though the differences are small. Further investigation of the applicability of the proposed
methodology to more earthquakes and a larger database is warranted.
Finally, Fig. 8 gives the impression that the results for the ve earthquakes deviate somewhat for higher values of the acceleration ratio.
This is an artifact of the logarithmic plot. Fig. 9 shows that, in a plot of
the normalized displacements themselves instead of the logarithms,
the absolute differences are smaller for large acceleration ratios and
larger for small acceleration ratios.
8. Use of results
If Eq. (16) is to be used in practice, engineers must have some way of
estimating its parameters. Usually the engineer will have an estimate of
the critical acceleration ac for a particular project. If the acceleration
ratio is to be used, there must be an estimate of amax, and this is true
for any of the methods discussed in this paper. One would expect that
values of amax would be part of the design criteria for a site. Values of
Ia can be computed from a suite of strong motion records from the
area or can be developed as an additional part of a PSHA.
One could select T by any of several reasonable and consistent
procedures and obtain reliable results provided that the users of the dimensionless plots and corresponding normalizing factor used that same
denition of T. First, design criteria often include design response spectra, and T can be estimated from the shape of the response spectra in the
absence of other data. That was the procedure adopted for this paper,
and users of the equations in this paper should be aware of that fact.
In almost all cases the response spectra had clear peaks, making it
easy to identify the corresponding period. This procedure has the
advantage that it does not require extraordinary manipulation of earthquake records. Second, an estimate can be obtained from the Fourier
spectra of typical records made in the area. Third, T could be estimated
as the inverse of the intensity of zero crossings in the strongest part of

Fig. 8. Results for Maule, Tocopilla, Valparaiso, Chi Chi, and Northridge records plotted as
in Fig. 7.

Fig. 9. Results for ve earthquakes in Fig. 8 plotted to arithmetic scale.

the accelerogram. Experimentation with the different ways of

estimating T indicates that the results from the inverse of the zero crossing intensity show slightly less scatter than those from the other
techniques.
Fourth, Rathje et al. (2004) identied four values of T that could be
used to characterize earthquake strong shaking: Tmthe weighted average of the periods in the discrete Fourier transform of the accelerogram
using the squares of the Fourier coefcients as weights, Tavgthe
weighted average of the periods in the 5% damped response spectrum
using the squares of the spectral accelerations as weights, T0the
weighted average of the periods in the 5% damped response spectrum
using the logarithms of the spectral accelerations as weights, and Tp
the period corresponding to the peak of the 5% damped response spectrum. They noted, T0 is most sensitive to the high frequency (low period) content of strong ground motions[and] Tm and Tavg best account
for the long period content of strong motions
Performing Fourier analyses of accelerograms or evaluating the frequency of zero crossings in the strongest portion of the ground motion
will often be beyond the scope that the design engineers are prepared to
undertake. In most practical cases visual examination of the response
spectra will provide a satisfactory approximate estimate of the range
of T, and a more precise estimate can be obtained by evaluating Tavg
from Rathje et al. (2004). The empirical results presented here are
based on the peak of the response spectrum.
As demonstrated in Appendix B, the slope of the regression line for
the mean of the logarithms of dimensionless displacement is a function
of (ac/amax) and is independent of the values of the normalizing parameters. This is not true for the other regression lines, such as the means
plus or minus one standard deviation, but it is true for the mean of the
dimensionless displacement. Since the value of T is in the normalizing
factor, this result implies that the choice of T affects the intercept or
vertical placement of the regression line but not its slope.
Once the design values of amax, Ia, and T have been established, the
design engineer can establish the critical value of ac at which the slope
will start to fail. In most cases this will correspond to a factor of safety
of 1.00. The engineer then computes the ratio ac/amax and enters Fig. 7
or Table 1 to nd the estimated normalized displacement. This is then
multiplied by the bracketed term in Eq. (16) to obtain the estimated
total sliding displacement. This procedure assumes that there is no
loss of strength with shearing and that amplication effects need to be
incorporated in the analysis.

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

An alternative case arises when it is desired to limit the sliding displacements to a value such as 100 cm. Then the designer divides the
limiting displacement by the bracketed term to obtain a dimensionless
displacement and enters Fig. 7 from the left to obtain the limiting
acceptable value of the acceleration ratio.
9. Conclusions
Applying the sliding-block analysis to strong motion timehistories
recorded in the Maule area during the 2010 earthquake, in the Tocopilla
area during the 2007 earthquake, and in Valparaiso during the 1985
earthquake indicates that the previously proposed empirical relations
between sliding displacement and acceleration ratio do not agree well
with computed results. The agreement is not signicantly improved
when the Arias intensity is incorporated in the empirical equations to
account for duration effects. It appears that the empirical relations do

Appendix A. Histograms for Chilean earthquakes

243

not account for the particular effects observed in subduction zone

earthquakes.
A new relation, Eq. (16), based on combining the expressions for
Arias intensity and displacements in a single pulse has been shown to
provide reproducible normalization of the computed sliding displacements. Furthermore, it is a rational extension of the analytical solution
for sinusoidal pulses. When applied to recordings in different parts of
Chile from three different earthquakes, the new relation gives very
similar results. Limited application to records from the Chi Chi and
Northridge events also give similar results.

