Professional Documents
Culture Documents
DESIGN OF EMBANKMENT DAMS TO WITHSTAND EARTHQUAKES - Fell (2004)
DESIGN OF EMBANKMENT DAMS TO WITHSTAND EARTHQUAKES - Fell (2004)
ABSTRACT
The ANCOLD Guidelines for the Design of Dams for Earthquake (ANCOLD 1998) were developed by a
working group convened by the author over the period 1993-1998.
Since the release of the guidelines there have been some important developments in the design of embankment
dams for earthquakes, and it has become apparent some aspects of the guideline could be explained better.
This paper outlines some of these issues with a view to encouraging the Profession to use up-to-date methods.
2. SELECTION OF THE DESIGN The author is of the view that the only logical
EARTHQUAKE approach is a risk based one, which properly
accounts for the conditional probability of the dam
*Professor of Civil Engineering, School of Civil and Environmental Engineering, The University of New South
Wales, Sydney 2052. r.fell@unsw.edu.au
10
Mw 0.0 and above
Mw 2.0 and above
Mw 3.0 and above
1 Mw 4.0 and above
Mw 5.0 and above
0.1
PGA
(g)
0.01
0.001
0.01 0.1 1 10 100 1000 10000 100000
Return Period (Years)
Figure 1 – Peak ground acceleration versus annual exceedance probability and earthquake magnitude for a site in
Australia
failing given the earthquake occurs, summed over For assessment of liquefaction, only earthquakes
the full range of earthquake loads. with magnitude 5 or higher should be included in
the earthquake loading (Youd et al 2001). The
same assumption has been shown to be acceptable
2.2 Effect of Earthquake for deformation analysis.
Magnitude on Seismic Where the available earthquake hazard analysis
Loading does not separate the effects of magnitude it is
worth having the analysis re-done. This will also
The example of “typical eastern Australian peak mean more up-to-date attenuation equations will be
ground acceleration Vs AEP” in Appendix B of the incorporated in the earthquake hazard analysis.
Guideline does not show the contributions of the
different magnitude earthquakes. Figure 1 shows an 3. RISK ANALYSIS APPROACH
example where this is done.
The section in the Guideline on Risk Analysis is
It can be seen that a major contribution to the largely superseded by the ANCOLD (2003)
earthquake hazard at high probabilities is from Guidelines for Risk Assessment. The method
earthquakes of very small magnitude. outlined in section 4.2.2 is correct in principle, but it
should be recognised that when assessing the
The motion from these small magnitude conditional probability (PBC) that the dam will
earthquakes is a very high frequency and will not breach for each of the ground motion ranges, this
contribute significantly to deformations. They do should be done for each of the potential failure
not cause liquefaction because of the high frequency paths eg.:
and small number of cycles of loading.
(a) For loss of freeboard due to liquefaction of safety is less than 1.0, freeboard is lost and breach
the foundation soil, consider: will occur.
Figure 2 – Limits in the particle size gradation curves separating liquefiable and unliquefiable soils as suggested
in 1985 by USNRC (USNRC 1985).
Figure 3 – Ranges of particle size gradation curves for mine tailings with low resistance to liquefaction as
suggested in 1985 by USNRC (Ishihara 1985, USNRC 1985).
Figure 4 – Particle size gradation of soils susceptible to flow liquefaction under static and earthquake loading
(Hunter and Fell 2003a, b).
Seed and Idriss (1982), based on data on - The ‘clay’ content (particles smaller than
liquefaction in China produced by Wang (1979, 0.005 mm) is ≤ 15% by weight.
1981), recognised that clayey soils could liquefy - The liquid limit is ≤ 35%.
provided their water content was high (relative to - The natural moisture content is ≥ 0.9 times
the liquid limit), the soils were not too plastic and the liquid limit.
not too much clay was present. They proposed that
liquefaction can occur only if all three of the These were included in the Guideline attributed to
following criteria are met: USBR (1989).
