Civil Engineering Dimension, Vol. 9, No.

2, 57–63, September 2007

ISSN 1410-9530

SEISMIC RISK ASSESSMENT OF LEVEES*

Dario Rosidi
Principal Technologist, CH2M HILL Corporation 155 Grand Avenue,
Suite 1000 Oakland, CA 94612, USA
Email: [email protected]

ABSTRACT
A seismic risk assessment procedure for earth embankments and levees is presented. The procedure consists of
three major elements: (1) probability of ground motion at the site, (2) probability of levee failure given a level of
ground motion has occurred and (3) expected loss resulting from the failure. This paper discusses the first two
elements of the risk assessment. The third element, which includes economic losses and human casualty, will
not be presented herein. The ground motions for risk assessment are developed using a probabilistic seismic
hazard analysis. A two-dimensional finite element analysis is performed to estimate the dynamic responses of
levee, and the probability of levee failure is calculated using the levee fragility curve. The overall objective of the
assessment is to develop an analytical tool for assessing the failure risk and the effectiveness of various levee
strengthening alternatives for risk reduction. An example of the procedure, as it applies to a levee built along the
perimeter of an island for flood protection and water storage, is presented. Variations in earthquake ground
motion and soil and water conditions at the site are incorporated in the risk assessment. The effects of
liquefaction in the foundation soils are also considered.
Keywords: Seismic analyses, risk assessment, logic tree, embankment, levee, dynamic response, slope
deformation

INTRODUCTION To help illustrating and discussing the procedure,

consider a hypothetical levee built along the
This paper discusses a seismic risk assessment perimeter of an island for flood protection and water
procedure for earth embankments and levees. The storage (reservoir). The interior soil stratigraphy in
overall objective of the assessment is to develop an the island consists of a surficial soft organic fibrous
analytical tool for assessing levee vulnerability peat layer, underlain by a silty sand aquifer. The
subject to seismic loads and for evaluating effect- sand aquifer, in turn, overlies stiff silty clay deposits,
tiveness of various levee strengthening alternatives. and it has the potential to liquefy when subjected to
A general risk assessment procedure consists of earthquake shaking. The thickness of the peat and
three major elements: (1) probability of ground sand layers varies from one part of the island to
motions at the levee site, (2) probability of levee another. The effectiveness of levee strengthening
failure given a level of ground motion has occurred techniques for reducing the risk is also illustrated.
and (3) the expected loss resulting from the failure.
This paper presents the first two elements of the risk
assessment. ANALYSIS APPROACH
The probabilistic seismic hazard analysis is com- Seismic risk of a levee is assessed through the
monly used to estimate the ground motions at a site. probability that the levee will experience one or more
To compute the levee dynamic responses and the failures during future earthquake. The likelihood of
estimated levee deformations, a two-dimensional a levee experiencing failure is assumed to be a
finite element analysis is performed. The results of function of seismic-induced slope deformation,
these analyses are then combined with the levee expressed in term of levee fragility. The assessment
fragility curve to provide estimates of the probability procedure, therefore, consists of three elements: (A)
of failure. earthquake ground motions, (B) levee deformations
and (C) expected levee damage (probability of levee
Because earthquake ground motions at a site are
uncertain and levee, soil and water conditions often failure). Discussions on these three elements are
vary over the levee site and time, the seismic risk is presented below.
calculated by considering the variation of these
factors. Earthquake Ground Motion
*Also published in the 11th Int. Conf. on Soil Dynamics & EQ Depending on its length and the propagation speeds
Engng. and the 3rd Int. Conf. on EQ Geotechnical Engng., January of the seismic waves, earthquake ground motion
07-09, 2004, Berkeley, California.
may vary significantly along the levee. If this is the
Note: Discussion is expected before November, 1st 2007, and will case, ground motions at various locations along the
be published in the “Civil Engineering Dimension” volume 10,
number 1, March 2008.
levee may need to be estimated. For a short levee
Received 5 March 2007; revised 11 April 2007; accepted 19 April system, a single estimate at the center of the levee
2007. length is often used for risk assessment.