Acknowledgments
Alejandro Contreras and Patricio Pineda helped in obtaining the records, and Dr. Rodolfo Saragoni made his papers available.

244

A. Urza, J.T. Christian / Engineering Geology 166 (2013) 237244

The bars over variables indicate the means. All the x terms are independent of the normalization factor. Each of the y terms is as follows:

Appendix B. Regression results

The dependent variables in the analyses presented in this paper
are the logarithms of the normalized displacement, that is each
value is log10(D A), where D is a displacement and A is a normalizing factor. In the normalizations used in this paper each value of
D depends on the parameters of the earthquake and on the ratio
ac/amax, but A is functionally independent of ac/amax. The following
notation is adopted:

n
n
1X
1X
Di
A
n i1
n i1
n
1X
Di A
Di A
n i1
n
1X
Di
Di
n i1

yi y Di A

B:4

1. There are n values of ac/amax, and each is identied by a subscript i.

Thus, i = 1,, n.
2. There are m accelerograms, each identied by the subscript j. Thus,
j = 1,, m.

Thus, each of the y terms in parentheses is also independent of the

acceleration ratio, so the slope of the regression line is independent of
the normalization factor.
The intercept of the regression line is 0:

Note that the normalization factor A proposed in the paper is (Ka/IaT)

in which Ka = amax/g, but it could be anything else so long as it does not
include dependence on the acceleration ratio ac/amax.
For each value of i and j, the data point log10(D A)ij = log10(Dij) +
log10(Aj) because A is independent of the acceleration ratio and thus
does not require a subscript i. For each of the i values of acceleration
ratio the mean of the logarithms of the normalized displacements,
which we will call yi for conciseness, is

0 y1 x

yi

m
m
m

 1X
  1X
 
1X
log10 Dij  A j
log10 Dij
log10 A j
m j1
m j1
m j1

B:1

To simplify further the notation on the right hand side of this equation, call the rst term Di and the second A, and call the acceleration
ratio xi, or

m
 
1X
log10 Dij
m j1
m
 
1X
log10 A j
A
m j1

Di

B:2

yi Di A
xi ac =a max i

We then t a straight line through the i points dened by xi and

yi. From linear regression theory the slope of the least-square error
t is
n
X
yi yxi x

i1

n
X
i1

B:3
2

xi x

B:5

In this expression y is clearly a function of the normalization factor,

so the intercept is not independent of the normalization factor.
If we now consider not the means of the normalized displacements
expressed by Eq. (B.1) but their variances or standard deviations, the
terms include products of the logarithms of Dij and Aj. Therefore, the
terms corresponding to yi cannot be separated as they were for the
means. Thus, we expect that the results do depend on the normalization
factors.
References
Ambraseys, N.M., Menu, J.M., 1988. Earthquake-induced ground displacements. Earthq.
Eng. Struct. Dyn. 16, 9851006.
Araya, R., Saragoni, G.R., 1984. Earthquake accelerogram destructiveness potential factor.
Eighth World Conference on Earthquake Engineering, San Francisco, CA, pp. 835842.
Arias, A., 1970. A measure of earthquake intensity, Seismic Design for Nuclear Power
Plants. In: Hansen, R.J. (Ed.), Massachusetts Institute of Technology Press, Cambridge,
MA, pp. 438483.
Arias, A., 1973. Earthquake intensity and smoothed earthquake spectra. Fifth World
Conference on Earthquake Engineering, Rome, paper 208.
Arias, A., 1993. La intensidad sismica como magnitud tensorial, discurso de incorporacin
como miembro de nmero de la Academia de Ciencias del Instituto de Chile.
Bray, J.D., Travasarou, T., 2007. Simplied procedure for estimating earthquake-induced
deviatoric slope displacements. J. Geotech. Geoenviron. Eng. ASCE 133 (4), 381392.
Crespellani, T., Madiai, C., Vannucchi, G., 1998. Earthquake destructiveness potential
factor and slope stability. Geotechnique 49, 411419.
Hsieh, S.-Y., Lee, C.-T., 2011. Empirical estimation of the Newmark displacement from
Arias intensity and critical acceleration. Eng. Geol. 122, 3442.
Jibson, R.W., 2007. Regression models for estimating coseismic landslide displacement.
Eng. Geol. 91, 209218.
Newmark, Nathan M., 1965. Effects of earthquakes on dams and embankments.
Geotechnique 15, 139160.
Rathje, E.M., Saygili, G., 2008. Probabilistic seismic hazard analysis for the sliding displacement of slopes: scalar and vector approaches. J. Geotech. Geoenviron. Eng. ASCE 134
(6), 804814.
Rathje, E.M., Saygili, G., 2009. Probabilistic assessment of earthquake-induced sliding
displacements of natural slopes. Bull. N. Z. Soc. Earthq. Eng. 42 (1), 1827.
Rathje, E.M., Faraj, F., Russell, S., Bray, J.D., 2004. Empirical relationships for frequency
content parameters of earthquake ground motions. Earthquake Spectra 20, 119144.
Saragoni, G.R., 2012. Earthquake Performance design of dams using destructiveness
potential factor. Second International Conference of Performance-Based Design in
Earthquake Geotechnical Engineering, Taormina, Italy.
Saragoni, G.R., Hart, G.C., 1974. Simulation of articial earthquakes. Earthq. Eng. Struct.
Dyn. 2, 249267.
Saygili, G., Rathje, E.M., 2008. Empirical predictive models for earthquake-induced sliding
displacements of slopes. J. Geotech. Geoenviron. Eng. ASCE 134 (6), 790803.
Yegian, M.K., Marciano, E.A., Ghahraman, V.G., 1991. Earthquake-induced permanent
deformations: probabilistic approach. J. Geotech. Eng. ASCE 117, 3550.

