The Statistics of Embankment Dam Failures and Accidents

Color profile: Generic CMYK printer profile
Composite Default screen
1000
The statistics of embankment dam failures and

accidents
Mark Foster, Robin Fell, and Matt Spannagle
Abstract: The paper describes the results of a statistical analysis of failures and accidents of embankment dams, specifically concentrating on those incidents involving piping and slope instability. The compilation of dam incidents includes details on the characteristics of the dams, including dam zoning, filters, core soil types, compaction, foundation
cutoff, and foundation geology. An assessment of the characteristics of the world population of dams was also carried
out. By comparing the characteristics of the dams which have experienced failures and accidents to those of the population of dams, it was possible to assess the relative influence of particular factors on the likelihood of piping and
slope instability.
Key words: dams, failures, piping, instability database.
Rsum : Cet article dcrit les rsultats dune analyse statistique des ruptures et accidents dans les barrages en terre,
se concentrant spcifiquement sur ces incidents impliquant la formation de renard et linstabilit des talus. La compilation des incidents de barrages inclut des dtails sur les caractristiques des barrages incluant le zonage du barrage, les
filtres, les types de noyau, le compactage, le rideau dtanchit de fondation, et la gologie de la fondation. Une
valuation des caractristiques de la population mondiale des barrages a galement t ralise. En comparant les
caractristiques des barrages qui ont t affects par des ruptures et des accidents avec celles de la population des barrages, il a t possible dvaluer linfluence relative que des facteurs particuliers ont sur la vraisemblance de renards et
de linstabilit des talus.
Mots cls : barrages, ruptures, renard, base de donnes sur linstabilit.
[Traduit par la Rdaction]
Foster et al.
1024
Introduction
Embankment dam engineering has evolved over many
centuries, with the major developments occurring since the
1940s with the development of soil mechanics and
geotechnical engineering. Some aspects are now readily analysed, e.g., the stability of the embankment slopes. Others,
e.g., piping failure through a dam foundation, remain more
difficult to quantify, and the measures taken in design and
construction are more experience based. It is particularly difficult to assess the safety of dams which do not meet modern
design and construction criteria, e.g., dams with no or inadequate filters.
Recognising the value of the historic performance of dams
in assessing dam safety, the International Commission on
Large Dams (ICOLD) has carried out extensive surveys of
dam incidents (ICOLD 1974, 1983, 1995). The ICOLD surveys are for large dams, a large dam being defined as a dam
which is more than 15 m in height (measured from the lowest point in the general foundations to the crest of the dam)
Received February 5, 1999. Accepted February 10, 2000.
Published on the NRC Research Press website on October 6,
2000.
M. Foster. URS, Level 3, 116 Miller St., North Sydney,
Australia 2060.
R. Fell. School of Civil and Environmental Engineering,
University of New South Wales, Sydney, Australia 2052.
M. Spannagle. Department of Land and Water Conservation,
GPO Box 39, Sydney, Australia 2001.
Can. Geotech. J. 37: 10001024 (2000)
I:\cgj\Cgj37\Cgj05\T00-030.vp
Monday, October 02, 2000 2:03:04 PM
or any dam between 10 and 15 m in height which meets one

of the following conditions: (i) the crest length is not less
than 500 m, (ii) the capacity of the reservoir formed by the
dam is not less than 106 m3, (iii) the maximum flood discharge dealt with by the dam is not less than 2000 m3/s, or
(iv) the dam is of unusual design.
ICOLD carried out analyses of the data compiled to determine the most common cause of dam incidents. Others, including USCOLD (1975, 1988), USCOLD Committee on
Dam Safety (1996), ANCOLD (1992), Charles and Boden
(1985), Olwage and Oosthuizen (1984), and Gomez et al.
(1979), have compiled data on incidents for various countries. There have been attempts to use the statistical analysis
of dam incidents to predict the likelihood of failure of dams,
including those by Silveira (1984, 1990), Blind (1983),
Serafim (1981a, 1981b), Tavares and Serafim (1983), Ingles
(1988), Gruner (1963, 1967), and Von Thun (1985). All of
these analyses are limited to the statistics of height, year of
construction, and only basic descriptions of the dam type.
For example, the embankment zoning classification is restricted to two categories in the ICOLD dam incident and
dam population databases, namely earthfill embankment
(TE) and rockfill embankment (ER).
Embankment zoning would be expected to have a significant influence on the likelihood of failure, particularly for
structural modes of failure which include slope stability and
piping. As is well recognised in dam engineering, and described in Fell et al. (1992), different embankment zoning
types have varying degrees of control of embankment seepage. These give varying degrees of control of the potential
2000 NRC Canada

Foster et al.
for piping failure through the embankment and foundation

and the pore pressures which affect slope stability.
This paper presents the results of a statistical analysis of
embankment dam incidents, specifically concentrating on internal erosion and slope instability. While the data can be
used in a quantitative risk assessment (QRA) framework,
they also provide useful insights into the factors which contribute to dam incidents and can therefore be used in a nonQRA, dam safety context. The study has been done as part
of a larger research project studying methods for estimating
the probability of failure of embankment dams for use in
QRA. This paper describes only part of the research project.
Foster et al. (2000) describe the application of these data to
estimating the relative likelihood of failure of embankment
dams by internal erosion and piping, Foster and Fell (2000a)
discuss the assessment of filters which do not satisfy modern
design criteria, and Foster and Fell (2000b) use event trees
to estimate the probability of failure of embankment dams
by internal erosion and piping. All components are described
in Foster (1999).
The principal components of the study were as follows:
(i) extension of the existing compilations of dam incidents to
include more details on embankment zoning, including the
presence or absence of filters, foundation geology, and embankment material characteristics such as core soil types and
compaction; and (ii) analysis of the dam incident database
and comparison to a dam population database to estimate
historic frequencies of failure for different modes of failure
and dam zoning types and identify factors, such as foundation geology types and core embankment characteristics, that
have an influence on the likelihood of embankment dam failure for piping and slope instability modes of failure.
Establishment of databases
Compilation of dam incident database
A list of dam incidents was compiled primarily from the
three ICOLD studies (ICOLD 1974, 1983, 1995) supplemented with additional incidents from the other existing
compilations, from the literature, and from the project sponsors. The criteria set for the selection of the dam incidents to
be entered into the database are (i) embankment dam failures
for all modes of failure for large dams; (ii) failures of embankment dams (not necessarily large) by piping and slope
instability; and (iii) accidents involving piping, slope instability, and seepage. The definitions of failures, accidents,
and incidents used are consistent with ICOLD (1983).
The first criterion was used to keep the dam failure and
dam population datasets consistent for proper statistical
analysis. The other two criteria were set to maximise the
amount of data for the detailed analysis of piping and slopeinstability failures.
Data on the dam and incident details were obtained from
(i) incident descriptions in ICOLD (1974, 1983) and other
compilations of dam incidents; (ii) published data from an
extensive search through the literature; and (iii) reports collected from the sponsor organisations and from the United
States Bureau of Reclamation (USBR), British Columbia
Hydroelectric and Power Corporation (BC Hydro), and the
Norwegian Geotechnical Institute.
1001
Information on the dam and incident details was extracted

from the data gathered and entered into a database called
ERDATA1.
The ERDATA1 database is divided into seven main categories: (i) dam details, e.g., dam name, country, height, year
constructed; (ii) dam zoning category, including dam zoning
and description of filters; (iii) foundation cutoff category;
(iv) foundation geology; (v) earthfill core characteristics;
(vi) incident details; and (vii) references. Sketches of the
dam zoning categories used are shown in Fig. 1. A full description of the variables in the ERDATA1 database is given
in Foster et al. (1998) and Foster (1999).
Population of embankment dams database
The population of embankment dams database is required
to determine whether an overrepresentation of a particular
dam characteristic in the incident cases, such as a particular
zoning type, is due to this characteristic being common in all
existing dams, or whether dams with this characteristic tend
to be more susceptible to dam incidents. The data from the
population of dams database are combined with those from
the incident database to estimate the frequencies of failure
and accidents for the various dam zoning types.
The ERDATA1 classification system has been used for
collating data on the population of dams, so the databases
are consistent. The ideal situation for a sound statistical
analysis would be to have the characteristics of the
ERDATA1 classification system for all the existing embankment dams as listed in the World register of dams (ICOLD
1984). However, there are insufficient data in the ICOLD
register to obtain data on dam zoning, foundation geology,
and other characteristics in the ERDATA1 classification system, so it was necessary to select sample populations of embankment dams in an attempt to represent the characteristics
of the world population. Sample populations of embankment
dams which were used include 356 dams in Australia, 44 in
New Zealand, 246 in the United States (from the USBR),
174 in Norway, and 642 described in papers in the ICOLD
congresses up to 1982, giving a total of 1462 embankment
dams, or about 13% of the total population.
These datasets were primarily selected due to the availability of sufficient data required for the ERDATA1 classification system. Information for the Australian and New
Zealand dams was obtained from the project sponsors and
from questionnaires sent out to dam owning authorities. The
zoning characteristics for the dams in the United States were
obtained from USBR (1994) and the foundation geology
characteristics from the Safety Evaluation of Existing Dams
(SEED) databooks held at the USBR office in Denver, Colorado. Data on the Norwegian dams were limited to the dam
zoning category which was obtained from K. Senneset (personal communication, 1996). Dams described in the ICOLD
congress papers from 1933 to 1982 were used to obtain a
representative set of dams to take account of the range of
trends in dam construction with time and for various countries.
Information was also collected from the literature, providing additional data on the distribution of dam zoning categories
in different parts of the world and describing general trends
in dam design. Literature sources included Snethlage et al.
(1958), Leps et al. (1978), Building Research Establishment
2000 NRC Canada

1002
Can. Geotech. J. Vol. 37, 2000
Fig. 1. Dam zoning categories.
(1990), Sherard (1953), ICOLD (1984), Skempton (1990),

Cooke (1984), ICOLD (1989), and Schnitter (1994).
The following approach was used for each of the main
categories: (i) dam zoning; (ii) foundation geology; and
(iii) core material geological origin, soil classification, and
compaction.
Dam zoning
The general approach used to make the estimates of the
world distribution of zoning types was to subdivide the data
into tables according to the country, dam height range
(<20 m, 2029 m, 3049 m, and >50 m), and construction
year period (before 1900, 19001929, 19301949, 195069,
and 19701986).
The dam zoning distribution estimates were made for
Australia, France, India, Japan (post-1950 only), New Zealand, Norway, United Kingdom, and United States. These
countries had sufficient data in the sample population available to give reasonably reliable estimates of the zoning characteristics of the population. The dams from remaining
countries were grouped into a category called other countries. Dams constructed in China, and in Japan prior to 1930,
were excluded from the analysis of the population due to the
lack of information in the literature on these dams and low
reported failure rates despite them making up a significant
proportion of the dam population.
The number of earthfill (TE) and rockfill (ER) dams for
each of the countries, construction periods, and height categories were obtained from the ICOLD register (ICOLD
1984). The information from the sample populations and
data gathered from the literature were used as a basis to
make estimates of the number of dams for each of the dam
zoning categories. A considerable degree of judgement was

required to make the estimates, but this was facilitated by
breaking down the data into the smaller units of construction
years and dam height ranges. The analysis was by a trial and
error process in which the estimated percentages were modified so that they reflected the expected trends in dam design
with time and dam height.
Table 1 shows the assessed distribution of dam zoning for
the world population of dams accounting for construction
period. Rockfill dams, comprising zoned earth and rockfill
dams, central core earth and rockfill dams, concrete face
rockfill dams, and rockfill with corewall dams, make up 21%
of the world population based on the dam zoning categories.
This is significantly higher than the proportion of rockfill
dams (ER), namely 9%, given by the ICOLD register
(ICOLD 1984). This difference is attributed to the differences in classification of rockfill dams of the ERDATA1 and
ICOLD systems. Zoned earth and rockfill dams with less
than 50% rockfill by volume are classified as TE by ICOLD.
Foundation geology
The distribution of foundation geology types was assumed
to be dependent on the spatial distribution of dams. Estimates were made for countries where sufficient data were
available from the sample population, namely United States,
India, United Kingdom, Canada, Australia, and New Zealand. This subset of countries was carried through into the
analysis of the dam incidents when comparing the distribution of foundation geology types of the accident and failure
cases to the population. The distributions of geology types
from the sample population database were determined for
each of the countries and were used as a basis together with
2000 NRC Canada