Civil Engineering Dimension

ISSN 1979-570X print © 2008 Petra Christian University
http://puslit2.petra.ac.id/ejournal 57
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

The earthquake ground motion is developed using a earthquakes, while the 1987 Whittier Narrows
probabilistic seismic hazard analysis (PSHA). This earthquake is selected to represent seismic events on
analysis procedure was originally developed by the local seismic sources.
Cornell [1] and Kulkarni et al. [2], and includes
many recent features/development. It takes into
accounts the uncertainties in the size, location and
rate of recurrence of earthquakes and in the
variation of ground motion characteristics at the site.
The major components or steps in a PSHA are:
1. Characterization of seismogenic earthquake
sources: the location, geometry and characteristics
of seismic sources or earthquake faults.
2. Specification of source recurrence relationship:
relationship that shows the annual recurrences of
earthquake of various magnitudes, up to the
maximum magnitude.
3. Evaluation of probability of distance to rupture:
this probability is assessed by considering the
geometry of the fault and relationships between
rupture dimension and magnitude.
4. Calculation of exceedance using ground motion
attenuation equations: the probability that the
ground motion parameter Z, from earthquake of a
certain magnitude occurring at a certain distance,
will exceed a specified level z at the site. Figure 1. Uniform Hazard Spectra for Return Periods of 43, 475,
5. Calculation of probabilistic seismic hazard: the and 2500 Years (Free-Field Stiff Soil)
mathematical procedure for combining the
components described in steps 1 through 4. Table 1. Summary of Earthquake Records Used
in the Dynamic Response Analysis
The seismic source model include faults that rupture
Recording Station
to the ground surface as well as those that do not Earth- Site Component Recorded
(i.e., blind thrust seismic sources). Random or Mw Distan-
quake Station Condi- (deg) PGA (g)
ce (km)
background seismic sources are also included. Proper tion
characterization of uncertainties in source para- 1987 Altadena-
Whittier 6.0 18 Eaton Soil 90 0.15
meters and ground motion attenuation models is Narrows Canyon
important in a PSHA; they can be incorporated 1992
7.3 64 Fort Irwin Soil 0 0.11
using a logic tree approach, where multiple values of Landers
parameters and ground motion models can be
considered and weighted. The results of a PSHA are The response spectral ordinates calculated from the
expressed in terms of Uniform Hazard Spectra recorded acceleration time histories show peaks and
(UHS) for various exceeding probabilities. Figure 1 valleys that deviate from the smooth UHS.
shows the calculated UHS for three ground motion Lilhanand and Tseng [3] proposed a method (later
return periods: a short period of 43 years, a moderate modified by Abrahamson [4]) to develop acceleration
return of 475 years and a long period of 2,500 years. time histories with overall characteristics that match
These UHS represent free-field motions at an the UHS. Using this method, the selected accelera-
outcropping site. tion time histories are spectrally matched to the
UHS. The 5%-damped response spectra calculated
Multiple earthquake acceleration time histories are from the modified motions are shown in Figure 2,
typically used as inputs to the levee dynamic together with the UHS for the return period of 475
response analysis. For the purpose of our discussion years. This figure indicates that the response spectra
and illustration, two earthquake time histories are calculated from the modified time histories closely
considered. They are selected from the 1992 Landers match the UHS.
earthquake (M= 7.3) recorded at Fort Irwin station
and the 1987 Whittier Narrows earthquake (M= 6.0) Levee Deformations
recorded at Altadena, Eaton Canyon station. Table 1
lists these recorded motions along with their closest The calculation of seismic-induced deformations of a
distances from the rupture planes and the recorded levee consists of the following two steps:
peak accelerations. The 1992 Landers earthquake is Step 1: Calculate site and levee dynamic responses.
selected to represent the larger and more distant Step 2: Estimate levee slope deformations.

58
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

from incremental sliding during the short periods

when earthquake inertia force in the critical sliding
mass exceeds the available resisting force.

The inertia force is calculated using the procedure

described above, and the resisting force is
represented by the yield acceleration (ky) calculated
for the sliding mass. The yield acceleration of a
sliding mass is defined as the horizontal acceleration
that will initiate slide movement. This method
involves the calculations of the displacement
(deformation) increment of a critical sliding mass at
each time step using the average horizontal
acceleration (kave) and the value of yield acceleration
(ky) calculated for the sliding mass.

The effects of liquefaction on the estimated slope

deformations can be incorporated through the
reduction in shear resistance along the critical slip
surface during earthquake shaking. This translates
into lower yield acceleration, ky, which in turn,
induces larger deformations.
Figure 2. Comparison of Response Spectra for Return Period
of 475 Years
Expected Damage
Step 1: Site and Levee Dynamic Responses As indicated earlier, the levee probability of failure is
assumed to be a function of levee slope deformations
The earthquake ground motions developed from a
PSHA that uses the standard attenuation models (i.e., the levee fragility is expressed as a relationship
are applicable for a free-field stiff soil or rock site. To between probability of levee failure and expected
account for the effects of local soils (fills, organic peat slope deformations). The conditions that define a
and sand) and levee on the ground motions levee failure include, among others, piping, over-
characteristics, a two-dimensional dynamic response topping, cracking, slumping and excessive settle-
analysis should be carried out. The analysis model ment. Example of a levee fragility curves is
should take into accounts the various layers of clayey presented in Figure 3.
and sandy soils encountered within and beneath the
levee system.