Engineering Geology 183 (2014) 241246

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Rational selection of critical acceleration factors for sliding stability

Alfredo Urza a, John T. Christian b,, Rodrigo Silva c, Antonio Bonani d
a

Prototype Engineering, Inc. 57 Westland Avenue, Winchester, MA 01890, USA

Prototype Engineering, Inc., 36E Seven Springs Lane, Burlington, MA 01803, USA
c
Compaia Minera Doa Ins de Collahuasi, Avda. Baquedano 902, Iquique, Chile
d
Gerente recursos mineros y desarollo, Dvision Salvador, CODELCO, Chile
b

a r t i c l e

i n f o

Article history:
Received 28 April 2014
Received in revised form 9 August 2014
Accepted 12 August 2014
Available online 20 August 2014
Keywords:
Acceleration
Arias intensity
Displacements
Earthquakes
Slope stability
Subduction zone

a b s t r a c t
The horizontal seismic loading coefcient is an essential input in evaluating the seismic adequacy of slopes, such
as those in open-pit mines and natural slopes. In some cases, the coefcient is established through dynamic nite
element analyses, which are time-consuming and require a new analysis for each facility, including a new suite of
accelerograms. The values of the coefcient are sometimes incorporated in design manuals, but the procedures
for establishing the values are seldom transparent. The usual situation is that the values arise from consensus, experience, and previous practice. In this paper, the UrzaChristian model for normalized sliding displacement has
been extended to develop the critical acceleration value corresponding to the probability of observing prescribed
amounts of sliding displacement. The method has been applied to two sets of data based on probabilistic seismic
hazard analyses. The results show that, to satisfy the criterion that there must be 0.1 probability of the sliding displacement exceeding 100 cm if a maximum credible earthquake (MCE) occurs, the critical acceleration must be
approximately 0.13 g. This means that a slope with these parameters in this environment must be stable enough
that a horizontal acceleration of 0.13 g is necessary to put it in a state of sliding motion. In the case of the operational basis earthquake (OBE), which is a much smaller ground motion, the criterion of 0.1 probability of
100 cm of sliding is achieved for a slope with a critical acceleration of 0.35 g.
2014 Elsevier B.V. All rights reserved.

1. Introduction
The horizontal seismic loading coefcient is an essential input in
evaluating the seismic adequacy of slopes, such as those in open-pit
mines and natural slopes. The coefcient is usually expressed as a fraction of the acceleration of gravity and is multiplied by the weight of a
potential sliding mass to give a horizontal force on the sliding mass.
When the force is applied as a static load, the result is called a pseudostatic analysis because it does not include actual dynamic behavior but
replaces it with an equivalent static problem. The static horizontal
load factor is usually designated Ka.
In some cases, the coefcient is established through dynamic nite
element analyses, but these procedures are time-consuming and require a new analysis for each facility, including a new suite of ground
motion records. Values of the coefcient are sometimes incorporated
in design manuals, but the procedures for establishing these values
are seldom transparent. The usual situation is that the Ka values arise
from consensus, experience, and previous practice. This paper proposes
a rational way to establish the seismic coefcient.

Corresponding author. Tel.: +1 781 272 3196.

E-mail addresses: alfredo.urzua@bc.edu (A. Urza), jtchrist36@comcast.net
(J.T. Christian), rasilva@collahuasi.cl (R. Silva), agbonani@CODELCO.cl (A. Bonani).

http://dx.doi.org/10.1016/j.enggeo.2014.08.011
0013-7952/ 2014 Elsevier B.V. All rights reserved.