Foster et al.
1003
Table 1. Estimated dam zoning (%) for the world population of embankment dams by construction period.
Zoning category
Before 1900
19001930
19301950
19501970
19701986
All years
Homogeneous earthfill
Earthfill with filter
Earthfill with rock toe
Zoned earthfill
Zoned earth and rockfill
Central core earth and rockfill
Concrete face earthfill
Concrete face rockfill
Puddle core earthfill
Earthfill with corewall
Rockfill with corewall
Hydraulic fill
Number of embankment dams
16
0
5
7
0
0
5
1
58
5
0.5
2
370
14
1
5
18
7
0
4
5
24
11
3
8
819
16
11
6
37
8
5
5
2
4
4
1
1
1167
9
18
7
37
10
8
5
2.5
0.5
2
1
0
4436
6
18
6
40
10
12
3
3
0
1
1
0
4400
10
15
6
36
9
8
4
3
5
2
1
1
11 192
the use of geology maps of the countries to make estimates

of the percentage distribution of foundation geology types.
Core material geological origin, soil classification, and
compaction
Estimation of the distribution of the geological origin of
the core material was made on the assumption that the core
geology would be influenced by the regional geology and
was therefore analysed by country. The distributions of core
geological origin types from the population database and the
foundation geology distributions described above were used
as a basis to make estimates of the core geology distributions for United States, Australia, New Zealand, United
Kingdom, Canada, India, Norway, and other countries. The
final world distribution was then obtained by applying a
weighted sum of the distributions from the countries based
on the number of large embankment dams in each country in
1982 (from ICOLD 1984).
The distribution of core soil types, classified by the
Unified Soil Classification System (USCS), was determined
directly from the percentages of core soil types from the
population database. It was assumed that there were sufficient data in the population database to give a representative
sample of core soil types. Typically, the core material comprises more than one soil type and so the percentages in the
distribution do not necessarily sum to 100%. The percentages were normalised such that they sum to 100% for comparison with the incident statistics.
Estimates of the distribution of the degree of compaction
of the core materials were made using the population database, assuming it to be dependent mainly on the period of
construction. Hydraulic fill and puddle core dams were not
included in the analysis because their compaction is inherent
in the dam zoning. Information provided by Sherard (1953)
and Skempton (1990) was used to adjust the database values.
An attempt was made to estimate the relative abundance
of the presence of dispersive soils in the cores of embankment dams due to their prominence in the piping failure
cases. There was insufficient information in the population
database on which to make such an estimate and so information from the literature was used. ICOLD (1990) lists parts
of the world which have experienced problems with
dispersive soils, and combined these countries have approximately 35% of large embankment dams. There is no basis on
which to estimate the proportion of dams in these countries

with dispersive soils actually present in the core, but the
value is probably less than 25%. If it is assumed that say 5
10% of the dams in these countries have dispersive soils
present in the core, then a very approximate estimate of the
proportion of dams in the world with dispersive soils is 2
4%.
Other dam characteristics
Estimates for the distributions of the other dam characteristics, such as foundation cutoff details, embankment and
foundation filters, and the location of conduits, were determined directly from the population database. Details are
given in Foster et al. (1998) and Foster (1999). The statistics
of year of construction, cumulative dam embankment years,
and dam heights were calculated from ICOLD (1984, 1983)
and used in the analysis. More details on the databases are
given in Foster (1999) and Foster et al. (1998).
Analysis methodology
There are two components of the analysis of the
ERDATA1 database, namely analysis of the overall statistics
of failure, and detailed analysis of piping and slope-instability
failures.
Mode of failure
The philosophy of the analysis of the ERDATA1 data was
to categorise dam accidents and failures into modes of failure as opposed to causes of failure. This is compatible with
the methods used in event-tree analysis.
The failure mode categories used are flood overtopping,
gatespillway failure, piping, slope instability, and earthquake. Piping failures were further subdivided into piping
through the foundation and piping from the embankment
into the foundation. Slope-instability failures were subdivided into upstream slides and downstream slides.
Dam incidents sometimes involve more than one failure
mode, so for example, development of piping through the
embankment may cause saturation of the downstream slope
which then initiates a downstream slide. In these cases, all
the modes of failure that were involved in the dam incident
are assigned in the database.
2000 NRC Canada

1004
Fig. 2. Stages of development of piping failure.
Overall statistics of failure

The frequencies of failure are estimated from the number
of dam incidents compared with either the total number of
large (i.e., >15, high) embankment dams (or dams of a particular zoning) up to 1986 to give the average frequencies of
failure over the life of the dam, or the total number of embankment dam-years, allowing for the estimated average age
of each zoning type (up to 1986) to give annual frequencies
of failure.
Analysis of piping and slope instability incidents
This analysis involved comparing the frequency of occurrence of the dam characteristics such as dam zoning type,
foundation geology, and embankment core type in the dam
incidents to that in the dam population. An overrepresentation of a particular feature, such as a particular dam
zoning or foundation geology type, in the dam incidents relative to the dam population suggests dams with that particular feature are possibly more vulnerable to that failure mode.
The analysis method utilises the concept that piping and
slope-instability failure modes can be broken down into several stages of development. Typically these stages are taken
to be initiation, progression, and breaching, as shown in
Fig. 2 for piping through the embankment. This concept has
been suggested by several authors, including Von Thun
(1996). Accidents involve initiation of piping, but the progression stage is limited and breaching does not occur.
Therefore, by comparing the characteristics of the accident
cases to the failure cases of a particular failure mode, it may
be possible to identify factors that influence the progression
stage of the failure mode.
The analysis of the data has kept all failures as one
dataset, rather than, for example, separating dams by zoning,
separating first-filling failures from those which occur later,
or separating failures due to piping around conduits from
other piping failures. An initial assessment of the data
showed there did not appear to be big differences in the geological and core material characteristics for these sets, but
this was not proven in a rigorous statistical way. Separation
of the data had the major disadvantage that already small
samples (of the dams which had failed or had accidents)
would become smaller.
Overall statistics of failure of embankment

dams
Table 2 gives the overall statistics of failure for all failure
modes, separating for all failures and failures during opera-
tions. The historical average frequency of failure of large

embankment dams is estimated to be 1.2% over the life of
the dam (136 embankment dam failures out of 11 192 large
embankment dams constructed up to 1986, excluding China
and Japan pre-1930). This reduces slightly to 1.1% over the
life of the dam for dam failures occurring only while the
dam was in operation. The historical annual probability of
failure of large embankment dams is estimated to be 4.5
104 per dam per year (136 embankment dam failures in an
estimated 300 400 embankment dam-years up to 1986). This
reduces slightly to 4.1 104 per dam year if failures occurring during construction are excluded. These figures would
reduce by about 30% if none of the dams in the 11 192 had
failed up to 1999.
Table 3 presents the statistics of failure by failure mode
and dam zoning. These overall statistics are useful if one
makes the assumption that the performance of dams in the
past is a reasonable prediction of what may happen in the future. It is apparent from the statistics combined with the conventional understanding of dam stability and piping that
factors such as dam zoning and core material properties have
an influence on the likelihood of failure or accidents and can
be used to get an idea of which dams are more or less likely
to experience stability and piping problems.
An analysis was carried out of the frequency of embankment dam failures for each mode of failure for the failures
occurring before and after 1950 (excluding failures during
construction). The failure statistics for dams constructed before and after 1950 are shown for all modes of failure and
structural modes of failure in Table 4. Structural modes of
failure are those involving piping, slope instability, or earthquake. The analysis showed the proportion of failures by
piping increases from 43% before 1950 to 54% after 1950.
Over the same period, the proportion of failures by flood
overtopping and appurtenant works modes of failure decreases from 53% to 41%. There is a significant reduction in
the proportion of failures due to sliding with time, reducing
from 7% before 1950 to only 1.5% after 1950.
The following sections present the outcomes of further
analysis of the incidents for piping and slope instability to
further assess the factors affecting the frequency of incidents
of piping through the embankment and slope instability. The
population database consists of over 11 000 dams, with over
300 000 dam-years of performance, so the analysis of the relationship of zoning to incidents is based on a large sample.
For the assessment of other factors, e.g., core material properties, there is a greater reliance on the characteristics of the
dams that have experienced incidents, and the sample size is
2000 NRC Canada

Foster et al.
1005
Table 2. Overall failure statistics for large embankment dams up to 1986, excluding dams constructed in Japan pre-1930 and in China.
No. of cases
% failures
(where known)
Average frequency of
failure (103)
Mode of failure
All
failures
Failures in
operation
All
failures
Failures in
operation
All
failures
Overtopping and appurtenant

Overtopping
Spillwaygate
Subtotal
46
16
62
40
15
55
35.9
12.5
48.4
34.2
12.8
47.0
4.1
1.4
5.5
3.6
1.3
4.9
Piping
Through embankment
Through foundation
From embankment into foundation
Subtotal
39
19
2
59
38
18
2
57
30.5
14.8
1.6
46.1
32.5
15.4
1.7
48.7
3.5
1.7
0.18
5.3
3.4
1.6
0.18
5.1
Slides
Downstream
Upstream
Subtotal
Earthquakeliquefaction
Unknown mode
Total no. of failures
Total no. of failures where mode of failure known
No. of embankment dams
6
1
7
2
8
136
128
11 192
4
1
5
2
7
124
117
11 192
4.7
0.8
5.5
1.6
3.4
0.9
4.3
1.7
0.54
0.09
0.63
0.18
0.36
0.09
0.45
0.18
Failures in
operation
12.2 (1.2%)
11.1 (1.1%)
Note: Subtotals and totals do not necessarily sum to 100%, as some failures were classified as multiple modes of failure.
smaller. For this reason it has been necessary to group all

the dams experiencing incidents together, rather than keeping them separate, for this part of the analysis.
Factors affecting the frequency of incidents

of piping through the embankment
Incidents have been classified as piping through the embankment if the incident involved any type of internal erosion process occurring primarily through the embankment
dam. Cases of piping along and into conduits through dams
are included. Cases where piping initiated at the embankmentfoundation contact are not included but analysed separately under piping from the embankment into the
foundation.
Dam zoning
The failure and accident statistics for piping through the
embankment are summarised in Table 5. It is evident that the
dam zoning categories with high average frequencies of failure by piping through the embankment tend to be the zoning
types with inherently poor control of seepage through the
embankment. Homogeneous earthfill dams, which have no
zoning of materials, have the highest frequency of failure,
nearly five times higher than the average of all dams combined. Other dam zoning categories with higher than average
frequencies of failure by this mode of piping are earthfill
with rock toe, concrete face earthfill, and puddle core
earthfill dams. These four zoning categories combined make
up nearly 80% of the failure cases but only 25% of the population. For homogeneous earthfill dams, earthfill dams with
rock toe, and concrete face earthfill dams, there are the same
number or more failures than accidents. It is possible that
many piping accidents have not been reported to the ICOLD
studies or in the literature, but this trend suggests that these