The computer program QUAD4M (Hudson et al. [5])

can be utilized for the analysis. QUAD4M is a two-
dimensional, plane-strain, finite element program for
dynamic response analysis. It uses the equivalent
linear procedure of Seed and Idriss [6] to model the
nonlinear behavior of soils. The softening of the soil
stiffness is specified using the shear modulus
degradation (G/Gmax) and damping vs. shear strain
curves. QUAD4M also incorporates a compliant base
(energy-transmitting base), which can be used to
model the elastic half-space.

The time histories of seismic-induced inertia forces

are calculated from the dynamic response analysis.
The average horizontal acceleration (kave) time Figure 3. Example of a Levee Fragility Curve
history acting on an identified critical sliding mass is
then calculated by summing the inertia forces and The expected levee damage from future earthquakes
dividing the sum with the mass of the slide. is calculated by integrating the probabilities of
failure estimated for the various conditions asso-
Step 2: Seismic-induced Slope Deformations ciated with levee and site soil conditions, subsurface
Seismic-induced permanent slope deformations of a and surface water levels and earthquake ground
levee can be estimated by the Double Integration motion and liquefaction scenarios. A logic tree
Method of Newmark [7]. This method is based on approach is commonly used for integrating the
the concept that deformation of a levee will result results of these conditions (see Figure 4).

59
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

To illustrate the application of the above risk (reservoir). The existing levee materials generally
assessment procedure, consider a hypothetical levee consist of peat and dredged sand, silt and clay.
system that was built along the perimeter of an Beneath the levee is a thick layer of peat with sandy
island for flood protection and water storage silt inter-layers. This peat is typically about 20 ft

Figure 4. A Logic Tree Approach for Integrating the Results of Various Assumptions

Figure 5. Typical Cross Section of Proposed Bench Alternative

Figure 6. Typical Cross Section of Proposed Rock Berm Alternative

60
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

(609.6 cm) to 40 ft (1,219.2 cm) thick in the fields The shear modulus degradation (G/Gmax) and dam-
away from the levee, but it has been highly com- ping curves of Kokusho [9] and Vucetic and Dobry
pressed under the weight of the levee. Underlying [10] are applied for the sandy soils (levee fill and
the peat, there is loose to dense sand stratum. In alluvium) and clays, respectively. For peat, the
some areas, this sand deposit is susceptible to relationships of Wheling et al. [8] are utilized. The
liquefaction under the expected earthquake shaking. dynamic soil properties used for the response
analyses are summarized in Table 2. Plots of the
Two cross sections representing different subsurface selected G/Gmax and damping vs. shear strain
conditions along the levee system are developed relationships are presented in Figure 7.
and used for the analysis: peat extends to elevation
–20 ft (-609.6 cm) in one section and extends to
elevation –40 ft (-1219.2 cm) in the second section.
The weights for these two sections used in the logic
tree of risk assessment are assigned based on the
percentages of the respective conditions encountered
along the levee system.

In addition to the existing levee condition, two levee

strengthening alternatives are also considered to
evaluate the effectiveness of these strengthening
alternatives. The two strengthening alternatives
involve:
1. Construction of a new bench (called bench
alternative), as shown in Figure 5.
2. Placement of a rock-berm (called rock-berm
alternative), as shown in Figure 6.

For the site and levee dynamic response analysis,

the shear and compressive wave velocities obtained
from a geophysical measurement at the site are
used. The relationship that relates maximum shear
modulus, over-consolidation ratio (OCR) and
effective pressure proposed by Wheling et al. [8] is
also utilized to account for the dependency of shear
modulus (or shear wave velocity) on effective
pressure.