The present paper describes a methodology and gives an example of

its application in a particular environment using relations developed
from accelerograms recorded in Chile during large earthquakes and
other parameters derived from seismic hazard studies for Chilean projects. Urza and Christian (2013) showed that the basic relations between the logarithms of normalized displacements and the
acceleration ratios evaluated from other sets of records (Northridge
1994, Chi Chi 1996) are close to those derived from the Chilean records,
but application of the proposed methodology for locations exposed to
different seismic hazards requires that comparable relations be developed from accelerograms recorded in that region and that the other parametersArias intensity, period, and peak ground accelerationalso
be evaluated for the particular local conditions. In other words, this
paper presents the results for one region and are not intended to
apply to all locations in the world. Comparable plots for other locations
must be based on seismological data for those regions.
Since Newmark (1965) described the sliding block analytical procedure, many researchers have investigated the consequences of applying
it in a variety of situations. Seed (1979) proposed using a design horizontal acceleration factor ac of 0.15 with a factor of safety greater than
1.15. Hynes-Grifn and Franklin (1984) noted that values of ac should
be related to the peak ground acceleration and indicated that 1 m was
an acceptable displacement criterion for dams. Stewart et al. (2003)
proposed a method for estimating ac for housing developments in

242

A. Urza et al. / Engineering Geology 183 (2014) 241246

Southern California. Bray and Travasarou (2009) also proposed a method for estimating ac in terms of acceptable displacement and seismic demand. Urza and Christian (2013) describe a number of other proposed
methods to estimate seismic displacement and, by implication, to estimate ac from the acceptable displacement.
2. The UrzaChristian analysis of seismic records
Urza and Christian (2013) computed the sliding displacement from
suites of strong-motion records made during three Chilean offshore
subduction zone earthquakes. The suite from the 1985 M = 7.8 earthquake was recorded in the Valparaiso region, the suite from the
2007 M = 7.7 earthquake in the Tocopilla region, and the suite from
the 2010 M = 8.8 earthquake in the Maule region. The suites represent
not only three different events but also three different regions of Chile:
Maule in the south, Valparaiso in the central region, and Tocopilla in the
north. Using the sliding block method (Newmark, 1965; Yegian et al.,
1991), Urza and Christian computed the sliding displacements for
each accelerogram and for a range of values of the ratio between the
critical acceleration for the slope (ac) and the maximum acceleration
in the record (amax).The critical acceleration is usually the horizontal
acceleration just necessary to reduce the pseudo-static factor of safety
to unity or less. There are several published relations between ac/amax
and estimated displacements, but the sliding displacements predicted
by those relations do not always correspond well with the displacements computed from the three sets of records (see Urza and
Christian, 2013 for references).
The proposed method started with the closed form solution for the
sliding-block displacement for cyclic input motion (Yegian et al., 1991):
h
i
2
DR f ac =amax amax NT

usually agreed that values of ac/amax N 0.5 are not of engineering interest (Franklin and Chang, 1977). When it is necessary to consider values
of ac/amax N 0.5 or to design against sliding displacements on the order
of centimeters, the present methodology could be extended to include
the non-linear portion of the function in Fig. 1, but at the cost of additional computational complexity.
The second input to the revised method is the Arias (1970) intensity,
dened as
Ia

t
0

a d

in which g is the acceleration of gravity, a is the ground acceleration, and

is the variable of integration equal to time. It is customary to integrate
over the entire duration of the strong motion record (t = length of record), but t can be any time after the start of the record so that Ia is actually a function of time. If there is no other indication, Ia is taken as
the value integrated over the entire record. The Arias intensity has
units of velocity and is usually expressed in meters per second (m/s)
or centimeters per second (cm/s).
For a sinusoidal acceleration pulse of period T and amplitude amax,
the Arias intensity is
Ia

2g

t
0

jaj d

2g

t
0

amax sin 2=T d

Integrating over a single period gives the Arias intensity for a single
pulse:
Ia

In this expression DR is the cumulative sliding displacement, ac and amax

are already dened, N is the number of cycles, and T is the period of the
cycle. The function f (ac/amax) is a dimensionless function that depends
on the shape of the input motion; different shapes of input pulses (rectangular, triangular, sinusoidal, etc.) yield different forms of the function.
Fig. 1 shows the function for sinusoidal input as originally developed by
Yegian et al. (1991) and independently conrmed by the present authors. A noted feature of the function f (ac/amax) is that, for ac/
amax 0.5, it is nearly linear against a logarithmic vertical scale. At larger
values of the acceleration ratio, the function drops off sharply, and it is

2g

2
a T
4g max

and for N pulses, with the ground acceleration expressed as Ka in units of

g (i.e., amax = Kag)
Ia

2
NgTK a
4

Substituting Eq. (5) into Eq. (1) and performing some algebraic manipulation yields
 
I T
DR hac =amax a
Ka

in which h(ac/amax) is a dimensionless function that incorporates the

shape of the accelerogram as well as constants such as /4. In other
words, a rational way to normalize the computed or estimated sliding
displacements in a dimensionless plot is to divide the computed values
by the bracketed term in Eq. (6), which has units of displacement.
Fig. 2 shows the result of applying the method to the three suites of
Chilean records. The gure was constructed by the following process:

Fig. 1. Sliding displacement during a single sinusoidal pulse (after Yegian et al., 1991).