Collins and Tjoumas (2003) point out that the liquefaction, even if not all criteria in (c) are
original Wang (1979, 1981) criteria were somewhat met.
different, being that soils which satisfied the
following criteria may liquefy: (e) If soils are close to the boundaries of these
criteria, and the consequences of
1. wc ≥ 0.9 wL, where wc is water content, liquefaction are great enough, err on the
wL is liquid limit side of caution and/or carry out laboratory
2. IL ≥ 0.75, where IL is Liquidity index, (wc tests.
– wP)/Ip
3. qu ≤ 50 – 70 kN/m2, where qu is
unconfined compressive strength 4.2 The Seed Method for
4. SPT N ≤ 4
5. Sensitivity St ≥ 4
Assessing Whether a Soil Will
Liquefy
They also point out that the Seed and Idriss (1982)
type criteria should only apply to shallow (less than The Guidelines outline the Seed method for
5 m) soils and, at higher stresses, the original assessing whether a soil will liquefy based on
criteria may be more meaningful. They quote Standard penetration Test data. This is based on
experience from Saluda dam where silty (ML) and Seed and De Alba (1986).
sandy (SM) soils outside the criteria were shown to
be liquefiable in laboratory tests. They recommend In 1996 a NCEER workshop on the evaluation of
the Seed and Idriss (1982) criteria not be applied to Liquefaction Resistance of Soils was attended by 21
non-plastic ML or SM soils, and suggest that such experts, who reached a consensus report on the state
soils, up to a plasticity index of 7, may be of the practice at the time. This is reported in
susceptible to liquefaction. NCEER (1997) and Youd et al. (2001).
Youd et al. (2001), quoting Robertson and Wride While there have been some further developments
(1998) suggest additional criteria if cone penetration since 1996, they have not significantly altered the
tests are being used. method detailed in Youd et al (2001).
From the above data it is concluded that: The Youd et al (2001) approach has some important
modifications compared to the Guideline and Seed
(a) Soils within the boundaries for flow and De Alba (1986), and it is the author’s opinion
liquefaction in Figure 4 are susceptible to that given the status of that paper, it should now be
flow liquefaction. adopted by the Profession in Australia.
(b) Finer soils, particularly mine tailings, which However further developments will occur, and the
may have clay size particles which are not literature should be read regularly to seek these.
clay minerals, and dredged fills may also be
subject to liquefaction under earthquake The main changes in the Youd et al (2001) approach
induced deformations. compared to the Guideline are:
(c) Soils which meet all of the following criteria (a) The stress reduction factor rd is shown to be
may be susceptible to liquefaction: uncertain and dependent on the depth below
- The ‘clay’ content (particles smaller than the ground surface.
0.005 mm) is ≤ 15% by weight.
- The liquid limit is ≤ 35%. (b) Corrections are applied to the SPT blow count
- The natural moisture content is ≥ 0.9 for borehole diameter, rod length, and
times the liquid limit. sampler liner as well as hammer efficiency.
(d) ML and SM soils, which have a plasticity In practice only the rod length correction and
index less than 7%, may be susceptible to hammer efficiency should come into the
calculation for conventional sized boreholes (e) There is a significant change to the correction
and SPT samplers with the inner liner in factor for earthquakes of magnitude less than
position. It should be noted that tests on two 7.5.
Australian free fall SPT hammers at Hume
Dam gave efficiencies CE of about 0.6, much Smaller magnitude earthquakes giving the
lower than the 0.8 to 1.3 for USA safety same peak horizontal acceleration are less
hammers with automatic trip-free fall (Youd likely to initiate liquefaction, because the
et al 2001). Why the Hume values were so earthquake will have fewer cycles of motion.
low is not known. The author is of the Larger magnitude earthquakes are more likely
opinion that for significant decision making it to initiate liquefaction. This is allowed for by
should be required that the hammer efficiency using a magnitude scaling factor (MSF) in the
is measured, because there will be a equation for factor of safety (FS) against
significant difference in the likelihood of liquefaction:
liquefaction of the hammer efficiency is 0.8
or 0.9, not the 0.6 usually assumed by default. FS = (CRR7.5/CSR)MSF
The assessment using CE of 0.6 is likely to be
conservative for virtually all Australian SPT where CSR = calculated cyclic stress ratio
hammers and drilling rig set-ups. generated by the earthquake shaking; and
CRR7.5 = cyclic resistance ratio for magnitude
(c) There is a small adjustment to the SPT clean- 7.5 earthquakes.