Tabel 2. Dynamic Material Properties

Moist Unit Shear Wve Modulus and
Description Weight (pcf) K2 max Velocity Damping
(ft/sec) Curves
Embankment Materials
New fills sand 120 80 - Sand1
Peat - free-feild Peat2
70 - See Note4
- under embankment Peat2
Foundation Materils
Sand (non-liquefied) 120-125 80 - Sand1
(liquefied) 120-125 - 300-400 See Note5
Clay 127 - 1000 Clay3
Note:
1. Relationships of Kokusho (1980), function of confining pressure
2. Relationships of Wehling et al (2001) Figure 7. Modulus and Damping Curves for Soils Used in
3. Relationships of Vucetic and Dobry (1991) for PI = 50
4. Shear wve velocity ws estimated using the following equations (Wehling Analysis
et al. 2001)
Some of the borings drilled within the levees and the
Gmax
vs = > 75 ft/sec underlying foundation soils encountered loose sand
ρ layers. The liquefaction potential analysis is
Gmax ⎡σ ' ⎤
0.87 performed to assess the liquefaction susceptibility of
= 75.7 ⎢ lc ⎥ OCR 0.65 these sandy layers using the standard SPT-based
Pa ⎣ Pa ⎦ procedure proposed by Youd and Idriss [11]. The
Where Pa and σ’lc are the atm ospheric and effective vertical pressures,
respectively
analysis results indicate the upper 5 ft (152.4 cm) of
5. For liquefied sand, no reduction in G is allowed and the damping is fix the sand layer beneath some areas of the levee
ed t 8%-10% of critical damping system is susceptible to liquefaction. Liquefaction

61
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

within these loose sandy deposits during earthquake Weights assigned to the reservoir and slough water
will increase the levee deformations, and hence the level scenarios are estimated based on the time
probability of levee failure. percentage of each scenario to occur annually.

The liquefied shear strength at small strains for the No comergence

Probability of Failure, %
loose sandy deposits is modeled using a shear wave
velocity of 300 ft (9144 cm)/sec. No shear modulus Existing
degradation is allowed for the liquefied soil condition.
The damping values are kept constant at 8% to 10%
Bench
of the critical damping value. The damage
Rock Berm
assessment considers the effects of liquefaction, and
the weights for the liquefaction scenarios used in the
logic tree (Figure 4) are selected based on judgment
and evaluation of field blow-counts recorded in the
sandy deposits.

Two operating water elevation scenarios are selected

to represent the normal fluctuation of surface Ground Motion Return Period, years
(reservoir) water and tidal water elevations near the
levee. These scenarios are as follows: Figure 8 - Probability of Failure for Existing, Bench, and Rock
Berm Alternatives
1. High Tide and Low Reservoir: a low storage
(reservoir) and high slough (tide) water at
Table 3 shows the contributions of the various
elevation +3.5 feet. (106.68 cm) This condition scenarios to the expected damages (probabilities of
was assumed to prevail 2/3 of the time (weight of failure) for the existing levee condition with peat
0.66). extending to elevation -20 ft (60 cm). The total failure
2. Low Tide and High Reservoir: high storage probability for this case is estimated to be about
(reservoir) water at elevation +4.0 feet (121.92 22.24 percent. The corresponding failure proba-
cm) and low slough (tide) water at elevation –1 bilities for the bench and rock-berm strengthening
foot (-30.48 cm). This condition was assumed to alternatives are calculated to be about 16.87 percent
prevail 1/3 of the time (weight of 0.34). and 16.06 percent, respectively (Tables 4 and 5).

Table 3. Probability of Failure of Existing Cross Section I (Peat At -20 Ft)

Probability Probability of Probability of Average Probability Probability of
Water Level Ground
Cross Section of Scenario Ground Liquefaction
Liquefaction of Failure for Failure in 50
Scenario Motion Level
(%) Motion (%) (%) Section (%) years (%)
Liquefaction 20 2.35 0.329
43 years 70
Non- Liquefaction 80 0.04 0.022
Cross Section I High tide low Liquefaction 70 95.00 16.625
100 475 years 25
(Peat at -20 ft) reservoir Non- Liquefaction 30 6.78 0.509
Liquefaction 95 95.00 4.513
2500 years 5
Non- Liquefaction 5 95.00 0.238
Sum of Failure
22.235
Probabilities (%)

Table 4. Probability of Failure of Cross Section I (Peat At -20 Ft) With Bench Alternative
Probability Probability of Probability of Average Probability Probability of
Water Level Ground
Cross Section of Scenario Ground Liquefaction Liquefaction of Failure for Failure in 50
Scenario Motion Level
(%) Motion (%) (%) Section (%) years (%)
Liquefaction 20 0.29 0.011
43 years 70
Non- Liquefaction 80 0.01 0.002
High tide low Liquefaction 70 14.68 0.848
33 475 years 25
reservoir Non- Liquefaction 30 0.23 0.006
Liquefaction 95 95.00 1.489
Cross Section I 2500 years 5
Non- Liquefaction 5 95.00 0.078
(Peat at -20 ft) Liquefaction 20 1.11 0.104
43 years 70
Non- Liquefaction 80 0.01 0.004
High tide low Liquefaction 70 95.00 11.139
67 475 years 25
reservoir Non- Liquefaction 30 0.09 1.005
Liquefaction 95 95.00 3.023
2500 years 5
Non- Liquefaction 5 95.00 0.159
Sum of Failure
16.867
Probabilities (%)