(1) A series of ten values of (ac/amax) was selected, starting with 0.05
and increasing in units of 0.05 to a nal value of 0.5.
(2) The values of amax, Ia, and T were computed for each
accelerogram.
(3) The sliding block analysis was performed for each value of (ac/
amax) and for each accelerogram. The analyses were run twice:
once with the signs of the accelerations in the original sense
and once with them reversed.
(4) The cumulative displacement in each case was divided by (IaT/
Ka) and the results plotted with a logarithmic vertical axis as dimensionless displacements DN.
(5) For each value of (ac/amax) the mean and the standard deviation
of the log10 DN were computed.
(6) After calculations had been completed for all values of (ac/amax),
straight lines were run through the ten sets of values of the mean
and the mean plus and minus one standard deviation of log10 DN.

A. Urza et al. / Engineering Geology 183 (2014) 241246

Fig. 2. Normalized displacement from three suites of recordings (after Urza and Christian,
2013)

243

Fig. 3. Normalized displacement including records from Chi Chi and Northridge.

The normalized data points fall very nearly on straight lines, so

the regression coefcients are 0.9 or better.
Urza and Christian (2013) plot the data points for all the Maule results and the lines for the mean and mean plus and minus the standard
deviation. They also present the regression lines for all the suites of
earthquake records and the composite regression lines.
The regression results from Fig. 2 can be expressed in the form
log10 DN 0 1 ac =amax

Table 1 shows the values of the coefcients for the three suites of records combined. The normalization procedure requires that a value of
the period T be established for each record. There are several ways to estimate the principal period of an accelerogram. They include the period
at the peak of the response spectrum, the most signicant period of the
Fourier spectrum, various averages of these periods, and the inverse of
the zero-crossing frequency. Each has advantages and disadvantages.
The most important consideration is that the same denition be used
in processing the records to develop the parameters of Eq. (7) and in
selecting the normalizing value used in Eq. (6). In the present case T,
was computed as the period at the peak of the response spectrum.
Fig. 3 shows the results when the method is applied to suites of records
from the three large Chilean earthquakes as well as those from the January 21, 1999, Chi Chi earthquake (M = 7.6) and the January 17, 1994,
Northridge earthquake (M = 6.7). The results are similar, although the
values of the parameters 0 and 1 differ by small amounts, reecting
the differences in local geology and in the style of the basic earthquake
motion.
To use this plot, one rst estimates the ratio ac/amax and then enters
Fig. 2 to obtain the normalized displacement. The three lines represent
the mean or estimated displacement, the mean plus one standard deviation, and the mean minus one standard deviation. The normalized displacement is then multiplied by (IaT/Ka) to obtain the displacement in
Table 1
Values of parameters for Eq. (1).
Case

Mean of log displacement

Mean plus SD of log displacement
Mean minus SD of log displacement

+0.322958
+0.398959
+0.246956

4.29435
3.97092
4.61778

Fig. 4. Replotted sliding displacements for OII landll (after Kavazanjian et al., 2013)

real units. If Ia is in centimeters per second, T is in seconds, and Ka is dimensionless, DR is in centimeters.

The log-linear relation between the horizontal acceleration and the
cumulative sliding displacement is consistent with the results from
other studies. For example, Fig. 4 shows the curves reported in Fig. 17
of Kavazanjian et al. (2013) re-plotted against a logarithmic vertical
axis. Each of the lines corresponds to one mode of failure and to one
earthquake accelerogram used for the analysis of the OII landll in
southern California. Thus, each line represents a single result rather
than a statistical analysis of a set of results, but the general shape of
the trends is clear.
3. Computational approach
The UrzaChristian results in Fig. 2 can be extended into a method
for estimating the critical value of the horizontal acceleration ac. Five

244

A. Urza et al. / Engineering Geology 183 (2014) 241246

parameters enter the calculation: ac, amax, Ia, T, and , which governs the
selection of a relation from Fig. 2. Because amax = Kag, Ka is not an independent parameter.
The rst step is to develop a computational procedure for calculating
the permanent horizontal sliding displacement DR from an input set of
ac, amax, Ia, T, and and then to use this routine repeatedly to evaluate
the statistical parameters of DR. The rst parameter, ac, is a function of
the geometry of the slope and the strength properties of the soils and
rocks comprising the slope. For the simple case of an innite slope inclined at an angle and composed of a frictional material with friction
angle , the critical acceleration is
ac g tan tan

More complicated geometries and soil properties will lead to more complicated relations for ac, which can be obtained from slope stability analyses, but the uncertainty in the value of ac will depend essentially on
the uncertainty in the geometry of the slope, the strength properties,
and the location of the phreatic surface. The statistical properties of ac
are independent of amax, Ia, T, and . Uncertainties in the values of amax,
Ia, and T must be established by analysis of seismic accelerograms combined with probabilistic seismic hazard analysis (PSHA). Fig. 2 shows
that, for a particular acceleration ratio, the normalized relation for Dr is
log-normally distributed with the means and standard deviations plotted
in the gure. The parameter is a standard normal variate that identies
the dimensionless equation for log10 Dr. In the case of a exible earth dam
or tailings structure, the embankment itself modies the input seismic excitation, changing the period of the motion, modifying the peak acceleration, and introducing damping. Many current procedures deal with these
effects by simply prescribing the motion at the crest. The present results
are based on the assumption that the input motions affect the rigid sliding
mass directly, but further study is required to arrive at denite
conclusions.
Once the procedure for calculating the dimensioned Dr is in place,
Monte Carlo simulation can be used to generate a statistical description
of Dr for the full range of input parameters. Experimentation revealed
that using 10,000 samples gives stable results, but identifying the values
of DR at low probabilities requires much larger suites of at least
1,000,000. Plots of the histograms of the values of Dr indicate that the central portion of the distribution is well represented by the log-normal distribution, but the tails are not. Therefore, the log-normal distribution
derived from the mean and standard deviation of the results should not
be used to estimate the displacements at the tails of the distribution. Instead, it is necessary to count the proportion of samples with displacements exceeding a given value.