sand cyclic resistance ratio curve for
magnitude 7.5 earthquakes. Curves are also Youd et al. (2001) present a discussion on
presented for Cone Penetration Tests. MSF and conclude that the original Seed and
Idriss (1982) factors are too conservative for
(d) There are new equations for fines content M < 7.5. They recommend MSF in the range
correction. shown in Figure 5 for M > 7.5.
Figure 5 – Magnitude scaling factors (MSF). Earthquake use “range of recommended MSF from NCEER
workshop” for M < 7.5; and Idriss for M > 7.5 (Youd et al. 2001, from Youd and Noble, 1997).
Liu et al. (2001) publish some additional (g) The NCEER workshop participants concluded
information relating to MSF. Their studies that “although curves relating Kα to α have
show MSF lower than the NCEER workshop been published (Harder and Boulanger, 1997),
values for M < 7.5. Given this, it would seem these curves should not be used by
prudent to use the lower bound of the NCEER nonspecialists in geotechnical earthquake
workshop values in Figure 5, and, for critical engineering or in routine engineering
studies, consider lower values after reviewing practice”.
Liu et al. (2001) and the most recent literature.
Harder and Boulanger (1997) suggested the
Given that in the Australian context, most of use of Figure 6 for soils with an effective
the seismic hazard is for earthquakes M < 7.5. confining stress less than 300 kPa. They point
This is a very important change, and will out that at significantly higher confining
reduce the estimated annual probability of stresses, sandy soils will be more contractive
liquefaction compared to the Seed and Idriss and low Kα may be appropriate. For high risk
(1982) factors given in the Guidelines. and critical projects on sloping ground or
where there are high initial static shear
(f) Youd et al (2001) provide a new graph for stresses, they recommend site-specific
estimating the overburden stress correction laboratory testing using high quality samples.
factor Kσ where in-situ effective stresses are
greater than 100kPa. In Figure 6:
α = τs/σ′vo
It should be noted that there is considerable
scatter in the input data to derive these curves. where τs = static shear stress; σ′vo = effective
They have been described as “minimal, or vertical stress.
conservative estimates” by Hynes and Olsen
(1999) and Youd et al. (2001), but
considerable care should be taken in their
application.
Figure 6 – Sloping ground correction factor Kα suggested by Harder and Boulanger (1997).
It is clear there is considerable uncertainty in There was a consensus that empirical corrections
the estimation of Kα. It seems that Kα will be relating Su(LIQ) to back-analysed field failures
less than 1.0 (meaning liquefaction is more represented an economical and reasonable means
likely) for relative densities lower than about for estimating the liquefied shear strength.
50%, so, marginal cases should be assumed
liquefiable if there are existing shear stresses, NSFW (1998) question the use of fines content
or laboratory tests and analysis of stresses corrections and recommend momentum effects be
should be carried out to assess the effects. accounted for in the analysis of case studies.
Idriss and Boulanger (2003) and Boulanger
(2003a, 2003b) present later information and Olsen and Stark (2002) have followed this
readers should review the latest literature. procedure in the analysis of 33 case studies of flow
liquefaction (excluding lateral spreading as
recommended by NSFW, 1998). They demonstrate
4.3 Post Liquefaction Shear that it appears that the Stark and Mesri (1992)
recommendation was generally conservative.
Strength
Given this, it is the author’s opinion that the Olsen
The Guideline recommends the use of the Stark and and Stark (2002) method should be adopted. This
Mesri (1992) method for assessing the residual uses (N1)60 and qc1 values (for SPT and CPT),
undrained strength, and this has been the common corrected to 100 kPa overburden stress, but with no
approach in Australia. fines content correction. It is also based on the mean
(N1)60 or qc1 for the liquefied zone, not the minimum.