62
 D. Rosidi / Seismic Risk Assessment o Levees / CED, Vol. 9, No. 2, 57–63, September 2007

Table 5. Probability of Failure of Cross Section I (Peat At -20 Ft) With Rock Berm Alternative
Probability of Probability of Average Probability Probability of
Water Level Probability of Ground
Cross Section Ground Liquefaction Liquefaction of Failure for Failure in 50
Scenario Scenario (%) Motion Level
Motion (%) (%) Section (%) years (%)
Liquefaction 20 0.01 0.000
43 years 70
Non- Liquefaction 80 0.01 0.002
High tide low Liquefaction 70 1.53 0.088
33 475 years 25
reservoir Non- Liquefaction 30 0.01 0.000
Liquefaction 95 95.00 1.489
2500 years 5
Cross Section I Non- Liquefaction 5 95.00 0.078
(Peat at -20 ft) Liquefaction 20 0.80 0.075
43 years 70
Non- Liquefaction 80 0.01 0.004
High tide low Liquefaction 70 95.00 11.139
67 475 years 25
reservoir Non- Liquefaction 30 0.06 0.003
Liquefaction 95 95.00 3.023
2500 years 5
Non- Liquefaction 5 95.00 0.159
Sum of Failure
16.061
Probabilities (%)

The calculated failure probabilities for the two 3 Lilhanand, K. and Tseng, W.S., Development
strengthening alternatives and the existing levee are and Application of Realistic Earthquake Time
compared in Figure 8, as a function of earthquake Histories Compatible with Multiple-Damping
ground motion return period. Design Spectra. Proceeding of the 9th World
Conference on Earthquake Engineering, Tokyo-
As expected, the strengthening is expected to reduce Kyoto, Japan, August 1988.
the risk of levee failure during the design life-time of 4 Abrahamson, N., Non-Stationary Spectral
the levee. The rock-berm alternative produces a Matching Program Personal Communication,
lower expected damage (probability of failure) than 1993.
the bench alternative. This is true because the rock-
berm alternative places the embankment over the 5 Hudson, M., Idriss, I.M., Beikae, M., User’s
existing levee, and therefore, makes use of the Manual for QUAD4M, Center for Geotechnical
stronger peat under the levee as opposed to the Modeling, Department of Civil & Environmental
weaker free-field peat. In addition, it provides a more Engineering, University of California, Davis,
stable slough side slope. California, May 1994.
6 Seed, H.B. and Idriss, I.M., Soil Moduli In
CONCLUSIONS Damping Factors For Dynamic Response
Analysis, Report No. EERC 70-10, University of
The seismic risk assessment procedure described in California, Berkeley, December 1970.
this paper provides a systematic procedure for
quantifying risk of a levee system, subject to 7 Newmark N.M., Effects of Earthquakes on
earthquake loading. The procedure can be used as Dams and Embankments, Geotechnique, Vol. 15,
an efficient/effective tool for decision making. The No. 2, June 1965.
procedure takes into account the uncertainties
8 Wheling, T.M., Boulanger, R.W., Arulnathan, R.,
associated with earthquake loading and soil and
Harder, L.F., and Driller, M.W., Nonlinear
water conditions. It can be used to evaluate various
Dynamic Properties of a Fibrous Organic Soil,
design alternatives, and to help identifying the
Journal of Geotechnical and Geoenvironmental
alternative that produces the lowest risk. Engineering, ASCE, 2001.
9 Kokusho T., Cyclic Triaxial Test of Dynamic Soil
REFERENCES
Properties for Wide Strain Range, Journal of
Soils and Foundations, 1980, Vol. 20, pp. 45-60.
1 Cornell, C.A., Engineering Seismic Risk Analysis
in Bulletin of the Seismological Society of 10 Vucetic, M. and Dobry, R., Effect of Soil Plas-
America, 1968, vol. 58, No.5, pp.1583-1606. ticity on Cyclic Response, Journal of Geotech-
nical Engineering, ASCE, 1991, Vol. 117, No.1.
2 Kulkarni, R. B., Sadigh, K., and Idriss, I. M.,
1979, Probabilistic evaluation of seismic expo- 11 NCEER, Proceeding of the NCEER Workshop on
sure, Proceedings, Second U. S. National Confe- Evaluation of Liquefaction Resistance of Soils,
rence on Earthquake Engineering, Stanford, edited by Youd, T.L., and Idriss, I.M., 1997,
California, August 22-24, 1979, pp. 90-99. Technical report NCEER-97-0022.

63