Table 3
Correlation matrix.

ac
amax
Ia
T

ac

amax

Ia

1.000
0.000
0.000
0.000

0.000
1.000
0.967
0.025

0.000
0.967
1.000
0.133

0.000
0.025
0.133
1.000

Saragoni (2009). The data are presented at the precision given in their
tables. Ruiz and Saragoni do not give values of the uncertainties for
these parameters, but the peak acceleration and Arias intensity in the
table are conservative values already incorporating a large part of the
uncertainty. The coefcients of variation (COV) should also represent
variability at or near a particular site and not the variability over an entire region. Taking all these factors into account indicates that values of
0.20 are appropriate for this example. If the methodology is applied to
an actual situation, both the conservatism in the estimates of the
means and the accuracy of the COVs must be addressed. Examination
of the suites of accelerograms indicates that a smaller COV of 0.1 is appropriate for the T parameter. Statistical analysis of the three suites of
accelerograms reveals the correlation structure among the logarithms
of these three parameters listed in Table 3. The correlations between T
and the other two parameters are small and probably do not affect the
results, but that between amax and Ia is too large to be ignored.
The values of the mean of ac were set to 0.025, 0.050, 0.075, 0.100,
0.150, 0.200, 0.250, 0.300, 0.350, and 0.400 g for the MCE analyses. However, the mean values did not exceed 0.150 g in the OBE analyses because the basic relation in Fig. 2 is not valid for values of ac/amax
greater than 0.5 and the mean of amax is 0.29 g.
5. Results
For the MCE, Fig. 5 shows the probability of exceeding the sliding displacement values. Each line corresponds to a particular mean value of ac.
Note that ac is expressed in units of the acceleration of gravity, g. An example use of this plot is to assume that we want to know the critical acceleration factor ac that would give a 0.01 probability of exceeding 100 cm of
sliding displacement. The plot shows that this corresponds to a mean ac
of approximately 0.19 g. The criterion of 100 cm or 3 ft of movement
seems to be commonly used in a variety of conditions such as earth
dams and waste landlls (e.g., Kavazanjian et al., 2013). Different criteria

4. Parametric values used

The analyses considered two seismic exposures: the maximum credible earthquake (MCE), and the operational basis earthquake (OBE).
Table 2 lists the values of the means and standard deviations of amax,
Ia, and T for the two exposures derived from Table 7.1 of Ruiz and

Table 2
Parametric values used in analyses.
Parameter

MCE

OBE

Mean

COV

Mean

COV

ac
amax
Ia
T

Varies
0.824 g
2003 cm/s
0.13 s

0.1
0.2
0.2
0.1

Varies
0.290 g
536.6 cm/s
0.13 s

0.1
0.2
0.2
0.1

Parameter

Mean

Standard deviation

0
1

0.101725
4.29435

0.160764
0.17603

Fig. 5. Sliding displacements for maximum credible earthquake (MCE).

A. Urza et al. / Engineering Geology 183 (2014) 241246

Fig. 6. Annual probabilities of sliding displacements for maximum credible earthquake

(MCE).

may be appropriate for other types of facilities or facilities in more critical

locations. This criterion is not an absolute number but must depend on
the specic circumstances for the project at hand. Jibson (2011) provides
insight into what is ultimately a management decision.
The probabilities plotted in this gure are those that exist, given that
the earthquake occurs. In other words, the probability of observing
more than 100 cm of sliding in the above example is 0.01 multiplied
by the probability of the MCE itself. The MCE is often selected to have
a 2% exceedance probability in 50 years or, equivalently, 4 104 per
year. Thus, if ac = 0.19 g, the annual probability of 100 cm of sliding
would be 4 106. Fig. 6 displays the annual probabilities of sliding incorporating the annual exceedance probability for the MCE.
Figs. 7 and 8 present the comparable results for the OBE. Fig. 7 shows
the exceedance probabilities under the assumption that the earthquake
happens. Fig. 8 shows the annual probabilities incorporating the recurrence relation of 10% in 50 years or 2 103 per year.

Fig. 7. Normalized displacement including records from Chi Chi and Northridge.