The Stark and Mesri (1992) method is based on
laboratory tests done by a number of researchers. Figures 7 and 8 show the back-analysed liquefied
Their recommended equation is:
shear strength ratio (Su(LIQ)/σvo′) where σ′vo is the
pre-failure vertical effective stress for CPT and SPT.
Sus/σvo′ = 0.0055 (N1)6oCS
Olsen and Stark (2002) indicate that the average
Where (N1)6oCS = SPT ‘N’ value corrected to 100
trend lines which are described by:
kPa effective overburden stress,
and 60% energy ratio, for clean Su(LIQ)/ σvo′ = 0.03 + 0.0143 (qci) ± 0.03, for qc1 <
sand. 6.5 mPa
Figure 7 – Liquefied shear strength ratio Su(LIQ)/ σvo′ based on normalized cone penetration test qc1 (Olsen and
Stark, 2002).
Figure 8 – Liquefied shear strength ratio Su(LIQ)/σvo′ based on normalized standard penetration test (N1)60, (Olsen
and Stark, 2002).
their view, at least for silty sands with > 12% fines,
it is reasonable to allow for the increase in Su(LIQ Baziar and Dobry (1995), Ishihara (1993) and
which would be indicated by the increase in σvo′ Cubrinovski and Ishihara (2000a, b) have
from the berm. For high risk projects they suggest investigated the boundary of flow liquefaction
laboratory consolidation tests be carried out to conditions. Figure 9 summarizes the outcomes.
confirm that it is parallel to the steady state line Readers should read the original papers before
applying this figure to serious decision making.
which is implicit in the assumption that Su(LIQ)/σvo′
is constant. Based on NSFW (1998) it would appear Saturated soils which are potentially liquefiable, but
unwise to allow for the strength increase from the which have not been subject to sufficient cyclic
increased effective stress of the berm for clean shear stress to reach a liquefaction condition, will
sands, and the Su(LIQ) obtained without the berm develop positive pore pressures during the cyclic
should be assumed to apply after the berm is built, loading of the earthquake. The amount of positive
unless sampling and laboratory testing indicates pore pressure will depend on many factors relating
otherwise. to the soil and the cyclic loading.
Whether the mean lines or upper and lower lines are For assessment of the post earthquake factor of
used for design, is for designers to assess, based on safety the pore pressure in these zones can be
the amount and quality of data available and the estimated from Figure 10, which relates the residual
consequences of failure. It would be wise to err excess pore pressure ratio Ru, at the end of the
somewhat on the conservative side given the earthquake loading to the factor of safety against
uncertainty in the estimated strengths. liquefaction, FS calculated from the simplified
method.
It should be noted that the residual undrained
strength (Sus) or liquefied shear strength (Su(LIQ))
only applies to contractive soils which are subject to
flow liquefaction.
Figure 9 – Comparison of flow liquefaction boundaries in terms of SPT (N1)60 for sands and silty sands from
monotonic laboratory undrained tests and earthquake triggered field case (Hunter and Fell, 2003a, b).
Figure 10 – Method for approximate estimation of the residual excess pore pressure Ru in non liquefied zones
(Marcuson et al., 1990).
This procedure is used by USBR and US Corps of recommends the US Corps of Engineers (1984)
Engineers. For more critical projects, carefully method be used as a screening tool.
planned and executed cyclic laboratory tests should
be used. This is virtually a waste of time and money in
Australian conditions where the MDE is likely to be
5. ANALYSIS OF SPT AND CPT less than 0.2 to 0.3g, so the pseudo static factor of
safety will be below 1.0. In reality it is the potential
DATA for deformations leading to loss of freeboard, and
cracking, leading to internal erosion and piping
It is the author’s opinion that it is best to analyse the
which are critical.
available SPT and CPT data on an individual test
(for SPT) or typical test values (for CPT) in each
Swaisgood (1998) gathered data on crest settlement,
borehole/probe. This allows the estimation of the
dam height, dam type, depth of alluvium in the
AEP of the earthquake peak ground acceleration at
foundation, earthquake magnitude and peak ground
which liquefaction will occur for each data point.
acceleration, and the focal distance of the dam to
These can in turn be contoured so that it can be
the earthquake.
readily seen when a viable post earthquake
mechanism through flow liquefied soil can be
Figure 11 shows relative crest settlement
developed.
(settlement/{dam height + thickness of alluvium in
the foundation}) versus peak ground acceleration
6. SEISMIC STABILITY AND (bedrock).Swaisgood (1998) recommended the
DEFORMATION ANALYSIS following equations to predict settlement:
OF EMBANKMENT DAMS
CS = SEF x RF
Figure 11 – Relative crest settlement versus peak ground acceleration (Swaisgood, 1998).
in which M is the magnitude of the earthquake (ML deformation to lead to overtopping. Eg. a 30 metre
or local magnitude below 6.5, and MS or surface high dam with 2m freeboard could withstand peak
wave magnitude at 6.5 or above), and PGA is the ground acceleration greater than 0.7g.
peak horizontal ground acceleration at the damsite
as a fraction of the acceleration due to gravity. This has also been demonstrated on a number of
projects by more detailed SHAKE/Newmark type
RF = 2.0 D-0.35 for earthfill dams analyses.
= 8.0 D-0.35 for hydraulic fill dams
= 0.12 D0.61 for rockfill embankments 6.2 Cracking Under Earthquake
in which D is the distance between seismic energy
Loading
source and dam, in kilometres.
It is suggested that the method of Pells and Fell
This method should only be used to give an (2003) be adopted to assess the likelihood of
approximate estimate of settlements, where there is transverse cracking. This is approximate but is
no potential for liquefaction or significant strain based on a fairly extensive database. Note that for
weakening. Account should be taken of the scatter embankments with foundation profiles leading to
in the data as shown in Figure 11 and in practice it differential settlement across the valley, cracking
is probably best just to use Figure 11. may already be present, or may initiate at low
What is clear is that for dams with normal freeboard seismic loading.
and without strain weakening clays, or liquefaction
issues, it is very unlikely there will be enough
Baziar, M.H. and Dobry, R. (1995). Residual Ishihara, K. (1985). Stabilityth of natural deposits
strength and large scale deformation potential of during earthquakes. Proc. 11 ICSMFE. Balkema,
loose silty sands. J. Geotechnical Engineering, Rotterdam, 321-376.
ASCE, Vol. 121, No.12, 896-906.
Ishihara, K. (1993). Liquefaction and flow failures
Boulanger, R.W. (2003(a)). Relating Kα to a during earthquakes. Geotechnique, Vol. 33, No. 3,
relative state parameter index. Accepted for 351-415.
publication in ASCE J. Geotechnical and
Geoenviromental Engineering. Liu, A.H., Stewart, J.P., Abrahamson, N.A. and
Moriwaki, Y. (2001). Equivalent number of
Boulanger, R.W. (2003(b)). High overburden stress uniform cycles for soil liquefaction analysis. J.
effects in liquefaction analysis, accepted for Geotechnical and Geoenvironmental Engineering,
publication in ASCE J. Geotechnical and ASCE, Vol. 127, No. 12, 1017-1026.
Geoenvironmental Engineering.
Marcuson, W.F., Hynes, M.E. and Franklin, A.G.
CDSA (1994). Canadian Dam Safety Association, (1990). Evaluation and use of residual strength in
Interim dam safety guidelines. seismic safety analysis of embankments.
Earthquake Spectra, Vol. 6, No. 3, 529-572.
Collins, S.A. and Tjoumas, C. (2003). FERC
oversight and input, Saluda dam seismic evaluation NCEER (1997). Proceedings of the NCEER
and remediation, Part 1, Proc. 23rd USSD workshop on evaluation of liquefaction resistance of
Conference, Charleston. US Society on Dams, 531- soils. Editors Youd, T.L. and Idriss, I.M. National
558. Center for Earthquake Engineering Research
technical report NCEER – 97-0022.
Cubrinovski, M. and Ishihara, K. (2000a). Flow
potential of sandy soils with different grain NSFW (1998). Shear strength of liquefied soils.
compositions. Soils and Foundations, 40 (4): 103- Proceedings of National Science Foundation
119. Workshop, Editors T.D. Stark, S. Molson, S.L.
Kramer and T.L. Youd. National Science
Cubrinovski, M. and Ishihara, K. (2000b). Flow Foundation, Earthquake Hazard Mitigation
potential of sandy soils. Proc. GeoEng. 2000, Program.
Technomic Publishing, Lancaster, 549-550.
Olsen, S.M. and Stark, T.D. (2002). Liquefied
Harder, L.F. Jr. and Boulanger, R.W. (1997). strength ratio from liquefaction flow failure case
“Application of Kα correction factors”. Proc. histories. Canadian Geotechnical Journal, Vol. 39,
NCEER Workshop on Evaluation of Liquefaction No. 3, 629-647.
Resistance of Soils, Nat. Ctr. For Earthquake Engrg.
Pells, S. and Fell, R. (2003). Damage and cracking Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I.,
of embankment dams by earthquake and the Castro, G., Christian J.T., Dobry, R., Finn, W.D.L.,
implications for internal erosion and piping.
Proceedings 21st International Congress on Large Harder, L.F., Hynes, M.E., Ishihara, K., Koester,
Dams, Montreal. ICOLD, Paris Q83-R17, J.P., Liao, S.S.C., Marcuson, W.F., Martin, G.R.,
International Commission on Large Dams, Paris. Mitchell, J.K., Moriwaki, Y., Power, MS.,
Robertson, P.K., and Wride, C.E. (1998). Robertson, P.K., Seed, R.B., and Stokoe, K.H.
Evaluating cyclic liquefaction potential using the (2001). Liquefaction resistance of soils: summary
cone penetration test. Can. Geotech. J., Ottawa, report from the 1996 NCEER and 1998
35(3), 442-459. NCEER/NSF Workshops on evaluation of
Seed, H.B. and De Alba, P. (1986). Use of the SPT liquefaction resistance of soils. J. Geotechnical and
and CPT tests for evaluating the liquefaction Geoenvironmental Engineering, ASCE, Vol. 127,
resistance of sands. In-Situ 86, Conference on use No. 10, 817-834.
of in-situ tests in geotechnical engineering.
Clemence, S.P. (ed). ASCE Geotechnical Special
Publication, No. 6, 281-302.
Seed, H.B. and Idriss, I.M. (1982). Ground motions
and soil liquefaction during earthquakes.
Monograph series, Earthquake Engineering
Research Institute, Berkeley, California.
Stark, T.D. and Mesri, G. (1992). Undrained shear
strength of liquefied sands for stability analysis.
JASCE Geotechnical Engineering, Vol. 118, No.
11, 1727-1747.
Swaisgood, J.R. (1998). Seismically-induced
deformation of embankment dams. 6th US National
Conference on Earthquake Engineering, Seattle,
Washington.
US Corps of Engineers (1984). Rationalising the
seismic coefficient method. Miscellaneous Paper
GL84-13.
USBR (1989). Design Standards. Embankment
dams, No. 13, Chapter 13. Seismic design and
analysis. US Bureau of Reclamation, Denver, Co.
USNRC (1985). Liquefaction of soils during
earthquakes. National Academy Press, Washington
DC.
Wang, W. (1979). Some findings in soil
liquefaction. Water Conservancy and Hydroelectric
Power Research Institute, Beijing.
Wang, W. (1981). Foundation problems in a
seismic design of hydraulic structures. Proceedings
joint US – PRC microzonation workshop, Harbin,
China, 15-1 to 15-13.
Youd, T.L., and Noble, S.K. (1997). Liquefaction
criteria based on statistical and probabilistic
analyses. Proc., NCEER Workshop on Evaluation
of Liquefaction Resistance of Soils, Nat. Ctr. for
Earthquake Engrg. Res., State Univ. of New York at
Buffalo, 201-215.