dams are more likely than other dams to fail (i.e., breach)
once piping initiates.
Embankment dams with downstream rockfill zones have a
particularly low incidence of failure due to piping through
the embankment. There is only one dam failure due to piping through the embankment for zoning categories with
downstream rockfill. This was Avalon Dam, a zoned earth
and rockfill dam which failed in 1904. It had no filter between the core and the dumped rockfill. The large number of
piping accidents but no failures of central core earth and
rockfill dams indicates that these dams have a low frequency
of failure because they are less likely to progress to breaching if piping initiates compared with dams with earthfill materials in the downstream zones. Review of the descriptions
of the accidents to rockfill dams suggest that this is probably
due to the inherent stability of the downstream rockfill zones
under large seepage flows.
Some specific points about the failures are as follows:
(1) Failures of homogeneous earthfill dams have generally
been associated with one or more of piping around conduits
passing through the dam (in nine cases), piping through
poorly compacted fill materials (in 12 cases), and piping
through dispersive fill materials (in four cases). The average
frequency of failure of homogeneous earthfill dams constructed prior to 1900 is about 10 times higher than that for
dams constructed after 1950.
(2) The failures of earthfill dams with filter drains have
generally been associated with piping through dispersive fill
materials around outlet conduits passing through the dam (in
three cases) or at the contact with concrete spillway structures (one case). Embankment filters were present in two of
the failures, and in both cases failure was attributed to piping
around the outlet conduit where locally there were no filters
2000 NRC Canada
Mode of failure
Piping
Dam zoning type
9.5
15
6.1
35.9
9.3
8.4
4.1
2.8
4.7
2.4
0.9
0.9
100
23
4
9
7
4
(17)
(2)
(9)
(5)
(3)
4 (1)
4
1
5
11
(4)
(0)
(4)
(3)
0 (0)
5
5
54
136
(3)
(3)
(13)
(66)
% of
failure
cases
28
5
11
9
5
(32)
(4)
(17)
(9)
(6)
Through
embankment
Slope instability
Through
foundation
From
embankment
into foundation
Downstream
slide
Upstream
slide
Earthquake
Overtopping
Spillway
gate
failure
Unknown
14
2
5
4
1
2
0
3
1
0
0
0
1
0
1
1
0
0
0
1
0
0
0
0
0
1
0
0
0
0
6
2
0
2
1
0
0
0
0
0
0
0
0
0
0
(8)
(0)
(8)
(6)
2
0
4
0
3
0
0
2
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
0
4
0
0
1
3
0
0
0
1
0 (0)
6 (6)
6 (6)
0
1
6
39
1
2
5
19
0
0
0
2
2
0
1
6
0
0
0
1
0
0
1
2
0
1
26
46
2
1
9
16
0
0
7
8
5 (2)
5
1
6
13
100 (100)
Note: The values in parentheses refer to statistics for structural modes of failure, comprising piping, slope instability, and earthquake modes of failure. The number of failure cases for the modes of failure do not necessarily sum
to the total number of failure cases because some dams were classified as multiple modes of failure.
2000 NRC Canada
Zoned earthfill
Zoned earth and
rockfill
Central core earthfill
and rockfill
Earthfill with concrete
corewall
Rockfill with concrete
corewall
Hydraulic fill
Other
Unknown
Total
% of
population
No. of
failure
cases

1006
Table 3. Failure statistics for large embankment dams by dam zoning categories (up to 1986).

Foster et al.
1007
Table 4. Average frequency of failure for large embankment dams constructed before and after 1950.
Class of dams
Dams constructed
before 1950
Dams constructed
after 1950
All dams
No. of large embankment dams constructed

No. of large embankment dam failures by all modes of failure
No. of embankment dam failures by structural modes of failure
Average frequency of failure over life of dam
Average frequency of failure over life of dam by structural modes of failure
Average annual frequency of failure by all modes of failure
Average annual frequency of failure by structural modes of failure
2356
88
41
3.8102
1.7102
8.6104
3.6104
8836
48
25
0.5102
0.3102
2.7104
1.6104
11 192
136
66
1.2102
0.6102
4.1104
2.0104
Table 5. Average frequency of failure due to piping through the embankment by dam zoning categories for large dams up to 1986.
Zoning category
No. of
failures
No. of
accidents
Zoned earthfill
Concrete (or other) face earthfill
Concrete (or other) face rockfill
Concrete corewall, earthfill
Concrete corewall, rockfill
Hydraulic fill
Zoning type unknown
All dams
14
2
5
4
1
0 (1)
2
0
4
0
0
0
7
39
9
1
5
9
7
19
1
1 (11)
10
2
2
3
6
75
Average
frequency
of failure
(103)
Average
frequency of
accident
(103)
Average annual frequency of

failure (106)*
First 5 years
of operation
After 5 years
of operation
16.0
1.5
8.9
1.2
1.2
(<1.1)
5.3
(<1)
9.3
(<1)
(<1)
(<1)
9.2
0.6
8.0
2.4
7.3
22.0
2.4
3.5
20.7
8.1
21.6
32.4
2086
189
1160
158
152
(<143)
691
(<130)
1205
(<130)
(<130)
(<130)
188
37
158
25
24
(<34)
75
(<17)
38
(<8)
(<13)
(<5)
3.5
6.7
453
56
*The percentages of failures by piping through the embankment occurring at the different times after construction are as follows: 49% during first
filling, 16% during the first 5 years operation, and 35% after 5 years operation. Calculations of annual frequencies of failure are made as follows: annual
frequency of failure (all years) = (average frequency of failure)/(average age); annual frequency of failure (first 5 years) = (average frequency of
failure)0.65/5; and annual frequency of failure (after 5 years) = (average frequency of failure)0.35/(average age 5).
Upper bound value of the average frequency of failure determined by assuming one dam failure.
Eleven accidents to concrete face rockfill dams involving leakages through the concrete face (not included in % statistics).
Assume average frequency of failure of < 1 103.
provided. For the other three failures, only foundation filters

were present.
(3) Earthfill dams with a rock toe have one of the highest
frequencies of failure by piping through the embankment.
The failure and accident cases have generally been the result
of outlet conduits passing through the dam, piping of the fill
materials into coarse rockfill materials, or piping through
cracks which formed through the dam over irregularities in
the foundation or steep abutments. Zoned earthfill dams have
a relatively low probability of failure; three of the four dams
failed on first filling and the other dam (Walter Bouldin
Dam) failed after 8 years operation. The construction periods for the dams that failed range from 1947 to 1975.
(4) In four of the seven accidents to zoned earth and
rockfill dams, no embankment filter was provided and core
materials were eroded into the downstream rockfill. The piping incidents at McMillan and Scofield dams demonstrated
the high discharge capacity of rockfill zones (in both cases,
dumped rockfill zones). At Scofield Dam, leakage during the
piping incident was estimated to be in the range 1400
5000 L/s through the rockfill. It took 2 days to stabilise the

flow through the dam by placing sandbags on the crest.
(5) There are a total of 21 piping accidents to central core
earth and rockfill dams, 15 of these involving piping of
broadly graded core materials of glacial origin into coarse or
segregated filters. At Matahina and Bullileo dams, no downstream filters were present and the core and transition materials were eroded into the downstream rockfill zone.
(6) In all three cases of failures of concrete face earthfill
dams, piping occurred around the conduits. The cause of
failure was attributed to problems associated with the connection of the conduit to the upstream face in two of the
cases and in the other case as a result of settlement and rupture of the conduit within the embankment.
(7) Concrete (or other impervious) face rockfill dams have
a low incidence of piping through the embankment, with no
failures and only two piping accident cases. Both accident
cases were attributed to internal erosion of the bedding layer
of the upstream face into the rockfill zone. Neither of the
two dams had concrete upstream faces: one was constructed
2000 NRC Canada

1008
Table 6. Geological origin of core material of dams with poor control of erosion experiencing piping incidents.
Dams with poor control of erosion
Dams with good control of erosion
Core geology
No. of
failures
No. of
accidents
No. of
failures
No. of
accidents
Aeolian
Alluvial
Colluvial
Glacial
Residual
Lacustrine
Marine
Volcanic
Unknown
No. of piping incidents
No. of soils
0
9
2
3
10
0
0
0
19
41
24
1
6
0
2
10
0
0
0
28
35
19
0
0
0
0
0
0
0
0
2
2
0
2
2
0
15
2
0
0
0
14
28
21
(0)
(38)
(8)
(12)
(42)
(0)
(0)
(0)
(100)
(5)
(31)
(0)
(11)
(53)
(0)
(0)
(0)
(100)
(0)
(0)
(0)
(0)
(0)
(0)
(0)
(0)
(100)
Population
with core
geology
(10)
(10)
(0)
(70)
(10)
(0)
(0)
(0)
7
22
6
22
38
2
<1
2
(100)
100
Note: Percentages are given in parentheses. Zoning categories with poor control of erosion through the dam are homogeneous earthfill, earthfill with
filter (foundation filter), earthfill with rock toe, zoned earthfill (with no filter), concrete face earthfill, and puddle core earthfill. Zoning categories with
good control of erosion through the dam are earthfill with filter (embankment filter), zoned earthfill (with filter), zoned earth and rockfill, and central core
earth and rockfill.
with a plastic membrane (Martin Gonzalo Dam), and the

other with a bituminous concrete membrane (Scotts Peak
Dam). There are also 11 cases involving leakages through
the concrete face in the dam incident database, but these are
classified as seepage accidents because they did not involve
piping of materials. An upper bound average frequency of
failure of <1.0 103, as shown in Table 5, for concrete face
rockfill dams is selected on the basis that the frequency of
failure would be expected to be lower than those of any of
the other dam zoning categories due to the effect of the
rockfill zone.
(8) Puddle clay core dams have relatively high average
frequencies of failure and accidents due to piping through
the embankment. Of the four failures of large dams, three
are early British dams constructed before 1865 and in each
case failure was attributed to piping of puddle core materials
into permeable shoulder fill following hydraulic fracturing
of the thin puddle cores. After the failure of Dale Dyke in
1864, nearly all puddle core dams constructed in the United
Kingdom used zoning of materials against the puddle clay
core (Skempton 1990). However, it is evident from the accidents that problems still persisted even after this period, but
they generally did not lead to failure of the dam. Deterioration of the dams appears to have been a factor in many of
the accident cases, with more than half of the piping accidents occurring after 50 years of operation. Failure of the
5 m high Clandeboye Dam, which occurred after 80 years of
operation, was attributed to collapse of a wooden conduit
that passed through the embankment. The majority of the accidents to puddle core dams involved either piping of thin
puddle cores into the shoulder fill or piping of fill materials
into cracks in old masonry conduits passing through the
dam. At Yan Yean Dam in Australia, a leak developed
through the puddle core after 92 years of operation due to
the presence of tree roots in the embankment.
(9) Hydraulic fill dams have a low incidence of piping
through the embankment, with no recorded failures and only
4% (three cases) of the accidents to large dams, but the population of these dams is small (less than 1% of the total population). An average frequency of failure of <1.0 103 is
suggested for hydraulic fill dams on the basis that the hydraulic filling process would tend to give good zoning of
materials similar to that of zoned earthfill dams.
(10) Only two dams, Zoeknog and Ghattara, are known to
have failed by piping through the embankment despite an
embankment filter being present. Both were earthfill dams
with a chimney filter, and in each case failure was attributed
to piping through dispersive backfill materials around the
outlet conduit. It is evident from the descriptions of these
cases that no filters were present to protect against piping
through the backfill material of the outlet conduit. Approximately 30% of accidents involving piping through the embankment occurred in dams with no embankment filters. In
21 of the 31 accidents where embankment filters were present, the initiation of piping was due to coarse or segregated
filters being placed adjacent to broadly graded core materials
(generally of glacial origin).
(11) About half of all piping failures and a quarter of accidents through the embankment are associated with the presence of conduits. Conduits influence the likelihood of piping
because they can increase the likelihood of initiation (particularly if there is poor compaction of the soil around the conduit) and progression (by providing a wall for the pipe to
form against). A review of the case data shows that virtually
all failures for which there are data occurred where poor
compaction was a contributing factor. Erosion into the conduit was a common cause of accidents but is not by itself
likely to lead to failure.
Core material properties
Geological origin
The statistics of failures and accidents for dams with different geological origin of the core material and for zoning
which gives good and poor control of erosion are shown in
Table 6. By comparing the percentage of failures and accidents to the percentage of the population, one can assess
whether the geological origin of the core has any influence
on the incidence of piping.
2000 NRC Canada

Foster et al.
1009
Table 7. Statistics of compaction of the core for the dams experiencing piping incidents.
No. of cases in the piping
incidents
Method of compaction
Failures
Accidents
No formal compaction
Rolled, modest control
Rolled, good control
Unknown
Total (where known)
11
13
5
17
29
8
25
15
35
48
% of
piping
failures
% of
piping
accidents
% of total
population
of dams
38
45
17
17
52
31
8
32
60
100
100
100
Table 8. Summary of data relating piping incidents to internal erosion control and compaction of the core material.
Class of dams
All dams (excluding hydraulic fill and puddle core)
Well compacted
No or poor compaction
Dams with limited internal erosion control
Well compacted
No or poor compaction
% of piping failures
% of the population
38
62
60
40
24
76
60
40
Note: Dam types with limited zoning are homogeneous earthfill, earthfill with foundation filter only, and earthfill with rock toe.
It is evident that dams with core materials of glacial origin

have experienced more piping accidents (i.e., initiation of
piping) than those built from other materials but have experienced fewer failures. The erodibility of some glacial soils
has been shown by pinhole tests (Ravaska 1997) to be similar to that of highly dispersive clays (category D1). It is possible that these glacial soils are more erodible because their
fine silt and clay size fractions are finely ground rock, rather
than more common clay minerals. The piping accidents involving glacial core materials have mostly occurred in dams
where embankment filters were present. Review of these
cases indicates problems have been attributed to the use of
coarse or segregated filters adjacent to the broadly graded
glacial core materials in central core earth and rockfill dams.
Failures involving piping of glacial core materials have only
occurred where no filters were present.
Dams with core materials of alluvial origin have experienced more than the average number of piping failures but
an average number of piping accidents. In all but one of the
nine failure cases involving alluvial soils, low-plasticity silts
(ML) or silty and clayey sands (SM, SC) were present.
Dams constructed of residual soils experience an average
number of piping failures. However, in three of the 10 piping
failures, the residual soils were dispersive, and therefore the
frequency of piping failures in nondispersive residual soils
tends to be less than the average.
Unified Soil Classification and dispersivity
The relationship between the classification of the embankment core materials and the incidence of piping through the
embankment was analysed. Dams constructed of lowplasticity silts (ML) experienced more piping failures than
average, those constructed of low-plasticity clays (CL) marginally less failures than average, and those constructed of
high-plasticity clays marginally less incidents than average,
but if dispersive soils are excluded dams experienced much
fewer incidents than average.
Dams constructed of dispersive clays are particularly susceptible to piping failures. Dispersive clays are recorded as
present in 18% (nine out of 51 cases) of the piping failures.
However, the actual proportion is likely to be higher, as the
majority of piping failures occurred prior to knowledge of
the nature of dispersive clays. All nine of the piping failures
where dispersive clays were known to be present occurred
on first (and usually rapid) filling of the reservoir. In six of
the cases, piping occurred around conduits or adjacent to
spillways. In one spectacular case, La Escondida Dam, 50
independent piping tunnels and eight breaches formed
through the dam on first filling.
Compaction of the core material
Tables 7 and 8 present the statistics of the incidence of
piping related to the compaction of the core material. The
piping incidents for hydraulic fill or puddle dams are not
considered in the analysis because these forms of compaction are directly related to the zoning of the dams, which is
analysed separately.
It is concluded that dams with no or poor compaction of
the core material experience many more piping incidents
than the average, and those with good compaction somewhat
less than the average. Dams with limited control of erosion
through the dam and with no formal compaction are more
likely to fail than experience accidents.

of piping through the foundation
Incidents are classified as piping through the foundation
where they involved any type of internal erosion process occurring primarily through the foundation of the dam. The incidents have been classified into categories depending on the
nature of the internal erosion as shown in Table 9. The ratio
of the number of accidents to the number of failures by piping through the foundation (77 accidents to 19 failures) is
significantly higher than that for piping through the embankment,
2000 NRC Canada

1010

Table 9. Incidence of piping through the foundation.
Type of piping
No. of failures
Piping through soil foundation

Dam foundation
Abutment
Reservoir foundation
Subtotal
Piping foundation soil into foundation rock
Dam foundation
Reservoir foundation
Subtotal
Piping through rock foundation
Piping spillway foundation
Piping foundation soil into relief wells toe drains
Sand boils in foundation
Unknown
Total
7
1
1
9 (43)
0
0
0
6
1
0
0
5
21
No. of accidents
16
10
3
29 (34)
(0)
(28)
(5)
(0)
(0)
(24)
(100)
5
1
6
7
10
9
22
2
85
(7)
(8)
(12)
(11)
(26)
(2)
(100)
Note: Percentages are given in parentheses.
Table 10. Average frequency of failure due to piping through the foundation by dam zoning types for large dams up to 1986.
Zoning category
No. of
failures
No. of
accidents
Zoned earthfill
Concrete (or other) face earthfill
Concrete (or other) face rockfill
Concrete corewall, earthfill
Concrete corewall, rockfill
Hydraulic fill
Unknown
All dams
2
0
3
1
0
0
3
0
0
2
0
1
7
19
9
5
2
14
6
7
2
0
0
1
0
7
17
70
Average
frequency of
failure (103)
3.0
Average
frequency of
accident (103)
10.4
11.2
3.9
3.9
4.6
7.6
9.8
5.8
11.8
Average annual frequency of

failure (103)
First 5 years
of operation
After 5 years
of operation
447
25
1044
59
88
6
1553
105
4.9
1768
68
15.7
91.8
2358
61
1.7
6.2
255
19
7.0
0.4
Note: The percentage of failures by piping through the embankment occurring at the different times after construction are as follows: 25% during first
filling, 50% during first 5 years of operation, and 25% after 5 year of operation. Calculations of annual frequencies of failure are made as follows: annual
frequency of failure (all years) = (average frequency of failure)/(average age), annual frequency of failure (first 5 years) = (average frequency of
failure)0.75/5, and annual frequency of failure (after 5 years) = (average frequency of failure)0.25/(average age 5).
suggesting that this mode of piping is less likely to progress

from the initiation stage of piping to breaching of the dam.
Self-healing of the piping processes by collapse of the pipe,
or self-limitation by the finite width of open joints in the
case of rock foundations, may partly explain the relatively
low number of failures by this mode of piping.
Dam zoning
Table 10 shows the statistics of failures and accidents by
piping through the foundation for each of the zoning categories and for all embankment dams combined. The low number of failure cases for foundation piping makes
comparisons between the dam zoning categories somewhat
doubtful, but the dam zoning categories with above-average
frequencies of failure are homogeneous earthfill, earthfill
with rock toe, concrete face earthfill, earthfill with corewall,

and hydraulic fill. These zoning categories would be expected to have relatively poor control of seepage and pore
pressures within the downstream foundation, and therefore it
is concluded that the zoning of the dam does appear to have
some influence on the frequency of failure by foundation
piping. The high incidence of piping failures for these zoning categories is not reflected in the accidents, so it is likely
the poor zoning of these dams contributes to the progression of piping. However, as discussed in the following sections, it is evident from the failures that other factors such as
the type of cutoff and foundation geology are more influential.
The significant proportion of rockfill dams in the foundation piping accidents (25%) but not in the failures suggests
2000 NRC Canada

Foster et al.
1011
Fig. 3. Foundation cutoff categories: (a) partially penetrating cutoff where cutoff does not extend to bedrock; (b) fully penetrating cutoff where cutoff extends to bedrock.
Table 11. Presence of foundation filters in piping through foundation accident and failure cases.
Presence of foundation filter
No. of failures
No. of accidents
No foundation filter
One foundation filter
Two foundation filters
Unknown
Total
13
1
0
7
21
33
23
2
27
85
the high seepage gradients allow initiation of piping, but the

rockfill prevents progression of piping to breaching.
Foundation filters
The effect of the presence of foundation filters is summarised in Table 11. Comparison with the population suggests
they are important in preventing failures but not accidents.
etrating. Here cutoff has been taken as cutoff trenches and

other cutoff types such as cutoff walls as shown in Fig. 3.
The number of foundation piping incidents in which the two
general types of foundation cutoff were present is listed in
Table 12.
The presence of grouting does not appear to have had a
significant influence in reducing the likelihood of initiation
of piping through rock foundations or piping of foundation
soils into foundation rock, with grouting carried out in approximately 50% of the incidents. There are five accidents
where piping occurred through the soil foundation despite
fully penetrating cutoffs being present. In these, foundation
erosion occurred through fractured foundation rock below
the cutoff (in two cases), due to an improper seal between
the cutoff wall and bedrock (in one case), or due to seepage
through the abutment under the cutoff (in two cases).
Foundation geology
Foundation cutoff
Foundation cutoff descriptors and other cutoff types used
in the ERDATA1 classification system have been simplified
for analysis of the foundation piping incidents into two
broad categories, namely partially penetrating and fully pen-
Piping in soil foundations

Foundation soil geology types that were involved in incidents of piping through soil foundations are shown in
Table 13 with the distribution of foundation geology types
Table 12. Incidence of piping through the foundation related to foundation cutoff types.
No. of failures
No. of accidents
Type of piping
Partially
penetrating
cutoff
Fully
penetrating
cutoff
Partially
penetrating
cutoff
Fully
penetrating
cutoff
Piping through soil foundation

Piping of foundation soil into rock
Piping through rock foundation
Foundation sand boils
Piping of foundation soil into drainage systems
Piping in spillway foundation
Unknown
Total
% of piping incidents
% of population
9
0
1
0
0
0
1
11
73
15
0
0
4
0
0
0
0
4
27
85
23
1
0
14
7
5
0
50
67
15
5
5
7
5
2
1
0
25
33
85
2000 NRC Canada

1012
Table 13. Soil geology types involved in foundation piping incidents.

Soil geology type
No. of
failures
No. of
accidents
Population
(%)
Alluvial
Glacial
Aeolian
Colluvial
Lacustrine
Residual
Volcanic
Unknown
Total no. of incidents
5
1
0
0
0
3
1
0
10
10
9
3
5
2
1
1
2
29
56
19
6
7
3
8
<1
100
(50)
(10)
(0)
(0)
(0)
(30)
(10)
(100)
(32)
(29)
(10)
(16)
(6)
(3)
(3)
(100)
Note: Percentages are given in parentheses. Where incidents have

involved piping of more than one soil geology type, all the soil types
have been counted (percentages are relative to the total number of soils,
where known). Percentages of the population are for the following
selected group of countries only: United States, India, Canada, Australia,
and New Zealand.
for the selected group of countries in the embankment dam

population.
The data suggest dams founded on glacial and colluvial
soils are more likely to experience piping accidents than
other dams but less likely to experience piping failures.
Therefore it appears that these soils are more susceptible to
the initiation of piping, but the soil properties and structure
give them good self-healing characteristics or a high likelihood of collapse of the pipes, so the progression of piping is
less likely when compared to other soils.
Table 13 also suggests dams with residual soils in the
foundation are more likely to experience piping failures than
the average. However, there are limited data, and in two of
the three failures involving piping of residual soils, namely
Baldwin Hills Reservoir and Laguna Dam, the residual soils
were dispersive. Residual soils are not common in the piping
accidents. Therefore dams with residual soils present in the
foundation are less likely to experience piping incidents in
the foundation unless the soils are dispersive.
Piping in rock foundations
Not unexpectedly, dams constructed on limestone rock
foundations appear to be particularly susceptible to piping
incidents. Review of the accident and failure cases with
limestone foundations shows that seven of the eight cases involved piping through infilled solution channels, joints, or
karstic cavities in the limestone. The other incidents were in
sandstone (one failure) and shale (one failure, one accident).
Piping of foundation soil into foundation rock
Foundation soils were piped into granite rock in three of
the five incident cases involving piping of foundation soil
into foundation rock. These three cases occurred in glaciated
valleys. Stress-relief effects on granite rocks due to glaciation may have contributed to fracturing of the bedrock in
these cases. These are described in Fell et al. (1992) and discussed further in this paper in relation to piping of embankment soils into rock foundations. In the other two cases, the
foundation soils were piped into a karstic limestone foundation (at Walter F. George Lock Dam) and into open fractured
sandstone (at Bad Axe Watershed Structure No. 33). In the
Table 14. Statistics of failures and accidents for piping from the
embankment into the foundation for large dams up to 1986.
Average annual
frequency
Failures
Incidents
(accidents and
failures)
Number
Average
frequency
First 5
years
After 5
years
2
22
1.8104
2.1103
1.9105
2.3104
3.9106
4.6105
Table 15. Characteristics of foundation cutoff for incidents of

piping from the embankments into rock foundations.
Cutoff
characteristics
Cutoff, shallow
Trench, deep
Descriptor,
unknown
Grouting
No grouting
Grouting
Unknown
Total
No. of
accidents
No. of
failures
% of piping
incidents
% of
population
9
9
6
2
1
0
52
48
66
34
12
9
3
24
2
1
0
3
54
46
42
58
latter case, vertical fractures, with joint openings 50300 mm

wide, were present in the foundation sandstone in the abutments of the dam.

of piping from the embankment into the
foundation
Incidents have been classified as piping from the embankment into the foundation in the cases where embankment
materials, including the cutoff trench fill, have been eroded
into the foundation of the dam due to seepage through the
foundation or embankment. Table 14 summarises the statistics of failures and accidents for piping from the embankment to the foundation. There are insufficient failure cases
(four in total) for this type of piping to assess the significance of the various factors on failures alone, so the analysis
uses failures and accidents and is likely to identify the factors that have the most influence on the frequency of initiation of piping.
The statistics given in Table 15 may underestimate the actual frequencies of piping from the embankment into the
foundation. All of the failure cases for this mode of failure
occurred after 1976, and it is likely that failures occurring
prior to this were classified by the investigators of the failures as piping through the embankment or foundation. The
well-investigated and well-publicised case of Teton Dam
failure in 1976 highlighted this mode of piping as a possible
mode of piping for other failures occurring after 1976.
Piping of embankment materials into rock foundations is
more common (24 accidents, three failures) than piping into
soil foundations (seven accidents, one failure), possibly
2000 NRC Canada

Foster et al.
1013
Table 16. Foundation rock into which core material piped.
Foundation rock
Sandstone
Sandstone and shale
Sandstone and limestone
Limestone
Granite
Quartzite
Schist
Tuff
Basalt
Unknown
No. of
failures
No. of
accidents
3
6
2
3
3
1
1
1
1
1
4
% of failures
and accidents
14
33
9
14
14
4
4
4
4
% in
population
21*
7
7
3
7
2
5
* For sandstone.
because of the presence of continuous, open joints in some

rock foundations. It is also likely to be easier to identify than
piping into soils.
Dam zoning
The analysis of incidents shows that dam zoning has little
influence on the initiation of piping. However, it would be
expected that dams with good downstream discharge capacity, e.g., free-draining rockfill, would be less likely to progress to breach compared with dams with downstream zones
with poor discharge capacity (e.g., zoned earthfill with silty
sandy gravel in the downstream zone).
Foundation filters
Foundation filters were present in six of the 31 accidents
and one of the four failures. In two of the accidents with
foundation filters, the accident was partly attributed to the
presence of coarse filters. All cases where foundation filters
were present involved piping into rock foundations, so it appears the seepage bypassed the filters, or the filters were
washed into the joints.
Foundation cutoff
All accidents and failures involving piping of embankment materials into rock foundations have fully penetrating
cutoff trenches to and into rock. The characteristics of the
foundation cutoff present in these incidents are summarised
in Table 15. The presence of deep, narrow cutoff trenches
and poorly treated open joints were believed to have been
important contributing factors in the failure of Teton Dam
and in accidents involving six puddle core dams. In all of
these cases, it was considered highly likely that hydraulic
fracturing within the cutoff trenches had contributed to the
initiation of piping. Even without fracturing, the gradients
are often high. These conditions can also make compaction
of the earthfill materials difficult and thereby potentially give
areas of poorly compacted earthfill materials against the
foundation.
The majority of incidents involving piping from the embankment into soil foundations have occurred in dams with
partially penetrating cutoff trenches with no other cutoff
present. Two accidents with fully penetrating cutoff trenches
involved puddle core earthfill dams with deep and narrow
trenches.
Erosion control at the corefoundation contact

Erosion-control measures, in the form of either concrete
covering or filter protection, at the embankmentfoundation
contact are present in only two of the incidents involving
piping from the embankment into the foundation. Both of
these are accidents involving piping into fractured bedrock.
At Brodhead Dam, glacial core materials were eroded into a
coarse foundation blanket layer which covered an area of
highly weathered and fractured shale bedrock. At Hallby
Dam, glacial core materials were eroded into granite bedrock despite the presence of foundation treatment in the
form of sealing of cracks and partial covering with concrete
slabs. The cause of this incident was attributed to the flushing out of montmorillonitic clay material in the open joints
of the granite bedrock.
Foundation geology
Table 16 presents the geology of the rock foundations into
which piping occurred. Interbedded sandstone and shale
foundations are relatively common in the incidents involving
piping into rock foundations. It is known that the presence
of interbedded layers of sandstone and shale can give rise to
open joints due to stress-relief effects, as described in Fell et
al. (1992). All six of the incident descriptions involving piping into interbedded sandstone and shale foundations noted
the fractured nature of the bedrock and the presence of open
joints. At Fontenelle Dam, open stress-relief joints in the
abutment were up to 300 mm wide, and at Cowm Dam very
wide (estimated to be at least 0.5 m), open stress-relief joints
are visible in the construction photographs of the foundation
bedrock. Piping of embankment materials into limestone
foundations was also relatively common. In two of the cases,
Wolf Creek Dam and Apa Dam, embankment materials were
piped into solution channels in the limestone foundation. At
Taibilla Dam, core materials were piped into open joints, up
to 100 mm wide, in the limestone foundation.
At least 10 of the 24 accidents involving piping into rock
foundations occurred in geological environments previously
affected by glaciation. The presence of highly fractured bedrock and open joints is noted in all 10 of these cases. Fell et
al. (1992) describe the open fractured nature of bedrock beneath glacial deposits at several sites. The formation of these
features is attributed to the effects of stress relief or
glacitectonic thrusting. The four accidents involving granite
2000 NRC Canada
Tuesday, October 03, 2000 10:28:27 AM

1014

Table 17. Statistics of failures and accidents for downstream slides on large dams up to 1986.
Average
frequency
Average annual frequency
Category
No. of
cases
First 5 years
After 5 years
Failures
Incidents (failures and accidents)
Incidents in operation
6
59
50
5.4104
5.3103
4.4103
4105
5.2104
4.4104
1.5105
1.2104
1.0104
Table 18. Incidence of downstream slide type and dam zoning (failure and accident cases combined).
Zoning type
Sloughing
Embankment
Embankment
and foundation
Unknown
Zoned earthfill
Zoned earthfill and rockfill
Earthfill with concrete corewall
Rockfill with concrete corewall
Hydraulic fill
Other
Unknown
Total
5
1
1
1
1
1
0
0
0
1
0
3
0
0
14
12
4
3
2
0
0
0
0
1
4
0
2
1
5 (1)
34 (1)
6 (1)
4
0
3
0
2
0
0
3
0
0
0
0
0
18 (1)
3
0
0
0
0
0
0
0
0
0
0
1
1
2
7
(1)
(1)
(1)
(1)
(4)
(2)
(1)
(1)
(1)
(5)
Note: Values in parentheses are the number of failure cases.
foundations were in geological environments that had been

affected by glaciation. The presence of horizontal stressrelief joints in the granite bedrock was found to be the cause
of the piping incident at Churchill Falls GJ-11A Dyke.
The three soil foundation types present in piping into soil
foundation, namely glacial (four accidents), colluvial (three
accidents), and alluvial (four accidents), are all characterised
by the potential for great variability, high permeability, and a
large range in grain sizes. In two of the accidents, the presence of open-work gravel layers in the foundation was
noted. Soil foundations of glacial and colluvial origin appear
to be overrepresented, which is not surprising because they
often have cobbles and boulders with large voids.
Embankment core characteristics
The effects of the geological origin, Unified Soil Classification, and compaction on the frequency of piping incidents
are similar to those for piping through the embankment.

of downstream sliding
Incidents have been classified as downstream slides where
the incident has involved any form of sliding movement of
the downstream slope of the dam. Dams which have shown
signs of incipient sliding of the downstream slope have also
been included. Downstream slides that have occurred to
dams during construction have been classified as accidents
unless they have resulted in uncontrolled release of reservoir
water. This is consistent with the definition used in the
ICOLD (1995) study.
The downstream slide incidents have been classified according to the location and type of sliding movement. These
are sloughing (progressive sliding of the downstream slope
due to seepage through the embankment), through the embankment (slide surface passes through the embankment
only), and through the embankment and foundation (base of
slide surface passes through the foundation).
Table 17 summarises the statistics of failures and accidents for downstream slides. The fact that there are relatively many accidents compared to failures may reflect the
fact that movement usually occurs slowly, giving warning of
a slope-instability problem and allowing remedial action or
drawing down of the reservoir, or the slide is simply too
shallow to directly release the reservoir.
Dam zoning and type of sliding
Table 18 summarises the types of sliding and how these
relate to the dam zoning. There are very few failures in total,
and of these only one (Utica Dam) involved failure through
the embankment, and one (Fruitgrowers Dam) failure
through the foundation. As might be expected, the type of
embankment zoning appears to have a significant influence
on the frequency of initiation of sliding of the downstream
slope. Dam zoning types with poor control of pore pressures
and seepage within the dam and foundation, such as homogeneous earthfill, earthfill with foundation filter only,
earthfill with rock toe, and earthfill with concrete corewall
dams, all have average frequencies of initiation of sliding
greater than the average of all the dams combined. The
overrepresentation of puddle core earthfill, earthfill with
concrete corewall, and hydraulic fill dams in the slide
2000 NRC Canada
Tuesday, October 03, 2000 10:28:27 AM

Foster et al.
1015
Table 19. Summary of data relating downstream slide incidents to the method of compaction of the
core material.
Compaction of core materials
No. of
rotational-type
slides
No. of
sloughing-type
slides
% of
incidents
% of
population
Hydraulic fill
Puddle
Unknown
Total no. of slide incidents
Total no. of incidents (where known)
2
6
10
8
7
19
52
33
3
0
6
2
0
3
14
11
48
30
21
100
8
32
60
100
incidents may also be a reflection of the lower standards of

design and construction (particularly the lack of compaction
by rollers) of these older dams compared with more modern
dam types. Dam types which inherently have relatively good
control of pore pressures and high, reliable shear strength in
the downstream zone, such as zoned earthfill, zoned earth
and rockfill, central core earth and rockfill, and concrete
face rockfill, tend to have a relatively low frequency of
downstream slide incidents compared with the dam population.
There is a recorded case of a failure of a recently constructed concrete face rockfill dam, Gouhou Dam, which is
believed to have failed by sliding of the downstream slope.
The incident is not included in the analysis because dams in
China are excluded from the accident and failure statistics in
this study. The cause of the failure is attributed to leakage
through the poorly constructed connection between the crest
wave wall and the concrete face which led to saturation of
the gravel fill downstream slope and probable sliding. Poor
zoning of the sandy gravel fill forming the body of the dam
is believed to have been a major contributing cause of the
failure.
Foundation geology
The relationship of downstream sliding incidents to foundation geology has been assessed and shows that sliding
through the foundation has occurred mainly on soil foundations (16 out of 18 cases). Sliding through rock has only
been noted in two cases. The type of rock, or origin of the
soil, has little influence, but the presence of high-plasticity
clays in the foundation does seem important, with these being present in 12 of the 18 incidents involving sliding
through the foundation. High-plasticity clays are generally
characterised by relatively low shear strength and, probably
more importantly, a large reduction from peak to residual
shear strength. The presence of fissured high-plasticity clays
was noted in two of the incidents involving sliding through
the foundation.
The presence of soft sedimentary rocks and residual soils
of sedimentary origin in the foundation appears to be a common feature of both upstream and downstream slide incidents where sliding has occurred through the foundation;
40% of upstream and downstream foundation slides combined (i.e., 14 out of 35 cases) involved sliding through
these materials. The presence of low-strength bedding features such as surface bedding shears in this geological envi-
ronment, particularly in interbedded sandstone and shales,

may explain why foundation slide incidents are so common.
Geological origin
The geological origin of the core material appears to have
some influence on the frequency of slide incidents, with
cores of lacustrine origin more susceptible, cores of glacial
origin less susceptible, and cores of residual and alluvial
soils neutral. Sloughing incidents are more likely with soils
of alluvial and glacial origin, and rotational slides are more
likely to occur in cores built of residual, glacial, and lacustrine soils.
Unified Soil Classification
Dams embankments comprised of high-plasticity materials
are much more likely than average to experience rotationaltype slide incidents. Half of the high-plasticity core materials were of either glacial or lacustrine origin, with the other
half unknown. No cases of sloughing-type incidents are
known to have occurred in dams with high-plasticity materials. Dams with embankment materials of low-plasticity clays
and silts (CL, ML) are also more likely to experience slide
incidents by both rotational- and sloughing-type slides. Approximately 70% of the cases with low-plasticity core materials are of either residual or alluvial origin.
Embankment core materials comprised of clayey or silty
sands and gravels (SC, SM, GC, GM) are shown to be less
likely to experience slide incidents. The lower proportion of
sandy materials in the embankment core materials of the
slide incidents is probably indicative of the higher shear
strength of these materials compared to the more clayey core
materials. They may be generally more permeable, facilitating dissipation of pore pressures and giving less chance of
contractive behaviour in poorly rolled material.
Compaction
Table 19 summarises the compaction of dams that have
experienced downstream sliding incidents. Hydraulic fill and
puddle core dams are not included because the core compaction is directly related to the zoning of the dam.
As might be expected, the degree of compaction of the
core materials appears to have a strong influence on the likelihood of initiation of sliding. Relatively poor compaction
conditions exist in the majority of the downstream slide
2000 NRC Canada

1016

Table 20. Statistics of failures and accidents for upstream slides on large dams.
Average
frequency
Average annual probability
Category
No. of
cases
First 5 years
After 5 years
Failures
Incidents (failures and accidents)
1
47
9105
4.2103
3106 all years

4.1104
1.0104
incident cases. The ratio of the percentage of incidents to

that of the population suggests dams which have no formal
compaction of the core materials are about seven times more
likely to experience slides than the average and about 25
times more likely than dams with good compaction of the
core. Poorly compacted core materials would be expected to
have considerably lower shear strengths and be more permeable compared with materials with poorly compacted clays.
The increased permeability of poorly compacted soils can
potentially allow wetting and softening of the core materials
and failure in undrained loading. Slides through the foundation appear to be less influenced by the degree of compaction of the embankment.

of upstream sliding
Incidents have been classified as upstream slides where
the incident has involved any form of sliding of the upstream
slope of the dam. The upstream slide incidents have been
classified according to the location and type of sliding movement in the same way as that for downstream slides. In addition, upstream slide types have been classified according to
whether or not the slide was initiated by a drawdown of the
reservoir.
Table 20 summarises the statistics of failures and accidents for upstream slides. There is only one large embankment dam that is known to have failed (i.e., breached) due to
sliding of the upstream slope. The incident description of
Kaila Dam (ICOLD 1974) describes settlement and piping
around the conduit giving rise to the slip of the entire upstream bank up to the corewall. It has been assumed the incident at Kaila Dam is a failure (i.e., was breached), which
is consistent with the classification given by both ICOLD
(1974) and ICOLD (1995).
The average frequency of failure occurring once sliding of
the upstream slope has initiated is one failure out of 46
cases, giving a ratio of approximately 2%. This is five times
lower than that for downstream slides. This is considered to
be an upper bound, as one might expect that a large number
of upstream sliding accident cases have not been reported in
the ICOLD studies or in the literature and the failure of
Kaila Dam might just as well be classified as a piping failure.
A total of 57% of all the upstream slide incidents were
initiated by a drawdown of the reservoir. The one failure
case does not appear to be a drawdown slide. Of the 19 incident cases which are not drawdown slides, 10 occurred during construction prior to reservoir filling. Therefore, 73% of
upstream slides which occurred in operation (i.e., excluding
construction slides) were initiated by drawdown of the reservoir. This helps explain why so few upstream slides have
caused breaches: the water level in the reservoir is low, so
Table 21. Incidence of upstream slide incidents.

Type of slide
No. of
failures
No. of
accidents
Sloughing
Through embankment only
Through embankment and foundation
Unknown
Total
0
0
0
1
1
1
26
17
2
46
freeboard is large. The number of cases for each of the three

different types of downstream slides are listed for the accident and failure cases in Table 21.
Sloughing-type slides are relatively uncommon in upstream slide incidents, with only one accident to Holmes
Creek Dam, which was constructed of very uniform, fine,
cohesionless sand. In this case, progressive sloughing of the
upstream face was initiated by an explosion and within 5 h
the sloughing had cut 9 m into the crest of the dam (Sherard
1953).
Dam zoning and type of sliding
An analysis of the relationship of dam zoning to upstream
slides shows the following embankment dam zoning categories which have average frequencies of initiation of upstream
slides higher than or similar to the average of all dams combined and the approximate ratios of the average frequencies
of the dam zoning categories to the average frequency of all
dams (for example, homogeneous earthfill dams are about
three times more likely to experience upstream slides than
all dams combined): homogeneous earthfill 3, earthfill with
rock toe 1, concrete face earthfill 1.5, puddle core earthfill
2, earthfill with concrete corewall 2.5, and hydraulic fill
12. All of these dam zoning types generally have earthfill
material in the upstream slope, making them more susceptible to upstream slides under drawdown conditions. The older
dams are also likely to have been poorly compacted.
All of the dams with rockfill zoning have a low incidence
of upstream slide incidents. This is related to the relatively
high shear strength and high permeability of rockfill materials. Only one of these three incidents was initiated by
drawdown of the reservoir (old Eildon Dam), which was a
concrete corewall dam with an earthfill core and rockfill
shoulders. The other two cases were a construction slide of a
central core earth and rockfill dam founded on soft fat
clays (Clendening Dam) and an upstream slide initiated by
an earthquake (La Calera Dam).
Zoned earthfill dams are less likely to experience upstream slide incidents than the population of dams by a factor of approximately 0.5. The majority of upstream slide
incidents to zoned earthfill dams have been foundation
slides, making up six of the nine incidents to this dam type.
2000 NRC Canada

Foster et al.
1017
Table 22. Summary of the factors influencing the frequency of failure by piping through the embankment.
General factors influencing frequency of failure
Factor
Much more likely
More likely
Neutral
Less likely
Much less likely
Zoning
Homogeneous earthfill,
earthfill with rock
toe, puddle core
earthfill
Concrete face
earthfill
Earthfill with
concrete corewall,
hydraulic fill
Earthfill with
filter, zoned
earthfill,
zoned earth
and rockfill
Central core earth

and rockfill;
concrete face
rockfill; rockfill
with concrete
corewall
Embankment filter
present
Glacial
Clayey and
silty gravels
(GC, GM);
low-plasticity
clays (CL)
High-plasticity clays
(CH)
Embankment filters
Core geological
origin
Core soil type
Alluvial
Compaction
Dispersive clays; lowplasticity silts (ML);

poorly graded and
well-graded sands
(SP, SW)
Conduits and other

locations of piping
Conduit through the

embankment
No embankment
filter
Aeolian, colluvial
Clayey and silty
sands (SC, SM)
Rolled, modest
control
Irregularities in
foundation or
abutment, steep
abutments
Foundation geology
The relationship of upstream sliding to incidents to foundation geology has been assessed and shows that the majority of the foundation slide incidents involve sliding through
soil foundations, making up 13 of the 17 cases. Sliding occurred through rock foundations in only four incidents,
namely at Bear Gulch Dam, where sliding was initiated on
three occasions, and at Fort Peck Dam.
The relationship between geological origin and compaction of the materials and incidence of slides is similar for
both upstream and downstream slides. Dams with core materials composed of clay materials are more likely to experience upstream slide incidents than the average. This is
evident for dams with low-plasticity clays (CL), which are
nearly two times more likely to experience slides than the
average. Embankment core materials comprised of clayey or
silty sands and gravels (SC, SM, GC, GM) are less likely to
experience slide incidents by both upstream and downstream
slides. This is probably indicative of the higher shear
strength of these materials compared with the more clayey
core materials.
Summary of the factors affecting the

frequency, timing, and location of piping
and sliding
Factors affecting the frequency of piping and sliding
Tables 2224 summarise the factors influencing the frequency of piping, and Tables 25 and 26 summarise the factors affecting the frequency of sliding. These tables have
been prepared from the analysis of the data, but also take ac-
Residual, lacustrine,
marine, volcanic
Well-graded and
poorly graded
gravels (GW,
GP); high-plasticity silts (MH)
Puddle, hydraulic
fill

No conduit through
the embankment
count of general dam engineering principles and the nature

of soil and rock environments. The allocation into the categories much more likely, more likely, neutral, less
likely, and much less likely is somewhat judgmental. It is
recognised that some factors may be surrogates for the other,
e.g., the geological environment and soil classification properties may be linked. However, in some cases, particularly
for older dams, there may be little other than the geological
environment available to assess the dams, so we have included both factors.
The analysis of dam zoning, filters, core properties, and
compaction from the dam failure statistics, as described in
the preceding sections, has assessed the factors individually.
An analysis was also carried out to assess what combination
of factors is more likely to lead to piping failure.
Dams which failed generally had several poor characteristics, such as limited zoning, no filters, poor compaction, and
erodible soils. Dams which suffered accidents commonly
had fewer of the poor characteristics and more of the good
characteristics. The results of this simple analysis indicate
that dams which have a combination of several poor characteristics are much more likely to fail by piping than those
with only one or two poor characteristics. As an example,
the presence of dispersive soils in itself may not necessarily
influence the likelihood of failure but is more likely to influence a dam with poor compaction around a conduit and homogeneous zoning.
Timing of incidents
The timing of piping incidents is summarised in Tables 2729 and Fig. 4. The frequency of piping failures is
significantly higher on first filling and early in the life of the
dam. There does, however, appear to be a trend for older
dams to also experience piping incidents. This may reflect
2000 NRC Canada

1018
Table 23. Summary of the factors influencing the frequency of failure by piping through the foundation.
General factors influencing frequency of failure
Factor
Much more likely
Zoning
Filters
Foundation type
(below cutoff)
Foundation cutoff
(soil foundation)
Foundation cutoff
(rock foundation)
Soil geology type
(below cutoff)
Rock type
(below cutoff)
More likely
Neutral
Less likely
Much less likely
Homogeneous earthfill,
earthfill with rock toe,
concrete face earthfill,
earthfill with corewall,
hydraulic fill
No foundation filter
present when required
Puddle core
earthfill
Earthfill with filter,

zoned earthfill
Zoned earth and

rockfill, central core
earth and rockfill,
concrete face rockfill,
rockfill with corewall
No foundation
filter
Erodible rock
Foundation filter(s)
present
Soil foundation
Shallow or no cutoff
trench
Dispersive soils,
volcanic ash
Limestone,
dolomite,
soluble rocks
(gypsum), basalt
Fully penetrating cutoff

wall
Residual
Non-erodible rock
Upstream blanket,
partially penetrating
cutoff wall
Cutoff trench
Aeolian,
colluvial,
lacustrine,
marine
Tuff, rhyolite, marble,

quartzite
Alluvial
Glacial
Sandstone, shale,
siltstone,
claystone,
mudstone,
hornfels,
agglomerate,
volcanic breccia
Conglomerate,
andesite, gabbro,
granite, gneiss,
schist, phyllite, slate
Fig. 4. Time after construction of incidents of piping through the embankment.
deterioration or the design and construction of the older

dams.
The time of the incident of piping in the embankment for
each of the dam zoning categories has been analysed for
failures and accidents. The results suggest that the time of
incident for failures may be related to dam zoning. Homogeneous earthfill, earthfill with filter, earthfill with rock toe,
and zoned earthfill dams all show a relatively high proportion of failures on first filling, ranging from 50% failures for
homogeneous earthfill dams up to 7080% for the other
three dam zoning types. Less than 25% of the failures in
concrete face earthfill and puddle core dams occurred on
first filling. For the accidents, less than half of the piping incidents occurred on first filling for most of the dam zoning
categories. Piping accidents of puddle core earthfill dams
tend to occur generally after many years of operation.
The time of sliding incidents is summarised in Tables 30
and 31. An assessment of the different failure types shows
that slides through the foundation tend to occur at an earlier
age than slides through the embankment; 44% of the foundation slides occurred during construction or on first filling
compared with only 15% of embankment slides. Sloughingtype slides also tend to be more frequent in the early stages.
The potential for preexisting shear surfaces or other weak
2000 NRC Canada

Foster et al.
1019
Table 24. Summary of the factors influencing the frequency of piping accidents and failures from the embankment into the foundation.
General factors influencing frequency of initiation of piping
Factor
Much more likely
More likely
Neutral
Less likely
Much less likely
Zoning
Appears to be
independent of
zoning
Appears to be independent of
presenceabsence of
embankment or
foundation filters
Appears to be
independent of
zoning
presenceabsence
of embankment
or foundation
filters
Appears to be
independent of
zoning
presenceabsence
of embankment or
foundation filters
Appears to be
independent of
zoning
Appears to be
independent of
presence
absence of
embankment or
foundation
filters
Shallow or no
cutoff trench
Appears to be
independent of
zoning
Appears to be
independent of
presence
absence of
embankment or
foundation
filters
Filters
Foundation cutoff
trench
Foundation type
Deep and narrow

cutoff trench
Erosion-control
measures of core
foundation
No erosion-control
measures, open
jointed bedrock, or
open-work gravels
Grouting of
foundations
Soil geology type
Rock type
Core compaction
Foundation
treatment
Founding on or
partly on soil
foundations
Founding on or
partly on rock
foundations
No erosion-control
measures
Erosion-control
measures
present
No grouting on
rock foundations
Glacial
Soil foundation only,

not applicable
Sandstone interbedded
with shale or limestone; limestone,
gypsum
Dolomite, tuff,
quartzite, rhyolite, basalt,
marble
Glacial
Dispersive clays; lowplasticity silts (ML);

poorly graded and
well-graded sands
(SP, SW)
compaction
Untreated vertical
faces or overhangs
in core foundation
Clayey and silty

sands (SC, SM)
Agglomerate, volcanic
breccia, granite,
andesite, gabbro,
gneiss
Aeolian, alluvial, colluvial lacustrine,
marine, volcanic
Well-graded and
poorly graded
gravels (GW, GP);
high-plasticity silts
(MH)
compaction
Colluvial
Core geological
origin
Core soil type
Average cutoff trench

width and depth
compaction
Irregularities in
foundation or
abutment, steep
abutments
Rock foundations
grouted
Residual
Sandstone, conglomerate
schist, phyllite,
slate, hornfels
Residual
Alluvial, aeolian,
lacustrine,
marine, volcanic
Shale, siltstone,
mudstone,
claystone
Clayey and silty

gravels (GC,
GM); lowplasticity clays
High-plasticity
clays (CH)
Appears to be
independent of
compaction
Careful slope
modification by
cutting, filling
with concrete
Appears to be
independent of
compaction
Careful slope
modification by
cutting, filling
with concrete
Note: The ranking is designed to place those rocks which commonly have open joints, e.g., due to stress relief, as much more likely and those with a
low likelihood of open joints as much less likely. Some rock types, assumed to occur in large masses, e.g., andesite, sometimes occur as flows, so may
be better in a much more likely category in that case.
zones in the foundation of low shear strength may explain

the tendency of foundation slides to initiate at an early age.
The effect of pore pressures generated in high-plasticity
clay foundations during construction may also explain the
tendency of foundation soils to initiate sliding early. In six
of the eight foundation slide incidents that occurred during
construction or on first filling, sliding took place through
high-plasticity and (or) soft clays in the foundation. Highplasticity and soft clays are generally characterised by low
permeability (giving slow dissipation of pore pressures), low
peak shear strength, and a large reduction in shear strength

from peak to residual once sliding is initiated. These all give
conditions conducive to initiation and progression of sliding.
Location of the initiation of piping
The location where piping originated for the piping
through the embankment and piping from the embankment
to the foundation was noted from the incident descriptions.
The number of cases where piping initiated for the different
types of location for both accidents and failures are plotted
2000 NRC Canada

1020
Table 25. Summary of the factors influencing the frequency of downstream slides, accidents and failures.
General factors influencing frequency of initiation of downstream slides
Factor
Much more likely
More likely
Neutral
Less likely
Much less likely
Zoning
Homogeneous
earthfill, earthfill
with corewall,
hydraulic earthfill
Earthfill with rock

toe
Concrete face
earthfill, puddle
core earthfill
Zoned earthfill, zoned

earth and rockfill,
central core earth
and rockfill,
concrete face
rockfill, rockfill
with corewall
Foundation type
Geology type
(foundation
slides)
Soil foundations
in foundation, i.e.,
marine, lacustrine
Residual soils of
sedimentary
origin and soft
sedimentary rocks
Core geological
origin
Core soil type
Lacustrine
All other geology

types (due to low
number of foundation slide cases)
Residual, alluvial,
colluvial, volcanic
Clayey sands (SC)
Core compaction
and silts (for rotational slides)
No formal
compaction
Rock foundations
Low-plasticity silts
and clays (ML,
CL)
Rolled, modest
control
Puddle, hydraulic
(accounted for by
zoning)
Glacial, aeolian
Clayey gravels
(GC)
Silty sands and

gravels (SM, GM)
Rolled, well
compacted (for
foundation
slides)
Rolled, well
compacted
(particularly for
embankment slides
and sloughing)
Fig. 5. Piping initiation location of the incidents of piping through the embankment.
in Figs. 5 and 6. The presence of conduits through the embankment and (or) through the trenches in the foundation
has an important influence on the initiation of piping
through the embankment for the reasons discussed earlier.
Other locations where piping has initiated, but much less
frequently than conduits, are at the contact between the em-
bankment and concrete structures such as spillways, over irregularities in the foundation or abutments, and over steep
abutments. These are all locations in the embankment which
are particularly susceptible to hydraulic fracturing and (or)
differential settlement and where difficulties can be experienced with compaction of the core materials.
2000 NRC Canada

Foster et al.
1021
Table 26. Summary of the factors influencing the frequency of initiation of upstream slides, accidents and failures.
General factors influencing frequency of initiation of downstream slides
Factor
Much more
likely
Zoning
Homogeneous
earthfill,
hydraulic fill
Foundation type
Geology type
(foundation
slides)
Soil foundations
High-plasticity
clays in foundation (i.e.,
marine,
lacustrine)
Glacial,
lacustrine
Core geological
origin
Core soil type
Core compaction
More likely
Neutral
Less likely
Much less likely
Concrete face earthfill,

puddle core earthfill,
earthfill with
corewall
Earthfill with rock

toe, rockfill with
corewall
Earthfill with
filter, zoned
earthfill
Zoned earth and

rockfill, central
core earth and
rockfill, concrete
face rockfill
Residual soils of sedimentary origin and

soft sedimentary
rocks
All other geology

types (due to
low number of
foundation slide
cases)
Residual, marine,
volcanic
High-plasticity
clays (CH)
Rock foundations
Alluvial
Low-plasticity silts and
clays (ML, CL)
No formal
compaction
Puddle, hydraulic
(accounted for
by zoning)
Aeolian, colluvial
Glacial
Clayey sands and

gravels (SC,
GC)
Rolled, well
compacted (for
foundation
slides)
Silty sands and

gravels (SM, GM)
Rolled, well
compacted (for
embankment slides
and sloughing)
Fig. 6. Piping initiation location of the incidents of piping from the embankment into the foundation.
Conclusions
Overall statistics
The analysis of the dam incidents in the ERDATA1 database has shown that structural modes of failure, i.e., those
involving piping, slope instability, or an earthquake, account
for approximately half of the failures to large embankment

dams. Piping failure account for most of these. The incidence of piping through the embankment is two times higher
than piping through the foundation and 20 times higher than
piping from the embankment into the foundation. Embankment slides account for only 4% of embankment dam
2000 NRC Canada

1022
Table 27. Time after construction of incidents of piping through the embankment.
% of cases (where known)
Time of incident after construction
No. of
failures
No. of
accidents
Failures
Accidents
During construction
During first filling
After first filling and during first 5 years
After first 5 years
Unknown
Total no. of piping cases
1
24
7
18
1
51
0
26
13
60
3
102
2
48
14
36
100
0
26
13
61
100
Table 28. Time after construction of incidents of piping through the foundation.
No. of
failures
No. of
accidents
Failures
Accidents
During construction
After first 5 years
Unknown
1
4
10
5
1
21
0
23
19
36
7
85
5
20
50
25
100
0
30
24
46
100
Table 29. Time after construction of incidents of piping from the embankment into the foundation.
No. of
failures
No. of
accidents
Failures
Accidents
During construction
After first 5 years
Unknown
0
2
2
0
0
4
0
6
8
16
1
31
0
50
50
0
100
0
20
27
53
100
No. of
accidents
No. of
failures
Failures
Accidents
During construction
After first 5 years
Unknown
2
2
0
7
0
11
9
7
15
30
1
62
18
18
0
64
100
15
11
25
49
100
Table 30. Time after construction of incidents of downstream slides.
Table 31. Time after construction of incidents of upstream slides.

No. of
failures
No. of
accidents
% of incident
cases
During construction
10
21

After first 5 years
Unknown
0
1
0
0
1
1
12
23
0
46
2
28
49
100
2000 NRC Canada

Foster et al.
failures. Upstream slide failures are uncommon, with only

one known failure of a large embankment dam. The frequency of failures for embankment dams constructed prior
to 1950 is about seven times higher than that for dams constructed after 1950.
Piping and slope instability
The detailed analysis shows the embankment zoning has
an important influence on the frequency of initiation of piping and slope instability, with the exception of piping from
the embankment into the foundation. Dams with poor control of seepage and pore pressures in the embankment and
foundation have generally experienced more problems associated with piping and slope instability compared with other
dam types. The dam types with these characteristics are homogeneous earthfill, earthfill with filter, particularly with
only foundation filter, earthfill with rock toe, concrete face
earthfill, earthfill with concrete corewall, and puddle core
earthfill.
Dams with downstream rockfill zones have a low frequency of failure by all piping modes of failure but a high
proportion of piping accidents, suggesting that piping is less
likely to progress to breaching in rockfill dams than in dam
types with downstream zones of earthfill. This is attributed
to the inherent stability of the downstream rockfill zone under large seepage flows.
From the results of this analysis, the type of geological
environment does not seem to have as strong an influence on
the frequency of failure by piping and slope instability as
one might expect. This may be due in part to the low number of incidents occurring in the foundation, making comparisons to the geology types in the dam population
inconclusive. However, dams in glacial geological environments appear to be susceptible to piping accidents. Dams
constructed of or founded on glacial soils are more likely to
experience piping through the embankment or through the
foundation, respectively. Dams founded on rock which have
been affected by glaciation appear to be more likely to experience piping from the embankment into the foundation.
A disproportionate number of the failures by piping occurred on first filling, and at least two thirds of failures by
all three modes of piping occurred within the first 5 years of
operation of the dam.
Acknowledgements
The support of the 17 sponsors of the research project,
Dams Risk Assessment Estimation of the Probability of
Failure, and the Australian Research Council is acknowledged. The sponsors of the project are ACT Electricity and
Water, Department of Land and Water Conservation, Electricity Corporation New Zealand, Goulburn Murray Water,
Gutteridge Haskins and Davey (GHD), Hydro Electric Commission Tasmania, Melbourne Water Corporation, NSW Department of Public Works and Services, NSW Dam Safety
Committee, Pacific Power, Queensland Department of Natural Resources, Snowy Mountain Engineering Corporation,
Snowy Mountain Hydro-electric Authority, South Australia
Water Corporation, Sydney Water Corporation (Australian
Water Technologies), and Western Australia Water Corporation. The assistance from other organisations that allowed
1023
access to their reports for the research project is also

acknowledged. These include the United States Bureau of
Reclamation (USBR), British Columbia Hydroelectric and
Power Corporation (BC Hydro), Norwegian Geotechnical
Institute, and Alberta Dam Safety.
References
ANCOLD. 1992. Status of dam safety in Australia. ANCOLD Bulletin, Issue No. 91, August 1992, pp. 929.
Blind, H. 1983. The safety of dams. Water Power and Dam Construction (May), pp. 1721.
Building Research Establishment. 1990. An engineering guide to
the safety of embankment dams in the United Kingdom. Report
CI/SfB 187, Building Research Establishment, Garston, U.K.
Charles, J.A., and Boden, J.B. 1985. The failure of embankment
dams in the United Kingdom. In Proceedings of the Symposium
on Failures in Earthworks. Thomas Telford, London, pp. 181
202.
Cooke, J.B. 1984. Progress in rockfill dams. Journal of
Geotechnical Engineering, ASCE, 110(10): 13831414.
Fell, R., MacGregor, J.P., and Stapledon, D.H. 1992. Geotechnical
engineering of embankment dams. A.A. Balkema, Rotterdam,
The Netherlands.
Foster, M.A. 1999. The probability of failure of embankment dams
by internal erosion and piping. Ph.D. thesis, School of Civil and
Environmental Engineering, University of New South Wales,
Sydney, Australia.
Foster, M.A., and Fell, R. 2000a. Assessing embankment dam filters which do not satisfy modern design criteria. Journal of
Geotechnical and Geoenvironmental Engineering, ASCE. (In
press.)
Foster, M., and Fell, R. 2000b. Use of event trees to estimate the
probability of failure of embankment dams by internal erosion
and piping. In Proceedings of the 20th ICOLD Congress on
Large Dams, Beijing, Question 76. Vol. 1, pp. 237260.
Foster, M.A., Fell, R., and Spannagle, M. 1998. Analysis of embankment dam incidents. UNICIV Report No. R-374, School of
Civil and Environmental Engineering, University fo New South
Wales, Sydney, Australia. ISBN 85841 349 3.
Foster, M., Fell, R., and Spannagle, M. 2000. A method for assessing the relative likelihood of failure of embankment dams by
piping. Canadian Geotechnical Journal, 37: 10251061.
Gomez Laa, G., Alonso Franco, M., and Hernandez, J.L.R. 1979.
Reflections on some incidents in Spanish dams. In Proceedings
of the 13th International Congress on Large Dams (ICOLD),
New Delhi, Vol. Q49, pp. 721740.
Gruner, E. 1963. Dam disasters. Proceedings of the Institution of
Civil Engineers, 24: 4760.
Gruner, E. 1967. The safety of reservoirs. In World dams today.
The Japan Dam Association, p. 104109.
ICOLD. 1974. Lessons from dam incidents. Complete edition. International Commission on Large Dams (ICOLD), Paris.
ICOLD. 1983. Deterioration of dams and reservoirs. International
Commission on Large Dams (ICOLD), Paris.
ICOLD. 1984. World register of dams. 3rd updating. International
Commission on Large Dams (ICOLD), Paris.
ICOLD. 1989. Rockfill dams with concrete facing. State of the art.
International Commission on Large Dams (ICOLD), Bulletin
70.
ICOLD. 1990. Dispersive soils in embankment dams. Review. International Commission on Large Dams (ICOLD), Bulletin 77.
2000 NRC Canada

1024
ICOLD. 1995. Dam failures statistical analysis. International Commission on Large Dams (ICOLD), Bulletin 99.
Ingles, O.G. 1988. A review of dam failure: past, present and future. In Flood insurance and relief in Australia. Edited by D.I.
Smith, H. Smith, and J.W. Smith. Australian National University, Canberra, pp. 159167.
Leps, T.M., Strassburger, A.G., and Meehan, R.L. 1978. Seismic
stability of hydraulic fill dams, Part I. Water Power and Dam
Construction (October), pp. 2736.
Olwage, H.J.W., and Oosthuizen, C. 1984. Lessons learnt from
South African dam incidents. In Safety of dams. Edited by J.L.
Serafim. A.A. Balkema, Rotterdam, The Netherlands, pp. 3339.
Ravaska, O. 1997. Piping susceptibility of glacial till. In Proceedings of the 19th International Congress on Large Dams, Florence. International Committee on Large Dams (ICOLD), Q73
R30, pp. 455471.
Schnitter, N.J. 1994. A history of dams the useful pyramids. A.A.
Balkema, Rotterdam, The Netherlands.
Serafim, J.L. 1981a. Elements for a thorough statistical analysis of
dam failures. Revista da Universidade de Coimbra, Coimbra,
Portugal.
Serafim, J.L. 1981b. Safety of dams judged from failures. Water
Power and Dam Construction (December), pp. 3235.
Sherard, J.L. 1953. Influence of soil properties and construction
methods on the performance of homogeneous earth dams. Technical Memorandum 645, United States Bureau of Reclamation
(USBR), Denver, Colo.
Silveira, A. 1984. Statistical analysis of deteriorations and failures
of dams. In Safety of dams. Edited by J.L. Serafim. A.A.
Balkema, Rotterdam, The Netherlands, pp. 5560.
Silveira, A. 1990. Some considerations on the durability of dams.
Water Power and Dam Construction (February), pp. 1928.

Skempton, A.W. 1990. Historical development of British embankment dams to 1960. In Clay barriers for embankment dams.
Thomas Telford, London, pp. 1552.
Snethlage, J.B., Scheidenhelm, F.W., and Vanderlip, A.N. 1958.
Rockfill dams: review and statistics. Journal of the Power Division, ASCE, 84(PO4): 1739 1 to 36.
Tavares, L.V., and Serafim, J.L. 1983. Probabilistic study on failure
of large dams. Journal of the Geotechnical Engineering Division, ASCE, 109(11): 14831486.
USBR. 1994. Maximum sections and earthwork control statistics
of earth-fill dams built by the Bureau of Reclamation. United
States Bureau of Reclamation (USBR), United States Department of the Interior, Denver, Colo.
USCOLD. 1975. Lessons from dam incidents, USA. Committee on
Failures and Accidents to Large Dams of the United States
Committee on Large Dams (USCOLD). American Society of
Civil Engineers, New York.
USCOLD. 1988. Lessons from dam incidents, USA. II. Committee
on Dam Safety of the United States Committee on Large Dams
(USCOLD). American Society of Civil Engineers, New York.
USCOLD Committee on Dam Safety. 1996. Anthology of dam
modification case histories. United States Committee on Large
Dams (USCOLD), Committee on Dam Safety, Denver, Colo.
Von Thun, J.L. 1985. Application of statistical data from dam failures and accidents to risk based decision analysis on existing
dams. Internal Report, United States Bureau of Reclamation
(USBR), Denver, Colo.
Von Thun, J.L. 1996. Understanding seepage and piping failures
the No. 1 dam safety problem in the west. In Proceedings of the
Association of Dam Safety Officials (ASDSO) Western Regional Conference, Lake Tahoe, Nev., pp. 124.
2000 NRC Canada

The Statistics of Embankment Dam Failures and Accidents

Transféré par

Droits d'auteur :

Formats disponibles

The Statistics of Embankment Dam Failures and Accidents

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

The Statistics of Embankment Dam Failures and Accidents

Transféré par

Droits d'auteur :

Formats disponibles

Color profile: Generic CMYK printer profile

Composite Default screen