245

There are several assumptions inherent in the analyses presented

here. First, the parameters for amax, Ia, T, and need to be selected carefully. The COV values used here are assumed values and need to be veried from PSHA results and the suite of accelerograms used for a
particular project. Second, the assumption that the normalized displacement is log-linearly related to the acceleration ratio is not valid for acceleration ratios larger than 0.5, but values greater than 0.5 are probably
not of engineering interest. Third, the sliding block analysis used here
does not account for strength degradation during sliding. The strength
properties of the sliding materials need to be corroborated by the geotechnical engineer.
Interpreting Figs. 5 through 8 requires some care. Fig. 5 shows that if
the acceleration necessary to set a slope in motion is ac = 0.15 g, there
is about 0.05 probability that the sliding will exceed 100 cm due to the
MCE. Fig. 7 shows that the same slope will have exceedance probability
that is much lower than this for the OBE; actually it is less than the
range of values for Fig. 7. This makes sense because the OE is a much
smaller earthquake.
However, if one asks what value of ac corresponds to 100 cm of displacement with a probability of 0.1, the answer from Fig. 5 for the MCE
is 0.13 g and from Fig. 7 about 0.035 g for the OBE. This seems counterintuitive. How could the smaller earthquake have a smaller critical acceleration? The answer is that the values of ac describe how much horizontal acceleration is needed to put the slide in motion. For a slope to be
in a condition that the small accelerations of the OBE would cause
100 cm of sliding, it must be near failure, and only 0.035 g of horizontal
acceleration will bring it to the point of sliding. The more robust slope of
the rst example, which requires 0.13 g of horizontal acceleration to
start sliding, must be shaken by the much larger MCE in order to develop 100 cm of displacement.
Figs. 6 and 8, which incorporate the annual recurrence rates of the
earthquakes, are useful for comparing risks. For example, Fig. 6 shows
that a slope with ac = 0.075 g has an annual probability of 2 104
for 100 cm of sliding due to the MCE. Fig. 8 shows that the same slope
has an annual probability of 2 106 for 100 cm of displacement due
to the OBE. In other words, the contribution of the OBE to the annual
risk is insignicant compared to the contribution of the MCE.

Fig. 8. Annual probabilities of sliding displacements for operational basis earthquake

(OBE).

246

A. Urza et al. / Engineering Geology 183 (2014) 241246

6. Conclusions
The UrzaChristian model for normalized sliding displacement has
been extended to develop a probabilistically based method of determining the critical acceleration value corresponding to the probability of observing prescribed amounts of sliding displacement.
The method has been applied to data developed from seismic hazard
studies for particular locations in Chile. Other sets of parameters can be
developed for other specic site conditions.
For the parameters used in the sample analysis and the requirement
that there be 0.1 probability of 100 cm of sliding displacement, a critical
acceleration value of 0.13 g is found for the MCE. This means that a slope
with these parameters in this environment must be stable enough that a
horizontal acceleration of 0.13 g is necessary to put it in a state of sliding
motion. In the case of the OBE, which is a much smaller ground motion,
the criterion of 0.1 probability of 100 cm of sliding is achieved for a slope
with a critical acceleration of 0.035 g.
Further analyses for different conditions and assumptions should be
carried out to understand better the implications of the model.
References
Arias, A., 1970. A measure of earthquake intensity. In: Hansen, R.J. (Ed.), Seismic Design
for Nuclear Power Plants. Massachusetts Institute of Technology Press, Cambridge,
MA, pp. 438483.

Bray, J.D., Travasarou, T., 2009. Pseudostatic coefcient for use in simplied seismic slope
stability evaluation. J. Geotech. Geoenviron. Eng. ASCE 135 (9), 13361340.
Franklin, A.G., Chang, D., 1977. Permanent Displacements of Earth Embankments by
Newmark Sliding Block Analysis. Report 5, Miscellaneous Paper S-71-17, U. SArmy
Corps of Engineers Waterways Experiment Station, Vicksburg, MS.
Hynes-Grifn, M.E., Franklin, A.G., 1984. Rationalizing the Seismic Coefcient Method.
Miscellaneous Paper No. GL-8413U.S. Army Engineer WES, Vicksburg, MS.
Jibson, R.W., 2011. Methods for assessing the stability of slopes during earthquakesa retrospective. Eng. Geol. 122, 4350.
Kavazanjian Jr., E., Matasovic, N., Bachus, R.C., 2013. 11th Peck Lecture: Predesign geotechnical investigation for the OII superfund site landll. J. Geotech. Geoenviron.
Eng. ASCE 139 (11), 18491863.
Newmark, Nathan M., 1965. Effects of earthquakes on dams and embankments.
Geotechnique 15, 139160.
Ruiz, S., Saragoni, R., 2009. Riesgo ssmico, espectros de diseo generacin de
accelerogramas articiales, Private Client Deposito de Relaves Pampa Pabelln,
Santa Ines de Collahuassi, Rosario, Ujima. S y S Ingenieros Consultores Ltda.
Seed, H.B., 1979. Considerations in the earthquake-resistant design of earth and rockll
dams. Geotechnique 29 (3), 215263.
Stewart, J.P., Blake, T.F., Hollingsworth, R.A., 2003. A screen analysis procedure for seismic
slope stability. Earthq. Spectra., EERI 19 (3), 697712.
Urza, A., Christian, J.T., 2013. Sliding displacements due to subduction-zone earthquakes.
Eng. Geol. 166, 237244.
Yegian, M.K., Marciano, E.A., Ghahraman, V.G., 1991. Earthquake-induced permanent deformations: probabilistic approach. J. Geotech. Eng. ASCE 117, 3550.

Technical Breakthrough Abstract

Displacements from the 2014 Iquique M8:2 Earthquake and

M7:7 Aftershock Added to a Sliding Displacement Model
Alfredo Urzua, Ph.D., M.ASCE1; and John T. Christian, Ph.D., Dist.M.ASCE2

Downloaded from ascelibrary.org by John Christian on 10/28/14. Copyright ASCE. For personal use only; all rights reserved.

DOI: 10.1061/(ASCE)GT.1943-5606.0001238
There are many empirical relations for estimating sliding displacements caused by seismic excitation of a mass resting on a rigid
plane. Usually the accelerograms from a suite of earthquakes are
input to a Newmark sliding block analysis, and regression analysis of the displacements against a set of parameters gives the
empirical relation. When the relations are applied to accelerograms from events not in the original database, the accuracy of
the predictions varies considerably (Urza and Christian 2013).
Combining an analytical expression for the Arias intensity of a
sinusoid with the sliding block displacement for sinusoidal input,
Urza and Christian (2013) obtained the following estimate of
displacements:

log10 DN b0 b1 ac =amax

DR DN

Ia T
Ka


(1)

where DR 5 sliding displacement (meters or centimeters); DN

5 normalized displacement; ac 5 critical horizontal acceleration
at which sliding starts; amax 5 maximum acceleration in the
accelerogram; b0 and b1 5 parameters found by linear regression;
Ia 5 Arias intensity of the ground motion (meters per second or
centimeters per second); T 5 fundamental period of the ground
motion (seconds); and Ka 5 peak acceleration (gravitational
force). Urza and Christian (2013) developed values of b0 and b1
on the basis of accelerograms from three large Chilean earthquakes: the 1985 M 5 7:8 earthquake in the Valparaiso region, the
2007 M 5 7:7 earthquake in the Tocopilla region, and the 2010
M 5 8:8 earthquake in the Maule region. The analysis for each
record (1) selected a value of ac =amax between 0.05 and 0.50 in
intervals of 0.05; (2) chose one of the time histories; (3) performed
the sliding block analysis for that combination of time history and
ac =amax ; and (4) repeated Steps 13 for all the other values of
ac =amax and time histories. Linear regression analyses (R2 5 0:892)
gave the dashed lines in Fig. 1. The same analysis on records from
the 1999 M 5 7:6 Chi Chi and 1994 M 5 6:7 Northridge earthquakes gave similar results, indicating that the model applies to
events with other source mechanisms.
The accelerograms from the 2014 M 5 8:2 Iquique earthquake
and a major aftershock with M 5 7:7 have become available. Applying the same algorithms to 46 records from the main shock and 28
from the aftershock provides the solid lines in Fig. 1. The normalized
displacements are slightly larger than those for the original suite and

have slightly greater scatter. Nevertheless, the results are close and
support the basic soundness of the model.
How close the new results are to those predicted by the earlier
lines can be appreciated by considering the estimated displacements
in centimeters. A report for a site in Chile proposes that the maximum credible earthquake has Ia 5 2,000 cm=s, T 5 0:13, and
Ka 5 0:8g. When these parameters and the original b0 and b1 values
are used in Eq. (1) for a slope with critical acceleration ac 5 0:2g,
the median computed sliding displacement is 50 cm. The b0 and b1
values from the 2014 Iquique earthquakes give 58 cm. In view of the
many recognized limitations of the sliding block model, these results
are remarkably close.

Implications
In summary, processing the accelerograms from the Iquique earthquake and a large aftershock gives estimates of computed sliding
displacements that are very close to those derived from the earlier
sets of records. This supports the accuracy and robustness of the
basic log-linear model.

References
Urza, A., and Christian, J. T. (2013). Sliding displacements due to
subduction-zone earthquakes. Eng. Geol., 166(Nov), 237244.

President, Prototype Engineering, 57 Westland Ave., Winchester, MA

01890. E-mail: prototypeengineering@comcast.net
2
Consulting Engineer, 36E Seven Springs Ln., Burlington, MA 01803
(corresponding author). E-mail: jtchrist36@comcast.net
Note. This manuscript was submitted on September 2, 2014; approved
on October 1, 2014; published online on October 27, 2014. This technical
breakthrough abstract is part of the Journal of Geotechnical and Geoenvironmental Engineering, ASCE, ISSN 1090-0241/02814002(1)/
$25.00.
ASCE

Fig. 1. Normalized sliding displacements from the Urza and Christian

(2013) model compared with normalized displacements from the 2014
Iquique earthquake: dashed lines are from Urza and Christian (2013);
solid lines are for the 2014 earthquakes; heavy lines are mean values;
light lines are means 6 SD

02814002-1

J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng.