INTERNAL EROSION OF EXISTING

DAMS, LEVEES AND DIKES, AND

THEIR FOUNDATIONS
BULLETIN 1XX
Volume 1: INTERNAL EROSION
PROCESSES AND ENGINEERING
ASSESSMENT

Photo: Courtesy of US Bureau of Reclamation

22 January 2013
 NOTICE – DISCLAIMER
The information, analyses and conclusions in this document have no legal force and
must not be considered as substituting for legally-enforceable official regulations. They are
intended for the use of expereinced professionals who are alone equipped to judge their
pertinence and applicability.
This document had been drafted with the greatest care but, in view of the pace of
change in sience and technology, we cannot guarantee that it covers all aspects of the topics
discussed.
We decline all responsibility whatsoever for how the information herein is interpreted
and used and will accept no liability for any loss or damage arising therefrom.
Do not read unless you accept this disclaimer without reservation.
Committee on Embankment Dams

Chairman:
Canada JEAN-PIERRE TOURNIER

Vice Chairman:
Russia VADIM RADCHENKO

Members:
Australia MICHAEL MARLEY
Austria PETER TSCHERNUTTER
Brazil CARLO VIOTTI
Bulgaria NETZO DIMITROV
China PROF NENG-HUI LI
Colombia ALBERTO MARULANDA
Egypt
Finland JUHA LAASONEN
France JEAN-JACQUES FRY
Germany MARKUS LIMBACH
Greece GEORGIOS DOUNIAS
India V K KAPOOR
Indonesia D JAWARDI
Iran NASSER TARKESH DOOZ
Italy PROF RUGGIERO JAPELLI
Japan T HORI
Norway HELGE SAXEGAARD
Pakistan K ALAMGIR
Portugal E MARANDA DAS NEVES
Slovakia MARIAN MISCIK
South Africa DANIE BADENHORST
Spain ANTONIO SORIANO
Sri Lanka L SOORIYABANDARA
Sweden INGVAR EKSTRÖM
Switzerland PETER BRENNER
Thailand A SRAMOON
Turkey NURETTIN PELEN
United Kingdom RODNEY BRIDLE
United States DAVID PAUL
Venezuela G MARTINEZ

Co-opted members:
Canada ERIC PELOQUIN
 TABLE OF CONTENTS
INTERNAL EROSION OF EXISTING DAMS, LEVEES AND DIKES, AND THEIR
FOUNDATIONS ........................................................................................... i
BULLETIN 1XX..................................................................................................................... i
Volume 1: INTERNAL EROSION PROCESSES AND ENGINEERING ASSESSMENT . i
TABLE OF CONTENTS ........................................................................................................ i
FOREWORD AND ACKNOWLEDGEMENTS ................................................................... i
EXECUTIVE SUMMARY ..................................................................................................... i
1. THE IMPORTANCE OF INTERNAL EROSION FOR DAM SAFETY, STATISTICS
OF FAILURE, AND THE OUTLINE OF THIS BULLETIN ..................... 3
1.1 THE PURPOSE, SCOPE AND OUTLINE OF THIS BULLETIN ................ 3
1.2 THE IMPORTANCE OF INTERNAL EROSION TO DAM SAFETY ........... 5
1.3 STATISTICS OF EMBANKMENT DAM FAILURES ................................... 6
2. AN OVERVIEW OF INTERNAL EROSION MECHANISMS ..................................... 8
2.1 A DESCRIPTION OF THE OVERALL PROCESS....................................... 8
2.2 THE MECHANICS OF EROSION – PARTICLE DETACHMENT .............. 9
2.3 THE FOUR MECHANISMS OF INITIATION OF EROSION ................... 10
2.3.1 Concentrated leaks ...................................................................................... 10
2.3.2 Backward erosion ....................................................................................... 14
2.3.2.1 Backward erosion piping ......................................................................... 14
2.3.2.2 Global backward erosion ......................................................................... 15
2.3.3 Contact erosion ........................................................................................... 15
2.3.4 Suffusion ..................................................................................................... 16
2.4 CONTINUATION AND FILTER ACTION ................................................. 17
2.5 PROGRESSION .......................................................................................... 18
2.6 DETECTION .............................................................................................. 18
2.7 INTERVENTION ........................................................................................ 20
2.8 BREACH ..................................................................................................... 21
3. INITIATION OF CONCENTRATED LEAK EROSION ............................................. 25
3.1 CONCENTRATED LEAK EROSION: THE OVERALL PROCESS ........... 25
3.2 INITIATION OF CONCENTRATED LEAK EROSION IN SITUATIONS
WHERE CRACKS, OPENINGS AND LOW STRESS ZONES MAY BE PRESENT IN AN
EMBANKMENT OR FOUNDATION................................................................................... 25
3.2.1 Cracking and hydraulic fracture due to cross valley differential settlement
of the core .................................................................................................................... 25

i
 3.2.2 Cracking and hydraulic fracture due to cross valley arching ..................... 26
3.2.3 Cracking and hydraulic fracture due to differential settlement in the
foundation under the core ................................................................................................. 26
3.2.4 Cracking and hydraulic fracture due to small scale irregularities in the
foundation profile under the core ..................................................................................... 26
3.2.5 Cracking and hydraulic fracture due to arching of the core onto the
shoulders of the embankment ........................................................................................... 26
3.2.6 Crack or gap adjacent to a spillway or abutment walls and where concrete
dams abut embankment dams ........................................................................................... 27
3.2.7 Internal erosion associated with conduits embedded in the embankment .. 27
3.2.8 Cracking or hydraulic fracture in poorly compacted layers in cohesive and
broadly graded till soils in embankments ......................................................................... 28
3.2.9 Cracking due to desiccation ........................................................................ 29
3.2.10 Transverse cracking caused by settlement during earthquakes ................. 29
3.2.11 Cracking or high permeability layers due to freezing ............................... 29
3.2.12 Internal erosion initiated by the effects of animal burrows and vegetation ..
................................................................................................................... 30
3.2.13 Relative importance of conduits, spillway walls cracking mechanisms, and
poorly compacted zones ................................................................................................... 31
3.3 ESTIMATION OF CRACK WIDTH AND DEPTH OF CRACKING ......... 32
3.4 THE MECHANICS OF EROSION IN CONCENTRATED LEAKS ............ 34
3.4.1 The procedure ............................................................................................. 34
3.4.2 The estimation of hydraulic shear stresses in cracks and erosion pipes ..... 34
3.4.3 Erosion properties of soils in the core of embankment dams - basic
principles .................................................................................................................... 35
3.4.4 Comparison of the hydraulic shear stress in the crack (  ) to the critical
shear stress which will initiate erosion for the soil in the core of the embankment (c) .. 37
3.5 DISPERSIVE SOILS AND THEIR IMPORTANCE IN CONCENTRATED
LEAK EROSION .................................................................................................................. 37
3.5.1 What are dispersive soils? .......................................................................... 37
3.5.2 How can dispersive soils be identified? ..................................................... 38
3.5.3 Effect of dispersion on erosion properties ................................................. 39
3.5.4 Effect of dispersion on filter design and assessment ................................. 39
4. INITIATION OF BACKWARD EROSION .................................................................. 40
4.1 BACKWARD EROSION PIPING LEADING TO FORMATION OF A PIPE
.................................................................................................................... 40
4.1.1 The overall process ..................................................................................... 40
4.1.2 The process in sandy dam or dike foundations........................................... 40
4.1.3 Backward erosion piping within a dam, levee or dike ............................... 42

ii
 4.1.4 Soils which are subject to backward erosion piping leading to formation of
a pipe .................................................................................................................... 42
4.2 METHODS FOR PREDICTING WHETHER BACKWARD EROSION
PIPING WILL INITIATE AND PROGRESS ........................................................................ 43
4.2.1 Terzaghi and Peck (1948) ........................................................................... 43
4.2.2 Method of Sellmeijer and co-workers at Deltares ...................................... 43
4.2.3 Hoffman’s Hydraulic Approach ..................................................................... 46
4.2.4 Russian Design Methods ............................................................................ 47
4.2.5 Other Methods ............................................................................................ 49
4.3 SOME FIELD OBSERVATIONS ................................................................ 50
4.4 GUIDANCE ON WHETHER THE OVERLYING SOIL WILL FORM A
ROOF TO THE PIPE ........................................................................................................... 51
4.5 THE PROCESS OF GLOBAL BACKWARD EROSION IN CENTRAL
CORE DAMS .................................................................................................................... 51
4.6 GLOBAL BACKWARD EROSION, LEADING TO UNRAVELING, AT A
SEEPING DOWNSTREAM SURFACE OF THE EMBANKMENT ..................................... 52
5. INITIATION OF CONTACT EROSION ...................................................................... 53
5.1 THE OVERALL PROCESS ......................................................................... 53
5.2 FAILURE MODES AND VULNERABLE DAM TYPES ............................. 53
5.3 METHODS TO ESTIMATE LOADS AND RESISTANCE TO CONTACT
EROSION .................................................................................................................... 54
5.3.1 Fine cohesionless soil below a coarse soil layer......................................... 55
5.3.2 Silt and clay erosion (particles < 75µm)..................................................... 57
5.3.3 Fine soil above a coarse soil layer .............................................................. 58
5.3.4 Influence of uniformity............................................................................... 58
6. INITIATION OF SUFFUSION ...................................................................................... 59
6.1 GENERAL DESCRIPTION OF THE PROCESSES ................................... 59
6.2 METHODS OF IDENTIFYING SOILS WHICH ARE INTERNALLY
UNSTABLE AND POTENTIALLY SUBJECT TO SUFFUSION (CRITERION 1:
GEOMETRIC AND CRITERION 2: FILLED OR UNDERFILLED VOIDS) ...................... 60
6.2.1 General requirements .................................................................................. 60
6.2.2 Some methods for assessing whether a soil is subject to suffusion............ 62
Kenney and Lau method ...................................................................................... 62
Wan and Fell adaptation of the Burenkova method ............................................ 64
Some general comments ...................................................................................... 65
6.3 ASSESSMENT OF THE LARGEST ERODIBLE PARTICLES IN
SUFFUSION .................................................................................................................... 66
6.4 ASSESSMENT OF SEEPAGE GRADIENT WHICH WILL CAUSE
SUFFUSION (CRITERION 3) ............................................................................................. 66

iii
 6.5 ALLOWING FOR THE EFFECTS OF SEGREGATION WHEN
ASSESSING SUFFUSION ................................................................................................... 67
7. CONTINUATION OR INTERRUPTION OF EROSION BY FILTER ACTION ........ 69
7.1 INTRODUCTION ....................................................................................... 69
7.2 CONTINUATION SCENARIOS ................................................................. 69
7.2.1 Internal erosion in the embankment ........................................................... 69
7.2.2 Internal erosion in the foundation ............................................................... 72
7.2.3 Internal erosion of the embankment at or into the foundation ................... 72
7.3 ASSESSMENT OF THE LIKELIHOOD OF CONTINUATION WHERE A
FILTER/TRANSITION ZONE IS PRESENT DOWNSTREAM OF THE CORE OR THERE
IS A DOWNSTREAM SHOULDER ZONE WHICH IS NOT CAPABLE OF HOLDING A
CRACK .................................................................................................................... 73
7.4 FILTER FUNCTIONS ................................................................................ 74
Filter Retention .................................................................................................... 74
No cohesion (filter will not hold a crack) ............................................................ 76
Self filtration or stability ..................................................................................... 76
Drainage .............................................................................................................. 76
Strength................................................................................................................ 77
7.5 BRIEF DESCRIPTION OF SOME OF THE AVAILABLE FILTER DESIGN
CRITERIA FOR CRITICAL FILTERS ................................................................................. 77
Particle Size Based Criteria-Sherard and Dunnigan (1989) ................................ 77
Methods based on constriction size or opening size ........................................... 78
Methods based on the permeability of the filter .................................................. 78
7.6 FOSTER AND FELL (1999a, 2001) METHOD FOR ASSESSING THE
LIKELIHOOD OF CONTINUATION OF EROSION FOR FILTERS AND TRANSITIONS
WHICH DO NOT MEET MODERN FILTER DESIGN CRITERIA .................................... 80
7.7 ASSESSMENT OF THE LIKELIHOOD OF CONTINUATION FOR
INTERNAL EROSION INTO AN OPEN DEFECT, JOINT OR CRACK IN THE
FOUNDATION, IN A WALL OR CONDUIT ....................................................................... 84
8. PROGRESSION OF EROSION..................................................................................... 85
8.1 THE TWO CONDITIONS FOR PROGRESSION ...................................... 85
8.1.1 Hydraulic condition .................................................................................... 85
8.1.2 Mechanical condition ................................................................................. 85
8.2 OVERALL APPROACH TO ASSESSING PROGRESSION OF
CONCENTRATED LEAK EROSION ................................................................................... 86
8.2.1 Assessing whether the soil will hold a roof to a developing pipe .............. 86
8.2.2 Assessing whether crack filling action will occur ...................................... 87
8.2.3 Assessing whether upstream flow limitation will occur ............................. 88

iv
 8.3 ASSESSING THE RATE OF DEVELOPMENT OF A CONCENTRATED
LEAK EROSION PIPE ......................................................................................................... 88
8.3.1 Rapid enlargement of concentrated leak erosion pipes and need for filters
or other works to prevent continuation and progression .................................................. 89
8.4 RATE OF PROGRESSION OF INTERNAL EROSION GENERALLY ...... 89
9. ENGINEERING ASSESSMENT OF THE VULNERABILITY OF A DAM TO
INTERNAL EROSION .............................................................................. 90
9.1 EIGHT STEPS TO FAILURE BY INTERNAL EROSION .......................... 90
9.2 APPROACHES TO ENGINEERING ASSESSMENT ................................. 91
9.2.1 Deterministic or Standards-Based Approach ............................................. 91
9.2.2 Engineering Judgment ................................................................................ 92
9.2.3 Quantitative Risk Assessment .................................................................... 92
9.2.4 Use of the three approaches ........................................................................ 93
9.3 ENGINEERING ANALYSIS PROCEDURE ............................................... 93
9.4 RESERVOIR AND SEISMIC LOADING CONDITIONS ........................... 94
9.4.1 The effect of reservoir level on internal erosion......................................... 94
9.4.2 Reservoir levels to be considered in the engineering assessment of internal
erosion .................................................................................................................... 95
9.4.3 Seismic loads to be considered ...................................................................... 95
9.5 DESIGN STANDARDS AND THE LIKELIHOOD OF FAILURE FROM
INTERNAL EROSION .......................................................................................................... 96
9.6 SAFETY MARGINS AND APPROXIMATIONS ......................................... 96
9.7 SOIL PROPERTIES IN RELATION TO INTERNAL EROSION ................ 97
9.7.1 Material susceptibility, stress conditions and hydraulic load ..................... 97
Hydraulic Fracture ............................................................................................... 98
9.7.2 The physical process ................................................................................... 98
Three classes of soil in relation to internal erosion ............................................. 98
Degree of saturation ............................................................................................ 99
Dispersion properties of the soil .......................................................................... 99
Effect of soil structure ....................................................................................... 100
Effect of testing method on erosion rate index .................................................. 100
Effect of shear strength of the soil ..................................................................... 100
9.8 THE IMPORTANCE OF HAVING RELIABLE INFORMATION UPON
WHICH TO MAKE THE ASSESSMENT OF INTERNAL EROSION ................................ 100
Geometric model ............................................................................................... 100
Geological model of the foundation .................................................................. 100
Geotechnical model of the embankment and foundations ................................ 101

v
 Hydraulic or seepage model .............................................................................. 101
Stress state in the dam and its foundation ......................................................... 101
Staged approach and collection of data ............................................................. 101
9.9 FAILURE MODES ANALYSIS ................................................................. 101
9.10 SCREENING OF POTENTIAL FAILURE MODES ................................. 105
9.10.1 Screening of PFM on the zoning of the dam and the properties of the core
of the embankment. ........................................................................................................ 105
(a) Zoning of the embankment .......................................................................... 105
(b) Screening of PFM on the properties of the core of the embankment .......... 105
9.10.2 Screening of PFM on foundation geology and properties ....................... 106
9.10.3 Screening of PFM on details of the embankment and conduits and
retaining walls................................................................................................................. 107
9.11 ANALYSIS OF POTENTIAL FAILURE MODES AT WHICH INTERNAL
EROSION COULD INITIATE, CONTINUE AND PROGRESS ........................................ 110
9.11.1 Some general principles .......................................................................... 110
9.11.2 Concentrated leak erosion-some aids to judgment .................................. 112
9.12 DECISION MAKING: ARE ANTI-EROSION MEASURES NECESSARY
OR WILL MONITORING BE SUFFICIENT? ................................................................... 112
10. REMEDIATION AND IMPROVEMENTS TO DAMS TO RESIST INTERNAL
EROSION ................................................................................................. 113
10.1 METHODS AVAILABLE .......................................................................... 113
10.2 BARRIERS ................................................................................................ 113
10.3 FILTERS ................................................................................................... 114
10.3.1 Filter blankets and filtered berms ............................................................ 114
10.3.2 Filter collars on conduits and spillway channels ..................................... 114
10.3. 3 Filter trenches ......................................................................................... 114
10.3.4 Filtered relief wells .................................................................................. 114
10.3.5 Multiple filters to provide drainage capacity........................................... 114
10.4 CONDITION OF CONDUITS .......................................................................... 115
11. SURVEILLANCE AND MONITORING .................................................................... 116
11.1 OBJECTIVES AND CHALLENGES OF SURVEILLANCE AND
MONITORING FOR TIMELY DETECTION OF INTERNAL EROSION ......................... 116
11.2 GENERAL PRINCIPLES .......................................................................... 117
11.2.1 The likelihood of detection and intervention........................................... 118
11.2.2 Some Information on the Rate of Internal Erosion ................................. 119
11.2.3 A review of methods of monitoring and detection and their applicability to
internal erosion modes .................................................................................................... 119

vi
12. METHODS FOR ASSESSMENT OF THE LIKELIHOOD OF FAILURE OF A DAM
BY INTERNAL EROSION ..................................................................... 125
12.1 INTRODUCTION ..................................................................................... 125
12.2 GENERAL PROCEDURE ........................................................................ 125
12.3 ESTIMATION OF LIKELIHOODS OF FAILURE FOR THE POTENTIAL
FAILURE MODES APPLICABLE TO THE DAM............................................................. 126
12.3.1 Some general principles .......................................................................... 126
12.3.2 Concentrated leak erosion-some aids to judgment .................................. 126
12.4 QUANTITATIVE AND SEMI QUANTITATIVE RISK ASSESSMENT
METHODS .................................................................................................................. 126
12.4.1 Methods based on historic databases of failures and incidents ............... 126
12.4.1.1 Interim Guide to Quantitative Risk Assessment for UK Reservoirs 126
12.4.1.2 UNSW Method ................................................................................. 127
12.4.2 Event tree methods .................................................................................. 128
13. TERMINOLOGY – INTERNAL EROSION PROCESSES – INTERNATIONAL
GLOSSARY ............................................................................................. 130
13.1 TERMINOLOGY USED IN RELATION TO INTERNAL EROSION........ 130
13.2 DEFINITIONS RELATED TO THE ANALYSIS OF THE PROBABILITY
OF FAILURE .................................................................................................................. 136
13.3 ICOLD DEFINITIONS OF LARGE DAM, FAILURE, ACCIDENT AND
INCIDENT .................................................................................................................. 136
REFERENCES .................................................................................................................. 138

vii
 FOREWORD AND ACKNOWLEDGEMENTS
Internal erosion and piping in embankments and their foundations is the main cause of
failures and accidents at embankment dams. For new dams, the potential for internal erosion
can be controlled by good design and construction of the core of the dam and provision of
filters to intercept seepage through the embankment and the foundations.
However many existing dams, dikes and levees were inadequately zoned and not
provided with filters or have filters or transitions which do not satisfy modern filter design
criteria, and are susceptible to internal erosion failure, with a risk of internal erosion
increasing with ageing. Accordingly, the reassessment of the safety of these dams is an
important issue.
This has been recognized by the European Club of ICOLD which devoted a Working
Group to the problem and much research has been carried out in recent years to better
understand the physical processes and mechanics of internal erosion. This has built upon the
knowledge gained from many years of earlier research and successful design of new dams.
The ICOLD committee on Embankment Dams believe that there are considerable
benefits in gathering together international experts and researchers on this topic. Thus the
chairman committed the members Jean-Jacques Fry (France) and Rodney Bridle (UK) to
work with the Working Group of the European Club and other experts on internal erosion to
write a bulletin and give guidance for practitioners responsible for assessing the safety of
water retaining structures.
We expect that the people in charge of dam safety will have free access to this
publication so it can be used effectively for the safety assessment of existing dams and dikes.
It will also be useful for those designing new dams and dikes.
As the state of the art of internal erosion of dams, dikes and their foundations is an
evolving science readers should look in the literature for new developments following the
production of this Bulletin.
I also wish to record special thanks to Dr Jean-Jacques Fry, Mr Rodney Bridle, and
Emeritus Professor Robin Fell (Australia) for undertaking much of the drafting of this
Bulletin, to members of the European Club Working Group for their inputs in developing the
Bulletin, and to members of the ICOLD committee for their contributions.

Jean-Pierre TOURNIER,
Chairman,
Committee on Embankment Dams
 EXECUTIVE SUMMARY
Internal erosion is one of the major causes of embankment dam failure. When
constructing new dams, protection against internal erosion is provided by zoning and by
providing filters (ICOLD Bulletin No. 95, ICOLD, 1994). However, many existing dams are
not adequately zoned and do not have filters and may therefore be vulnerable to internal
erosion. Others have filters not designed and/or constructed to modern standards, they too
may be vulnerable to internal erosion.
Internal erosion occurs when soil particles within an embankment dam or its foundation
are carried downstream by seepage flow. It starts when the erosive forces imposed by the
hydraulic loads exceed the resistance of the materials in the dam to erosion. The erosive
forces are directly related to reservoir water level.
Many internal erosion problems become apparent on first filling but most occur many
years after first filling. This is almost certainly because the dam is subject for the first time to
high water levels which may initiate erosion in the upper parts of the dam. After long periods
of flood flows, for example, which cause the reservoir to rise higher on the dam and/or to
subject it to greater gradients and velocities of flow than it has experienced before. It may be
because ageing causes deterioration of the dam, particularly of the pipes and culverts through
dams but also due to on-going settlements under repeated reservoir draw downs, or from the
effect of earthquakes. Another factor may be that some internal erosion mechanisms progress
slowly in an episodic fashion under cyclical reservoir loading, taking years and even decades
before the effects of the episodes of erosion show any of the symptoms of internal erosion.
The Bulletin is in two volumes. Volume 1 deals predominantly with internal erosion
processes and the engineering assessment of the vulnerability of a dam to failure or damage
by internal erosion, with a brief oversight of monitoring for and detection of internal erosion
and remediation to protect dams against internal erosion. It includes a comprehensive listing
of the Terminology used in internal erosion. Many references are also given, including links
to many from an ICOLD internal erosion webpage.
Volume 2 gives more details of internal erosion investigations, and appropriate testing,
monitoring and detection, and remediation, and gives case histories.
This Volume 1 gives a statement of the problem, explaining why internal erosion is a
threat to existing dams and the importance of assessing the vulnerability of individual dams
to it. It then goes through the overall process of erosion from initiation, through continuation
(or arrest) of erosion, through progression, and on to breach, unless erosion is detected early
enough by appropriate monitoring systems, to allow timely intervention to halt or slow the
development of a breach and failure.
The four mechanisms through which internal erosion is initiated are described: erosion
in concentrated leaks, backward erosion, contact erosion, and suffusion. The Bulletin then
describes the circumstances in which erosion, once initiated, will continue or be arrested by
the filtering capability of the materials in the dam and foundations. Filtering processes, filter
properties and filter rules are explained in order that the filtering capability can be assessed.
The rate at which erosion continues, if it is not arrested, is discussed. In many cases, the rate
is such that failure could occur too rapidly to give time to protect lives and property
downstream, and some positive action should be taken, usually retrofitting filters, to prevent
erosion.
i
 An engineering assessment process is proposed. Assessment by the deterministic
approach, by engineering judgment augmented by qualitative and semi-quantitative risk
assessment; or by quantitative risk assessment, is described. Critical design loads relate to
critical water levels. Systematic consideration of potential failure modes is required,
eliminating those in which internal erosion could not continue to failure, because of fill and
foundation soil properties, the zoning of the dam and other factors, as detailed in the Bulletin.
The remaining potential failure modes must then be examined. Information is given in
the Bulletin to determine the erosion resistance, the hydraulic gradients and the soil properties
at which erosion will initiate. Information is also given on how to assess the filtering
capability of the materials in the dam and foundation and their ability to stop erosion, allow a
limited amount of erosion which would then stop, or allow erosion to continue unchecked.
Some information on remediation methods and monitoring is given to assist in decision
making where the assessment has not given a definite conclusion on whether damaging
erosion will or will not occur at the dam. Quantitative risk assessment methods are briefly
described. Such methods may also be used to resolve uncertainty, and would be advisable in
the case of high hazard dams and high remediation costs to confirm and justify the need for
and the scale of the proposed remediation.

ii
 1. THE IMPORTANCE OF INTERNAL
EROSION FOR DAM SAFETY,
STATISTICS OF FAILURE, AND THE
OUTLINE OF THIS BULLETIN
1.1 THE PURPOSE, SCOPE AND OUTLINE OF THIS
BULLETIN
Internal erosion occurs when soil particles within an embankment dam or its foundation
are carried downstream by seepage flow.
Internal erosion is a major cause of dam failures and of incidents that threaten the safety
of dams. Most internal erosion failures have occurred on first filling as weaknesses in the
dam or its foundation were ‘discovered’ by the rising water. However, failures do occur after
first filling and internal erosion remains a threat to existing dams because:
 They have not been subjected to, or designed to resist, extreme loads such as
extreme water level and earthquakes (which can cause settlement
cracking),Cracking occurs as a result of cycling of water level, differential
settlement and desiccation,
 Ageing causes deterioration, particularly deterioration of conduits, spillways and
other structures through dams, at which internal erosion may be initiated
 They may not be protected against internal erosion by filters, or if filters or
transition zones are present, they may not have been designed to modern standards
and may be ineffective.
The implications of internal erosion and assessment of the vulnerability of dams to it
have been addressed in recent years by risk assessment methods pioneered in Australia, North
America and Europe (e.g. ICOLD, 2005; Fell & Fry, 2007; Hartford & Baecher, 2004;
Brown & Gosden, 2004; Bowles et al, 2003). The end product of the methods is an estimate
of the probability of failure of the subject dam. When coupled with a dambreak study from
which likely loss of life and cost of damage from a failure can be estimated, an assessment of
whether the subject dam is safe enough can be made based on assessments of acceptable
societal risk. If it is not, the effects of safety works and improved monitoring can be assessed
and implemented as necessary.
In developing the risk assessment approach researchers have improved understanding
of the mechanisms that can result in dam failure by internal erosion. This improved
understanding, when added to existing knowledge of seepage forces and hydraulic conditions
in an embankment dam and its foundations, makes it possible to assess the ability of a dam to
resist the forces imposed on it that may cause internal erosion to initiate, continue and
progress to failure. This bulletin is based on the application of this improved understanding of
the mechanics of internal erosion to assess whether existing dams will resist erosion, whether
safety works are required to resist it or whether improved monitoring will be a sufficient
safeguard against it.
The objective of the bulletin is to give practical guidance for dam owners, dam
engineers and regulators on dealing with the threat of internal erosion at existing dams. It
explains the mechanics of internal erosion and how to apply them to make an engineering
___________________________________________________________________________
3
analysis of the ability of dam to withstand extreme internal erosion loads. It recommends a
staged iterative approach to the analysis, gradually improving data inputs if necessary to
reach safe conclusions. It advises broadly on safety works should they be needed and on
appropriate monitoring regimes.
It advises on the selection of critical internal erosion loads and notes that these occur
when water level is highest, the same as that occurring during the passage of the critical
flood.
The engineering analyses are intended to be used independently, but quantitative risk
assessments will complement them, and for important dams it is recommended that both
approaches be used.
The Bulletin was written largely for assessing existing dams, but the principles may be
applied to new dams. The Bulletin was written mainly for large dams, but is applicable to
small dams, dikes, levees and their foundations.
The Bulletin includes:
Statistics on dam failures and accidents caused by internal erosion
An overview of the internal erosion process, including the need to consider the
complete process from application of the load, to initiation of erosion and through to
the potential for continuation and progression of erosion and breach of the dam.
Details of methods which are currently available for assessing if concentrated
leak erosion, backward erosion, contact erosion and suffusion will initiate and the
ability of dams to resist erosion if it has initiated, and a summary of on-going research
into these processes.
Advice on engineering assessment of the vulnerability of dams to internal erosion
by three approaches: deterministic; engineering judgment backed by qualitative and
semi-quantitative risk assessment; and quantitative risk assessment. Advice is also
given on identifying Potential Failure Modes and screening out improbable ones.
Advice on remedial measures which may be taken if needed to make the dam
resistant to internal erosion; and on methods for monitoring and surveillance of internal
erosion to assist in decision-making if the results of the engineering analyses are not
clear.
Brief descriptions of quantitative risk assessment methods are given and it is
suggested that such methods may also be used to resolve uncertainty and would be
advisable to justify the need for and scale of remediation at high hazard dams.
Terminology commonly used in the internal erosion process, and in risk
assessment of internal erosion.
The sound assessment of potential internal erosion mechanisms has to be based on
adequate subsurface exploration, assessment of data from site investigations prior to the
construction of the dam, construction records, and analysis of the data from instruments
within the dam and its foundation, and include field and laboratory testing. A staged approach
is recommended, collecting and applying more data gradually if needed to decide if safety
works are needed to make the dam adequately safe.
Volume 2 of the Bulletin provides details of laboratory tests which measure properties
of erosion and filter capability of soils; monitoring methods including several newly
developed techniques, and information on case studies relevant to the subject.

___________________________________________________________________________
4
1.2 THE IMPORTANCE OF INTERNAL EROSION TO
DAM SAFETY
Internal erosion is an important safety issue for large and small dams, dikes and levees
as shown by the statistics of historic of failures and incidents which are discussed below.
Internal erosion occurs when soil particles are carried by seepage in embankments,
foundations, from embankment to foundation, around and into conduits through
embankments and adjacent walls supporting embankments.
The statistics of embankment dam incidents show that internal erosion is a significant
cause of incidents and to a lesser extent failure for older dams. Incidents include new or
increased seepage and leakage, sinkholes and accelerating settlements of the dam. There is
also evidence that incidents and failures are more likely along the interface between the
embankment fill and structures such as culverts and spillway walls. An understanding of
internal erosion is important to dam owners as this:
 Is necessary to understand and interpret the behavior of the dam

 Is necessary to assess the safety of the dam, and whether it is safe enough or whether safety
works are needed to make it so

 Should govern the dam safety surveillance and monitoring regimes

The cause of these incidents can include poor compaction and incorrectly designed
filters which even for many relatively modern dams may not be adequately designed or
constructed to control internal erosion. This is particularly important in the upper parts of
dams where cracking is most likely to occur and filters are often omitted or were not
constructed to modern standards, or were not included at the interface with structures. For
many dams revised flood forecasts indicate the upper parts may now be required to hold the
reservoir under conditions for which they were not designed.
Four mechanisms of initiation have been identified and are described in the Bulletin.
The assessment of filters and transition zones and downstream zones to act as filters is well
understood and in many dams it can be demonstrated that these will provide adequate filter
protection even if they do not meet modern filter design criteria. Tools to carry out an
assessment of the ability of the dam to resist internal erosion are also included in this
Bulletin.
Many new and potentially useful methods for monitoring and detecting internal erosion
and seepage have been developed in recent years, and are described in the accompanying
Bulletin on Detection and Monitoring of Internal Erosion. These and long established
methods such as visual inspection, leakage measurement and monitoring of pore pressures
with piezometers are valuable tools in keeping dams safe by detecting internal erosion early
in the process.

___________________________________________________________________________
5
1.3 STATISTICS OF EMBANKMENT DAM FAILURES
Based on the records of dam incidents and the dam register in ICOLD (1974, 1983,
1995), Foster et al, (1998, 2000a) analyzed the statistics of dam failures for large dams
constructed between 1800 and 1986, excluding dams constructed in Japan pre 1930 and in
China (adapted from Foster et al, 1998, 2000a). This is summarized in Table 1.1 below:

Table 1.1 Statistics of embankment dam failures

Failure Mechanism Erosion Embankment Sliding
External erosion
Mode of failure Internal erosion Static Instability Seismic instability
(overtopping)
% over the world 48% 46% 4% 2%
% over the world 94% 6%

It can be seen that internal erosion has been responsible for about half of embankment
dam failures where the mode of failure is known. It is approximately equal in importance to
failures by overtopping in floods due to inadequate spillway capacity and malfunction of
gates and other outlets leading to overtopping.
By comparison embankment slides and failures due to earthquakes account for only 4%
and 1.7% of embankment dam failures in operation respectively.

Table 1.2.Historical frequencies of failures and accidents in embankments of large

dams constructed from 1800 to 1986, excluding dams constructed in Japan pre 1930, and in
China (adapted from Foster et al, 1998, 2000a)
AROUND
CASE TOTAL IN DAM CONDUITS OR
WALLS
Internal erosion failures 36 19 17
Internal erosion accidents 75 52 23
Seepage accidents with no detected erosion 36 30 6
Total no. of failures and accidents 146 101 46
Population of dams 11192 11192 5,596
Historical frequency for failures and accidents 0.013 0.009 0.0082
Proportion of failures and accidents on first fill 36%
Proportion of failures and accidents after first 64%
fill
Historical frequency for first fill 0.0032 0.0030
Historical frequency after first fill 0.0058 0.0052
Historical annual frequency after first fill 4 4
2.2 x 10 2.0 x 10

As Table 1.2 shows, overall about 1 in 80 embankment dams have failed or

experienced an accident by internal erosion. There are about 4 failures per 10,000 dams per
year from all causes about 2 of the failures are from internal erosion.
The largest number of failures has occurred in the embankment. Nearly half of these are
associated with conduits which penetrate through the embankment, or walls which support
the embankment.
There are many more accidents than failures. As defined by ICOLD (see Terminology),
accidents are incidents which have been prevented from becoming failures by immediate
remedial measures, including possibly drawing down the water in the reservoir. There are a
similar number of accidents relating to internal erosion in the foundation as in the
embankment.

___________________________________________________________________________
6
 For all internal erosion modes about two-thirds of all failures and about half of
accidents occur on first filling or in the first 5 years of operation. This means however that
around half of all incidents have occurred in dams after 5 years of operation.
Foster et al (1998, 2000a) found that nearly all internal erosion failures in the
embankment occurred when the reservoir level was at the highest level ever, or within one
meter of that level. For internal erosion in the foundation the reservoir level was not as
important.
There is evidence from the ICOLD data in Table 1.3 below that the later statistics are
better reflecting improved design and construction methods. However the problems are on-
going.
Table 1.3: Failure statistics for embankment dams excluding China from 1970 to 1989
(adapted from Foster et al, 1998, 2000a)
Modes of failure Ratio of failures to the Ratio of failures to the
number of dams 1970-1979 number of dams 1980-1989
Internal erosion 0.0020 0.0016
Overtopping and appurtenant structure failure 0.0026 0.0019
resulting in overtopping
Sliding 0.0004 0.0001

There has not been a systematic study of international failure and accident statistics
since the ICOLD (1995) study. However there have been studies for individual countries. Li
(2012) provides statistics of dam failures in China.

Brown and Gosden (2008) estimated that in the United Kingdom’s population of about
2,500 dams registered under the Reservoirs Act, there are about 1,600 incidents for every
recorded failure. About 60% of the incidents appear to be related to internal erosion, with
typically two serious erosion incidents a year. It should be noted that the definition of
incident used in this study is not the same as in ICOLD (see Terminology). It includes
incidents which are unlikely to have resulted in failure. The “severe incidents” however
would classify as ICOLD accidents.
Engemoen and Redlinger (2009) and Engemoen (2011, 2012a) carried out detailed
analyses of internal erosion incidents for the Bureau of Reclamation’s 220 embankment
dams. They found that the successful performance of a dam does not preclude it from having
an internal erosion incident; and that 99 dams had experienced incidents, of which only one
was a failure; 53 involved definite particle transport, and the remainder were excessive
seepage or sand boils. Of the 99 incidents, 9 were in the embankment, 70 in the foundation, 6
from embankment into a conduit, 5 were into or along a conduit and 11 were into toe or
under drains. The dams store water for irrigation purposes and are usually filled to ‘normal
high pool level’ (to with 7-12 feet of crest) at the beginning of each year. Thus, “full” or
“high” pool levels occur in most years. They found that internal erosion can occur at any time
in a dam’s life; but it is likely that most incidents occurred when the pool was relatively close
to normal high level. They developed annualized statistics of failures and incidents from
these data. The definition of “incident” used by these authors is not the same as the ICOLD
definition. Schaefer et al (2011) describe some recent internal erosion incidents in USACE
dams.

___________________________________________________________________________
7
 2. AN OVERVIEW OF INTERNAL
EROSION MECHANISMS
2.1 A DESCRIPTION OF THE OVERALL PROCESS
Failures and incidents by internal erosion of embankment dams and their foundations
are categorized into three general failure modes, which are;
 Internal erosion through the embankment, which includes internal erosion associated with
through-penetrating structures, such as conduits associated with outlet works, spillway walls or
adjoining a concrete gravity structure supporting the embankment.

 Internal erosion through the foundation, and

 Internal erosion of the embankment into or at the foundation. Including (a) seepage through the
embankment eroding material into the foundation, or (b) seepage in the foundation at the
embankment contact eroding the embankment material.

The process of internal erosion may be broadly broken into four phases:
 initiation of erosion,

 continuation of erosion,

 progression to form a pipe or occasionally cause surface instability (sloughing), and

 Initiation of a breach.

This is shown in Figure 2.1(A) for internal erosion through the embankment initiated
by a concentrated leak. Similar processes apply for piping through the foundation, and from
the embankment to the foundation and are shown in Figure 2.1(B) and (C).

INITIATION 
CONTINUATION 
PROGRESSION 
BREACH
Concentrated leak Continuation of Enlargement of Breach
forms, erosion erosion concentrated leak mechanism forms
initiates along walls of
crack
(A) INTERNAL EROSION IN THE EMBANKMENT INITIATED BY EROSION IN
A CONCENTRATED LEAK

___________________________________________________________________________
8
 INITIATION 
CONTINUATION PROGRESSION BREACH

 
Leakage exits from the Continuation of Backward erosion Breach
foundation and backward erosion in progresses to mechanism
erosion initiates form a pipe forms
(B) INTERNAL EROSION IN THE FOUNDATION INITIATED BY
BACKWARD EROSION

INITIATION 
CONTINUATION 
PROGRESSION 
BREACH
Leakage exits the core into Continuation of Backward erosion Breach
the foundation and erosion progresses to form mechanism
backward erosion initiates a pipe. Eroded soil forms
as core erodes into the is transported in the
foundation foundation
(C) INTERNAL EROSION FROM THE EMBANKMENT TO FOUNDATION
INITIATED BY BACKWARD EROSION
Figure 2.1 - Models for the development of failure by internal erosion (Foster and Fell
1999b)

2.2 THE MECHANICS OF EROSION – PARTICLE

DETACHMENT
The first condition for internal erosion to occur is particle detachment. Water seeping
through the dam or flowing in cracks must provide sufficient energy to detach particles from
the soil structure.
The nature of the soil in the dam determines its vulnerability to erosion. Three classes
have to be distinguished:
 Non plastic (cohesionless) soils such as silts, sands and gravels. These may collapse when
saturated under flooding, will not sustain a crack when saturated, and are relatively easily eroded.
As non-plastic soils become coarser they progressively require more energy to initiate erosion.
These soils are subject to backward erosion, contact erosion or suffusion depending on their
particle size distribution.

 Plastic soils, such as clays, clayey sands, and clayey sandy gravels are generally more resistant to
erosion than cohesionless soils. These soils are subject to concentrated leak erosion and contact
___________________________________________________________________________
9
 erosion. Clay soils will hold a crack even when saturated. Higher energy is generally required to
detach particles from cracks within a cohesive fill but the particles removed are small and easily
carried through the crack. Backward erosion and suffusion cannot occur in plastic soils under the
gradients normally experienced in dams and their foundations.

 Dispersive soils are plastic (clay) soils in which the chemistry of the seeping water causes
dispersion (de-flocculation) of the clay flocs, breaking them down into smaller, easily eroded,
particles.

More details are given in Section 9.7.2, including the influence of the degree of
saturation on soils when used as fill materials in dams.

2.3 THE FOUR MECHANISMS OF INITIATION OF

EROSION
Initiation of erosion occurs in four mechanisms:
 Concentrated leaks

 Backward erosion

 Contact erosion

 Suffusion

2.3.1 Concentrated leaks

Where there is an opening in plastic soil, and sometimes in unsaturated silt, silty sand
or silty sandy gravel, through which concentrated leakage occurs, the sides of the opening
may be eroded by the leaking water.
Such concentrated leaks may occur through a crack caused by differential settlement
during construction of the dam or in operation, by hydraulic fracture due to low stresses
around conduits or in the upper parts of the dam due to differential settlement, or through
desiccation at high levels in the fill. Frost action also creates cracks in dam crests. Figures
2.2, 2.3 and 2.4 show different situations where cracks and gaps may occur.
Concentrated leaks may also occur because of collapse settlement of poorly compacted
fill in the embankment, around conduits, and adjacent to walls. They may also occur due to
the action of animals burrowing into levees and small dams and tree roots rotting in dams and
forming holes.

___________________________________________________________________________
10
 Figure 2.2 Situations in which cracks may form from desiccation or from cross valley
differential settlement. Such settlement may lead to tension cracks, or tearing at steep
abutment slopes or steps; or the lateral strain resulting from the settlement may result in low
stress zones in which hydraulic fracture may occur.

___________________________________________________________________________
11
 Figure 2.3 Situations which may cause differential settlement in the foundation of dams
leading to cracking, lateral strains and low stress zones subject to hydraulic fracture (Fell et al
2008, based on Sherard et al, 1963)

___________________________________________________________________________
12
 spillway embankment crest

potential for gap to form

as dam fill settles

da
m
se f
ttle ill
s

dam foundation

potential for gap to form

as dam fill settles
spillway embankment crest

da ttle
change in slope
1
m s
se

fil
1
l

foundation

Figure 2.4 Situations where a gap may form between the dam fill and spillway wall (a)
Steep foundation adjacent to spillway wall; (b) Change in slope of the retaining wall. (Fell et
al, 2005)

___________________________________________________________________________
13
2.3.2 Backward erosion
There are two kinds of backward erosion:
 Backward erosion piping, at the back (upstream) end of a very thin pipe below a “roof”. It mainly
occurs in foundations, as in Figure 2.5, but may occur within embankments.

 Global backward erosion, leading to development of a near-vertical pipe in the core of an

embankment as in Figure 2.7.

Both occur in non-plastic soils.

2.3.2.1 Backward erosion piping

Figure 2.5 Backward erosion piping at back (upstream) end of a thin pipe below a
“roof” (Koenders and Sellmeijer 1992)

The erosion process begins at a free surface on the downstream side of a dam or dike.
This free surface may be in a ditch or other excavation penetrating into the eroding non-
plastic (cohesionless) soil or may form by first heaving of the plastic (cohesive) strata
overlying the non-plastic soil.
The process progresses beneath the dike or dam. For this to occur, the dike or dam, or
the cohesive strata must form a roof for the eroding “pipe”. The presence of backward
erosion piping is often exhibited by the presence of sand boils at the downstream side of the
dam or dike. Figure 2.6 shows an example of a sand boil.

Figure 2.6 Sand boil on the land side of a dike.

Backward erosion piping occurs where critically high hydraulic gradients at the toe of a
dam erode particles upwards and backwards below the dam through small erosion channels
and flow velocity can transport the eroded particles downstream.

___________________________________________________________________________
14
2.3.2.2 Global backward erosion
There is a potential for global backward erosion in narrow or even reasonably wide
sloping or partially sloping core dams, such as that shown in Figure 2.7, if they are not
properly protected by filters or a transition zone.

Reservoir Surface Sinkhole Developed by Backward

Erosion

Figure 2.7 Potential mode of ‘global backward erosion’ leading to formation of a

sinkhole in a sloping core.

Particles are detached at the downstream surface of the core which is inadequately
protected by the filter or transition zone. The progression of the erosion process is assisted by
gravity, and there is no need for a cohesive soil layer to form the roof for the pipe. It may be
one of the causes of sinkholes in dams constructed of glacial tills.
The possibility of global backward erosion occurring in central core dams is discussed
in Section 4.5 and the possibility of global backward erosion leading to unraveling of
downstream slopes is dealt with in Section 4.6.

2.3.3 Contact erosion

Contact erosion occurs where a coarse soil such as a gravel is in contact with a fine soil,
and flow parallel to the contact in the coarse soil erodes the fine soil. For example, flow
through gravel alluvium in the foundations of dam or dike may erode the base of an overlying
silt layer, or erosion of the finer layers of soil in a core may occur into a coarse gravelly layer
formed by segregation during construction. Figure 2.8 shows examples.

___________________________________________________________________________
15
 Figure 2.8: Possible location of contact erosion initiation. a) Homogeneous dam with
layered fill due to segregation during construction and a coarse foundation soil. b) Zoned dam
with potential for contact erosion at high reservoir levels above the core and for erosion into
coarse layers in the foundation (Beguin et al, 2009).

2.3.4 Suffusion
Suffusion occurs when water flows through internally unstable widely graded or gap
graded non plastic soils. Glacial tills are sometimes vulnerable to suffusion. Some fills and
filters in dams also have very broad or gap-gradings, and contain excessive fines content.
The small particles of soil are transported by the seepage flow through the pores of the
coarser particles. The coarser particles are not transported and the effective stresses are
largely transferred through the matrix of the coarser particles.
Figure 13.3 (see Internal Instability in Terminology) shows how the “finer soil” fraction
is defined.
Suffusion results in an increase in permeability, greater seepage velocities, and
potentially higher hydraulic gradients, possibly accelerating the rate of suffusion. Stable
situations may be reached with some of the finer fraction remaining in equilibrium with the
seepage stresses. Suffusion may re-commence during cycling periods of water loads or
during higher reservoir or river water level.
Suffusion occurring within an embankment core or the foundation of a dam may also
lead to some settlement of the embankment.
A filter constructed of internally unstable materials will have a potential for erosion of
the finer particles in the filter, rendering the filter coarser and less effective in protecting the
core materials from erosion.
It should be noted that segregation of broadly graded or gap graded non-plastic soils
during placement in the dam may create layers which are internally unstable even though the
average grading of the soil is internally stable.

___________________________________________________________________________
16
2.4 CONTINUATION AND FILTER ACTION
Erosion once initiated will continue unless the eroding forces are reduced or the
passage of the eroded particles is impeded in some way.
Dam engineers have known since the 1950s and 1960s that the most efficient way of
stopping the erosion process is to zone the dam and to provide filters. Filters were then
mainly designed with the Terzaghi criteria (Terzaghi and Peck, 1948).
Figure 2.9 shows the types of embankment dam zoning.

core
downstream zone
of rockfill Puddle core

0. Homogeneous earthfill 4. Zoned earth and rockfill 8. Puddle core earthfill

Embankment filter rockfill concrete corewall

and/or rockfill
Foundation filter earthfill
core

1. Earthfill with filter 5. Central core earth and rockfill 9. Earthfill with corewall

concrete concrete corewall

facing
Rock toe rockfill
earthfill
Max 0.2H

2. Earthfill with rock toe 6. Concrete face earthfill 10. Rockfill with corewall

concrete
core facing hydraulic fill core
downstream zone
of sand/gravel rockfill

3. Zoned earthfill 7. Concrete face rockfill 11. Hydraulic fill

Figure 2.9 - Dam zoning categories. Some of these are only found in older dams.

These can be grouped in regards to their capability of providing control for internal
erosion in the embankment (Table 2.2). This is based on the provision for or lack of filters,
and the statistics of failure and incidents in Foster et al (2000a). It does not take account of
the details of the dam zoning, design and construction so is only a general guide.

Table 2.2: Classification of likelihood of internal erosion associated with the dam zoning
Likelihood of Control for internal erosion Dam zoning and category number
Internal Erosion
A Large Little or no control Homogeneous earth fill (category 0);
Earth fill with rock toe (category 2).
B Moderate Some control of internal erosion Zoned earth fill (3);
depending on detail of zoning and Zoned earth and rock fill (4);
filter capability. Puddle core(8);
Hydraulic fill (11).
C Low Moderate control of internal erosion Concrete face earth fill (6);
depending on the filter capacity and Concrete face rock fill (7);
details of the core wall or face slab. Concrete core earth fill (9);
Concrete core rock fill (10).
D Very Low Good control of internal erosion Earth fill with filters (1);
subject to good details of zoning and Central core earth and rock fill (5).
filter design.

___________________________________________________________________________
17
 What has been recognized (Foster (1999a) and Foster and Fell (2001)) is that filters and
transition zones which are coarser than required by design methods based on particle size will
often be quite effective in controlling erosion. Even downstream rock fill and sand/gravel
zones which were not designed as filters may provide some protection against erosion
continuing.
Depending on the ratio of particle and pore sizes, the erosion will either:
 not continue, (no-erosion); or

 stop after only minor erosion, (some erosion); or

 stop only after a significant amount of erosion, (excessive erosion); or

 will continue (continuing erosion)

2.5 PROGRESSION
Progression is the phase of internal erosion where:
a) For concentrated leak erosion, the erosion in the crack or concentrated leak leads to development
of a pipe.

b) For backward erosion the erosion process extends upstream from the point of initiation, and a
network of small erosion channels forms beneath the soil or embankment providing the roof to
the erosion pipes. If these small erosion channels reach the reservoir or river, then a pipe forms.
In global backward erosion, sinkholes (vertical pipes) form.

c) For contact erosion the erosion of the finer soil into the coarser soil continues. This may in
particular cases lead to development of a pipe in the finer soil.

d) For suffusion, some of the finer fraction is eroded leaving the coarse matrix of the soil. No pipe
is formed but the permeability of the soil may be increased significantly.

2.6 DETECTION
The likelihood that a particular failure path can be detected, and if so, whether it is
possible to intervene (e.g. by lowering the reservoir level), or carry out repairs to prevent the
dam breaching is usually best considered as two questions:
1. Will this failure path be detected?

2. Will intervention and repair be possible in the time available?

The likelihood of detection and successful intervention and repair is dependent on a

number of factors including:
a) The category of internal erosion i.e. internal erosion in the embankment, the foundation or
embankment to foundation.

b) The mechanism of initiation of internal erosion – concentrated leak, backward erosion, contact
erosion or suffusion.

c) The breach mode – gross enlargement of a pipe, instability of the downstream slope, slope
instability from static liquefaction (which may include increase of pore pressure and sudden
collapse in an eroded zone), unraveling or sloughing of the downstream slope, overtopping (e.g.

___________________________________________________________________________
18
 due to settlement of the crest from suffusion and/or due to the formation of a sinkhole from a
pipe in the embankment).

d) The nature of and the geometry of the materials in the foundation.

e) The zoning of the embankment, and the materials in the embankment.

f) The reservoir level at the time of the internal erosion incident, and how rapidly it can be drawn
down.

g) The type and frequency of monitoring and surveillance at the dam and the training of the staff to
recognize a developing internal erosion and piping incident.

h) The ability to get trained personnel out to the site in the event of an internal erosion incident.

i) The ability of those responsible to be able to direct emergency release of the reservoir

j) The availability of materials and equipment to intervene and carry out repair works.

The signs of internal erosion detected by visual inspection related to the type of
initiation phenomena are:
 Internal erosion by concentrated leak is in most cases detected by the emergence of leakage or
water containing eroded particles on the downstream face, around conduits or at the downstream
toe of the dam.

 Backward erosion, contact erosion and suffusion are usually detected by the presence of sand
boils at the toe of the dam, or other signs of eroded soil. Suffusion may also be evidenced by
emergence of leakage or water on the downstream slope rather than sand boils, unless the
suffusion is in the foundation materials.

 For contact erosion and suffusion, the common evidence is settlement of the crest of the dam or
levee.

Sinkholes may be present in all types of internal erosion. If sinkholes are the observed
signs, the erosion has already progressed and the time to failure may be small. It is far
preferable to detect the initiation of erosion earlier by regular surveillance, monitoring
including leakage, pore pressure measurement, temperature measurement and acoustics.
Anomalies monitored by pore pressure or seepage measurements may be evidence of
internal erosion under constant water load. Such anomalies include:
 random changes of piezometric head or seepage rate,

 irrecoverable trends.

These classical methods are more likely to be effective for backward erosion, suffusion
and contact erosion. Concentrated leak erosion may progress rapidly in many situations.
Many less direct means of detecting seepage are now available. The most promising is
temperature measurement which can be used to infer localized flow. Fiber optic cables
facilitate data collection and make it possible to cover large parts of the dam. Remote sensing
options also offer great potential in detecting whether the seepage has caused erosion. These
will be discussed in detail in Volume 2 of the Bulletin.

___________________________________________________________________________
19
 In many situations it is necessary to consider the capability of monitoring to manage the
risks. In some situations it may be necessary to carry out remedial works to achieve tolerable
risks rather than rely on monitoring and surveillance.

2.7 INTERVENTION
Intervention to prevent the progression of internal erosion and piping and breach can
take several forms including:
(i) Drawing down the reservoir level using spillway gates or outlet valves.
(ii) Installing pressure relief wells in the foundation of the embankment.
(iii) Building reverse filters over “boils” or areas where eroding material is
emerging from the foundation of the embankment.
(iv) Building a weighting berm to reduce the likelihood of heave, or slope
instability, or unraveling.
(v) Dumping granular material (sand/gravel/rockfill) into the upstream side of
sinkholes to try to block them.
More than one of these measures may be used together. Which is applicable or feasible
will depend on the particular circumstances of the dam.
It should be recognized that there may be reluctance on the part of the reservoir owner
or operator to release water given the lost revenue that may result, or if release of reservoir
water is likely to result in property damage and loss of life, for example if levee banks
downstream of the dam are likely to be overtopped by the flood resulting from release of the
water.
Table 2.3 can be used to assess the likelihood that, given the concentrated leak is
detected, intervention is not successful. This can be done for various reservoir levels. It is not
practical to cover all the possible scenarios. In making this assessment consideration should
be made for the failure mode and location of the developing pipe.
If erosion is occurring without becoming catastrophic or if a dam appears vulnerable to
internal erosion, it may be necessary to carry out works to inhibit its development.
Emergency works to interrupt progress to a breach are commonly done by providing
filtered drainage at seepage points. This allows the seeping water to flow safely whilst
trapping eroded particles and gradually inhibiting the development of erosion. Sometimes the
width of the dam is increased by adding a filtered berm. This has the effect of decreasing the
hydraulic gradient and the seepage energy available. It also loads the materials through which
seepage and erosion is occurring thereby increasing the effective stress and reducing the
permeability and consequently reducing the energy available to sustain erosion.
Backward erosion may be counteracted by adding filtered toe weights to increase the
downward load to be greater than the upward water pressure, thereby reducing gradient
below critical and preventing the development of the backward erosion pipes and sand boils.
In some cases, a small amount of coarse materials is put into and over the leak before
the proper filter materials are added as the flows velocities can be too high for stable
placement of filter sized particles. In many cases, multi-staged filter/drains as well as other
weighting materials are needed to achieve stability.
Other interventions seek to block seepage, diaphragm walls and clay cores, for
example, but these have the disadvantage that if they are not continuous, concentrated
seepage and leakage can occur at gaps in the impermeable membrane and this may cause
erosion locally and worsen the situation.

___________________________________________________________________________
20
 In the short to medium term increased surveillance and/or increased monitoring can be
used as an intervention, until additional data is collected. For old dams where little is known
about the internal geometry and soils and risk to the public is low to medium this may be the
most effective means of managing risk.
Table 2.3 –Likelihood that given the concentrated leak is detected, intervention is not
successful (after Fell et al, 2008)
Time for
Likelihood of
Development of
What can be done intervention not
Concentrated Leak
succeeding
to Initial Breach
< 3 hrs There is too little time to successfully intervene regardless of the 0.99
failure mode
3 to 12 hrs In most cases it will be impractical to intervene successfully in this 0.9 to 0.99
amount of time. Only in cases where there is a straight forward
method of intervention, and there are personnel, equipment and
materials available will intervention be successful.
12 to 24 hrs In many cases it will be impractical to intervene successfully in this 0.85 to 0.99
time. Only in cases where there is a straight forward method of
intervention, and there are personnel, equipment and materials
available will intervention be successful; or it is a small storage
which can be drawn down to stop the failure mode.
1 to 2 days In many cases it will be impractical to intervene successfully in this 0.7 to 0.95
time. Only in cases where there is a straight forward method of
intervention, and there are personnel, equipment and materials
available will intervention be successful; or it is a small storage or
medium storage with large gate discharge capacity so that the
reservoir can be drawn down to stop the failure mode.
2 to 7 days In some cases it will be practical to intervene successfully in this 0.2 to 0.9 (a)
time. In cases where there is a straight forward method of
intervention, and there are personnel, equipment and materials
available; or it is a small storage or medium storage with large gate
discharge capacity allowing the reservoir to be drawn down to stop
the failure mode.
Weeks or months In some cases it will be practical to intervene successfully in this 0.1 to 0.8 (a)
time. Where there is a straight forward method of intervention, and
there are personnel, equipment and materials available and large
resources intervention has a fair chance of being successful; or it is
a small or medium storage with large gate discharge capacity
allowing the reservoir to be drawn down to stop the failure mode
Note: (a) Use values less than 0.5 only if there is a high degree of confidence in the assessed value.

2.8 BREACH
In these circumstances, the entire process of internal erosion has been concluded,
detection and intervention has failed, and the question is whether the dam will breach by one
of the five mechanisms listed below, or whether the internal erosion process will stabilize.
The breach phenomena are listed in order of their observed frequency of occurrence. The first
four are shown schematically in Figures 2.10 (a, b).
 Gross enlargement of the pipe

 Overtopping (e.g. due to settlement of the crest from suffusion and/or due to the formation of a
sinkhole from a pipe in the embankment).

 Slope instability of the downstream slope

 Unraveling of the downstream face.

___________________________________________________________________________
21
 Static liquefaction which is a form of slope instability which may include increase of pore
pressure and sudden collapse in eroded zone.

COLLAPSE OF PIPE LEADING TO PORE PRESSURES INCREASE

WITH INTERNAL EROSION SLOPE INSTABILITY OF
LOSS OF FREEBOARD DOWNSTREAM SHOULDER

OUTFLOW

ENLARGING PIPE

ORIGINAL PORE
PRESSURES

Breach by pipe enlargement Breach by slope instability

Figure 2.10(a) Potential breach (failure) phenomena-pipe enlargement and slope
instability

CREST SETTLES & EROSION PIPE OUTFLOW EXCEEDS

DAM OVERTOPS FINE FRACTION ROCKFILL CAPACITY;
MIGRATES ROCKFILL DISLODGES
CAUSING SLOPE
INSTABILITY

C
LAY
CORE

SUFFUSIVE COARSE FRACTION

ZONE REMAINS

Breach by overtopping by settlement Breach by unraveling of the downstream face

Figure 2.10(b) Potential breach (failure) phenomena-overtopping by settlement and
unraveling of the downstream face, Fell and Fry (2007)

The mechanics of breaching and the vulnerabilities of various dam types to breaching
are not dealt with in detail elsewhere in this bulletin, and the following paragraphs summarize
the situation.
An important aspect of breaching is the time it takes from the first observable
concentrated leak to when the breach forms because that determines the time available to
warn people downstream and to take action to prevent breach or alleviate the downstream
impacts of a breach. More is said about this in Fell et al (2001, 2008).
Breach initiation is almost certain to occur if continuing erosion is expected or there are
no filters and detection and intervention have failed. Exceptions to this will be if the reservoir
has small capacity and the reservoir level drops below the inlet of the developing pipe before
a breach mechanism has time to develop.
Breach initiation is less likely if some or excessive erosion is expected. For these cases,
the seepage flows that develop through the dam or foundation will be limited by the filtering
material.
For most dam types and failure modes, the likelihood for breach development will be
dominated by one or two of the potential breach mechanisms. Breach mechanisms will not
necessarily be applicable to some dam zoning types or modes of piping and can be ignored.
Table 2.4 lists breach mechanisms depending on the dam zoning type for internal
erosion in the embankment due to a crack or poorly compacted zone. Table 2.4 also applies to
internal erosion in a soil foundation. Figure 2.9 in Section 2.4 shows the dam zoning types.
___________________________________________________________________________
22
 Table 2.4 does not apply to failure modes involving open or in filled defects and
solution features in rock foundations because the leakage flows may exceed the capacity of
even free draining rockfill.

Table 2.4 Screening of breach mechanisms for internal erosion through the
embankment, internal erosion in soil foundations, and from embankment into foundation
(Fell et al, 2008)
Breach Mechanisms
Dam Zoning Type Gross Enlargement Slope Instability Sloughing or Sinkhole
(refer to Figure 2.9) Unraveling Development
0 Homogeneous *  Exclude, except if 
earthfill downstream fill is
cohesionless
1 Earthfill with filters *  Exclude, except if 
downstream fill is
cohesionless
2 Earthfill with rockfill *  Exclude, except if 
toe downstream fill is
cohesionless
3 Zoned earthfill Exclude, except if  Exclude, except if 
downstream fill can downstream fill is
support a roof cohesionless
4 Zoned earthfill and Exclude, except if  * 
rockfill downstream fill can
support a roof
5 Central core earth Exclude, except if Exclude, except if * 
and rockfill (or gravel downstream fill can existing dam has
shells) support a roof marginal stability
6 Concrete face  * Exclude, except if 
earthfill downstream fill is
cohesionless
7 Concrete face Exclude Exclude, except if * Exclude
rockfill (including dam is gravel of
gravel fill) low permeability
8 Puddle core earthfill *  Exclude, except if 
downstream fill is
cohesionless
9 Earthfill with Exclude * Exclude, except if 
corewall downstream fill is
cohesionless
10 Rockfill with Exclude Exclude, except if * 
corewall existing dam has
marginal stability
11 Hydraulic fill Exclude, except if  * 
downstream fill can
support a roof
Key: Exclude: Not a potential breach mechanism
 Potential breach mechanism
* Potential breach mechanism: usually the more critical one

___________________________________________________________________________
23
 Internal erosion by the process of suffusion is usually unlikely to lead to the formation
of a pipe through the dam or its foundation, and hence the probability of breach by gross
enlargement is low.
However there are situations where suffusion can lead to high pore pressures at the toe
of a dike and this can cause hydraulic fracture or heave of an overlying berm. Breach by
slope instability or sloughing/unraveling are usually the more critical mechanisms for
suffusion, although the estimated probabilities for breach are usually relatively low for this
mode of internal erosion.

___________________________________________________________________________
24
 3. INITIATION OF CONCENTRATED
LEAK EROSION
3.1 CONCENTRATED LEAK EROSION: THE
OVERALL PROCESS
Whether erosion will initiate in concentrated leaks in a dam or its foundation depends
on whether the stresses in the dam will result in a crack or low stress zone subject to
hydraulic fracture and geometrical and hydraulic criteria within the crack:
 Whether there is a crack (or flaw) below the reservoir level formed by differential settlement
during or after construction, hydraulic fracture, desiccation, or collapse of a poorly compacted
layer of soil in the embankment or around a conduit sited to pass through the embankment.

 Given there is a crack, whether the forces imposed on the sides of the crack by the water flowing
through the crack are sufficient to initiate erosion.

Whether in the absence of filters or where filters are too coarse, the erosion will
progress to form a pipe depends on:
 Whether the water flowing through the crack will cause the soil on the sides of the crack to swell,
closing the crack, or reducing the width so the forces imposed on the sides of the crack by the
water flowing through the crack are insufficient for erosion to progress.

 Whether the soil will hold open the pipe (“will it hold a roof”).

 Whether upstream zones will limit the erosion process or crack stopping occurs.

Also important is the rate of erosion and enlargement of the pipe because this
influences whether the leak can be detected and intervention taken to lower the reservoir or
stop the erosion process before breach occurs.

3.2 INITIATION OF CONCENTRATED LEAK

EROSION IN SITUATIONS WHERE CRACKS,
OPENINGS AND LOW STRESS ZONES MAY BE
PRESENT IN AN EMBANKMENT OR
FOUNDATION
3.2.1 Cracking and hydraulic fracture due to cross valley
differential settlement of the core
As an embankment dam is constructed the partially saturated compacted soil in the
embankment consolidates and settlement occurs. Where the valley sides are steep and/or have
steps in the profile, such as shown in Figure 2.2 differential settlements occur due to the
variations in the height of the embankment, and these can lead to tensile or low stress zones
in which cracks may form. This has been recognized for some time by Sherard (1973, 1985,
1986), Høeg et al (1998), Kjaernsli et al (1992) and others. Hydraulic fracture through these
low stressed zones as the dam is filled or under flood conditions is an associated phenomenon
___________________________________________________________________________
25
as discussed by Sherard et al (1972), and Sherard (1985, 1986). For most dams, about 80% to
90% of the total settlement occurs during construction (Hunter, 2003, Hunter and Fell, 2003),
so the stresses set up in the construction phase largely control the likelihood of low stress
zones and cracking.
It should be recognized that these stresses are inevitable even in well designed and
constructed dams and are not caused by poor design or construction practice.
Embankments with abutments steeper than about 45 degrees and particularly steeper
than 60 degrees are likely to be susceptible to cracking, particularly for high dams.
Embankments with a step in the foundation profile where the height of the embankment
above the step is less than half the maximum embankment height and the width of the step is
greater than the lower embankment are also susceptible to cracking.

3.2.2 Cracking and hydraulic fracture due to cross valley arching

If the valley in which the dam is constructed is narrow and steep, cross valley arching
can occur and the vertical stresses are shed onto the sides of the valley. This can lead to a
situation where hydraulic fracture can occur. Fell et al (2008) suggest that cross valley
arching is most likely to occur if the width of the valley base is less than a quarter of the dam
height, and the valley sides are steeper than about 60 degrees. It is unlikely to be an issue if
the width of the valley base is greater than three quarters the dam height and the valley sides
are flatter than 45 degrees.

3.2.3 Cracking and hydraulic fracture due to differential

settlement in the foundation under the core
Figure 2.3 which is adapted from Sherard et al (1963) shows foundation conditions
which are likely to lead to differential settlement and cracking or low stress zones conducive
to hydraulic fracture. Zones which may lead to differential settlement greater than 0.5% of
the dam height, with steep changes in the foundation profile are most likely to suffer cracking
and hydraulic fracture. Differential settlements of less than 0.2% of the dam height spread
over some distance are unlikely to lead to cracking and low stress zones.

3.2.4 Cracking and hydraulic fracture due to small scale

irregularities in the foundation profile under the core
Small scale irregularities in the foundation of the core can lead to cracking or low
stresses conducive to hydraulic fracture, e.g. NGI (1984). For cracking or low stresses to
occur the small scale irregularities need to be persistent over all or most of the distance across
the core, and have steps greater than about 3% to 5% of the embankment height.
There are examples of these irregularities being formed by constructing haul roads
across the core (Newman and Foster, 2006), and steps in slope correction concrete (Gillon,
2007).

3.2.5 Cracking and hydraulic fracture due to arching of the core

onto the shoulders of the embankment
As detailed in Bui et al (2004, 2005) and Fell et al (2008) this is most likely to be a
problem for cores which are very narrow - core width less than 0.25 the embankment height,
and for soils subject to collapse compression on saturation (poorly compacted soil placed dry
of optimum moisture content). It is unlikely to be a problem for cores which are wider than
0.5 to 1.0 the embankment height and the core is well compacted at around optimum
moisture content. It is not sufficient to cause hydraulic fracture to just have stiff shoulders
and a less stiff core. There needs to be differential movement of the core after construction.
___________________________________________________________________________
26
Differential expansion between the core and shoulders as a reservoir refills after being drawn
down may result in hydraulic fracture in cores (e.g. Johnston et al, 1999).

3.2.6 Crack or gap adjacent to a spillway or abutment walls and

where concrete dams abut embankment dams
Cracking or a gap may form adjacent to walls due to the earthfill settling away from the
wall during and after construction. Cracks or gaps may also form at dam abutments. Figure
2.4 shows some situations where this is likely to occur.
Well designed walls with uniform contact slopes flatter than about 0.25H:1V are
unlikely to have gaps form. Vertical or over-hanging walls are likely to have gaps.
Cracking and gaps may form due to deformations of flexible and / or under-designed
retaining walls (e.g. designed for active rather than at-rest earth pressures.
Where there is a wrap-around junction between a concrete gravity dam and an
embankment dam, differential settlements similar to those described above may occur. Fell et
al (2008) give details.

3.2.7 Internal erosion associated with conduits embedded in the

embankment
Many internal erosion and piping failures and incidents occur where conduits are
embedded in the embankment.

Figure 3.1 Some causes of internal erosion around conduits: (A) Inadequate compaction
due to the presence of cut-off collars; (B) Inadequate compaction under pipe; (C) Cracking in
soil or extremely weathered rock in the sides of a trench; (D) Cracks due to differential
settlement; (E) Use of corrugations or other roughening of the surface of cut-off collars or
concrete surround. (Fell et al 2005).
___________________________________________________________________________
27
 Figure 3.1 shows some of these features. The conduit facilitates the initiation of internal
erosion by:

 Causing stress distributions due to the stiff conduit and its less stiff surrounding soil which lead
to low principal stresses and hydraulic fracture. This is discussed by Sherard et al. (1972) and
Charles (1997). This can occur on the sides of culverts which are constructed in a trench. Sherard
et al. (1972) point out that it can also occur where the concrete culvert, or the concrete surround
to a pipe, has a sharp corner. In this case piping can be expected above the culvert.

 Drying of the soil in sides of the trench in which the culvert is placed during construction can
also cause cracks which allow initiation of piping.

 Making compaction of soil difficult, particularly if collars are provided at close intervals or the
concrete is formed with corrugated steel sheet or other non-smooth formwork, preventing
compaction of the soil adjacent to the conduit. It can also be difficult to compact the soil
surrounding the conduit if the space between the conduit and the sides of the trench in which it is
placed is small and compaction by rollers is not possible.

 Poor compaction is likely to lead to collapse settlement of the soil on saturation forming a gap
adjacent to the conduit. Note that it is the continuity of the potential defect caused by a conduit
and the excavation through the dam which is so critical to the initiation of internal erosion.
Internal erosion may also occur into open joints or other defects in conduits or into
conduits which have deteriorated due to corrosion.
FEMA (2006) “Conduits through Embankment Dams, best practices for design,
construction, problem identification and evaluation, inspection, maintenance, renovation and
repair” provides a comprehensive coverage of the issues relating to conduits in embankment
dams. Fell et al (2008) provide some guidance on how to assess the likelihood of these
features leading to initiation of erosion.

3.2.8 Cracking or hydraulic fracture in poorly compacted layers

in cohesive and broadly graded till soils in embankments
It is well documented, e.g. Sherard (1973), Foster et al (2000a), that internal erosion
and piping occurs in poorly compacted cohesive soils. This is particularly so for dispersive
soils. The mechanism is potentially of two types:
 The soil behaves as a series of clods with openings between the clods in which water passes.

 The soil collapses on saturation forming a flaw (open pathway) in which the water flows. The
flaws may be the result of collapse deformation alone or be caused by hydraulic fracture when
the minimum principal stress is reduced locally by the deformation and becomes less than the
pore pressure.

This is most likely where there is poorly compacted soil against a conduit or abutment
but is possible within layers of soil, particularly if they were too thick to be effectively
compacted by the compaction equipment used. Sherard (1973) gives examples of this.
Lawton et al (1992) consider the collapse settlement issue.
The large number of piping and sinkhole incidents experienced in Swedish central core
earth and rock fill dams (Nilsson, 2007 a, b) constructed with broadly graded glacial till cores
may have been primarily initiated by collapse settlement of the core which was lightly
compacted in layers up to 800mm thick, some of which were close to structures and sheet
piled cut-offs.
___________________________________________________________________________
28
 Segregation in broadly graded fills may also lead to cracking through local collapses or
hydraulic fracture. More is said about soils subject to segregation in Section 6.5.

3.2.9 Cracking due to desiccation

Desiccation cracking is most likely to be an issue in climates with less than 250mm
annual rainfall, in high plasticity cores, and where there is no surface layer over the core.
Experience in excavating into the crest of embankments is that if there is a road pavement of
granular road base or a rockfill or other non-plastic layer at least 300mm thick cracking is not
generally observed even in seasonal climates with extended dry periods. The likelihood of
cracking is further reduced if the road pavement is sealed with asphalt, concrete or bitumen
seal.
Desiccation cracking does not commonly persist to a great depth so only becomes an
issue for reservoir levels nearing the crest level.
Desiccation cracking may also occur on seasonal shut down surfaces during
construction, or on the surface of the first stage of dams which are built in two stages. Good
construction practice would be to remove the desiccated soil but this was not always done.
It should be noted that in some situations the vertical desiccation cracks in the crests of
embankments form hexagonal columns, which are cut at frequent intervals by horizontal
cracks, to form ‘blocks’. Such circumstances have led to failures, as reported by Marsland
and Cooling (1958) and by Dyer, Utili and Zielinski (2007), because the desiccated blocks
are easily eroded and overtopping may occur when the reservoir water level rises too rapidly
for re-saturation and swelling to take place.

3.2.10 Transverse cracking caused by settlement during

earthquakes
When embankment dams experience a large earthquake they often settle and spread in
the upstream-downstream direction. Many exhibit longitudinal cracking, and some transverse
cracking. Some examples are Austrian Dam (Harder et al, 1991; Forster and MacDonald,
1998); Guadeloupe Dam (Harder et al, 1991) and Lexington Dam (Fong and Bennett, 1995).
Pells and Fell (2002, 2003) analyzed case data for dams not subject to liquefaction which
showed that dams which settled more than about 1.5 % of their height were almost certain to
exhibit transverse cracks; those which settled between 0.5% and 1.5% had about a 20%
chance of exhibiting transverse cracks; and dams which settled between 0.2% and 0.5% had
about a 5% chance of transverse cracking.
These cracks are in the upper part of the embankment and whether they are a potential
initiator of internal erosion depends on the reservoir level and depth of cracking.
If liquefaction occurs the deformations are likely to be large and the likelihood of
cracking greater. These may be estimated by numerical analyses methods.

3.2.11 Cracking or high permeability layers due to freezing

The effects of frost on embankment dams are described in Vuola et al (2007). The
effects of frost are to cause extra water to be drawn into the soil by capillarity action. This
causes ice lenses to form with associated heave and potentially cracking and / or loosening of
the soil. The cracking is most likely to be longitudinal but may also be transverse at the crest
of the dam. Figures 3.2, 3.3 and 3.4 show the potential effects.
Vuola et al (2007) provide an equation for estimating the depth of frost penetration.
There are a number of references which provide information on the susceptibility of
soils to frost penetration. These include Wallace (1987), Holtz and Kovacs (1981), Vuola et
___________________________________________________________________________
29
al (2007) and USACE (1984) which gives the most specific data. Core materials which are
most susceptible to freezing and ice lens formation include silts, clayey silts, silty sands, silty
gravels and clayey sands and gravels with a plasticity index < 12.

Figure 3.2 Effects of frost on an embankment dam (Vuola et al, 2007)

Figure 3.3 Damage phenomena occurring in embankment dams from frost thaw (Vuola
et al, 2007)

Figure 3.4 Effect of frost against a spillway structure (Vuola et al, 2007)

3.2.12 Internal erosion initiated by the effects of animal

burrows and vegetation
Animal burrows in the embankment or levee can lead to a situation where there are
nearly continuous holes through the embankment or situations like those shown in Figure 3.5
where high gradients between holes may result in initiation of erosion.
FEMA (2005a) “Impacts of animals on Earthen Dams, FEMA Report 473” provides a
comprehensive coverage of the issues.

___________________________________________________________________________
30
 Figure 3.5 potential impacts of burrowing animals on internal erosion in embankments
and levees (FEMA 2005a).

Vegetation growing on dams and levees can lead to situations which may lead to a
potential concentrated leak and initiation of erosion. The effects include:
 Decaying roots that create seepage paths.

 Roots penetrating into open joint and cracks in foundation rock, potentially creating seepage
paths.

 Root penetration of conduit joints and joints in concrete structures and opening the joints to allow
erosion into or out of the conduit or wall.

 Clogging embankment under-drain systems.

 Interfering with effective visual inspection.

FEMA (2005b) Technical Manual for Dam Owners, Impacts of Plants on Earthen
dams, FEMA Report 534 provides information and guidance on how to manage this problem.
Wray (2006) prepared a CIRIA guide to management of the European Rabbit.

3.2.13 Relative importance of conduits, spillway walls

cracking mechanisms, and poorly compacted zones
As detailed in Table 1.2, about half of internal erosion incidents in embankments have
historically involved conduits through the embankment and spillway walls supporting the
embankment. These are related to poor compaction, differential settlement and arching
effects.
Excluding conduits and spillway walls the relative proportion of incidents in the Foster
et al (1998, 2000a) ICOLD database caused by cracking and poorly compacted zones is:
 Differential settlement cracking – 36/57 = 63%.

 Poorly compacted and high permeability zones – 21/57 = 37%.

A further assessment of the case study information results is given in Table 3.1:

___________________________________________________________________________
31
 Table 3.1 Relative proportion of incidents from the different causes of crack formation
(Fell et al, 2008, based on Foster and Fell, 1998, 2000a)
No of cases % of cases
Differential settlement, cross valley 20 35%
Differential settlement, cross section 6 11%
Differential settlement, foundation 4 7%
Cracking mechanisms
Differential settlement, embankment staging 0 0%
Desiccation cracking 3 5%
Closure section 3 5%
Total for cracking mechanisms 36 63%
Poorly Compacted and Foundation contact 5 9%
High permeability
zones In embankment 16 28%
Total for Poorly Compacted and High Permeability Zones 21 37%

3.3 ESTIMATION OF CRACK WIDTH AND DEPTH

OF CRACKING
As described above, cracks form in the crests of embankments from desiccation, from
settlement resulting from consolidation, creep or earthquake shaking or in poorly compacted
and high permeability zones.
Settlement causes spreading of the embankment; this results in longitudinal cracks,
parallel to the centerline of the dam, but settlement resulting in transverse cracks passing
through the crest of embankments from the upstream face to the downstream face are sites
where concentrated leak erosion may occur. It is necessary to estimate the width and depth of
such cracks (and the hydraulic gradients in them) to ascertain, using the methods in Section
3.4, if the hydraulic shear force will or will not exceed the hydraulic shear strength of the soil
in the walls of the crack, thereby initiating, or not initiating, internal erosion.
Transverse cracks form as a consequence of vertical settlement, which has the effect of
elongating the crest. The elongation will be uniform, giving rise to numerous fine cracks, if
the foundation profile is such that sharp local differences in settlement will not occur. If the
foundation profile is irregular (see Figures 2.2 and 2.3, for example) differential settlement
will occur, giving rise to locally greater elongation of the crest, and locally wider cracks in
the crest, above the irregularities in the foundation profile. Estimates of crack width may be
made by applying these principles, but there will be many approximations and uncertainties.
The width and depth of cracks which may be present in the embankment may be more
reliably estimated using methods detailed in Fell et al (2008) and Fell et al (2007). The
methods require assessing the likelihood of cracks forming, including by hydraulic fracture,
and are briefly described in Section 12.3.2. However, they are based on a review of the
literature on observed cracking including Sherard (1973), Talbot (1994) and Lawrence
(2002), and the results of numerical modeling by Bui et al (2004, 2005) which assists in
assessing the likely depth of cracking. The values estimated by these methods are also
approximate and this uncertainty should be allowed for in assessing internal erosion.
Assessment of crack widths and depths requires careful consideration and analysis,
probably by more than one method. Table 9.7 gives information on situations in which cracks
or concentrated leaks are unlikely to form in embankments. If cracks seem likely to form, the
methods and references mentioned above should be used to estimate crack dimensions. To
assist in checking such estimates, Table 3.2 gives some examples of the likely maximum
crack depths and widths formed as a result of cross valley differential settlement or
___________________________________________________________________________
32
differential settlement in the foundation using these methods. It should be noted that it
includes examples of situations where both narrow cracks and wide cracks may be formed,
including referring to Figures 2.2 and 2.3 to show how different foundation profiles and
different foundation conditions cause different amounts of settlement and lateral strain,
resulting in cracks of different widths and depths.

Table 3.2 Examples to assist in assessing the likely maximum crack depths and widths
formed as a result of cross valley differential settlement or differential settlement due to soil
in the foundation
Description Situations similar Maximum Likely Maximum Likely Crack Width
to those shown in Depth of Cracking at Depth Shown Below Dam
parts of: Below Dam Crest Crest
1.5m 3m 6m
Uniform abutment without 3m 2mm 1mm
benches, gentle abutment
slopes < 30o, dam less than
15m high
Wide bench low in Figure 2.2 5m 20mm 5mm
abutment, moderate
abutment slopes 30o < to <
45o; dam 15m to 30m high
Wide bench high in Figure 2.2 7m 40mm 20mm 2mm
abutment, steep abutment
slopes 45o < to < 60o ; dam
30m to 60m high
Wide bench near crest in Figure 2.2 10m 60mm 35mm 7mm
abutment, very steep
abutment slopes > 60o;
dam >60m high
Rock or uniform soil 3m 2mm 1mm
foundation, gentle slopes <
30o on sides of
compressible zone, dam
less than 15m high
Shallow soils, differential Figure 2.3 5m 20mm 5mm
settlements of less than
0.2% of embankment
height, moderate slopes
30o < to < 45o on sides of
compressible zone ; dam
15m to 30m high
Moderate depth of soils, Figure 2.3 10m 60mm 35mm 7mm
differential settlements of
0.2% to 0.5% of
embankment height; steep
slopes 45o < to < 60o on
sides of compressible
zone; dam 30m to 60m
high
Deep soils, differential Figure 2.3 12m 90mm 50mm 10mm
settlements > 0.5% of
embankment height, very
steep slopes > 60o on sides
of compressible zone; dam
>60m high

___________________________________________________________________________
33
3.4 THE MECHANICS OF EROSION IN
CONCENTRATED LEAKS
3.4.1 The procedure
The procedure for assessing whether erosion will initiate in a crack or flaw is to:
Estimate the hydraulic shear stresses in the crack for the reservoir level under
consideration, taking account of the geometry of the core of the embankment and the
assumed crack dimensions and location relative to the reservoir surface so the flow gradient
can be determined.
Compare this hydraulic shear stress to the critical shear stress which will initiate
erosion for the soil in the core of the embankment (c) at the degree of saturation of the soil
on the sides of the crack. In doing this take account of the dispersion properties of the soil
and the chemistry of the seepage water.
There is always some uncertainty regarding the input parameters so the analysis should
check the sensitivity to the assumptions made.
The following sections give an outline of this process.

3.4.2 The estimation of hydraulic shear stresses in cracks and

erosion pipes
Wan (2006), Wan and Fell (2002, 2004, a, b) and Fell et al (2008) give details on how
to estimate the hydraulic shear stresses in cracks and cylinders.
They give the following equations for estimating the hydraulic shear stress on the
surface of a cylindrical pipe, or parallel sided transverse crack in an embankment. The
assumptions are:
 Linear head loss from upstream to downstream

 Steady uniform flow along the crack

 Zero pressure head at the downstream end

 Uniform frictional resistance along the surface of the crack or cylindrical pipe

 Driving force = frictional resistance.

(a) Cylindrical pipe: (b) Vertical transverse crack

gH f d  w gH f 2W
  w 
4L 2( H f  W ) L

where  = Hydraulic shear stress in N/m2

w = Density of water in kg/m3
g = Acceleration due to gravity = 9.8m/s2
Hf = Head loss in pipe or crack due to friction in meters
L = Length of pipe or crack base in meters
d = Diameter of the pipe in meters
W = Width of crack in meters
Bonelli and Brivois (2008) have developed a generalized equation for the rate of
enlargement.

___________________________________________________________________________
34
3.4.3 Erosion properties of soils in the core of embankment dams -
basic principles
It is important to reiterate that the resistance to initiation of erosion of the core to
concentrated leak erosion is characterized by the critical shear stress, and the rate of
enlargement of the pipe in the progression phase is characterized by the erosion coefficient.
The erosion properties of soils for concentrated leak erosion can be determined by
rotating cylinder tests e.g. Arulanandan and Perry (1983), Chapius (1986a, b), Lim (2006),
Lim and Khalili (2010); Slot Erosion Tests (SET), and the Hole Erosion Tests (HET) e.g.
Wan (2006), Wan and Fell (2002, 2004a, b), Bonelli et al (2007) and Jet Erosion Tests, e.g.
Hanson (1990, 1991) Hanson and Cook (2004). These tests allow determination of the critical
shear stress for initiation of erosion, and the erosion rate. At this time the most widely used
tests are the Hole Erosion Test and the Jet Erosion Test. More details on these tests are given
in Volume 2 of the Bulletin.
Wan and Fell (2002, 2004a, b) and Wan (2006) expressed the erosion rate in the form
of an Erosion Rate Index, IHET defined by:
IHET=-log(Ce) (3)
Ce is the Coefficient of Soil Erosion, and is derived from the Hole Erosion Test in
which:
ε● = Ce (τ – τc) where: ε● is the erosion rate per unit area
τ is the shear stress
τc is the Critical Shear Stress for initiation of erosion
The representative erosion rate index ĨHET is the hole erosion index IHET for soil
compacted to a density ratio of 95% of standard maximum dry density at optimum moisture
content.
Soils can be classified into 6 groups according to their Representative Erosion Rate
Index, ĨHET. The 6 groups are as shown in Table 3.3. Note that this is a logarithmic scale and
the rate of erosion of soils varies by up to five orders of magnitude.

Table 3.3 Descriptors for erosion rates of soils (Wan 2006)

Representative Erosion
Group No. Description
Rate Index
1 <2 Extremely rapid
2 2–3 Very rapid
3 3–4 Moderately rapid
4 4–5 Moderately slow
5 5–6 Very slow
6 >6 Extremely slow

In the absence of laboratory test values the Representative Erosion Rate Index (ĨHET)
can be related approximately to soil properties. Table 3.4 has been developed from test data
to give a first approximation to the likely range of ĨHET for different classifications of non-
dispersive soils.

___________________________________________________________________________
35
 Table 3.4 Representative erosion rate index (I HET ) versus soil classification for non-
dispersive soils based on Wan and Fell (2002)
Unified Soil Classification Representative Erosion Rate Index (I HET )
Likely Minimum Best Estimate Likely Maximum
SM with < 30% fines 1 <2 2.5
SM with > 30% fines <2 2 to 3 3.5
SC with < 30% fines <2 2 to 3 3.5
SC with >30% fines 2 3 4
ML 2 2 to 3 3
CL-ML 2 3 4
CL 3 3 to 4 4.5
CL-CH 3 4 5
MH 3 3 to 4 4.5
CH with Liquid Limit < 65% 3 4 5
CH with Liquid Limit > 65% 4 5 6
Notes: (1) Use best estimate value for best estimate probabilities. Check sensitivity if the outcome is strongly
dependent on the results.
(2) For important decisions carry out Hole Erosion Tests, rather than relying on this table which is approximate.

The Critical Shear Stress is related to the Erosion Rate Index. The approximate
estimates and likely range of Critical Shear Stress (c) in Table 3.5 have to be used, with
caution, when Hole Erosion Test values are not available.

Table 3.5 Approximate estimates and likely range of Critical Shear Stress (c) versus
Hole Erosion Index (I HET ) (Fell et al 2008)
Hole Erosion Critical Shear Stress for initiation of erosion (c) Pa
Index Non Dispersive Soil Behavior Dispersive Soil Behavior
(I HET )
Best Estimate Likely Range Best Estimate Likely Range

<2 2 1 to 5 1 0.5 to 2
2 to 3 2 1 to 5 1 0.5 to 2
3.5 5 2 to 20 2 1 to 5
4 25 10 to 50 5 2 to 10
5 60 25 to 100 5 2 to 10
6 100 60 to 140 5 2 to 10
Note. To be used with caution. For important decisions carry out Hole Erosion Tests to determine the Critical
Shear Stress (c)

It is emphasized that it is better to carry out a series of Hole Erosion Tests at varying
heads or to use the method of Bonelli et al (2007); Bonelli and Brivois (2008) to define the
critical shear stress (c) than to rely on these relationships.
Marot et al (2011) performed a series of Hole Erosion Tests and Jet Erosion Tests on
different fine-grained soils, covering a large range of erodibility. It is shown that with
commonly used methods erosion coefficient and average critical shear stress are different
with the JET and with the HET. Moreover the relative soils classifications yielded by the two
erodimeters are not exactly the same. Based on the energy method, an erosion resistance
index is determined for both the JET and HET apparatus and a classification of surface
erosion resistance is proposed. For both apparatus, values of erosion resistance index are
roughly the same for each soil and a single classification of soil erodibility is obtained.
Other points to note are the effect of degree of saturation of the soil and the effect of
dispersion. By distinguishing dispersive behavior from non-dispersive behavior, Regazzoni
and Marot (2011) proposed an expression of the erosion resistance index as a function of
three physical parameters: compaction, saturation ratio and difference between clay water
content and liquid limit.
___________________________________________________________________________
36
3.4.4 Comparison of the hydraulic shear stress in the crack (  ) to
the critical shear stress which will initiate erosion for the soil in
the core of the embankment (c)
To assess the likelihood erosion will initiate in a crack the estimated hydraulic shear
stress is compared to the initial shear stress for the soil taking account of the moisture content
of the soil in the core, and the chemistry of the water in the reservoir. When doing this
assessment allowance should be made for the uncertainty in the calculations and properties.
For an average gradient of 0.5 in a 2mm wide crack the hydraulic shear stress is 5Pa,
for a 5mm wide crack 12Pa, and for a 20mm wide crack 50Pa. Hence from Table 3.4 for
dispersive and other highly erodible soils even narrow cracks will be sufficient for erosion to
initiate.
It is important to recognize that erosion in cracks or flaws is not a result of
establishment of a flow net, but is due to flow in open cracks. This occurs quickly as the
reservoir rises into the cracked zone. Steady state and transient flow net analyses are
irrelevant to this process.

3.5 DISPERSIVE SOILS AND THEIR IMPORTANCE

IN CONCENTRATED LEAK EROSION
3.5.1 What are dispersive soils?
Soils in which the clay particles will detach from each other and from the soil structure
without a flow of water, and go into suspension are termed dispersive clays.
The clay particles in many clay soils exist in accumulations known as flocs. In these
soils changes in pore water chemistry brought about by seepage through a dam, for example,
can cause the flocs to break down, de-flocculate, and effectively become smaller, more easily
eroded, particles.
The dispersivity of a soil is directly related to its clay mineralogy. In particular soils
with a high exchangeable sodium percentage such as Na or Ca montmorillonite present, tend
to be dispersive, while kaolinite and related minerals (e.g. halloysite) are non dispersive.
Soils with illite present tend to be moderately dispersive.
The dispersivity depends also on the pore water chemistry. Low pore water salt
concentrations lead to greater dispersivity and high salt concentrations can suppress
dispersion in susceptible soils. Hence percolation of a saline soil with fresh water can lead to
dispersion.
There have been many failures of dams constructed of dispersive clays. Mitchell (1993)
indicates these have been mostly low to medium plasticity (CL and CL-CH) clays that
contained montmorillonite. Fell et al (2005) indicate that their experience is similar, that is
the most susceptible soils are not those with a high plasticity, but those with limited clay size
fractions sufficient only to give low to medium plasticity.
Sherard et al. (1976) however tested some soils as dispersive in the pinhole test with %
passing 0.005 mm greater than 50%. The better performance of the higher plasticity clays
probably relates to greater resistance to erosion and greater likelihood for cracks to close as
the soil swells.
Sherard et al. (1976) indicate that based on their tests, soils with less than 10% finer
than 0.005 mm may not have enough clay to support dispersive piping.

___________________________________________________________________________
37
3.5.2 How can dispersive soils be identified?
There are several laboratory tests which can be used to determine the dispersivity of a
soil:
a) Emerson class number; Standards Australia (1997), test AS 1289, 3.81 and USBR (1979). The
test is carried out in distilled water, but may be repeated in water from the dam, or groundwater.
This often gives significantly different results due to the presence of dissolved salts in the water
(higher salt content gives less dispersive results). The soils are graded according to class, with
Class 1 being the highly dispersive, Class 8 non dispersive. Soils with Emerson Class 1 to 4
needs to be treated with extra caution in dam construction

b) Soil Conservation Service test, also known as the double hydrometer test, or percent dispersion
test; ASTM 2001 test D42291-99). Sherard et al. (1976) indicate that soils with a percent
dispersion greater than 50% are susceptible to dispersion failure in dams, and those with a
percent dispersion less than 15% are not susceptible. This test, and the ‘triple’ hydrometer test in
which the soil is also dispersed in reservoir water, compares the dispersion in water to the
dispersion that occurs when a chemical dispersant is used purposely to deflocculate the soil as it
is in standard laboratory particle size distribution procedures.

c) Pinhole dispersion classification, also known as the pinhole test, or Sherard pinhole test (ASTM
1998, test D4647-93). Soils which tested as D1 and D2 were found by Sherard et al. (1976) to
have suffered piping failure in earth dams, and severe erosion damage by rainfall in
embankments and natural deposits while those with ND1 and ND2 classification had not. As for
the Emerson class number, the results are dependent on the chemistry of the water used for the
test (the standard test uses distilled water).

d) Chemical tests. Based on correlation with many dam failures and soil from dams which have
leaked continuously (without any filters to control erosion) and not failed (Sherard et al., 1976)
proposed to determine the dispersivity of soil based on the percent sodium and sodium
adsorption ratio. These are defined as:

Na +
Percent sodium = x 100
Total dissolved salts

Na +
or = x 100
Ca ++ + Mg ++ + Na + + K +

Na+
Sodium adsorption ratio (SAR) =
1  1/2
2(Ca++ + Mg++)
 
+ ++ ++
in which Na , Ca , and Mg are measured in milli-equivalents per liter of saturation extract.
Details are given in Sherard et al. (1976).
Most authors consider that it is necessary to use more than one test to ascertain the
dispersivity of a soil. Sherard and Decker (1977) suggest that four tests should be used: Soil
Conservation Service, pinhole, Emerson and chemical test. They were of the opinion that the
pinhole test was best.
More details on these tests and the properties of dispersive soils can be found in Fell et
al (2005).
While laboratory tests are a useful way of identifying dispersive soils, much can be
determined by observing the behavior of the soils in the field, e.g.:
___________________________________________________________________________
38
 The presence of deep erosion gullies and piping failure in existing small farm dams usually
indicates the presence of dispersive soils.

 Erosion of road cuttings, tunnel erosion along gully lines and erosion of weathered or clay
infilled rock joints may indicate potentially dispersive soils.

 The presence of cloudy water in farm dams and puddles of water after rain indicates dispersive
soils.

3.5.3 Effect of dispersion on erosion properties

See Section 3.4.3.

3.5.4 Effect of dispersion on filter design and assessment

See Section 7.5

___________________________________________________________________________
39
 4. INITIATION OF BACKWARD
EROSION
4.1 BACKWARD EROSION PIPING LEADING TO
FORMATION OF A PIPE
4.1.1 The overall process
Backward erosion piping occurs in non-plastic soils. It mainly occurs in foundations but
may occur within embankments.
The erosion process begins at a free surface on the downstream side of a dam or dike as
shown in Figures 2.5 and 4.1.

Figure 4.1 Geometry of Delft backward erosion piping model (Sellmeijer et al. 2012)
This free surface may be in a ditch or may form by cracking due to heaving of the
cohesive strata overlying the cohesionless soil as shown in Figure 4.1, or in some cases may
be the seeping surface on the downstream face of the dam, or in other cases the stream bed
downstream of the dike.

4.1.2 The process in sandy dam or dike foundations

The backward erosion process progresses beneath the dike or dam. For this to occur the
dike or dam or cohesive strata must form a roof for the eroding “pipe”. The presence of
backward erosion piping is often exhibited by the presence of sand boils at the downstream
side of the dam or dike.
Sellmeijer and co-workers from Delft Hydraulics and Delft Geotechnics Laboratories in
The Netherlands carried out more than 70 backward erosion piping tests. The first were in
flumes and are reported in de Witt et al (1981), Silvis (1991), Weijers and Sellmeijer (1993)
___________________________________________________________________________
40
and Technical Advisory Committee (1999). The tests were mostly on fine to medium sands,
with a few tests on medium to coarse sands. The sands were uniform with uniformity
coefficients Cu = 1.58 to 3.53. Early tests were on small scale models (base length 0.8 m), but
later tests were on very large models. These experiments and those carried out by Townsend
et al (1988) at University of Florida model backward erosion in the foundation of dikes,
levees and dams.
These experiments showed that backward erosion initiates in the slot through the strata
overlying the eroding soil representing a crack or drainage ditch excavated through the strata,
and progresses in multiple small “channels” rather than a single “pipe”. The channels are
quite small. The height of the channels is typically 4 to 10 (d 15 ); that is often less than 2mm.
For any head less than a critical head, the development of the channels stops. If the head is
increased, erosion begins again. Figure 4.2 shows some examples. The critical head occurs
when the length of the channel (l) is about 0.3 to 0.5 of the flow path length L. For heads less
than this the progression of the pipe reaches a stable condition. For heads greater than the
critical head the piping channel extends upstream and breaks through to the reservoir. The
erosion then progresses rapidly as erosion in an open pipe. For these experiments the rate of
progression of the pipe was relatively uniform until the length approaches about 40% of the
total seepage path. It then accelerates. The piping progressed about 6 meters in an hour in the
largest of the experiments.
In 2009 and 2010 further experiments were carried out at Deltares. These are reported
in Van Beek et al (2010, a, b), Sellmeijer et al (2012) and Van Beek et al (2012b). The largest
of these was virtually full scale. From these experiments Van Beek et al (2012b) refined the
description of the backward erosion process as follows:
1. Phase 1 Seepage occurs in the permeable strata. In the experiments there were no confining
strata.

2. Phase 2.Backward Erosion. At the beginning of the process there is rearrangement of grains,
individual grain movements, and formation of small channels. The process reaches equilibrium
for the hydraulic head applied. Very small amounts of sand are transported, in the order of cubic
centimeters in this phase of the process. With an increase in the hydraulic head to the critical head
sand is transported continuously. A variety of erosion patterns is observed in the small and
medium scale experiments. In the large scale experiment sand boils are formed, and in the small
and medium scale experiments craters are formed. These characterize the reaching of the critical
head. The flow barely increases in this phase of the process. At a head greater than the critical
head the erosion does not cease and the erosion rate is in the order of cubic decimeters per hour.
The rate of erosion increases with increasing head.

3. Phase 3. Widening of the channel. As soon as the pipe reaches the upstream side a pressure surge
occurs in the pipe. In the small scale experiments this in turn causes a large amount of sand to be
eroded rapidly. In the medium and large scale experiments blockages caused by local collapse of
the roof of the pipe take place and the widening process takes longer. The widening pipe develops
from the upstream to downstream. The flow and sand transport to the exit point do not increase
significantly. The widening process took up to a few days in the large scale experiments. When
the widening pipe has almost reached the downstream side the sand transport and flow increases
suddenly. The situation can change from sand boils to this condition without warning.

4. Phase 4 Failure and Breakthrough. Failure occurs soon after the widening phase is complete, but
can be delayed due to collapse of the levee causing the first pipes to close.

___________________________________________________________________________
41
 Figure 4.2 Delft Hydraulics laboratory piping flume test model: typical test results
showing the development of the length of the tunnel as the head increases (Weijers and
Sellmeijer (1993).

4.1.3 Backward erosion piping within a dam, levee or dike

For backward erosion piping to occur within a dam, levee or dike embankment the
following criteria would have to be met:
1. The soil where pipes develop would need to be non-plastic.

2. There would have to be a free surface at the downstream of the dam core at which erosion
could initiate; e.g. if there is no filter or an inadequately filtered contact between the core and
a transition zone.

3. The core or some layer within the core would have to form a roof for the pipe to progress.
This might occur at the phreatic surface if the partially saturated soil above the surface was
silty, or if more cohesive strata were layered in the core.

4.1.4 Soils which are subject to backward erosion piping leading

to formation of a pipe
The experience in USA and Europe is that backward erosion piping mostly occurs in
the foundations of levees, dikes and dams where the eroding soil is fine to medium grain size
sand, with a uniformity coefficient Cu < 3.
The participants at the Aussois Workshop (Fell and Fry 2007) considered that at
gradients likely to occur within a dam or its foundation backward erosion is probably
restricted to non plastic soils or soils with only limited plasticity. For practical purposes, Fell
et al (2008) have concluded that based on the available data, the results of Wan and Fell
(2004c, 2007, 2008) tests on internal instability and their experience and judgment, soils with
Plasticity Index > 7 should be considered not subject to backward erosion at the gradients
experienced in dams and their foundations.

___________________________________________________________________________
42
4.2 METHODS FOR PREDICTING WHETHER
BACKWARD EROSION PIPING WILL INITIATE
AND PROGRESS
4.2.1 Terzaghi and Peck (1948)
Terzaghi and Peck (1948) show that backward erosion piping will initiate when a heave
or zero effective stress condition occurs in cohesionless soils at the downstream toe of a dike,
levee or dam. They do not consider the conditions under which the backward erosion will
progress to form a pipe except to recognize that it is necessary for a roof to be formed in an
overlying stratum or by the levee.
The Terzaghi and Peck (1948) philosophy has been the dominating influence in design
of levees (dikes) and dams, particularly in the USA (e.g. USACE, 2000, 2005, Wolff, 2002).
Design methods have concentrated on avoiding the heave or blow-out condition. It has
however commonly been assumed, at least implicitly, that if backward erosion initiates to
form a sand boil it will progress to form a pipe, at least under repeated loading from
successive floods.
The Delft and University of Florida experiments (see below) show that this is not
necessarily the case and that the progression of the pipe may stabilize. However the latest
experiments by Deltares indicate that once sand boils form the critical gradient at which
progression continues has probably been exceeded and backward erosion will progress unless
the gradient is reduced by the river level dropping or the tail water level being raised, e.g. by
building sand bag levees around the sand boil.

4.2.2 Method of Sellmeijer and co-workers at Deltares

Sellmeijer (1988), Sellmeijer and Koenders (1991), Koenders and Sellmeijer (1992)
developed a mathematical model for backward erosion piping based on the experiments
carried out at Delft Laboratories in the 1980’s.
This model was refined using the results of the Deltares testing and is presented in Van
Beek et al (2010) and Sellmeijer et al (2012).
The critical gradient is determined as a product of three contributions: resistance factor,
scale factor and geometrical shape factor. The improvement is realized by the outcome of 38
small-scale tests, applying the multivariate method. The refined equations for the critical
gradient at which backward erosion will progress are:

___________________________________________________________________________
43
 H 1
  FR FS FG
L c
0.02
 p
0.35 0.13
 RD  U   KAS 
FR   tan       
w  RDm   Um   KAS m 
0.6
d d 
FS  70  70m 
3  L
 d 70 
0.28
 0.04
 D   D 2.8
FG  0.91    L  1
L

Where:
H [m] : hydraulic head across structure
L [m] : seepage length (= base length of the embankment)
D [m] : thickness of sand layer under the embankment
c [-] : erosion coefficient
FR [-] : resistance factor
FS [-] : scale factor
FG [-] : geometrical shape factor
RD [%] : relative density

U [-] : uniformity coefficient Cu = d 60 / d 10

KAS [%] : roundness

d70 [m] : soil particle diameter for which 70% by weight of the soil is finer

 'p [kN/m3]: submerged unit weight of soil particles 9.8(G-1)

G [t/m3]: soil particle density

 w [kN/m3]: unit weight of water

 [-] : Whites drag coefficient

 [DEG]: bedding angle (angle of repose) of sand

 [m2] : intrinsic permeability


Where: = k
g
2
 [m /sec]: kinematic viscosity
g [m/sec2]: gravity
k [m/sec]: hydraulic permeability
H, D and L are defined in Figure 2.5 in Chapter 2. In these equations the variables are
normalized by the mean values in the data set. For the data set used by those authors the
mean values are as detailed in Table 4.1.
Figure 4.3 can be used to assess critical gradients for soils within the limits in Table
4.1. Note that the effect of the FG term is included in the curves on Figure 4.3.
___________________________________________________________________________
44
Figure 4.3 Critical gradient for various FR * FS values and embankment dimensions. H, D and
L are defined in Figure 2.5. As an example, for FR * FS = 0.100, D/L = 1.0, critical gradient at
which backward erosion will progress to form a pipe back to reservoir, is H/L = 0.10.
(Courtesy of Dr Hans Sellmeijer and Vera van Beek).
Sellmeijer et al (2012) indicate that their data set assumes a bedding angle of 37°, and a
Whites coefficient of 0.25. For water at 20o Celsius,  = 1.02 x 10-7 (k) where  is m2, k is in
m/sec. They indicate that they ignore the roundness and Uniformity Coefficient terms as they
do not contribute significantly.
They indicate that the refinements have been determined from the small scale tests and
it is not altogether clear if there may be a scale effect, so that the outcome for large structures
may not be properly modeled. They also indicate that the equations should only apply within
the limits of the parameters during testing. These limits are given in Table 4.1.

Table 4.1 Limits of the Sellmeijer et al (2012) method

Parameter Minimum maximum mean
RD 34 % 100 % 72.5 %
U 1.3 2.6 1.81
KAS 35 % 70 % 49.8 %
d70 150 μm or 430 μm or 207 μm or
1.5E-04m 4.3E-04m 2.07E-04m

They found that the equations predicted the large scale experiment behavior quite well
for the fine grained soil, but that it was not so accurate for the soil with a d70 = 260 μm.
There are very significant scale effects. The smaller the structure, the higher the critical
gradient. In the tests the value of the scale factor FS ranged from 0.134 for the large scale
IJKdijk tests, to 0.192 for the medium-scale tests and 0.421 for the small-scale tests.
It should be noted that Sellmeijer et al (2012) emphasize that the above equations do
not include any margin of safety and that for design rules they are likely to apply factors of
safety.
Van Beek et al (2012a) extended the earlier work to make it applicable to the prediction
of piping in multi-layer foundations.
___________________________________________________________________________
45
4.2.3 Hoffman’s Hydraulic Approach
Dr Gijs Hoffmans, from Deltares, Delft, Netherlands re-analyzed the results of the Delft
experiments (Hoffmans, 2013). He, like Sellmeijer, recognizes that for backward erosion to progress,
the particles being released from the initiating zone at the upstream end of the piping channel must be
transported down the channel. Hence it is the interaction of the gradients at the backward eroding
head of the piping channel, and the flow gradient in the piping channels which controls the process.
The latter is controlled by the gradients of flow into the pipes but also the permeability of the soil as
this affects the flow quantity in the pipes.
Table 4.2 gives Hoffmans’ equations and lists the variables. The most important variables in the
Hoffmans approach are hydraulic conductivity (k), particle sizes d50, d15 and the critical pipe height
(ℓz,c), and a coefficient αH. The latter two are determined from the experiments.

Table 4.2 Overview of Hoffmans’ piping equation (from Hoffmans 2013)

 H1  H 2  c g  am , c d15 
3

 d 
2

  1  c  50
L   Re,  L  Re KD

with

   D 2 g  d15  
3
2

 exp    
c f am , c

L   L    Re, 
 

Symbol Indicative range Comment

d15 0.08 mm < d15 < 0.32 mm d15 for which 15% of the particles is finer than d15

d50 0.13 mm < d50 < 0.75 mm mean particle diameter

D 0.1 m < D < 50 m thickness of the sand layer

D*  d50  g /  2 
1
3
4 < D* < 16 dimensionless particle diameter

g g = 9.81 m/s2 acceleration of gravity

(H1 – H2)c/L 0 < (H1 – H2)c/L < 1 critical hydraulic gradient

H1 1 m to 10 m sea or river level

H2 usually taken as 0 m water level in ditch or at soil surface

K 10-5 m/s < K < 10-3 m/s hydraulic conductivity (see also Eq. 2.37)

ℓc 0 < ℓc/L < 1 ratio of critical pipe length and seepage length

Re ℓRe = 18.2∙10-6 m geometrical length scale (see also Section A17)

L 0.3 m < L < 200 m seepage length

f f = 5 geometrical groundwater coefficient (Section A17)

___________________________________________________________________________
46
 Re,ℓ Re,ℓ = 6.8 geometrical pipe coefficient (see also Section A17)

Δ = ρs/ρ – 1 Δ = 1.65 (sand) relative density

ν 10–6 m2/s < ν < 1.4·10–6 m2/s kinematic viscosity

ℓam,c = 0.2(D*)-0.3 0.08 < ℓam,c < 0.16 critical Shields parameter (for laminar flow)

if ℓ = 0 then the equilibrium is stable (or no erosion occurs);

if 0 < ℓ < ℓc then the equilibrium is unstable (or occasional erosion occurs);

if ℓ  ℓc then there is no equilibrium anymore (or erosion occurs until the dike fails).

4.2.4 Russian Design Methods

Belkova et al (2011), Zhilenkov et al (2011) and Rumyantsev et al (2012) give
summaries of Russian standards for design for seepage flow in the embankment and
foundations of dams. Radchenko et al (2012) give more details. They indicate that these
standards require that numerical analyses of seepage in the dam body, foundation and river
banks are made, including accounting for non-uniform and anisotropic geological conditions.
The standards require that the following parameters of seepage flow should be
determined in studies and analyses:
 position of phreatic surface in the dam body and banks;

 seepage flow rate through the dam body, foundation and banks;

 heads (or hydraulic gradients) of seepage flow in the dam body, foundation and zones of outflow
into the drainage or downstream of the downstream toe, in the zones of contact of soils of
different soil parameters and at the boundaries of impervious elements.

In seepage analyses of thawing earth dams constructed in the northern construction and
climatic zone, parameters of seepage flow should be determined.
Possible variation of seepage parameters with time should also be taken into account.
To determine what is termed the “seepage strength” of the system as a whole and its
separate elements, two concepts, those of overall and local seepage strength, are used:
The “seepage strength” of the dam body and impervious elements is estimated on the
basis of numerical analyses and laboratory testing of soils under the hydraulic gradients
expected to be acting in the structural elements of the dam. The analyses are made using the
highest value of head (reservoir water level) expected to occur at the dam. The stress-strain
state of the dam and its foundation are also taken into account.
Using the results the seepage strength is checked, and the dam layout, in particular the
location of drainage, is adjusted if necessary to satisfy the design criteria.
For the dam body, the following are determined:
 permeability factor k;

 unit water absorption q, for soils in the freezing zone

___________________________________________________________________________
47
 parameters of soil seepage strength (local and averaged critical hydraulic gradients Jcr and Jcr,m
and critical seepage rates Vcr)

If necessary, structural solutions are adjusted after numerical checks of the foundation
strength.
The standards require that in the foundations the following should be prevented:
 loss of overall seepage strength of non-rock foundations;

 damage to, or ineffective operation of, impervious elements in the foundation, which causes
unacceptable loss of water, flooding, saturation of slopes, etc.;

 non-uniform displacements of foundation sections, which might cause failure of impervious

elements;

 increase of pore pressures or seepage flow rate.

As a result, the following should be determined:

 overall seepage strength of non-rock foundations;
 local seepage strength of rock and non-rock foundations in cases when piping channels may
occur.
The main parameters used for the estimation of seepage strength are hydraulic gradient
(average or local), or seepage flow rate.
The criterion of seepage strength is the following inequality:
1
J est ,m  J cr ,m
n
Where:
Jest,m = average hydraulic gradient acting in the analyzed seepage zone;
 n = reliability factor which depends on the social and economic consequences of failure of the dam;
Jcr,m = average critical hydraulic gradient taken on the basis of soil testing in conditions representing
operational conditions at the dam.

In preliminary analyses, values of Jcr,m may be taken using available similar results or
from Table 4.3 (for soils in dam zones) and Table 4.4 (for soils embedded in structure
foundation).

Table 4.3. Values of average critical hydraulic gradients Jcr,m

Apron Diaphragm / Dam body
Soil
core
Clay, clay concrete 15 12 8-2
Sandy clay 10 8 4-1.5
Clayey sand 3 2 2-1
Medium sand - - 1
Fine sand - - 0.75

___________________________________________________________________________
48
 Table 4.4. Values of Jcr,m for soils embedded in structure foundation
Soil Calculated averaged critical hydraulic gradient
Fine sand 0.32
Average coarse sand 0.42
Coarse sand 0.48
Sandy clay 0.60
Loam 0.80
Clay 1.35

The values of average critical gradient are determined depending on the physical and
mechanical properties of soils and the methods of soil placement, higher values of Jcr,m are set
for denser soils.
In the estimation of local seepage strength, the following conditions are used:
1 Vcro
J est  J cr , or V n
n n
Where:
Jest – actual hydraulic gradient determined by numerical analyses, modeling or from field observation
data;
Jcr, Vcro – local critical hydraulic gradients and seepage flow rates in pores determined in target zones by
numerical methods or by testing piping strength of soils in laboratory or field conditions;
J, V – actual local hydraulic gradients and seepage flow rates determined in the same zones by seepage
analysis;
n – porosity.

4.2.5 Other Methods

Schmertmann (2000) developed a method based on backward erosion piping tests in
flumes at University of Florida. These were carried out in a number of sets, including those
by Townsend et al (1988) who carried out 15 tests. Those tests were carried out on a range of
soils from fine to medium sands, up to coarse sand/fine gravel mixes.
The tests were different to the Delft tests in that a starter pipe was formed at the
downstream end. The head differential was increased progressively to keep the tunnel
progressing upstream until the critical head was reached causing the tunnel to form a pipe to
the upstream source.
Most of the tests (32 out of 39) which Schmertmann (2000) relied upon were carried
out on sands with a uniformity coefficient less than 3.2. The other 7 tests were carried out on
gap-graded soils and a well graded soil with uniformity coefficients around 6.
Schmertmann (2000) found that the critical average gradient i pmt above which
backward erosion would occur was related to the uniformity coefficient Cu (d60/d10) of the
soils tested, as follows:
i pmt = 0.05+0.183(Cu-1)
He also plotted the Delft tests and found a similar correlation. However there were few
tests on soils with a Cu >3.
The method involves quite large corrections for scale effects. Given it is based on little
data in the larger uniformity coefficient range, and some of these may be affected by internal
instability, caution should be exercised in using the method for soils with Cu > 3.

___________________________________________________________________________
49
4.3 SOME FIELD OBSERVATIONS
The US Army Corps of Engineers (USACE) have a shared responsibility with state and
local governments and local levee boards for managing the levee systems along major USA
Rivers including the Mississippi. USACE carried out extensive underseepage investigations
from the 1930s to the 1960s covering the middle and lower Mississippi River, from Alton,
Illinois to New Orleans, Louisiana. As recorded by Sills and Vroman (2007), these early
studies resulted in the development of current USACE levee design methods which were
validated during decades of field observations and laboratory tests.
More recent investigations (Wolff 1974 and 2002, Shannon and Wilson Inc, 1995,
Wilson 2003, and Glyn and Kuszmaul, 2010) have observed the ageing levee systems’
performance. Some observations from these include:
In any one flood the levees in one USACE District, e.g. St Louis, may experience many
hundreds of sand boils in the foundations. However there are few documented cases of
breaching of the levees from piping. This is influenced by the “flood fighting” efforts of the
Districts and the Levee Board personnel. They build sand bag rings and in some cases sub-
levees around the sand boils to stop the flow of sand (but not the flow of water). The few
documented breaches may support the laboratory tests (see 4.2.1 above) which show that
erosion initiates at lower average gradients than the higher critical gradients required to
progress the erosion to form a continuous pipe from the land side to the river side. More
analysis of the field data is required to investigate the conditions at which erosion causing
sand boils progressed to erosion causing levee failure.
Sills and Vroman (2007), Wolff (2002) and Glynn and Kuszmaul (2004) report that
there are some cases of levees which have sand boil activity occurring at successively lower
average gradients (lower river stages). Glynn and Kuszmaul (2004) show that greater sand
boil activity occurred in the 1995 flood than the 1993 flood along the levees studied, even
though the river stage was lower in 1995. In 1995, 37% of the sand boils occurred at
locations close to the 1993 ones. The phenomenon of possibly decreasing gradients at which
successive backward erosion to form sand boils occurs was discussed in the early
investigations by USACE, but not quantified. It was not investigated by the laboratory tests
described in 4.2.1 above. Sills and Vroman (2007) suggest that it may arise because after
sand boil episodes, the erosion pipes are not re-filled, and probably remain open, thereby
increasing the permeability/porosity of the soil, and reducing the gradient at which backward
erosion commences during subsequent high water level events.
USACE (1956) and Wolff (2002) also show that local geology has an important
influence on the occurrence of sand boils. Sand boils are more likely to occur where swales
from point bar deposits cross the levee at an angle and concentrate seepage at the toe. Glynn
et al (2012) conducted statistical analyses of sand boil locations and found that the parameters
most significant to sand boil locations are: geological conditions, landside confining layer
thickness, effective grain size coefficients and previous sand boil locations. Using logistic
regression of these significant parameters, empirical models were developed for calculating
the probability of sand boils developing at selected levee segments. The logistic regression
modeling method can be used to develop a unique statistical model for predicting sand boil
activity for any levee system. More research is planned to apply the modeling method to
rivers located in different geologic regions.

___________________________________________________________________________
50
4.4 GUIDANCE ON WHETHER THE OVERLYING
SOIL WILL FORM A ROOF TO THE PIPE
Section 8.2.1 and Table 8.1 give guidance on whether the overlying strata will form a
roof for the backward erosion pipe.

4.5 THE PROCESS OF GLOBAL BACKWARD

EROSION IN CENTRAL CORE DAMS
The available evidence on the occurrence of global backward erosion (see Figure 2.7 in
Section 2.3.2.2) in central core dams is that this internal erosion mode will only occur at
relatively high gradients except when the core consists of cohesionless, internally unstable
soil subject to suffusion. This evidence includes:
a) Laboratory tests by Sun (1989), Marot et al (2007) and Bendahmane et al (2008) confirmed that
backward erosion could occur in more cohesive soils, but initiated at very high gradients which
were not likely to occur in dams or their foundations. The tests done at the VNII Vodgeo
Laboratory and reported by Istomina (1957) showed that the critical hydraulic gradient of a soil
with a very low plasticity and liquid limit equal to 14 is higher than 10 at moisture content close
to the Optimum Proctor. They showed that the critical hydraulic gradient not only depends on the
plasticity but on the soil consistency as well.

b) Bendahmane et al (2008) carried out tests which showed that soils consisting of 10% kaolin /
90% fine sand initiated backward erosion at gradients between 90 and 140 depending on
confining pressure. For kaolin contents 20% and 30% no backward erosion occurred at gradients
as high as 100. They observed what they called suffusion with gradients as low as 5, but noted
that this value should not be taken as generally applicable.

c) Moffat and Fannin (2011) and Moffat et al (2011) indicate that there is an initial movement of
finer particles at gradients of about 4 for the cohesionless soils they tested and more extensive
movement of finer and some coarser fraction soils at higher gradients, from about 10 to 30 in the
soils tested. These movements of particles are accompanied by a reduction in volume as particles
are eroded from the soil through inadequate filters.

d) Marot et al (2007) and Bendahmane et al (2008) tested clay soils and Sail et al (2011) a non
plastic gap graded soils with 40% finer fraction. They observed some particle movement at a
gradient of about 5 and major movement at much higher gradients.

e) Some recent tests on glacial till soils from an Australian dam showed that for vertical downward
flow to model the condition in Figure 2.7 what appears to be backward erosion occurred in
broadly graded cohesionless soils at a gradient of 9. Erosion had progressed for about 40 days at
a gradient of 5 and may have reached the failure condition without the increased gradient if the
duration of the test had been longer. This raises questions as to time effects for this type of
backward erosion as it seems the process may be very slow. More testing is required to explore
this phenomenon.

There are no known reliable methods for predicting global backward erosion in broadly
graded cohesionless soils such as glacial tills in the cores of dams, and for development of
vertical piping as shown in Figure 2.7. For these situations it is recommended that
specifically designed laboratory tests should be carried out.

___________________________________________________________________________
51
4.6 GLOBAL BACKWARD EROSION, LEADING TO
UNRAVELING, AT A SEEPING DOWNSTREAM
SURFACE OF THE EMBANKMENT
This is quite a different mode which may occur in dikes or small dams constructed only
of silty sand gravel. There is a potential for particles to be removed from the downstream
slope by gravity and hydraulic seepage gradient under the through flow breakout line, where
seepage emerges from the downstream face or toe. The failure mode is one of unraveling of
the downstream slope rather than formation of a pipe through the embankment.
It is very likely that unraveling will occur if the particle diameter is under a critical size
associated to the discharge rate. This can be assessed by the methods Solvik (1991, 1995),
and the revised method by EBL Kompetanze AS (2005). The following equation is
recommended to describe the initiation of erosion of dam toe subject to through flow:
d 50  0,43S 00, 43 q 0,78
Where d50 is rock size in meters, S0 slope of rockfill and q unit discharge in m3/s/m.
If suffusion occurs in the material used to construct the dike this process may be
accelerated.

___________________________________________________________________________
52
 5. INITIATION OF CONTACT
EROSION
5.1 THE OVERALL PROCESS
Contact erosion occurs where a coarse soil such as a gravel is in contact with a fine soil,
and flow parallel to the contact in the coarse soil erodes the fine soil. For example flow
through gravel alluvium in the foundations of dam, levee or dike may erode the base of an
overlying silt layer, or erosion of the finer layers of soil in a core may occur into a coarse
gravelly layer formed by segregation during construction. Figure 5.1 shows examples.
Particles of the finer layer may be destabilized by the water flow and transported
through the pores of the coarser layer parallel to the interface.
This phenomenon requires two conditions. First, the coarse layer has to be
geometrically open to the other layer, that is to say, to have pores sufficiently large so that
fine particles can pass through them. Second, the hydraulic conditions must be such that the
flow velocity is sufficient to detach particles and to transport them.

5.2 FAILURE MODES AND VULNERABLE DAM

TYPES
All interfaces between different soils which exist in the dam or in the foundation may
be a vulnerable to contact erosion (Figure 2.8). Nevertheless, locations with possible high
velocities in the coarser layer and high particle size grading contrast between layers are the
most likely to experience contact erosion. These characteristics often occur at the interface
between the core and a gravelly foundation. For example, at the zoned dikes on the Rhone
River, with clayey silt cores and gravel shoulders, some twenty cases of leakage associated
with development of a sinkhole or subsidence have been reported (CFGB, 1997). The process
starts with contact erosion at the interface between the silt and the gravel, often at the contact
between the fill and the foundation or in the foundation. This causes a cavity within the fill,
and then the pressure drops around that cavity, causing the roof collapse. Those materials fill
in the lower part of the cavity and enlarge it at the top; then the roof of the new cavity is
decompressed and, step by step, the cavity progresses along a practically vertical chimney
towards the crest or the upstream face (Figure 5.1 a). Leakage flows decrease after the
materials fall, and then progressively build up again, carrying away the fallen material until
the next occurrence. In most cases the sinkhole volume is about 1 to 2 m3.
Contact erosion may also appear between any granular layer (filter, drain, riprap) and a
fine soil in contact with that layer. For example, in the case of abnormally high water level in
the reservoir rising above the top of the core, flow can pass through a riprap layer or a
gravelly shoulder zone over the top of the core and can erode the core.
Other consequences of contact erosion are possible (Figure 5.1). First, depending of the
mechanical properties of the core, the cavity created by contact erosion may not collapse and
can be a beginning for backward erosion piping (consequence b). This process has been
identified in hydraulic flume tests at scale 1/1, where it appeared that the downstream part of
the opening was held open by hydraulic fracture. Second, a weaker zone, less dense because
of the development of contact erosion, can cause a loss of stability (consequence c). Finally,
the eroded fine particles may clog the permeable layer and increase pore water pressure
which may result in instability of the downstream slope (consequence d).

___________________________________________________________________________
53
 Figure 5.1 Consequences of contact erosion. Black arrows indicate a groundwater flow
through a more permeable layer (light grey) under a less permeable dam (dark grey). a)
sinkhole daylights b) beginning of backward erosion piping c) creation of a weaker zone
initiating instability d) clogging of the permeable layer and increase of pore water pressure.
(Beguin, 2011).

5.3 METHODS TO ESTIMATE LOADS AND

RESISTANCE TO CONTACT EROSION
The two conditions needed for initiation of contact erosion are:
1. Geometrical condition: pores of the coarse layer have to be sufficiently large to allow
particles to pass through. Some authors have proposed criteria in the specific context of
contact erosion but classic filter criteria can also be used (Section 7.5, ICOLD 1994).

2. Hydraulic condition: the flow velocity has to be sufficient to detach the particles and also to
transport them.

If both conditions are fulfilled, contact erosion is likely to occur.

Different authors who have studied contact erosion e.g. Brauns (1985), de Graauw et al
(1983), Wörman et al. (1992), Den Adel et al (1994), Guidoux et al (2010), proposed
expressions for the hydraulic conditions which imply the detachment, depending on each
configuration, and the transport of particles. These laws have been established based on
experimental results. These authors noticed that close to the geometrical limit, the hydraulic
loading for erosion initiation is increasing. They noted a transition zone of combined
influence where the hydraulic loading needed to initiate contact erosion is higher than in the
domain of pure hydraulic influence as shown in Table 5.1.
Table 5.1 is derived from Brauns (1985), Wörman (1992) and Den Adel et al (1994).
They found that for D 15 / d 85 ratios less than in the third column, there is geometrical filtration
whatever the hydraulic loading, so contact erosion could not occur. For D 15 / d 85 ratios more
than in the fifth column hydraulic loading controls erosion and there is no filtration effect. In
between these two limits both geometric and hydraulic factors control erosion.

___________________________________________________________________________
54
 Table 5.1 Domain of geometrical and hydraulic influence. (courtesy of R.Beguin, from
Brauns, 1985, Wörman, 1992 and Den Adel, 1994).

Note: D 15 is the particle size of the coarser soil for which 15% is finer; d 85 is the particle size of the finer soil
for which 85% is finer. n is porosity. For DH see 5.3.2 below.

The second boundary between “Geometrical and Hydraulic condition” and “Hydraulic
condition”, defines if the coarse layer grading has, or has not, an influence on the hydraulic
criteria for erosion initiation. For example, Brauns (1985) obtained that in the “Hydraulic
condition” domain, the critical velocity for a fine soil can be calculated without taking into
account the coarse soil grading. (This is valid in its experimental range 25<D 15 / d 85 <57). In
the “Geometrical and Hydraulic condition” domain, for 7.5<D 15 / d 85 <25, the critical velocity
will be also function of the coarse soil grading. In a similar manner, erosion laws proposed by
Den Adel (1994) and Wörman (1992) are valid in their “Hydraulic condition” domain, where
the influence of the coarse layer on the initiation of erosion can be neglected.
For one particular fine soil, the “critical” gradient, which corresponds to the gradient in
the coarse layer parallel to the contact at which the erosion initiates, can vary by one order of
magnitude, depending of the permeability of the coarse layer. However, in the same tests, the
“critical” Darcy velocity for erosion initiation does not significantly depend on the coarse
layer permeability, and is only related to the fine soil resistance to erosion. Therefore Darcy
velocity has been chosen by the majority of authors as a good indicator of the hydraulic
loading.
The hydraulic conditions for contact erosion depend on the configuration considered.

5.3.1 Fine cohesionless soil below a coarse soil layer

This configuration has been widely studied (Istomina, 1957; Pravedny, 1966, Brauns,
1985; Bezuijen et al., 1987).
In the case of sand, erosion particles are mainly transported as bed load and authors
conclude that we can use classical river erosion criteria empirically adapted to the case of
contact erosion. In consequence, they propose laws based on Shields (1936) criterion. One
conclusion of these works is that the diameter of coarse layer particles has a weak influence
on the critical velocity.
Experimental results (Beguin 2011) are shown in Figure 5.2 with other results based on
the work of other researchers.

___________________________________________________________________________
55
Figure 5.2 Critical velocities for contact erosion of sand above and below gravel (courtesy of Dr Remi Beguin)

______________________________________________________________________________________________________________________
56
 Experimental results range between 0.01m/s and 1m/s for the critical Darcy velocity,
and a minimum seems to appear for particles of diameter 10-4 m (0.1mm, 100µm).
Brauns’ (1985) law is the simplest formula to use and gives a good approximation for
sand:
s  w
U crit  0.65.nD ( ) gd50
w
with n D the porosity of the gravel layer, ρs (kg/m3) the density of the sand particles, ρw
(kg/m3) the water density, d50 (m) the median diameter of sand grading curve.
For loose-packed materials (say d > 10-4 m), Hoffmans (2013) proposed an equation,
based on Darcy’s law, the Poiseuille flow, the shear stress approach of Shields (1936), the
Shields Number ℓam,c of sediment transport in laminar flow and some geometrical
assumptions regarding the pipes, giving the critical Darcy velocity U crit :

U crit  1
2 Rem,c  am,c gd15
Thus Ucrit depends on the maximum critical Reynolds number and the critical bed shear
velocity. Considering the laminar flow conditions, where the Reynolds filter number (Ref) is
smaller than 10, it is noted that Hoffman’s equation is comparable with the method as
proposed by Guidoux et al (2010) (see 5.3.2 below).
To predict the rate of erosion evolution, three models of sand erosion have been
developed which allow estimation of the amount of transported sand as a function of the
hydraulic loading (Wörman, 1992; Den Adel, 1994; Scheuermann, 2002). However these
models have been validated only for the soils tested by their authors, and they have to be used
with caution.

5.3.2 Silt and clay erosion (particles < 75µm)

Guidoux et al (2010) carried out some experimental tests of contact erosion with silt
and clay. He adapted Brauns’ (1985) law with an empirical parameter to take into account the
adhesive forces. To consider widely or gap graded soils, he also proposed to use the effective
diameter dH of the fine soil instead of the d50 (Kozeny, 1953). The effective diameter is
defined by:
1
 m Fj 
dH    ,
 j 1 d 
 j 
Where Fj is the percentage of the fraction of diameter dj in the grading curve of the soil.
This diameter conserves the specific surface of the initial grain size distribution.
Guidoux et al (2010) proposed that the critical Darcy velocity U crit can be determined
from:
   w    
U crit  0.65.nD  s  gd H 1  2 
 
 w   dH 
The empirical parameter β is a function of the cohesive properties of the soil, selected
as 5.3x10-9m2 by Guidoux et al (2010). The other variables are as for Brauns (1985) above.
This formula gives a reasonable fit to the experimental data for these finer soils. It cannot be
applied for very cohesive soils which exhibit different erosion behavior linked not to particle
diameter, but to other more relevant properties such as the clay mineralogy and degree of
saturation.
___________________________________________________________________________
57
5.3.3 Fine soil above a coarse soil layer
In this configuration experimental data is limited. The phenomenon is complex and
cannot be directly linked to river erosion. In consequence, a safe hydraulic condition for
contact erosion in this configuration should be based on the transport velocity for the
particles, considering that detachments occur whatever the flow velocity is. For example,
Goltz et al (2009) proposed a law for particle transport, based on Muckenthaler’s approach
(Muckenthaler, 1989; Perzlmaier, 2007) comparing sedimentation velocity of particles and
pore velocity.
Schmitz (2007) carried out experimental tests of erosion of silt layers above coarse
layers. In contrast to the previous configuration (fine soil below coarse soil) he noticed an
influence of the confining stress on the critical velocity. For higher vertical stresses on the
sample, he measured higher critical velocities. Except for one value, the critical velocities
measured are of the same order of magnitude as in the other configuration, between 1cm/s to
10cm/s (Figure 5.2). He proposed a law for critical velocity depending on results of vane
shear strength tests on the soil, but comparison of its model with results is not very
convincing. This is to be expected as the critical hydraulic shear stress for clay soils is not
related to the shear strength, as discussed in Section 3.4.
Beguin (2011) carried out six tests on silt embankments over gravel. The results are
included on Figure 5.2. The critical velocities were more scattered than those for the gravel
above silt configuration, but of a similar order of magnitude. It seems that, even in the silt
above gravel configuration, where erosion might be expected to be initiated when silt
particles ‘fall’ into the gravel, the initiation of erosion is dependent on the transport of
particles, not by detachment.

5.3.4 Influence of uniformity

When contact erosion occurs in a widely graded fine soil, segregation processes usually
influence the process. Fines particles are eroded preferentially because they are more
sensitive to hydraulic loading. Coarser particles are also likely to clog the pores of the coarse
layer and generate a filter layer at the surface of the fine layer. This can occur for a fine layer
above or below the coarse soil and tends to result in a decrease of the erosion rate with the
development of this filter layer. In consequence, the geometrical and the hydraulic condition
for erosion may be fulfilled for contact erosion but after a certain amount of eroded soil
which can be excessive for dam safety or not, erosion stops.

___________________________________________________________________________
58
 6. INITIATION OF SUFFUSION
6.1 GENERAL DESCRIPTION OF THE PROCESSES
Suffusion occurs when water flows through widely graded or gap graded cohesionless
soils such as alluvium of a large river, colluvium in the bed of rivers in mountainous areas,
embankment cores constructed of glacial origin soils, and in filters which have very broad or
gap gradings or excessive fines content.
The small particles of soil are transported by the seepage flow through the pores of the
coarser particles. The coarser particles are not transported and the effective stresses are
largely transferred through the matrix of the coarser particles.
For suffusion to occur, the following three criteria, geometric criterion, stress criterion
and hydraulic criterion, have to be satisfied:
 Criterion 1: The size of the finer soil particles must be smaller than the size of the constrictions
between the coarser particles, which form the basic skeleton of the soil.

 Criterion 2: The amount of finer soil particles must be less than enough to fill the voids of the
basic skeleton formed by the coarser particles. If there are more than enough finer soil particles
for void filling, the coarser particles will be “floating” in the matrix of fine soil particles, instead
of forming the basic soil skeleton.

 Criterion 3: The velocity of flow through the soil matrix must impose a high enough stress to
overcome the stresses imposed on the particles by the surrounding soil and to move the finer soil
particles through the constrictions between the larger soil particles.

Figure 13.3 (see Internal Instability in Terminology) shows how the “finer soil” fraction
is defined, in this case as the point of inflection of the particle size distribution plot. This may
also be determined by experiments to determine the percentage of finer fraction which gives
the maximum density, that is which just fills the voids in the coarser fraction. Soils with more
fine soil are over-filled, those with less are under-filled. ICOLD (2008) says more about
under- and over-filled soils.
Suffusion results in an increase in permeability, greater seepage velocities, and
potentially higher hydraulic gradients, possibly accelerating the rate of suffusion. Stable
situations may be reached with some of the finer fraction remaining in equilibrium with the
seepage stresses. Suffusion may re-commence during cycling periods of water loads or
during higher reservoir or river water level.
Suffusion occurring within an embankment core or the foundation of a dam may also
lead to some settlement of the embankment. A filter constructed of internally unstable
materials will have a potential for erosion of the finer particles in the filter, rendering the
filter coarser and less effective in protecting the core materials from erosion.
It should be noted that segregation of broadly graded or gap graded non-plastic soils
during placement in the dam may create layers which are internally unstable even though the
average grading of the soil is internally stable.

___________________________________________________________________________
59
6.2 METHODS OF IDENTIFYING SOILS WHICH ARE
INTERNALLY UNSTABLE AND POTENTIALLY
SUBJECT TO SUFFUSION (CRITERION 1:
GEOMETRIC AND CRITERION 2: FILLED OR
UNDERFILLED VOIDS)
6.2.1 General requirements
Figures 6.1 and 6.2 show particle size distributions of some soils which have been
found to be internally unstable in laboratory tests. It will be noted that there are gap graded
and broadly graded soils. The Kenney and co-authors’ soils are sandy gravels, whereas the
Wan and Fell soils are silty sandy gravels. Soils 14A and 15 had 11% and 21% kaolin in
them so were slightly plastic.
As pointed out by Kenney and Lau (1985, 1986) for a soil to be internally unstable and
subject to suffusion the percentage of finer fraction (finer than the point of inflection of the
particle size plot) must be smaller than the available void space. They suggested this lay
between 20% of the total soil for well graded soils and 30% for narrow graded soils. Wan and
Fell (2004c, 2007) showed that this could theoretically be as high as 40% but in their sample
it was between 22% and 33% for broadly graded soils and 29% and 38% for the gap graded
soils.
For a greater proportion of finer soil the coarse particles are surrounded by the finer
particles. Such soils are not subject to suffusion but may be subject to global backward
erosion.
For practical purposes Fell et al (2008) have concluded that based on the data described
available soils with Plasticity Index > 7 should be considered not subject to suffusion at the
gradients usually experienced in dams and their foundations. If for some particular reason the
gradient is higher than about 4 then soils with a Plasticity Index  12 should be considered
for suffusion. This was considered to be a somewhat conservative approach.
It must be recognized that the types of soils which are potentially susceptible to
suffusion are also potentially subject to segregation during placement in the dam or dike.
Therefore when assessing whether the material in the dam or dike is potentially suffusive, a
range of gradations should be considered which covers the potential effects of segregation
(Section 6.5).

___________________________________________________________________________
60
 SAND GRAVEL
FINES
FINE MEDIUM COARSE FINE COARSE
100 100

90 90
Percentage Passing by Weight, F (%)

80 80

70 70

60 60

50 50
10 A2 B1
40 40
15
30 14A B2 30

20 20
C1
10 D1 10
A3
0 0
0.002 0.005 0.075 0.425 2.0 4.75 19 75
Grain Size,d (mm)

Figure 6.1 (a) Soil samples tested as being internally unstable (suffusive) by Wan and
Fell (2004c, 2007)

SAND GRAVEL
FINES FINE MEDIUM COARSE FINE COARSE
100 100
RD
90 90
Percentage Passing by Weight, F (%)

80 80
3R
70 70
11
60 60

50 50
2R
40 6 40
7
5
30 30
1, 1A
20 20

10 13 10
4R 9
0 0
0.001 0.005 0.075 0.425 2.0 4.75 19 75
Grain Size, d (mm)

Figure 6.1 (b) Soil samples tested as being internally stable (not suffusive) by Wan and
Fell (2004c, 2007)

___________________________________________________________________________
61
 SAND GRAVEL
FINES FINE COARSE
FINE MEDIUM COARSE
100 100
Kenney & Lau (1984, 1985), Kenney et al. (1983)
90 Kenney et al. (1984) 90
Percentage Passing by Weight, F (%)

80 80

70 D 70

60 60

50 50

40 40
A
Ys
30 30
AD Y
20 20
AC X
10 10

0 0
0.001 0.005 0.075 0.425 2.0 4.75 19 75 100
Grain Size, d (mm)

Figure 6.2 Soil samples tested as being internally unstable (suffusive) by Kenney &
Lau (1984, 1986) and Kenney et al. (1985)

6.2.2 Some methods for assessing whether a soil is subject to

suffusion.
There are a number of methods available to determine whether a soil is subject to
suffusion. The following are some of the more widely used and / or later methods. It is
suggested that the method or methods used are those which were developed for soils most
closely matching the soil being assessed. For important decisions it may be necessary to carry
out tests on the soils under consideration given the uncertainty in the methods currently
available. Messerklinger and Straubhaar (2011) report on a series of such tests carried out on
potential filter materials for Goescheneralp dam in Switzerland.

Kenney and Lau method

The Kenney and Lau (1985, 1986) method plots values of F, the mass fractions smaller
than selected grain sizes D (taken from a conventional particle size distribution plot) for the
soil against H, the mass fractions smaller than 4D. For particles to move there must be a
deficiency in the mass of particles in the range D to 4D. Figures 6.3 and 6.4 show the method.
If the soil plots to the right of the boundary of Figure 6.4 (H<F), it is likely to be
internally unstable provided that it satisfies the rules on limiting finer fraction; that is F, 0.2
for widely graded soils and F, 0.3 for narrow graded soils.
The method was developed for filter and transition zones with less than 5% fines
passing 0.075mm.

___________________________________________________________________________
62
 Figure 6.3 Method of characterizing the shape of a grading curve (Kenney and Lau,
1985).

Figure 6.4 Method of assessing internal instability by Kenney and Lau (1985), with the
revised criteria from Kenney and Lau (1986). Legend: WG, soils widely graded (Cu > 3) in
the range F= 0.2-1.0; NG, soils narrowly graded (Cu < 3) in the range F= 0.3 to 1.0.

___________________________________________________________________________
63
Wan and Fell adaptation of the Burenkova method
Wan and Fell (2004c, 2007) found that the Burenkova (1993) method gave reasonable
assessments of whether a soil was internally unstable when used for the soils they had tested.
However the method does not give a clear-cut boundary between internally stable and
unstable soils in the data set. To model this logistic regression was used by Wan and Fell
(2004c, 2007) to define contours of equal probability of internal instability. The contours and
the logistic equations for silt-sand-gravel and clay-silt-sand-gravel mixtures are shown on
Figure 6.5 and those for sand-gravel soils are shown on Figure 6.6.

10
Data point (447, 17.2)
plotted out of range
Data source: P is the probability, predicted by logistic
Kenney et al. (1983) regression, that a soil is internally unstable
Kenney et al. (1984) if it is plotted along the respective
8 Kenney & Lau (1984, 85) dotted line P .
Lafleur et al. (1989) Sun (1989) data and UNSW data points B1,
Burenkova (1993) D1 not included in the logistic regression.
Skempton & Brogan (1994)
Chapuis et al. (1996) P = exp(Z)/[1 + exp(Z)]
h' = d90 /d60

UNSW Z = 2.378 LOG(h") - 3.648 h' + 3.701

6 Internally stable soil samples are
represented by hollow symbols
(e.g. , , etc.), and
internally unstable soil samples are
represented by solid symbols
(e.g. , , etc.). D1
4 0.05
B1 0.10
0.30
0.50
0.70
0.90
0.95
2

1 10 100 1000 4000

h" = d90 /d15

Figure 6.5 Contours of the probability of internal instability (potentially suffusive) for
silt-sand-gravel and clay-silt-sand-gravel mixtures with a plasticity index less than 13% and
less than 10% clay size fraction (% passing 0.002 mm) (Wan and Fell, 2004c, 2007).

___________________________________________________________________________
64
 10
Data source: P is the probability, predicted by logistic
Kenney et al. (1983) regression, that a soil is internally unstable
Kenney et al. (1984) if it is plotted along the respective
Kenney & Lau (1984, 85) dotted line P .
Lafleur et al. (1989) Sun (1989) data are not included in
8
Burenkova (1993) the logistic regression.
Skempton & Brogan (1994)
Chapuis et al. (1996) P = exp(Z)/[1 + exp(Z)]
UNSW (only data points 4R, Z = 3.875 LOG(h") - 3.591 h' + 2.436
h' = d90 /d60

A3 and C1)
Internally stable soil samples are
6
represented by hollow symbols
(e.g. , , etc.), and
internally unstable soil samples are
represented by solid symbols
(e.g. , , etc.).

4
0.05
0.10
0.30
0.50
0.70
4R 0.90
2 C1 0.95
A3

1 10 100 200
h" = d90 /d15

Figure 6.6 Contours of the probability of internal instability (potentially suffusive) for
sand-gravel soils with less than 10% non-plastic fines passing 0.075mm (Wan and Fell,
2004c, 2007).

Some general comments

Wan and Fell (2004c, 2007) found that for the silty sandy gravel soils they tested, the
Kenney and Lau (1985, 1986) method was too conservative, identifying the soils shown in
Figure 6.4 as internally unstable when they tested as stable. In view of this it is apparent that
the Kenney and Lau (1985, 1986) method is too conservative for silty sand gravel soils and it
is suggested it only be applied to the sand gravel soils for which it was developed. Wan and
Fell (2004c, 2007) also found that the Istomina (1957), Sherard (1979) and Sun (1989)
methods were conservative for these silt sand gravel soils.
Rönnqvist (2007, 2008, 2009, 2010) applied Kenney and Lau (1985, 1986) to assess the
internal stability of the filters and core grading of a number of existing moraine (glacial till)
core dams, a number of which had exhibited signs of internal erosion, and others which had
not. He found a correlation between dams with historic performance of internal erosion and
internal instability of the filter and the core. However this was related to the fact that the
filters of the dams showing internal erosion were coarser than the no-erosion filters. Most
were in the “some erosion” or “excessive erosion” range as defined by Foster and Fell (2001).

___________________________________________________________________________
65
6.3 ASSESSMENT OF THE LARGEST ERODIBLE
PARTICLES IN SUFFUSION
(a) Geometric criteria
Wan (2006) and Wan and Fell (2004c) give details of a method to determine what
fraction of the soil will be eroded. Fell et al (2008) have found that in practical terms it can be
assumed that 50% of the finer fraction as defined by the point of inflection of broadly graded
soils and the fine limit of the gap in gap-graded soils is eroded, and the particle size
distribution re-plotted.
They suggest that if this becomes critical, laboratory tests should be carried out on the
internally unstable soil.
Salehi Sadaghiani and Witt (2012) and Salehi Sadaghiani et al (2012) used a statistical
approach to identify mobile grains and the supporting skeleton, and for considering the
geometrical filtration response of widely gap graded soils. They used the method to re-assess
the internal stability of 20 grain size distributions from the literature, agreeing with the
findings of Wan and Fell (2008) in all but two cases.
(b) Hydraulic criteria.
Goltz et al (2009) are developing hydraulic criteria based on constriction opening size.
Trials using these criteria have so far been carried out on limited test data.

6.4 ASSESSMENT OF SEEPAGE GRADIENT WHICH

WILL CAUSE SUFFUSION (CRITERION 3)
Skempton and Brogan (1994) showed that erosion, which they described as
‘segregation piping’, will begin in internally unstable cohesionless soils at seepage gradients
lower than the Terzaghi critical gradient (which is approximately one, depending on soil
density or zero effective stress, when the pore pressure in a permeable soil below the
interface equals the total pressure applied by an impermeable soil above and ‘heave’ or
‘uplift’ occurs).
Wan and Fell (2004c, 2007) found that internally unstable soils eroded with upward
gradients of 0.8 or less, with several less than 0.3. There is a general trend that soils with a
higher porosity begin to erode at lower hydraulic gradients. Loose, higher porosity soils
began to erode at gradients less than 0.3. Soils with plastic fines required higher gradients to
begin to erode. Gap-graded soils tended to begin to erode at lower gradients than non gap-
graded soils with the same fines content.
Monnet (1998) correlated the hydraulic gradient initiating failure measured by
Skempton and Brogan (1994) to the measured permeability, showing that the hydraulic
criterion governs the value of hydraulic gradient moving finer particles.
Li and Fannin (2008) proposed a hydro-mechanical criterion deduced from tests on
unstable materials. Starting from the fact that seepage failure occurs in stable soils under the
Terzaghi critical hydraulic gradient, iCT, Fannin and Li deduced the critical gradient, iC, for
any unstable soil, from the proportion of the effective stress  sustained by the fines:

'
iC   .iCT  
W

___________________________________________________________________________
66
 where γW and γ' are the water and buoyant soil specific weight and  is the reduction
factor of the vertical effective stress 'V carried by the finer particles in the internally unstable
soil as was first proposed by Skempton & Brogan (1994). The finer grains carry a reduced
portion of the effective stress, 'f:
 'f   . V'
The reduction stress factor  is related to the mobility of the finer particles and is
dependent on the geometric criterion d'85/O50, where d'85 is the d85 of the finer fraction of soil
and O50 is the effective constriction size of the coarse fraction, as follows:
α = 3.85(d'85/O50) – 0.616
This method needs more test data on internally unstable soils which are subject to
suffusion so the variables can be calibrated.
Moffat and Fannin (2011) have further developed these concepts but the test data is for
soils which are not internally unstable as defined in this Bulletin.
With a specific centrifuge bench, Marot et al (2012) analyzed the influence of hydraulic
gradient on initiation and development of suffusion. As for the backward erosion process, the
characterization of suffusion depends on the length of the seepage path by using the hydraulic
gradient concept. The difference in the results is in the magnitude of critical hydraulic
gradient and magnitude of erosion rate. For an increase of tested specimen length by a factor
2, the value of critical hydraulic gradient can be multiplied by a factor of 0.6 and rate of
erosion can double. This influence of seepage path length may be due to the probability for a
detached particle to be filtrated or not. This filtration can induce clogging and thus a
modification of the seepage. An energy analysis of interstitial fluid has been developed. This
analysis leads to a linear correlation between expended power by fluid flow and rate of
erosion. By integrating over the time, the eroded mass is linearly correlated to the energy
dissipation. The suffusion characterization does not depend on specimen length by using the
energy analysis.
Chang et al (2012) found that increases in deviatoric stress lowered the gradient at
which suffusion initiated and skeleton deformation occurred when compared to one
dimensional seepage tests.
As there are no generalized methods for accurately predicting the critical seepage
gradient laboratory tests should be carried out which carefully simulate the field conditions.
When selecting the particle size distribution of the soils to be tested allowance should be
made for the potential effects of segregation of the materials during placement in the dam. It
may be necessary to construct specific test equipment because of the large particles involved.

6.5 ALLOWING FOR THE EFFECTS OF

SEGREGATION WHEN ASSESSING SUFFUSION
Soils which are subject to suffusion are also subject to segregation in stockpiles prior
to placement and during placement on the embankment. Segregation is also a problem for
broadly graded soils such as those in Figure 13.3 and in filters.
Characteristics which make it more likely that silt-sand-gravel soils will segregate on
placement include:
 A broad grading, particularly with maximum particle size >75 mm.

 A low percentage of sand and fine gravel sizes (<40% finer than 4.75 mm).

___________________________________________________________________________
67
 Poor construction practices e.g. end dumping from trucks, high lift heights and poor control of
stockpiling operations.

 Soils placed dry, rather than moistened prior to placement.

These criteria are based on Sherard and Dunnigan (1989) and Kenney and Westland
(1993). Milligan (2003) reports on the effect of wetting in preventing segregation, based on
Sutherland (2002) who investigated the mechanisms of segregation.
Some information on segregation may be available from particle size distributions
taken during construction, from photographs of placement during construction, or from
samples taken from test pits in the embankment.
Segregation can become critical because it may result in layers of coarser soil which are
internally unstable whereas the average gradation is not. A segregated filter can be coarser than
what is reported in for example the dam records, and streaks or pockets of gravel/cobble-sized
materials can exist.

___________________________________________________________________________
68
 7. CONTINUATION OR
INTERRUPTION OF EROSION BY
FILTER ACTION
7.1 INTRODUCTION
After the internal erosion process has been initiated, by one of the four mechanisms
described in the previous chapters, it must continue if it is to become a threat to the dam.
Erosion may be interrupted or stopped by filter action in the soils in the fill and
foundation materials downstream of the point where it initiated. Filtering occurs when the
sizes of pore spaces in the soil through which seeping water containing eroded particles is
passing are such that the eroded particles are trapped, thereby stopping the erosion from
continuing. Filters are normally coarser and more permeable than the filtered soil (called the
‘base soil’) and the seeping water, no longer containing eroded particles, drains away
downstream.
The zoning of dams is important in preventing or allowing erosion to continue and
various scenarios are discussed below. In many existing dams filter zones are included
downstream of the core zone, as filter blankets on foundations and in other vulnerable
locations. In others, there may be no formal filter zones, but the zoning may include materials
such as transition zones that may act as filters to vulnerable zones. If erosion has initiated in
homogeneous dams, the seepage containing eroded particles will not pass through any filter
zone, and erosion will therefore continue (although it may not progress, as discussed in the
following chapter).
Formal filters in old dams may not have been designed to modern methods, and other
potentially filtering zones will not have been designed as filters. Their ability to act as filters
can be assessed using modern filter design methods, but modern methods (described below)
lead to very safe filters, as is appropriate given the vital role of filters in protecting dams
against internal erosion. However, soils which are coarser or more permeable than filters
designed to modern criteria often provide a substantial filtering capability. The means of
assessing such filtering potential is described below.

7.2 CONTINUATION SCENARIOS

7.2.1 Internal erosion in the embankment
There are a number of scenarios which may be applicable depending on the zoning of
the dam and the failure mode under consideration. Example sketches are shown in Figure 7.1.
 Scenario 1: Homogeneous zoning with no fully intercepting filter.

 Scenario 2: Downstream shoulder of fine grained cohesive material which is capable of holding a
crack/pipe. Soils which are capable of holding a crack or pipe are:

– well compacted shoulder (shell), containing > 5% plastic fines; or

– poorly compacted shoulder (shell), containing >15% plastic fines

– well compacted shoulder, containing > 30% non plastic fines

___________________________________________________________________________
69
 – poorly compacted shoulder, >30% non plastic

 Scenario 3: Filter/transition zone is present downstream of the core or a downstream shoulder

zone which is not capable of holding a crack/pipe. This includes earthfill dams with a chimney
filter.

 Scenario 4: Erosion into a crack or open joint (e.g. open joint or crack in a conduit or adjoining
concrete structure).

 Scenario 5: Erosion into a toe drain and under-drains.

The manner in which the likelihood of continuation of erosion for these scenarios is
considered differently:
Scenario 1: There is no potential for filtering. Continuing erosion is certain.
Scenario 2: The issue for this scenario is whether the crack/high permeability feature
that is present through the core is continuous through the downstream shoulder, or if not,
whether it can find an exit. This depends on the following factors:
 The mechanism causing the concentrated leak, in particular whether it also causes cracking in the
shoulder.

 The material characteristics and width of the downstream shoulder zone.

Scenario 3: Assess the likelihood of the filter or transition zone being effective using
filter design criteria as detailed in Section 7.3. For existing dams where many filters do not
satisfy modern no-erosion design criteria it is suggested that a four way split for filtering
behavior is considered (see Section 7.6):
 Seals with No Erosion – the filtering material stops erosion with no or very little erosion of the
material it is protecting. The increase in leakage flows is so small that it is unlikely to be
detectable.

 Seals with Some Erosion – the filtering material initially allow erosion from the soil it is
protecting, but it eventually seals up and stops erosion. Leakage flows due to piping can be up to
100 l/s, but are self healing.

 Seals with Excessive Erosion – the filtering material allows erosion from the material it is
protecting, and in the process permits large increases in leakage flow (up to 1000 l/s), but the
flows are self healing. The extent of erosion is sufficient to cause sinkholes on the crest and
erosion tunnels through the core.

 Continuing Erosion – the filtering material is too coarse to stop erosion of the material it is
protecting and continuing erosion is permitted. Unlimited erosion and leakage flows are likely.

This approach often shows that filters which do not satisfy modern filter design criteria
will after some erosion of the soil being filtered eventually seal with medium size particles
eroded from the protected soil. Provided the embankment can accommodate the leakage that
occurs up to the time the filters seal, the dam will not fail. In some cases erosion will initiate
again adjacent to the original area due to changed leakage pathways, causing a second
incident.

___________________________________________________________________________
70
 Scenarios Examples
Scenario 1:
Homogeneous zoning with 1 1
no fully intercepting filter
HOMOGENEOUS EARTHFILL EARTHFILL WITH TOE DRAIN,
INTERNAL EROSION ABOVE TOE DRAIN
2
3 3
1

INTERNAL EROSION ABOVE CORE AND FILTERS

Scenario 2: 1A 1 1A
1A 2
Downstream shoulder of 1A
1
fine grained cohesive
material which is capable of ZONED EARTHFILL WITH COHESIVE SHELLS INTERNAL EROSION ABOVE FILTER ZONE,
COHESIVE DOWNSTREAM SHELL
holding a crack/pipe.
2
Scenario 3: 1
2

3
Filter/transition zone is 1 3
1
present downstream of the
core or a downstream EARTHFILL WITH CHIMNEY FILTER ZONED EARTHFILL WITH CHIMNEY FILTER

shoulder zone which is not

capable of holding a 3 3
1
crack/pipe.
ZONED EARTHFILL WITH GRANULAR SHELLS

Scenario 4: SOIL

Piping into an open defect, SOIL

joint or crack. CONDUIT

OPEN JOINTED ROCK

EROSION INTO AN OPEN CRACK EROSION INTO OPEN JOINTS IN

OR JOINT IN A CONDUIT OR WALL ROCK FOUNDATION

Scenario 5:
1 1
Erosion into a toe drain

INTERNAL EROSION THROUGH THE INTERNAL EROSION THROUGH THE

EMBANKMENT INTO A TOE DRAIN FOUNDATION INTO A TOE DRAIN

Figure 7.1 Scenarios for continuation of internal erosion in an embankment (Fell et al,
2008)

Scenario 4: For erosion to continue through an open defect, the defect needs to be
sufficiently open to allow the soil particles surrounding the defect to pass through it. The
suggested procedure for examining this is given in Section 7.7.
Scenario 5: This scenario is applicable if the failure path under consideration involves
a seepage path that exits into a toe drain which could lead to continuing erosion of the
embankment or foundation materials. The assessment considers the design and construction
details of the toe drain, whether filter criteria are met, and the observed condition of the toe
drain (from video or external inspections).

___________________________________________________________________________
71
7.2.2 Internal erosion in the foundation
The likelihood of continuation of erosion should be estimated by assessing the
likelihood that the exit will be a filtered or unfiltered exit.
a) Given the exit is unfiltered the likelihood of continuing erosion will be 1.0.

b) Given the exit is filtered, estimate the likelihood for Continuing Erosion using the method
described in Section 7.3. The probability of continuation will be the product of the likelihood of a
filtered exit and the likelihood assessed considering the filters.

It needs to be recognized that low permeability strata beneath horizontal drains may
prevent them working effectively, Figure 7.2 shows an alluvial foundation where the lower
permeability strata (A and E) will prevent the seepage in the most permeable sand and gravel
strata (B, and D) from flowing into a filtered exit in the horizontal drain.

7.2.3 Internal erosion of the embankment at or into the

foundation
a) Erosion into open joints in the foundation. The likelihood of continuation depends on the joint
openings being sufficiently open and continuous to allow erosion of the core materials.

b) Erosion into coarse grained soil foundation. The likelihood of continuation depends on the
coarse grained soils being sufficiently coarse and continuous to allow erosion of the core
materials

These assessments are often difficult to make because the amount of information on the
foundations is often limited. If there are soils overlying the exits to the open joints or the
coarse grained soils these may act as filters to the erosion.
The assessment of filtering should be carried out as detailed in Sections 7.3 and 7.7.

Figure 7.2 Schematic example of an embankment founded on a stratified alluvial soil

where much of the seepage flow will be to an unfiltered exit (Fell et al, 2005)

___________________________________________________________________________
72
7.3 ASSESSMENT OF THE LIKELIHOOD OF
CONTINUATION WHERE A
FILTER/TRANSITION ZONE IS PRESENT
DOWNSTREAM OF THE CORE OR THERE IS A
DOWNSTREAM SHOULDER ZONE WHICH IS
NOT CAPABLE OF HOLDING A CRACK
The following steps should be followed when assessing the likelihood of continuation:
 Gather the available information on particle sizes of the core or foundation and the filter /
transition materials. This may include data from construction control testing, testing on
samples taken from the dam and its foundation, data from investigations of borrow areas
used for core and filter materials.

 Plot the particle size distributions for the core material and the filters or transitions which
are protecting the core. If the maximum particle size of the core material is >4.75 mm, then
re-grade the core grading such that the maximum size is 4.75 mm. If the base soil is gap
graded, then re-grade the base soil grading on the particle size that is missing (i.e. at the
point of inflection of the grading curve).

 Check if the filter/transition zone will hold an open crack. Modern filter design rules
require less than 5% non-plastic fines content and these will not hold a crack. Fell et al
(2008) suggested Table 7.2 to assist in assessing the likelihood of holding a crack based on
published information and field performance. Table 7.1 was developed based on laboratory
tests carried out by Park, (2005) and case data from Foster (1999), Foster and Fell (1999b).

Table 7.1 Likelihood for Filters with Excessive Fines Holding a Crack (Fell et al 2008)
Probability of holding a crack
Fines Content
Fines Plasticity
% Passing 0.075 mm Compacted Not compacted

5% 0.001 0.0002
7% 0.005 0.001
Non plastic (and no
12% 0.05 0.01
cementing present)
15% 0.1 0.02
>30% 0.5 0.1
Plastic (or fines 5% 0.05 0.02
susceptible to cementing) 7% 0.1 0.05
12% 0.5 0.3
≥ 15% 0.9 0.7
Note: Fines susceptible to cementing for filters having a matrix predominately of sand sized particles (e.g. filters
derived from crushed limestone).

 Assess also if the materials used in the filters may be subject to cementing; e.g. they are
constructed of limestone or dolomite. Fell et al (2005), Schaefer et al (2011) and USBR
(2011) discuss this and other potential mechanisms for cementation and crack holding.

 Assess if the filter / transition materials are likely to be subject to segregation taking
account of the specified and/or measured particle size distribution, the method of
placement, and the width of the filter/transition zone. Adjust the particle size distributions
to allow for segregation if it is likely to occur. The methods of USDA SCS (1994), Kenney
and Westland (1993) and Sherard and Dunnigan (1989) may be used to assist in this
___________________________________________________________________________
73
 assessment. Milligan (2003) makes important observations on the occurrence and
prevention of segregation in filters and fills.

 Assess if the filter / transition materials are internally unstable, and if so adjust the particle
size distributions to account for this using the procedure outlined in Section 6.3.

 Use modern filter design criteria to assess whether the filter/transition will prevent
continuation. There are a number of filter design criteria which may be used. These are
outlined in Section 7.5. Which method is used will be a matter of personal choice. For
filters or transitions which are coarser than required by the filter design criteria the Foster
and Fell (1999a, 2001) method as described below may be used because it allows
assessment of filters which are too coarse to satisfy modern no-erosion design criteria. Fry
(2006, 2007) proposed a hydraulic approach of the concentrated leak development through
the core-shoulders system.

 Check for a “blow-out” condition. In cases where there is limited depth of cover over the
filter/transition zone, assess the potential for blow out by comparing the seepage head at
the downstream face of the core to the weight of soil cover.

 Check whether the filter or transition and the zones downstream will be sufficiently
permeable to perform their required drainage function. This is related to the particle size
distribution of the filter and fines content and plasticity, but more importantly to whether
there are two filters or another downstream zone which will act as a drain.

7.4 FILTER FUNCTIONS

Five filter functions govern the capability of providing control for internal erosion.
1. Retention. The filter limits the transport (erosion) of particles of the base soil it is
protecting.
2. Self Filtration or stability. The filter is internally stable and not subject to finer
particles eroding from within the filter, and will not hold a crack due to cementation
or the presence of plastic fines.
3. No cohesion so the filter will not hold a crack.
4. Drainage. The filter is sufficiently permeable to transport the water flowing
through the base soil into drainage layers or to the toe of the dam.
5. Strength. The filter transfers the stresses within the dam without being crushed and
therefore becoming finer.

Filter Retention
The basic concept of filter retention design is to design the filter so that the voids in the
filter are sufficiently small to prevent erosion of the base soil.
A further basic concept, inherent in filter design and explained by Sherard et al (1984),
is that the base soil will provide a degree of ‘self-filtering’. Hence in Figure 7.3 (a), in a well
graded base soil, the coarser particles in the base soil are prevented from eroding into the
filter and they in turn prevent the medium sized particles in the base soil from eroding and the
medium sized particles in the base soil prevent the fine particles in the base soil from eroding.
If the base soil is gap-graded or graded concave upwards, there is a deficiency of medium
sized particles, as shown in Figure 7.3 (b). Then self filtering does not occur and the fine

___________________________________________________________________________
74
particles in the base soil will erode through the coarse particles. In these situations for the
filter to be successful in controlling erosion, it must be able to control the erosion of these
finer particles.
For dispersive soils the clay particles do not flocculate to form aggregated particles of
clay so finer filters are required than for non-dispersive soils of the same particle size
distribution.
For filters and transitions which are subject to suffusion the finer fraction may erode
from the filter and result in a coarser filter less able to prevent continuation of erosion.
Four methods are used to achieve the retention function:
 Geometric Criterion
1. Methods based on particle size distribution: These were first developed by Terzaghi
and later refined, e.g. methods based on Sherard & Dunnigan, (1989). For design
purposes the D15F is most commonly used to define this void size and D85B the
representative size of the base soil. This is determined on the fraction of the soil
passing a 4.75mm sieve.
2. Methods based on the controlling constriction size or equivalent opening size: e.g.
Brauns and Witt (1987); Schuler and Brauns (1993, 1997); Lafleur et al (1993);
Giroud (2003); Indraratna and Raut (2006); Indraratna et al (2007) and Raut and
Indraratna (2008). Giroud (2003) and Indraratna’s research team’s methods allow
for the effects of relative density on the filter design. Generally, these methods
relate the controlling constriction size to a % of fine passing sizes.
 Hydraulic Criterion
3. Methods that include an element of hydrodynamics such as the Russian standards
(see 4.2.4 and Pavchich, 1961, ICOLD, 1994).
4. Methods based on permeability which is regarded as being correlated to the
constriction size: e.g. Vaughan & Soares (1982), Vaughan and Bridle (2004),
Delgado et al, (2006).
Which is used is a matter of personal preference and experience. The methods above
are mainly used to design no-erosion filters in critical filter situations.
In some applications such as in long dikes it may be very costly to use such filters and
filters which allow some erosion (or “open filters”) may be used. These can be designed
using the methods of Bakker (1987), Den-Adel et al. (1994) and Verheij et al (2012).
However great care should be taken in applying this approach, particularly to situations
where, if the filter fails and erosion occurs, it is not practical to repair the damage.

___________________________________________________________________________
75
 Figure 7.3 Filtering and self filtering concepts, (a) well graded base soil, (b) gap
graded, or concave graded soil, deficient in medium sized particles (Fell et. al. 2005).

No cohesion (filter will not hold a crack)

This function is to confirm that the filter will collapse at locations where an erosion
pipe or a crack has initiated. The ability not to hold a crack is commonly achieved by limiting
the fines content (% passing 0.075mm) and requiring that the fines are non-plastic.
There is some evidence that even with these precautions very dense compacted filters
may hold a crack (Redlinger et al, 2011).
It should be recognized that cementing of filters may occur if the filter is composed of
carbonate rock particles such as limestone or dolomite particularly for the sand fraction (Fell
et al, 2005), and that even small amounts of silt in broadly graded silty sandy gravel may
result in the filter being able to crack. This is discussed further in Section 7.3.1 and in Table
7.2.

Self filtration or stability

Self filtration and stability are achieved by checking against internal instability as
described in Section 6.2.2.

Drainage
This is achieved by limiting the fines content (% passing 0.075mm), and in some
design manuals relying on relationships between D15F / D15B,to ensure the filter is sufficiently
more permeable than the base soil to perform the drainage function. This is also related to the
zoning of the dam, and in particular whether drainage filter zones or rockfill zones are
provided.

___________________________________________________________________________
76
Strength
This is achieved by specifying durability and crushing strengths of the aggregates.
Some more details on filters are provided in Section 7.5. More information is also
available in the ICOLD Bulletin on Embankment Dam Granular Filters and Drains, (ICOLD,
1994).

7.5 BRIEF DESCRIPTION OF SOME OF THE

AVAILABLE FILTER DESIGN CRITERIA FOR
CRITICAL FILTERS
Particle Size Based Criteria-Sherard and Dunnigan (1989)
Sherard and Dunnigan (1985, 1989) recommended the following steps:
For all soils with a gravel component (except Group 3 below), the filters should be
designed on the grading of that part of the soil finer than 4.75 mm. This is an important step
and designed to overcome problems with very broadly graded soils which will not self filter.
Base soil gradations are determined by sieve tests and hydrometer tests using deflocculant
such as sodium hexametaphosphate.
Impervious Soil Group 1 (fine silts and clays): For fine silts and clays that have more
than 85% by weight (of the portion finer than the 4.75 mm sieve) finer than the 0.075 mm
sieve, the allowable filter for design should have D15F < 9 D85B.
Impervious Soil Group 2 (sandy silts and clays and silty and clayey sands): For sandy
(and gravelly) impervious soils with 40% to 85% by weight (of the portion finer than the 4.75
mm sieve) finer than the 0.075 mm sieve, the allowable filter for design should have D 15F ≤
0.7 mm.
Impervious Soil Group 3 (sands and sandy gravels with small content of fines): For
silty and clayey sands and gravels with 15% or less by weight (of the portion finer than the
4.75 mm sieve) finer than the 0.075 mm sieve the allowable filter for design should have D15F
< 4 D85B where D85B can be the 85% finer size of the entire material including gravels.
Impervious Soil Group 4: For coarse impervious soils intermediate between Groups 2
and 3 above, with 15% to 40% passing the 0.075 mm sieve, the allowable filter for design is
intermediate, inversely related linearly with the fines content and can be computed by straight
line interpolation. As an example (Figure 7.4) for an impervious sandy soil with 30% of silty
or clayey fines and D85B = 2 mm, the allowable filter for design is in between the value of
D15F = 0.7 mm (for soils of Group 2) and D15F = 4 x (2) = 8 mm (for soils of Group 3), and is
calculated as follows:
40-30
D15F = (8-0.7) + 0.7 = 3.6mm
40-15

___________________________________________________________________________
77
 Figure 7.4 Determination of allowable D15F for filters for impervious Soil Group 4
(having between 15% and 40% finer than 0.075 mm sieve). (Sherard and Dunnigan, 1985)

These criteria have been used to develop Design Guidelines such as USDA-SCS (1994)
and USBR (2004). They supersede the methods of Terzaghi and Peck (1948) which are not
conservative for coarse base soils and too conservative for fine soils.

Foster and Fell (1999a, 2001) tested more soils and added these to the Sherard and
Dunnigan test data and found that the Sherard and Dunnigan criteria were not sufficiently
conservative for dispersive soils. For soils testing as D1 or D2 in the Pinhole test in distilled
water Foster and Fell (2001) suggested using D15F < 6 D85B for Group 1 soils, and D15F ≤
0.5 mm for Group 2 soils, and to move the boundary between Group 1 and 2 soils from 40%
to 35% fines passing 0.075mm.

Methods based on constriction size or opening size

There are a number of these methods, which generally relate the controlling
constriction size to a % of fine passing sizes. Some are described in ICOLD (1994): e.g.
Brauns and Witt, (1987). Others include Schuler and Brauns, (1993, 1997); Lafleur et al
(1993); Giroud (2003, Terzaghi lecture, 2010); Indraratna and Raut (2006); Indraratna et al
(2007) and Raut and Indraratna (2008). The methods of Giroud and Indraratna’s research
team allow for the effects of relative density on the filter design.

Methods based on the permeability of the filter

There are a number of methods relating the permeability of a filter to its filtering
capability. This is potentially useful in assessing the vulnerability of a dam to internal erosion
because if the permeability of the zone adjacent to a core were known, its ability to filter and
protect the core against erosion could be assessed. The difficulty with this approach is that
permeability is very variable, both locally and widely even in one fill zone, and measuring it
in-situ is also subject to some difficulty. However, as the discussion above on applying
particle size filter design criteria shows, assessing filtering capability requires both analysis
and judgment, some permeability based methods are included.
Vaughan and Soares (1982), Vaughan and Bridle (2004) describe the “perfect filter”
method which was developed so that the filter itself would be able to prevent erosion of the
finest particles in the base soil. It was pointed out that the finest ‘particles’ present in non-
dispersive clay soils are often flocs, accumulations of finer particles. Flocs are usually
dispersed (de-flocculated) when samples are prepared for the sedimentation stage of a particle
size analysis, but the floc size for the base soil is determined by hydrometer sedimentation
tests with no dispersant added (the ‘double-dispersion’ test). The water used for the test
___________________________________________________________________________
78
should be of the same chemistry as the water to be seeping through the dam since the floc
size is related to the degree of dispersion which is controlled by the water chemistry. A more
conservative approach is to prepare samples in distilled water also (the ‘triple dispersion’
test). Samples can be tested as a dilute suspension in filter experiments.
Vaughan & Soares (1982) carried out a series of tests to determine the sizes of floc
trapped by filters of varying permeability. These are summarized on Figure 7.5. The open
points show effective filters and the solid points show ineffective filters. For example, a floc
of 100 microns (0.1 mm) effective diameter would be retained by a filter with a permeability
of about 10-2 m/s or less.
The Vaughan & Soares (1982) equation linking permeability of filter to the particle size
that the filter will trap is as follows:
k = 6.7 *10-6 * δR 1.52
where: δR = size of smallest particle retained in microns (10-6 mm)
k = permeability of filter (m/s)
This can be transposed as follows:
δR = 2.54 * 103 *(k)0.658
The “perfect filter” method is very conservative because it does not rely on self filtering
(Bridle 2007, 2008 and Bridle et al 2007). It will often give finer filters than particle size
based methods, but conversely if the permeability of a potential filter zone downstream of the
core, say, satisfies the Vaughan & Soares criterion, erosion will not continue.

Figure 7.5 Vaughan & Soares (1982) relationship between filter permeability and size of floc
trapped

Delgado et al (2006) carried out extensive filter tests to determine no-erosion filter
(NEF) criteria based on the percentage of the base soil passing the 0.075mm sieve, and the
permeability of the filter.
They combined these data with that of Sherard et al (1984), Foster and Fell (1999a) and
others to produce Figure 7.6. For the tests done elsewhere Delgado et al (2006) determined
the filter permeability using the relationship from Sherard and Dunnigan (1989) (k =
0.35 (D15) 2 where (D15) is in millimeters, and k is in cm / sec). For the University of Granada
(UGR) tests Delgado et al (2006) determined the permeability by laboratory tests, thereby
taking account of the degree of compaction of the filter.
#
___________________________________________________________________________
79
 Success (external data)
1,E+03 Intermediate (external data)
Failure (external data) Soil group 2 Soil group 1
Success (All UGR NEF tests)
Intermediate (All UGR NEF tests)
Failure (All UGR NEF tests)
1,E+02 Boundary filter, (All UGR NEF tests)
Boundary filter lower limit All UGR NEF tests
Boundary filter regression line, (All UGR NEF tests)
Filter Permeability kF [cm/s]

1,E+01

1,E+00

1,E-01

1,E-02

lower limit:
kF = 0.004 cm/s
1,E-03
40 50 60 70 80 90 100
Base Soil P0.075B [%]

Figure 7.6 Delgado et al (2006) filter design criteria for non dispersive soils with 40%
to 100% passing 0.075mm sieve relating percentage passing 0.075mm of the base soil to
filter permeability in cm / sec.
Delgado et al (2006) show two design criteria, one the “Boundary Filter Lower Limit of
all UGR NEF tests, and the second “Boundary Filter Regression Line” which is fitted to
base-filter combinations characterizing the transition from no failure to failure. They do not
specifically say which of these should be used but suggest that users take account of all the
data in the Figure 7.6 so the uncertainty in the criteria is apparent.
Delgado et al (2012) used NEF tests to examine the influence of plasticity, mineralogy,
water content, hydraulic gradient and additives including aluminum sulfate on internal
erosion of clayey soils protected by granular filters. As expected the aluminum sulfate
increased floc size and thereby increased filter size. Base soils at higher water content
required finer filters but the other factors had no significant effect on filter requirements.

7.6 FOSTER AND FELL (1999a, 2001) METHOD FOR

ASSESSING THE LIKELIHOOD OF
CONTINUATION OF EROSION FOR FILTERS
AND TRANSITIONS WHICH DO NOT MEET
MODERN FILTER DESIGN CRITERIA
Foster and Fell (1999a, 2001) analyzed the tests in the NEF equipment with filters
which were too coarse to satisfy no-erosion filter criteria but which did eventually seal as
coarser particles from the base soil were eroded and collected on the surface of the filter to
form a finer filter which stopped the erosion process. They also considered the tests to define
the filter grading at which erosion would continue, called the continuing erosion boundary.
From these data and the test data from Sherard et al (1984) they proposed the criteria in
Table 7.3 for the excessive and continuing erosion boundaries.

___________________________________________________________________________
80
 Information from case histories of poor filter performance suggests the potential
maximum leakage flows that could develop due to piping are as follows:
 Filters falling into the Some Erosion category – up to100 l/sec:

 Filters falling into the Excessive Erosion category – 100 to 1000 l/sec.

 Filters falling into the Continuing Erosion category – flows of 1000 l/sec and increasing.

Figure 13.2 in Terminology presents the ‘no’, ‘some’, ‘excessive’ and ‘continuing’
erosion concept diagramatically.
Fell et al (2008) recommend that the maximum leakage flows listed above be used in
the assessment of the probability of a breach mechanism developing. Lower leakage flows
are likely if upstream flow limitation occurs, but the factors would need to be carefully
considered and justified if they were to be relied on in the assessment.

Table 7.3 Excessive and Continuing erosion criteria (Foster and Fell 1999a, 2001).
Base Soil Proposed Criteria for Excessive Erosion Proposed Criteria for
Boundary Continuing Erosion
Boundary
Soils with DB95<0.3 mm DF15 > 9 DB95
Soils with 0.3<DB95<2 DF15 > 9 DB90
mm
Soils with DB95>2 mm DF15 > the DF15 value which gives an For all soils:
and fines content >35% erosion loss of 0.25g/cm2 in the CEF test DF15 > 9DB95
(0.25g/cm2 contour line in Figure 7.7)
Soils with DB95>2 mm DF15 > 9 DB85
and fines content <15%
Soils with DB95>2 mm DF15 > 2.5 DF15 design,
and fines content 15-35% where DF15 design is given by:
DF15design=(35-pp%0.075 mm) (4DB85-
0.7)/20+0.7
Notes: (1) Criteria are directly applicable to soils with DB95 up to 4.75 mm. For soils with coarser particles
determine DB85 and DB95 using grading curves adjusted to give a maximum size of 4.75 mm.(2). CEF is
Continuing Erosion Filter test.

___________________________________________________________________________
81
 Mud Mountain
40mm
30
Range of DF15 for dams with
poor filter performance
25

Juklavatn Secondary
Juktan

Churchill Falls FF11

Average DF15 for dams with Contours of
poor filter performance Erosion Loss
Filter DF15 (mm)

20
from Filter
Range of DF15 for dams with Tests
good filter performance
15

Viddalsvatn
1.0g/cm2

Balderhead
0.5g/cm2

Tarbela AD1
No Erosion
Rowallan
10 Hyttejuvet 0.25g/cm
Boundary

Wreck Cove
2
for Soil Group 2 Hills Creek 0.1g/cm2
5 DF15=0.7mm
Brodhead
Songa
0
0 5 10 15 20 25 30 35 40 45 50 55 60
Core material % fine - medium sand (0.075 - 1.18mm)

Figure 7.7 Comparison of erosion losses measured in filter tests to dams with poor and
good filter performance (Foster 1999, Foster and Fell 2001).

This approach is particularly useful when assessing existing dams as many have a filter
or transition which is coarser than required for no-erosion filters. Many of these dams have
rockfill or permeable zones downstream which can cope with some leakage flow without
failing before the filters eventually seal.

It should be noted that for soils with a small percentage of sand size particles (in
Sherard and Dunnigan Group 1) there is often not a large difference between the no-erosion,
excessive erosion and continuing erosion D15 sizes. There are insufficient sand size particles
in the soils to be eroded and seal on the filter and form an effective filter for the base soil.
Delgado et al (2006) found the same applies to their permeability based criteria.
In practice, the core and filter or transition in a dam have a range of particle size
gradings, and it is important to consider the implications of this. To do this the steps listed
below should be followed, producing the results of the type shown in Figure 7.8:
 Evaluate the DF15 values for the No Erosion (e.g. in accordance with Sherard & Dunnigan,
1989), Some Erosion (i.e. low end of Excessive), Excessive Erosion and Continuing Erosion
boundaries using Table 7.3. Do this for the finest base soil grading, the average base soil grading
and the coarsest base soil grading. Plot the DF15 values for these boundaries on the grading
curve limits of the filter/transition material (see Figure 7.8 for an example). Use the grading
curves for the filter/transition zone adjusted for segregation and suffusion if applicable;

 Estimate the probabilities for Some, Excessive and Continuing Erosion for each representative
base soil grading. Estimate the proportion of the filter/transition grading that fall into each of the
particular erosion categories based on the plot of filter/transition grading curves versus Filter
Erosion Boundaries. More detailed guidance on how to do this is given in Fell et al (2008);

 Assess the potential leakage flows that could develop if piping were to initiate in the core based
on the guidance given above;

___________________________________________________________________________
82
 Assess the implications of this leakage on the stability of the downstream slope and whether
unraveling might occur.

It should be noted that as for the example in Figure 7.8, the filters can be significantly
coarser in dams than those required for no-erosion filters, and still eventually seal with only
some erosion of the protected soil, and relatively small leakage flows.
The principles of this process are applicable regardless of which method is being used
to assess the suitability of the filter or transition.
An example of the application of these principles is given in Foster (2007).

100

90 Zone 1- Fine Grading

Zone 1 - Average Grading

80

Zone 1 - Coarse Grading

70 Filter erosion boundaries for the Average core grading
Filter erosion boundaries for the Fine core grading
Percent Passing

60 Filter erosion boundaries for the Coarse core grading

Zone 1 - Core Material
re-graded to maximum size of 4.75 mm
50
Zone 3- Filter
Transition
40

30

20
0.7 2.5 36 38

No 0.7 Some 2.0 Excessive 30 Continuing

10
Erosion Erosion Erosion Erosion

0
0.001 0.01 0.1 1 10 100 1000
Seive Size (mm)

Assessment of Zone 1 core against no erosion, excessive erosion and continuing erosion criteria
Core Base soil sizes (mm) No Erosion Excessive Continuing
Gradation Erosion Erosion
% passing % fine-medium sand
DB85 (mm) DB95 (mm) 0.075mm (0.075 - 1.18mm) DF15 (mm) DF15 (mm) DF15 (mm)
Fine Grading 1.9 3.3 50 25 0.7 2 30
Average 2.4 4 41 29 0.7 2.5 36
Coarse Grading 2.5 4.2 35 30 0.7 2.6 38

Figure 7.8 Example of plot showing filter/transition gradings compared to Filter

Erosion Boundaries (Fell et al 2008)

___________________________________________________________________________
83
7.7 ASSESSMENT OF THE LIKELIHOOD OF
CONTINUATION FOR INTERNAL EROSION
INTO AN OPEN DEFECT, JOINT OR CRACK IN
THE FOUNDATION, IN A WALL OR CONDUIT
There are no commonly adopted criteria for assessing the likelihood of continuation for
this scenario. Fell et al (2008) suggest use of the criteria in Table 7.4 which have been
determined by assuming that the Foster and Fell (1999a, 2001), continuing erosion criteria
apply to erosion into an open defect, joint or crack, and that the crack width is equivalent to
the filter opening size of the voids between the particles in a filter. The filter opening size has
been shown by Sherard et al (1984) to be D15 / 9.

Table 7.4 Continuing Erosion criterion for erosion into an open defect (Fell et al, 2008)
Erosion Condition Comparison of Soil Gradation to Joint/Defect
Opening Size (JOS)
Continuing erosion (CE) JOSCE = D95 surrounding soil
Notes: JOSCE = Joint/defect opening size that would allow continuing erosion of the
surrounding soil.
D95 should be based on the average soil grading after re-grading on 4.75mm
particle size.

___________________________________________________________________________
84
 8. PROGRESSION OF EROSION
8.1 THE TWO CONDITIONS FOR PROGRESSION
Erosion, once initiated and not arrested by filter action in the dam, will progress under
two conditions: one a hydraulic condition and the other a mechanical condition:

8.1.1 Hydraulic condition

For erosion to progress, hydraulic conditions must be suitable. Water seeping through
the dam must be doing so with sufficient velocity to provide sufficient energy or drag force to
continue to transport particles along openings and off the external surfaces of the dam in a
continuous process. This happens for the four initiating mechanisms when:
 For concentrated leak erosion, as the pipe enlarges the hydraulic shear stresses increase, so the
erosion will progress unless the reservoir is drawn down to reduce the gradient.

 For backward erosion, whether the erosion progresses upstream towards the reservoir or river
depends on the seepage gradient within the pipe. There is a critical gradient above which the
particle will be detached and a critical flow velocity above which the soil particles will be
transported in the pipe.

 For contact erosion, the situation may be as for concentrated leak erosion or erosion may not
progress if for example the finer soil clogs the coarse soil.

 For suffusion, provided the eroded soil is carried away, the hydraulic condition will become
increasingly able to erode the finer fraction as the permeability increases.

The progression can stop if the flow in the concentrated leak pipe or in the soil
experiencing concentrated leak erosion, contact erosion or suffusion, is limited by head losses
in the upstream or downstream zones. In this case an equilibrium situation may develop
where the eroding forces become equal to or less than the resisting forces. In backward
erosion, progression of the pipe stops if the gradient is lower than the critical gradient.

8.1.2 Mechanical condition

For erosion to progress, the mechanical conditions must be suitable. Water seeping
through the dam must be doing so because the crack is sustained by hydraulic fracture or
because the pipe or the cavity through which eroded particles are being transported does not
collapse. Plastic (cohesive) soils, when both saturated and partially saturated, can sustain
openings by the soil acting as a roof or wall to the pipe. Non-plastic silts, sands, and gravels
will in general not “hold a roof” as the roof will collapse when they become saturated.
However, partially saturated and high fines content non-plastic soils may hold a roof, along
the phreatic surface, for example, but the roof, which in this situation is sustained by pore
pressure suction, may collapse on saturation.
Progression may also stop if particles from a zone upstream of the core are transported
in the developing pipe and eventually seal the filter. This action is sometimes called crack
stopping. For it to occur there must be a filter or transition downstream of the core to trap the
eroded particles. An example of this is Matahina dam (Gillon, 2007).

___________________________________________________________________________
85
8.2 OVERALL APPROACH TO ASSESSING
PROGRESSION OF CONCENTRATED LEAK
EROSION
The likelihood of concentrated leak erosion progressing depends on whether:
 The soil will “hold a roof” over a pipe.

 “Crack filling” action will not stop the erosion process.

 Flow in the developing pipe will not be restricted by an upstream zone or hydraulic losses
in upstream and downstream zones (or for example a concrete face slab)

It should be noted that in the absence of crack filling and flow limitation, or lowering of
the reservoir to reduce the gradient of flow in the developing pipe, erosion will progress. This
is because the hydraulic shear stresses in the pipe increase as the pipe enlarges.

8.2.1 Assessing whether the soil will hold a roof to a developing

pipe
For internal erosion and piping through the dam or piping from the embankment into a
rock foundation, the core must be capable of holding the roof of a pipe.
Based on case studies (Foster 1999), (Foster and Fell 1999b), the most important
factors are:
 The fines content of the soil (% passing 0.075 mm) – ≥ 15% fines likely to be able to hold
a roof regardless of whether the fines were non plastic or plastic.

 Whether the soil was partially saturated or saturated.

Other factors which were considered to be likely to have an influence were degree of
compaction (loose soil would be less likely to support a roof to a pipe than dense), and
reservoir operation (cyclic reservoir levels were more likely to cause collapse than steady).
Fell et al (2008) developed Table 8.1 based on these case data and taking into account
the results of testing by Park (2005) which showed sandy gravels with 5 to 15% non-plastic
fines collapsed quickly when saturated. Sandy gravels with 5% cohesive fines collapsed after
some time, but very slowly with 15% cohesive fines. The table adds detail to the general rule
that non-plastic silts, sands and gravels cannot support a roof, except when partially saturated,
and plastic (cohesive) clays and silts can support a roof when saturated or partially saturated.
In the table a likelihood of 1.0 means that the soil is certain to support a roof, very small
likelihoods mean that the soil will not support a roof.

___________________________________________________________________________
86
 Table 8.1 Likelihood of a soil being able to support a roof to an erosion pipe (Fell et al
2008)
Soil Classification Percentage Plasticity of Moisture Condition Likelihood of
Fines Fines Supporting a
Roof
Clays, sandy > 50% Plastic Moist or saturated 1.0
clays (CL, CH, CL-
CH)
ML or MH >50% Plastic or Moist or saturated 1.0
non-plastic
Sandy clays, 15% - 50% Plastic Moist or Saturated 1.0
Gravely clays,
(SC, GC)
Silty sands, > 15% Non plastic Moist 0.7 to 1.0
Silty gravels, Saturated 0.5 to 1.0
Silty sandy
gravel (SM, GM)
Granular soils 5% to 15% Plastic Moist 0.5 to 1.0
with some cohesive Saturated 0.2 to 0.5
fines (SC-SP, SC-SW,
GC-GP, GC-GW)
Granular soils 5% to 15% Non plastic Moist 0.05 to 0.1
with some non plastic 0.02 to 0.05
fines (SM-SP, SM-SW,
Saturated
GM-GP, GM-GW)
Granular soils, < 5% Non plastic Moist and saturated 0.0001
(SP, SW, GP, GW) Plastic Moist and saturated 0.001 to 0.01
Notes: (1) Lower range of probabilities is for poorly compacted materials (i.e. not rolled), and upper
bound for well compacted materials.
(2) Cemented materials give higher probabilities than indicated in the table. If soils are cemented,
use the category that best describes the particular situation.

8.2.2 Assessing whether crack filling action will occur

For crack filling to occur requires a granular zone upstream of the core, with particles
of a size which can be transported by the water flowing into the crack or pipe; and a
downstream filter transition zone or rockfill which is sufficiently fine to act as a filter to these
particles. Case studies where crack filling was evident include Matahina Dam, (Gillon, 2007),
Swedish dams (Nilsson, 2007a, b), and others in Foster and Fell (1999b). These are typified
by sinkholes appearing above the upstream filter zone which is taken as evidence of material
being washed into the pipe.
Crack filling action is only possible for central and sloping core earth and rockfill (or
gravel) shoulders. Whether it will be effective is dependent on compatibility of particle sizes
of the granular soils upstream of the core and in the downstream filter transition, and then the
compatibility of the downstream transition with in-filled material from upstream, and the
core.

___________________________________________________________________________
87
8.2.3 Assessing whether upstream flow limitation will occur
Upstream flow limitation may occur where there is a relatively fine grained granular
material (fine rock fill, or sandy gravels) upstream of the core, or where there is a concrete
face slab or concrete core wall.
Fell et al (2008) gives guidance on the assessment of the likelihood of upstream flow
limitation being effective.
Both upstream and downstream zones can generate such high hydraulic losses
according to the permeability of the filter that the erosion rate in the crack passing through
the core can be stopped. Concentrated leak erosion through a 2 mm wide crack of the core of
a 50 m high zoned dam without a filter was modeled by Fry (2007). This showed that there
are three behaviors: erosion does not initiate, erosion initiates and stops, and erosion does not
stop.
These behaviors are mainly dependant on the critical shear stress of the core which
controls the initiation of erosion and the stabilization of erosion and the permeability of the
upstream shoulder controlling the head loss at the borders of the core. Fry (2007) showed that
for shoulder permeability lower than 10-3 m/s the final discharge rate stabilized below 10 l/s
for the case modeled. This figure is an indication of the benefits of upstream flow limitation
but the actual ‘stabilized’ discharge will vary in individual situations.

8.3 ASSESSING THE RATE OF DEVELOPMENT OF A

CONCENTRATED LEAK EROSION PIPE
If analysis using the information in the paragraphs above shows that a concentrated leak
will be sustained and the roof will not collapse or the flow rate drop because of upstream and
downstream restrictions, the concentrated leak erosion pipe will enlarge.
From the Hole Erosion Index of the eroding soil and average hydraulic gradient along
the pipe it is possible to estimate the rate at which a piping hole will enlarge. This is
important because it is a significant factor in assessing the likelihood of successful
intervention to stop the piping process should an emergency occur. The results are
summarized in Figure 8.1 assuming (a) Unrestricted potential for erosion (i.e. no flow
limitation, continuing erosion condition); (b) Initial pipe diameter of 25mm; (c) Zero critical
shear stress which is conservative, particularly for IHET >3.5; (d) Reservoir level remains
constant. It will be seen from Figure 8.1 that the time for erosion to progress is very
dependent on the soil erosion properties, expressed by the Erosion Rate Index, IHET, see
Tables 3.3 and 3.4.

___________________________________________________________________________
88
 10000
Gradient along pipe = 0.1
Time for Pipe to Enlarge to 1m diameter (hours)

Gradient along pipe = 0.15

1000 Gradient along pipe = 0.2
Gradient along pipe = 0.5

100

10

0.1
1 2 3 4 5 6
Hole Erosion Index of Core Material

Figure 8.1 Approximate time for pipe to enlarge from 25mm to 1 m diameter The time
to erode to 2 m diameter is about 20% greater (Courtesy of M Foster)

8.3.1 Rapid enlargement of concentrated leak erosion pipes and

need for filters or other works to prevent continuation and
progression
Note that Figure 8.1 shows that the time for the pipe to enlarge is only a matter of hours
in most soils, and even in the most resistant of soils (Erosion Rate Index 5, best estimate, see
table 3.4) enlargement occurs in only 100-500 hours, or 4 days to 3 weeks. These are short
periods but would provide time to issue warnings and take precautions to alleviate some of
the impacts of failure in an emergency. However, for the less erosion resistant soils the times
to enlarge are so short that, other than at dams where the consequences of failure are minimal,
some action, such as providing filter protection, should be taken to prevent erosion from
continuing and progressing.

8.4 RATE OF PROGRESSION OF INTERNAL

EROSION GENERALLY
Sections 11.2.2 and 11.2.3, particularly Tables 11.1, 11.2 and 11.3, give general
information on the rate of progression of internal erosion. In many situations, the rate is rapid.
Therefore, as with the progression of concentrated leak erosion discussed above, if erosion is
intiated and continues, and is not arrested by the filtering capability of the dam, some
remediation, such as providing filters, will be required.

___________________________________________________________________________
89
 9. ENGINEERING ASSESSMENT OF
THE VULNERABILITY OF A DAM TO
INTERNAL EROSION
9.1 EIGHT STEPS TO FAILURE BY INTERNAL
EROSION
This chapter deals with the framework for carrying out an engineering analysis of the
vulnerability or resistance of a dam to internal erosion. It follows the process described in
Chapter 2 and examines the loads that may cause internal erosion and their effects at all
locations where parts of the overall internal erosion process may occur. In effect it examines
failure modes - the steps along routes through which the internal erosion process may lead to
failure – and analyses whether the mechanical properties of the material in the dam will allow
internal erosion to occur or whether they will resist it.
The framework for making the engineering analysis of the mechanics of internal
erosion has been adapted from those developed for risk assessment which goes through the
process from application of loads through to breach. The entire process must be examined
because the probability of erosion proceeding or stopping must be estimated at each step in
order to arrive at an overall probability of failure from internal erosion. Each step in the
process is summarized in Chapter 2. The entire framework is summarized in Table 9.1:

Table 9.1: The eight steps of the framework for assessment of the probability of failure
by internal erosion for a dam adapted from Fell and Fry (2007)
STEPS MATTERS TO BE CONSIDERED
1 Loading Hydrostatic or reservoir loads: Frequent water level, rare flood and dam
safety flood.
Seismic: a range of loading including OBE and MDE (ICOLD, 2011)
2 Location of Embankment: Upper portion; Lower portion; Along conduit; Adjacent to
initiation of wall;
erosion Over changes in slope in the cross valley foundation profile; At construction
features such as haul roads and river closure sections;
Foundation: Valley; Abutment
Embankment to Foundation: Valley; Abutment
3 Initiation Erosion Mechanism: Concentrated leak; Backward Erosion; Contact
Erosion; Suffusion
Whether erosion will initiate under the seepage gradients :
4 Continuation Assess whether filters, transition zones or downstream zones will prevent
(Filtration) erosion continuing:
Particle size methods: This can provide no erosion; excessive erosion;
continuing erosion limits. If there is no filter erosion will continue.
Equivalent opening size methods
Permeability based methods.

___________________________________________________________________________
90
 STEPS MATTERS TO BE CONSIDERED
5 Progression For concentrated leak and contact erosion:
Will a developing pipe stay open?
Upstream and downstream flow limitation taking account of erosion
properties.
For backward erosion, and suffusion:
Is the critical gradient or velocity reached for erosion to progress?
6 Detection Piping process:
Monitoring type and frequency:
Surveillance frequency:
7 Intervention Piping process:
Personnel availability and training:
Equipment and materials availability:
Weather and the impacts of flooding on access:
The erosion rate and its impact on the time to available intervene.
Whether the reservoir can be drawn down in time to prevent breach.
8 Breach Gross enlargement of the pipe
Overtopping (e.g. due to settlement of the crest from suffusion and/or due to
the formation of a sinkhole from a pipe in the embankment)
Slope instability of the downstream slope
Unraveling of the downstream face.
Static liquefaction

9.2 APPROACHES TO ENGINEERING ASSESSMENT

To assess if internal erosion could progress to failure, it is necessary to assess each of
the eight steps described in Table 9.1.
First, all the Potential Failure Modes (PFM) are identified, and those not likely to
progress to failure of the dam screened out as detailed in Section 9.10 and 9.11.
The remaining Potential Failure Modes are then assessed. To assess the initiation of
concentrated leak erosion, the width and depth of cracks, or the dimensions of openings, is
required in addition to a knowledge of soil properties and estimates of hydraulic gradients.
Guidance on the estimation of crack dimensions is given in Section 3.3. Initiation of failure
by backward erosion can be assessed with a knowledge of soil properties and water level,
because backward erosion is intiated when the hydraulic gradient exceeds some critical value
(Figure 4.4). Initiation of contact erosion can be assessed from a knowledge (or estimate) of
permeability and hydraulic gradient (to give Darcy velocity) (Figure 5.2). Suffusion may
initiate in soils with unfavourable grading at low hydraulic gradients (Chapter 6).
Three approaches are proposed to define the applied hydraulic loads and the means of
analysing whether or not internal erosion could progress to failure, as follows:

9.2.1 Deterministic or Standards-Based Approach

In the Deterministic or Standards-based approach the reservoir level is taken to be the
spillway design flood level (or the level at which the hydraulic gradient would be greatest if
the passage of the design flood raises tailwater levels). The maximum design seismic load is
also assumed to apply.

___________________________________________________________________________
91
 It is assumed that concentrated leak erosion will initiate in the embankment under the
design flood level and that cracks in which erosion would intiate would be formed under the
maximum design seismic load. For backward erosion, suffusion and contact erosion
calculations at spillway design flood level will determine whether or not initiation of erosion
will occur. The engineering assessment then examines whether continuation, progression, and
breach could occur.

9.2.2 Engineering Judgment

The Engineering Judgment Approach, which may include qualitative and semi-
quantitative risk assessment, as the engineer may judge necessary, attempts to enumerate all
the factors involved. This includes estimating crack widths and depths, using methods and
knowledge from the engineer’s experience and the guidance in Section 3.3. A range of crack
dimensions should be considered. The information in Section 3.4 can then be used to assess
whether or not the dam would resist the intiation of concentrated leak erosion. If the engineer
judged, or analyses had indicated, that openings resulting from hydraulic fracture, in low
stress zones, for example, would be present, the potential for erosion through them could also
be examined.
As in the Deterministic Approach, the analysis is carried out assuming the reservoir
level is the spillway design flood level, and that the maximum design seismic load applies. If
the assessment showed that concentrated leak erosion could occur, the engineering
assessment then examines whether continuation, progression, and breach could occur.
Whether backward erosion, contact erosion or suffusion will initiate, continue and progress to
breach will have been determined from the Deterministic Approach.
As a further aid to judgement, the situation at reservoir water levels lower than the
spillway design flood could be examined, where the zoning of the embankment changes, or
where seepage has been known to increase significantly, for example. The approach then
becomes semi-quantitiative because the likelihoods of these reservoir levels being reached
can be estimated from the hydrology and the spillway characteristics. Using these data
engineering judgment can be used to qualitatively assess the safety of the dam. This may
include assessing the likelihood of successful detection and intervention in the event that
internal erosion initiates.
The method of engineering judgement augmented by qualitative and semi-quantitative
risk assessments is also a preliminary to quantitative risk assessment, and may assist in
deciding if the particular circumstances at a dam necessitate additional investigation by
quantitative risk assessment.

9.2.3 Quantitative Risk Assessment

For the Quantitative Risk Assessment approach calculations are carried out as for the
Engineering Judgment approach but estimates are made of the likelihoods of whether
cracking or hydraulic fracture may occur, and the depth and width of cracks or other defects
in which concentrated leak erosion may initiate estimated. The likelihood of initiation of
erosion is estimated from these, the seepage flow gradients in the cracks, and the soil erosion
properties. These likelihoods and the likelihood of cracking are estimated by engineering
judgment by person experienced in such matters, aided by published information and
manuals. In critical cases, these elements of quantitative risk assessment could be used to
investigate concentrated leak erosion alone. The likelihood for all 8 steps in the internal
erosion process are then combined to estimate the likelihood of failure of the dam. This is
done for each PFM. The methods are introduced briefly in Chapter 12.

___________________________________________________________________________
92
9.2.4 Use of the three approaches
Which approaches are used is for the Dam Engineer to decide in consultation with the
Owner and Regulator.
As many dam engineers are not familiar with risk based methods they will prefer to use
Deterministic or Engineering Judgement Approaches.
It should be recognized that the deterministic approach is inherently conservative
because it assumes that cracks are present in which concentrated leak erosion will initiate.
A major objective of the bulletin is to provide information to support engineers in
making analyses and judgments about the vulnerability of dams to internal erosion. The
engineering judgment approach is intended to provide the means to make a reality check on
what may be over-conservative conclusions from the deterministic approach, or to show that
the uncertainties are such that quantitative risk assessment is also required to reach
responsible conclusions as to the dam’s resistance or vulnerability to internal erosion and the
need for remediation.
The methods outlined recognize the difficulty and uncertainty in estimating crack
depths and widths in concentrated leak erosion. Such uncertainties can best be examined in a
risk based framework. There will be dams where a staged approach is warranted, with a
progression from the Deterministic or Standards Based Method, to Engineering Judgement,
and then to Quantitative Risk Assessment.
It is recommended that the engineering analyses and assessment be carried out
iteratively in stages, making use initially of all readily available information about the dam,
its geometry and the foundations, making further investigations only if more data would
make it possible to make clearer and more definite decisions on the vulnerability or resistance
of the dam to internal erosion.

9.3 ENGINEERING ANALYSIS PROCEDURE

The full procedure for an engineering analysis of the vulnerability of a dam and its
foundations to internal erosion is:
(1) Assemble all the relevant information about the dam and its foundations as
described in Section 9.8.
(2) Determine the loading conditions for the dam resulting from the reservoir and
earthquakes as described in Sections 9.4, 9.5 and 9.6 and Section 3 in relation to
concentrated leaks.
(3) Carry out a detailed Potential Failure Modes Analysis (PFMA) as described in
Section 9.9.
(4) Screen out PFM which are not applicable to the dam. Some guidance on how this
may be done is given in Section 9.10. Screening may be done on the zoning of the
dam and the foundation conditions, or a more detailed understanding of the factors
influencing the initiation and continuation of erosion.
(5) For the remaining PFM analyze the ability of the soils in the dam to resist the
maximum applied loads (at the highest water level, and in the widest crack or
opening in the case of concentrated leak erosion). Make staged assessments and
collect and use more data as necessary to reach definite conclusions on the ability of
the dam to resist erosion or not. Use engineering judgment and similarities of PFM
as far as possible to reduce the number of PFM examined in detail. Some guidance
on how this may be done is given in Section 9.11.
(6) For PFM where definite conclusions cannot be reached and in cases where
extensive remedial works are necessary to make the dam resistant to erosion, use
___________________________________________________________________________
93
 qualitative or quantitative risk analysis methods to estimate the likelihood of failure
taking account of the frequency of the applied reservoir and seismic loads and
whether the risk is acceptable and expenditure should or should not be made. Some
guidance on how this may be done is given in Section 9.12 and Chapter 12.
It can be seen that the scope of the analysis is defined by the number of realistic failure
modes and the procedure therefore concentrates first on eliminating unrealistic failure modes
and then examining realistic modes in detail, going through stages if the analyses show more
data to be needed to reach reliable conclusions.
If the engineering analysis plainly shows that internal erosion could progress and the
consequences of failure were severe, works to make the dam safe against internal erosion
would be required. On completion of the works, a monitoring regime proportionate to the
consequences of failure would be required.
If the properties of the dam and its foundation were such that the engineering analysis
plainly shows that internal erosion could not initiate or progress, no works would be
necessary, but a monitoring regime proportionate to the consequences of failure would be
required.
In cases where the properties of the dam and its foundation were such that it was not
clear from the engineering analyses, even after applying more data in several iterations, that
internal erosion could or could not initiate and progress, decisions about safety works or no
safety works and the stringency of monitoring regimes would have to be made on the basis of
the assessed likelihood of failure and the severity of the consequences of failure. In such
cases, decisions could be usefully supported by making quantitative risk assessments, as
described in Chapter 12.
Remedial works to make dams safe against internal erosion are discussed in Chapter
10. Information on appropriate monitoring regimes is given in Chapter 11.

9.4 RESERVOIR AND SEISMIC LOADING

CONDITIONS
9.4.1 The effect of reservoir level on internal erosion
The statistics about dam failures resulting from internal erosion in the embankment
(Chapter 1) show that most failures of existing dams occurred when the reservoir water level
was at or close to its highest ever level. This may be because the upper parts of dams are
rarely tested by high water levels, consequently weaknesses at high levels in the dam are
found only during rare periods of high water level.
However many incidents within the embankment have occurred at normal operating
reservoir levels and it should also be noted that for internal erosion in the foundations most
failures and accidents occur at other than maximum reservoir levels. Hence the assessment of
the safety of the dam should not be only for “extreme” reservoir levels, but over the full
range.
The upper parts of dams are often more vulnerable to internal erosion than elsewhere.
They may be zoned without filters as shown in the examples in Figures 9.3 and 9.4. The
upper parts are also more likely to be affected by cracking and hydraulic fracture due to
differential settlements within the dam and its foundation during and after construction. Also,
the upper parts of an embankment dam may dry out and desiccation cracks may provide
routes in which erosion occurs.

___________________________________________________________________________
94
 It should also be noted that during periods of high water level the entire dam may be
subjected to higher hydraulic loads than ever before. The high pore pressures may cause
hydraulic fracture to occur leading to cracks and openings through the dam fill in places, such
as beside culverts, which were previously intact. Note that hydraulic fractures may close and
re-open as pore pressures change, seasonally, for example. Note also that because of cycles of
loading (from wet and dry seasons, for example), the stress state in embankment dams is not
constant over time. Consequently cracks or low stress zones through which erosion may
initiate could form at any time in dams which have been resistant to erosion previously.

9.4.2 Reservoir levels to be considered in the engineering

assessment of internal erosion
The reservoir levels to be considered depend on which approach is being followed:
Deterministic approach: The highest reservoir water level (or hydraulic gradient)
expected during the passage of the spillway design flood.
Engineering Judgment: The highest reservoir water level expected during the passage
of the spillway design flood; and reservoir levels at which there are changes in zoning,
the level of geological features which occur in the foundation, levels at which there
have been recorded anomalous seepage or piezometer readings.
Quantitative Risk Assessment: As for the Deterministic approach, plus pool of record
(historic high reservoir level), and embankment crest level (if the spillway design flood
will overtop the dam).
Depending on the geometry of the dam zoning and the characteristics of the floods and
reservoir, the most vulnerable parts of the dam may be reached by relatively high likelihood
floods (as in Figure 9.3(b) or lower likelihood as in Figure 9.3(a)). This should be taken
account of in making the assessment of the safety of the dam.

9.4.3 Seismic loads to be considered

The seismic loads to be considered depend on which approach is being followed:
Deterministic approach: The maximum design earthquake.
Engineering Judgment approach: The maximum design earthquake, and seismic loads
which may cause minor cracking, are very likely to cause transverse cracking, or likely
to cause liquefaction in the dam foundation or within the embankment
Quantitative Risk Assessment approach: As for the Engineering Judgment approach.
For seismic loading, a coincident pool (i.e. reservoir level at the time of the earthquake)
must be considered. The selection of this pool depends on several factors. For dams where
the joint occurrence of earthquake and floods are high, the reservoir levels to be considered
may be as described above. Otherwise, high reservoir levels may be screened out based on
the low joint loading probability, and the evaluation of seismic failure modes can be based on
a “sunny day” or “normal” operating level, often defined as a 90-percent duration exceedance
pool elevation (i.e., the reservoir level that is exceeded 90 percent of a given year).

___________________________________________________________________________
95
9.5 DESIGN STANDARDS AND THE LIKELIHOOD OF
FAILURE FROM INTERNAL EROSION
It should be noted that if the spillway can safely pass the Probable Maximum Flood,
and the dam can resist the internal erosion load applied by the peak water level during the
passage of the PMF flood, it is safe to PMF standards from both floods and internal erosion.
If the dam could be shown to resist the earthquake of the same probability of
occurrence as the PMF, it could be shown to be resistant to overtopping, internal erosion and
seismic stability to PMF standards.
PMF is a deterministic criterion, similar to MCE (maximum credible earthquake). The
probability of occurrence of deterministic criteria is unknown, although expected to be very
low.
It should be recognized that it is inconsistent to design a spillway upgrade for an
existing dam to PMF standards, and not provide a similarly low likelihood of failure by
internal erosion; e.g. by operating the dam without filters or with inadequately controlled
internal erosion in the foundation.
The above observations about co-incidence of probability of failure from earthquake,
flood and internal erosion can be more plainly made by using probabilistic, not deterministic,
criteria. Thus, if the spillway is designed to pass the 1 in 10,000-year flood safely, and found
or improved to resist internal erosion during the peak water level or highest hydraulic
gradient as the 1 in 10,000-year flood passes, and found or improved to be stable during the 1
in 10,000-year earthquake, the probability of failure from any of the three main threats to
embankment dams will be at most 1 in 10,000.

9.6 SAFETY MARGINS AND APPROXIMATIONS

Engineers should recognize that estimates of the flood and earthquake characteristics
are approximate, as are estimates of the erosion characteristics of the materials in the dam
subject to seepage flows. They should exercise judgment when drawing conclusions based on
the application of this Bulletin.
Estimates of crack widths could be critical in establishing whether or not the dam is
safe or not from erosion. Engineers may judge that the resistance of the dam to erosion
should also be examined for a range of crack widths and for varied soil resistance parameters.
The state of the art in estimating crack depths and widths is limited and there is
considerable uncertainty in the estimates. Fell et al (2008) give guidance but the uncertainty
remains. This is a reason why dam engineers may chose to use deterministic, engineering
judgment and qualitative or semi quantitative risk assessment approaches as detailed in
Section 9.2 rather than methods based on the physics of erosion and / or quantitative risk
assessment methods.
Many internal erosion failures initiate through deterioration or shortcomings of
conduits and spillways through dams. For example, if holes form in conduits, seepage
velocities towards the hole increase and may be sufficient to erode the fill and erosion may
continue and progress in the ‘pipe’ provided by the conduit. Analysis to check if deterioration
of the conduits or spillways could lead to erosion can be made and in some circumstances
may show that rapid erosion to failure could not go undetected. Generally however it is
necessary to take precautions to avoid the ill-effects of deterioration or design shortcomings.
In many circumstances, knowledge of the soil parameters, soil behavior and the applied
loads makes it possible for an analysis of the mechanics of internal erosion to show that
___________________________________________________________________________
96
erosion could not occur. In others, the approximations and unknowns in the analysis of the
mechanics of internal erosion make it uncertain whether or not internal erosion could occur.
In such cases, decisions on whether to err on the side of safety and install safety works or to
trust to a monitoring regime to give adequately early warning of the onset of internal erosion
must be a matter of engineering judgment. The judgment will have been supported by the
analysis of the mechanics of internal erosion described in this bulletin. It could also be
supported by quantitative risk assessment, using methods described in Chapter 12.

9.7 SOIL PROPERTIES IN RELATION TO INTERNAL

EROSION
9.7.1 Material susceptibility, stress conditions and hydraulic load
As stated in the Terminology, internal erosion occurs when soil particles within an
embankment dam or its foundation, are carried downstream by seepage flow.

MATERIAL
SUSCEPTIBILITY
Internal instability
Filter incompatibility
Void space
Free surface
Low plasticity
Soil distress
Suffusion Cracking
Suffusion Bridging
Piping
Contact erosion
Boiling
Seepage velocity Low stress
Hydraulic gradient Vibration
Pore pressure Hydraulic Arching
fracture

CRITICAL Heave CRITICAL

HYDRAULIC STRESS
LOAD CONDITION

Figure 9.1 Venn diagram illustrating interaction of geometric, hydraulic and

mechanical susceptibilities of soils to initiation of internal erosion (Garner and Fannin, 2010)

Garner and Fannin (2010) use the Venn diagram in Figure 9.1 to illustrate that erosion
initiates when an unfavorable coincidence of material susceptibility, stress conditions and
hydraulic load occurs. The three conditions are:
Material Susceptibility: the potential for soil to experience loss of a portion of its finer
fraction, as a consequence primarily of grain size and also shape of the grain size distribution
curve;

___________________________________________________________________________
97
 Critical Stress Condition: the inability to resist internal erosion due to the magnitude of
effective stress, with recognition that stress varies spatially and/or temporally within the body
of the dam; and,
Critical Hydraulic Load: the hydraulic energy required to invoke a mechanism of
internal erosion, by means of seepage flow through the dam.
Some forms of erosion initiate when only two factors coincide, and the central subset
describes a zone within the dam that is susceptible to all three factors. The combination of
Material Susceptibility, Hydraulic Loading and Critical Stress gives rise to the release or
detachment, and transport of soil grains.

Hydraulic Fracture
An important issue in internal erosion is the stress conditions resulting from cross
valley differential settlement, differential settlement due to compressible soils in the
foundation, or arching in narrow cores which results in hydraulic fracture. This occurs when
the pore pressure (u) exceeds the minimum principal stress (σ3) (strictly speaking the pore
pressure should exceed the minimum principal stress plus the tensile strength of the soil).
When the orientation of the minimum principal stress is unfavorable, the resulting fractures
(cracks) provide sites for concentrated leak erosion. The fractures will remain open, in all soil
types, until the pore pressure drops, as may occur as a result of drainage through the
fractures, or the minimum stress increases, as may occur through settlement or creep.
Fractures which close may re-open when the stress state or pore pressure changes. Such
changes may occur when reservoir level rises, or seasonally, through rainfall on the
downstream slopes in wet weather for example.

9.7.2 The physical process

The first condition for internal erosion to occur is particle detachment. Water seeping
through the dam or flowing in cracks or concentrated leaks must be doing so with sufficient
velocity to provide sufficient energy to detach particles from the soil structure. The nature of
the soil in the dam determines its vulnerability to erosion. Three classes have to be
distinguished:

Three classes of soil in relation to internal erosion

 Non plastic soils such as silts, sands, silty sands, and silt, sand, gravel soils. These
collapse when saturated under flooding, will generally not sustain a crack when
saturated, and are relatively easily eroded. As cohesionless soils become coarser
through silts, sands, gravels and cobbles they progressively require more energy to
initiate erosion. Erosion resistance is related to particle weight, and in some cases
the stress state, and involves detachment of individual particles. These soils are
subject to backward erosion, contact erosion or suffusion depending on their
particle size distribution.
 Plastic soils, such as clays, clayey sands, and clayey sandy gravels are generally
more resistant to erosion than cohesionless soils. These soils are subject to
concentrated leak erosion and contact erosion. Clay soils will hold a crack even
when saturated. Non plastic soils with a high silt content will sustain a crack when
partially saturated and may do so when saturated. Higher energy is generally
required to detach particles from cracks or concentrated leaks within a cohesive fill
but the particles thus removed are small and easily carried through the crack.
Erosion resistance is related to contact forces between the water flowing in the

___________________________________________________________________________
98
 crack or concentrated leak, and the critical shear stress of the soil on the sides of
the crack.
Backward erosion and suffusion cannot occur in these soils under the gradients
normally experienced in dams and their foundations but may occur if local
gradients are very high.
 Dispersive plastic soils are soils in which, because of their clay mineralogy and the
water chemistry, erosion will initiate in cracks or concentrated leaks under very
low hydraulic stresses and gradients.

Degree of saturation
 Wan and Fell (2002, 2004a, b), and Lim (2006), Lim and Khalili (2010) found that
most clay soils tested have significantly higher erosion rate indices (slower
erosion) and higher critical shear stresses when saturated than at the partially
saturated compaction condition. There was however less dependence on the degree
of saturation for silty soils. This is an important finding because it means that once
the core of a dam constructed of clay soil is saturated and consolidated, it may have
a slower rate of erosion, and a higher critical shear stress.
 Just as important is that this does not apply to silty sand cores such as decomposed
and residual granites.
 From a practical point of view, it is therefore better to compact cohesive soils to
the normal requirements for dam cores; e.g. to 98% standard maximum dry
density, on the wet side of optimum, because the erosion resistance increases with
the degree of saturation of the soil. However if filters are provided, the difference
in behavior of soils compacted wet or dry of optimum is minor.
 It should be noted that clays in flood embankments and near the crest of dams can
become desiccated. Vertical shrinkage cracks form hexagonal blocks cut at
frequent intervals by horizontal cracks, and the resulting ‘blocks’ are readily
eroded. More information is given in Section 3.2.9.

Dispersion properties of the soil

 Soils which show dispersive behavior; soils classifying as Emerson Crumb Class 1
or 2, and Pinhole Dispersion D1 and D2, will have a very low critical shear stress if
the eroding fluid is sufficiently free of salts which might otherwise suppress
dispersion (deflocculation). That is if the eroding water has low salts content. It
should be noted that under flood conditions the salts content of the water in the
reservoir is likely to drop, so tests done in reservoir water may be un-conservative.
If in doubt with dispersive soils it is best to assume the reservoir water will not
inhibit dispersion and rely on the results of tests using distilled water. The effect of
reservoir water, distilled water and dispersants (e.g. sodium hexametaphosphate)
on grain size distribution is most obviously illustrated by ‘triple dispersion’ particle
size distribution tests as in Figure 3.6.
 Lim (2006) and Lim and Khalili (2010) showed that for rotating cylinder tests the
Erosion Rate Index is not greatly affected by whether the soil is dispersive after the
initially rapid part of the erosion process. So the major effect of dispersion is on the
critical shear stress at which erosion initiates, not the rate of erosion.

___________________________________________________________________________
99
Effect of soil structure
 Benahmed and Bonelli (2012), Lim (2006) and Wahl et al (2008) have all noted
that soil structure has an important effect on the erosion properties. They find that
erosion rates are significantly higher for the same soil if the soil is compacted dry
of optimum moisture content and the soil forms aggregated particle, and / or micro-
cracks. These allow erosion of blocks of the soil rather than of individual particles.
This is one of the reasons why higher erosion rates are measured in JET than HET
(see Volume 2), as the HET test is stopped with a relatively small hole diameter
not allowing the “blocks” of soil to dislodge from the sides of the hole. This
behavior was also noted in rotating cylinder tests by Lim (2006).
 Note also the effects of desiccation on soil structure, described in the paragraph
above on degree of saturation and in Section 3.2.9.

Effect of testing method on erosion rate index

 The effect of test method is discussed in Bulletin 2.

Effect of shear strength of the soil

 Contrary to what many have assumed, the erosion rate and critical shear stress are
unrelated to the effective stress or undrained shear strength of the soil. Soils with
similar undrained strengths can have an erosion rate 100 or even 1000 times
different; and critical shear stresses ranging from 1Pa to 100Pa depending on
dispersion properties, clay mineralogy, and degree of saturation.

9.8 THE IMPORTANCE OF HAVING RELIABLE

INFORMATION UPON WHICH TO MAKE THE
ASSESSMENT OF INTERNAL EROSION
It is essential that the assessment of internal erosion of a dam be based on the best
available information. It is best carried out in association with assessing other failure modes
such as slope instability as there may be interactions between failure modes.
The data to be gathered includes:

Geometric model
The geometric model compiles the details of the internal geometry of the dam. Is it
zoned, and what are the geometric details of the zones and the foundation contact of the fill
within the various zones?

Geological model of the foundation

The geological model identifies the lithology and stratigraphy of the soil and rock,
structural features (tectonic, position and orientation of discontinuities: joints, faults, etc) and
geological background (earthquake, geomorphology). The susceptibility to erosion of the
soils overlying rock; and in some cases the erosion characteristics of weathered or soft rock
and width of joint openings, and the presence of features such as karst openings, are the key
factors.
The hydrogeology of the foundation should also be understood.

___________________________________________________________________________
100
Geotechnical model of the embankment and foundations
The geotechnical model is the compilation of the mechanical properties required for
internal erosion assessment and stability analysis, including soil classification and particle
size, permeability, bulk density, and preferably data on the erosion properties of soils based
on laboratory tests. These are related to the geometric model of the embankment and the
geological model of the foundation. The data is obtained from site investigations prior to
construction, construction control testing, construction photographs, and site investigations of
the existing dam.

Hydraulic or seepage model

It is fundamental to understand the hydrodynamic situation in the dam and its
foundations. The routes of seepage paths through and beneath the dam can be determined
from piezometric readings. Seepage and hydraulic gradients through the dam and its
foundations in normal operating and high reservoir levels should be considered.

Stress state in the dam and its foundation

This may be required for detailed assessments of suffusion and global backward
erosion, and for aiding in the prediction of the likelihood and depth of cracking and hydraulic
fracture.

Staged approach and collection of data

In many dams not all this information is available so the assessment has to be made
using what is available. If critical information, such as the particle size distributions of the
core and the filter or transition, is missing it may be necessary to carry out investigations into
the dam or its foundations to obtain it. If initial analysis does not give a definite result, further
investigations to collect more data or carry out more tests may be required as part of the
staged approach to assessing whether or not the dam can or cannot resist internal erosion.

9.9 FAILURE MODES ANALYSIS

A potential failure mode is a sequence of events starting from an initiating mechanism,
such as a defect, flaw or seepage path in the dam or its foundation, which may lead to an
uncontrolled release of the reservoir. After compiling the best available information, and
considering the geotechnical model of the dam or dike and foundation, the dam safety
assessment team should go through a discussion of all potential failure modes and develop a
thorough understanding of the sequence of events and the potential location of seepage and
erosion paths through the embankment and foundation. The sequence of events and seepage
pathways should be documented by annotating cross sections and longitudinal sections of the
embankment and its foundation to help visualize the failure path. The development of the
failure modes should consider the following:
 Potential initiating mechanisms for each of the failure locations;

 Zoning of the embankment, including the configuration and properties of internal filter and
drainage measures;

 Foundation geology and stratigraphy; and properties and

 Filtered and unfiltered exit points of seepage.

Failures and incidents by internal erosion of embankment dams and their foundations
are categorized into four general failure paths, which are;
___________________________________________________________________________
101
  Internal erosion through the embankment,

 Internal erosion alongside through-penetrating structures, such as conduits associated with

outlet works, spillway walls or other adjoining a concrete gravity structure supporting the
embankment.

 Internal erosion through the foundation, and

 Internal erosion of the embankment into or at the foundation. Including (a) seepage
through the embankment eroding material into the foundation, or (b) seepage in the
foundation at the embankment contact eroding the embankment material.

It can be seen that with four general failure paths, four types of initiation of internal
erosion, and several zones and layers of differing materials in the dam and foundation, the
number of potential failure paths at a single dam soon becomes large. Useful first steps
therefore are to consider the soil types in the dam and their vulnerability to the four types of
erosion and the zoning in the dam to identify where the types of erosion the dam to which the
dam seems most vulnerable could occur. This often soon identifies the most likely locations
and the most likely initiating mechanisms of erosion, leading promptly into Steps 2 & 3 of
the framework for analysis of the mechanics of internal erosion, Table 9.1.
The possibility of overlooking potentially important failure modes is reduced by
considering the particular details of the dam and its appurtenant structures, such as details of
walls retaining the embankment, conduits through the embankment, by assembling
construction photographs and reports and inspecting the dam as part of the failure modes
assessment. It is also reduced by having the failure modes assessment done by a team which
includes the engineer and geologist most familiar with the dam, dam operating and
surveillance staff, and facilitated by a person experienced in failure modes analysis.
Figure 9.2 gives examples of possible locations of initiation of internal erosion.
For many dams, it is more likely that features likely to lead to initiation of internal
erosion are in the upper part of the dam. This is because cracking due to differential
settlement over large scale irregularities in the foundation profile is more likely to be present
near the crest. Continuation is also more likely because often the detailing of the dam design,
or as built, will give no or little filter protection. Figure 9.3 gives examples of this with
increased likelihood of failure by internal erosion under flood loading near the crest of the
dam because the earthfill core and filters were not taken to the crest of the dam.
Figure 9.4 is an example where there is a significantly different probability of internal
erosion above the berm than below because the upper part is essentially a homogeneous dam
while the sandy gravel in zone 2 may act as a filter. This is best managed in the analysis by
considering internal erosion above the berm (the upper part of the embankment) separately to
below the berm (the middle and lower parts of the embankment).

___________________________________________________________________________
102
 1. Spillway wall interface
2. Adjacent to conduit
3. Crack associated with steep abutment profile
4. Desiccation on top of core
5. Embankment to foundation
6. Foundation (if the foundation is soil or erodible rock)
7. Embankment through poorly compacted layer, crack, (or by backward erosion if the
core is cohesionless)
Figure 9.2 Example of possible locations of initiation of internal erosion (Fell and Fry, 2007)

___________________________________________________________________________
103
 Figure 9.3 - Examples of embankment crest details which may result in relatively high
likelihood of internal erosion (Fell et al, 2008)

Flow path
A

Flow path
B

Figure 4.3 - Example of an embankment with significantly different likelihood of

Figure 9.4 - Example of an embankment with significantly different likelihood of

internal erosion above and below the top of the downstream berm. (Fell et al, 2008)

___________________________________________________________________________
104
9.10 SCREENING OF POTENTIAL FAILURE MODES
9.10.1 Screening of PFM on the zoning of the dam and the
properties of the core of the embankment.
(a) Zoning of the embankment
The zoning of an embankment and in particular the presence or absence of filters which
satisfy modern design criteria, and are constructed to a high standard, has a significant effect
on the likelihood of failure by internal erosion. Figure 2.9 and Table 2.2 in Section 2.4,
summarized in Table 9.3 below, show typical embankment zoning and their relative
vulnerability related to the degree of filtering provided. These are generalizations but it can
be said that any dam in the first group A is very vulnerable because there is no filtering.
Internal erosion in Groups B dams may be limited by the hydraulic head losses through the
upstream and downstream fill and by the filtering capability of the fills. Group C dams are of
low vulnerability because of the concrete upstream slab or core wall, subject to good
detailing and no damage. Those in the fourth group D have in principle a very low
vulnerability because filters are provided, but the detail of zoning is critical. However in
reality in dams with formal and informal filters, the details of the filters are critical if internal
erosion is to be controlled, and dams should not be regarded as safe from internal erosion just
because their general zoning is favorable.

Table 9.3 Screening by dam zoning

Vulnerability to Potential for control of internal Dam zoning and category
internal erosion erosion number
A Very vulnerable Little or no control Homogeneous (unzoned) earth
fill (Category 0)
Earthfill with rock toe
(Category 2)
B Vulnerable Some control of internal erosion Zoned earthfill (3)
depending on detail of zoning and Zoned earth and rock fill (4)
filter capability Puddle core (8)
Hydraulic fill (11)
C Low vulnerability Moderate control of internal Concrete face earthfill (6)
erosion depending on the filter Concrete face rock fill (7)
capability and details of core wall Concrete core earth fill (9)
or face slab Concrete core rock fill (10)
D Very low Good control of internal erosion Earthfill with filters (1)
vulnerability subject to good details of zoning Central earth core and rock fill
and filter design (5)

(b) Screening of PFM on the properties of the core of the embankment

Table 9.4 shows PFM which can be screened out for internal erosion in the
embankment. If these criteria are met the only PFM which have to be considered are
concentrated leak and contact erosion PFM.

___________________________________________________________________________
105
 Table 9.4 Screening of PFMA for internal erosion in the core of the embankment
Exclude the Failure Mode if the Following Conditions are
Initiating Mechanism
Satisfied
Backward erosion in the core Exclude if the soil has a Plasticity Index ≥ 7.
Suffusion in the core (1) Exclude if the soil has a Plasticity Index ≥ 7.
OR
(2) Exclude if the soil is not gap-graded. If the soil is gap-
graded or broadly graded, exclude if the proportion of the
finer fraction of a non-plastic soil is more than 40% of the
total mass of the soil.

9.10.2 Screening of PFM on foundation geology and

properties
Tables 9.5 and 9.6 show how PFM for internal erosion in the foundation, and from
embankment to foundation can be screened based on the geology of the foundation and the
cut-off provided for the core of the embankment.

Table 9.5 Screening of PFM for internal erosion in the foundation (adapted from Fell et
al, 2008)
Exclude the Failure Mode if the Following Conditions are
Initiating Mechanism
Satisfied
All modes of internal erosion Exclude if the soil layer beneath the dam is isolated by a
of the foundation (backward cut-off trench founded in non-erodible rock.
erosion, suffusion, erosion in a
crack)
Backward erosion in a soil in Exclude if:
the foundation (1) The foundation soil has a Plasticity Index ≥ 7.
OR
(2) If the soil layer with PI  7 layer is not continuous
below the embankment (i.e. it terminates beneath the dam)
Suffusion in a soil in the Exclude if:
foundation (1) The foundation soil has a Plasticity Index ≥ 7.
OR
(2) The proportion of the finer fraction of a non-plastic soil
is more than 40% of the total mass of the soil.
OR
(3) If the soil layer with a PI  7 is not continuous below
the embankment (i.e. it terminates beneath the dam)
Erosion in a crack in soil in the Exclude if the foundation soil is non plastic.
foundation

___________________________________________________________________________
106
 Table 9.6 Screening of PFM for internal erosion from the embankment to the
foundation and contact erosion (adapted from Fell et al, 2008)
Exclude the Failure Path if the Following Conditions are
Initiating Mechanism
Satisfied
Internal erosion of the Exclude if:
embankment into or at a rock (1) Rock foundation below the core is comprised of rock
foundation containing closed rock defects (<1 mm wide) or defects
open less than 3D 95 ,of the fine limit of the core
OR
(2) Rock foundation below the core has been adequately
treated (e.g. shotcrete, slush grouting, mortar or concrete
treatment)
Internal erosion of the Exclude if:
embankment into or at a soil (1) Soil foundation below the core is comprised of fine
foundation grained soils with greater than 12% fines (fraction finer than
No 200 sieve (0.075mm)), and the soil does not contain
macrostructure such as root holes, relic joints or solution
features.
OR
(2) Soil foundation below the core is comprised of sands
(SP or SW) which are filter-compatible with the
embankment materials (i.e. satisfy the No Erosion criteria).

9.10.3 Screening of PFM on details of the embankment and

conduits and retaining walls.
The likelihood of an embankment experiencing initiation of concentrated leak erosion
will be low provided that the conditions described in Table 9.7 are ALL satisfied, AND the
soil in the embankment is in Erosion Group 3 or 4, moderately erodible or erosion resistant,
as determined from Table 9.8.
The likelihood of an embankment experiencing initiation of concentrated leak erosion
at conduits and retaining walls will be low provided that the conditions described in Table 9.9
are ALL satisfied, AND the soil in the embankment is in Erosion Group 3 or 4, moderately
erodible or erosion resistant, as determined from Table 9.8.
These are based on quantitative assessments in Fell et al (2008).
This is not to say these PFM cannot occur and for high consequence of failure dams
they should not be excluded without more detailed analysis.

___________________________________________________________________________
107
 Table 9.7 Screening of PFM for conditions in which a crack or concentrated leak in the
embankment is unlikely (adapted from Fell et al, 2008)
Exclude the Failure Path if the Following Conditions are
Initiating Mechanism
Satisfied
Transverse cracking due to Exclude if the embankment abutments are flatter than 15°,
cross valley differential and the embankment height is uniform across the valley.
settlement
Transverse cracking due to Exclude if the width of valley to dam height ratio W v /H > 2.
cross valley arching
Transverse cracking due to Exclude if there is no compressible soil in the foundation
differential settlements in the below the core
foundation beneath the core
Cracking in the crest due to Exclude if the reservoir level is below the likely depth of
desiccation by drying desiccation cracking under all conditions including during
extreme floods.

Cracking due to earthquake Exclude if the MDE is below a peak ground acceleration of
0.2g, AND
The embankment abutments are flatter than 15 degrees, and
the height is uniform across the valley; AND
The soils in the embankment and its foundation are not
susceptible to liquefaction.
Transverse cracking at the Exclude if the persistence of the irregularity across the width
foundation contact due to of the core is less than 50% of the core base width
small scale irregularities in
the foundation profile under
the core
Poorly compacted or high Exclude if:
permeability layer in the All soils are very well compacted with lift thicknesses less
embankment than 200mm, with good documentation and records; This
means:
(1) For plastic soils (Plasticity Index > 7), ≥98% standard dry
density ratio, moisture content 2% dry to 1% wet of OWC;
(2) For non plastic soils and soils with PI  7, >75% relative
density.

___________________________________________________________________________
108
 Exclude the Failure Path if the Following Conditions are
Initiating Mechanism
Satisfied
Poorly compacted or high Exclude if:
permeability layer on the (1) Contact soils are well compacted on a regular foundation
core-foundation contact surface with good documentation and records
OR
(2) Uniform or regular rock surface or surface treated with
shotcrete or concrete to correct slope irregularities, and soils
well compacted (contact soil compacted using special
compaction methods (e.g. rubber tires, use more plastic
material, compaction wet of OWC).
OR
(3) Uniform well compacted soil foundation, with good
mixing, bonding and compaction of contact fill.
OR
(4) Compacted soil foundation
Poorly compacted or high Exclude if;
permeability layers in the The climate is such that temperatures do not fall below
crest due to freezing freezing point except possibly overnight or for a day or two.
OR
If the reservoir stage being considered is below the likely
depth of freezing

Table 9.8 Erosion resistance of soils related to classification and dispersivity

Erosion Soil Group Soil Classification
1.Extremely erodible All dispersive soils; Sherard pinhole classes D1 and D2; or
Emerson Crumb Class 1 and 2.
AND
SM with <30% fines
2.Highly erodible SM with > 30% fines; SC with < 30% fines; ML; SC with
>30% fines; CL-ML;
3.Moderately Erodible CL; CL-CH; MH; CH with Liquid Limit <65%
4.Erosion resistant CH with Liquid Limit > 65%

___________________________________________________________________________
109
 Table 9.9 Screening of PFM for internal erosion around and into conduits through the
embankment or adjacent a wall supporting the embankment core for which internal erosion is
unlikely
Poorly compacted or high Exclude if:
permeability zone around a (1) There is no conduit passing through the embankment,
conduit through the OR
embankment (2) The conduit is totally embedded in a trench excavated in
non-erodible rock, backfilled to the surface with concrete
Erosion into a (non- Exclude if:
pressurized) conduit (1) There is no conduit passing through the embankment,
OR
(2) Careful internal inspection of conduit showing no
evidence of open joints or cracks.
Poorly compacted zone Exclude if there is no spillway or abutment wall in contact
associated with a spillway or with the embankment
abutment wall
Crack/gap adjacent to a Exclude if there is no spillway or abutment wall in contact
spillway or abutment wall with the embankment

9.11 ANALYSIS OF POTENTIAL FAILURE MODES

AT WHICH INTERNAL EROSION COULD
INITIATE, CONTINUE AND PROGRESS
9.11.1 Some general principles
Many PFMs will have been excluded by use of the criteria in the tables above. Some
iterations may have been necessary to collect the data needed to make full use of the tables. It
is now necessary to examine the remaining PFMs and analyze if the erosive forces are
sufficient to overcome the resistance of the soils in the dam and foundation and initiate
erosion, and whether the erosion could continue because no filtering capability is present, and
progress because cracks and openings could be sustained. If PFMs are identified through
which the dam’s properties are insufficient to resist erosion, it will be necessary to consider
remedial works to prevent it. If no PFMs along which erosion occurs are identified,
monitoring regimes will be required, focused on identification of the onset of erosion,
particularly along what seem to be the most vulnerable PFMs.
There are some general principles which should be applied when assessing the potential
for erosion leading to failure to occur along PFMs. These are:
(1) Consider the eight steps listed in Table 9.1 for the engineering analysis to establish
whether the dam’s properties are sufficient to resist erosion or whether remedial
works to protect the dam against erosion are needed.
(2) Take account of the fact that some PFM may be through weaknesses at high levels in
the dam, not yet tested by high water levels; e.g. concentrated leak erosion above the
filters in Figure 9.3 (a) and (b). Other PFM at low levels in the dam will be most
severely loaded when the water level in the reservoir is highest during the spillway
design flood.
___________________________________________________________________________
110
 (3) For concentrated leak erosion in the embankment or in a plastic soil foundation
consider whether a crack or low stress zone subject to hydraulic fracture may occur
for the PFM, whether the crack will extend below the reservoir and the erodibility of
the soil (Sections 9.9, 9.10.3 and 3.3 and 3.4, and Chapter 3 generally)
(4) For backward erosion piping in the foundation assess whether the reservoir level
during the spillway design flood will cause heave of any layer overlying the erodible
strata or if there is an open ditch or drain into which erosion can occur without heave.
Assess if the hydraulic gradient at spillway design flood water level and other
reservoir levels under consideration is at or above the critical gradient at which
erosion will initiate and progress for the erodible soil layer. (Section 4.2, Chapter 4).
(5) For global backward erosion in the embankment estimate the likely critical gradient
by comparison with published data, and compare this to the gradient in the
embankment at spillway design flood water level and other reservoir levels under
consideration. Carry out laboratory tests if the design gradient is approaching the
critical gradient (Sections 4.5 and 4.6).
(6) For contact erosion at the embankment-foundation interface, estimate the Darcy
velocity at spillway design water level and other reservoir levels under consideration
and compare this to the velocity at which contact erosion initiates (Section 5.3,
Chapter 5).
(7) For suffusion in the embankment core or filters, or in the foundation, assess the
likelihood that the soil is internally unstable and subject to suffusion. If it is, assess
the critical gradient at which erosion will initiate based on published data or
laboratory tests. Then assess the annual probability that these critical gradients will be
reached related to reservoir level. Assess the likely gradation of the soil after
suffusion and the effect on permeability and seepage flows (Section 6.3, Chapter 6).
(8) Assess whether filters, transition zones or downstream zones will prevent erosion
continuing. For filters and transitions which do not satisfy modern design criteria,
assess the likelihood that the filters will eventually arrest erosion using the no-erosion,
some erosion, and excessive erosion concepts (Chapter 7).
(9) Assess whether erosion will progress or whether upstream flow limitation will arrest
the erosion (Chapter 8).
(10) Assess the time for development of erosion from when it can be detected to
when a breach may develop. Then assess the likelihood of detection and successful
intervention taking account of the mode of internal erosion, the zoning of the
embankment, and the management factors which influence whether detection and
intervention will be possible (Sections 2.6 and 11.2.3).
(11) Assess whether failure may be averted by the erosion process self limiting,
e.g. by the reservoir draining, even if detection and intervention is unsuccessful
(Section 2.8).

It should be noted that whether or not internal erosion leads to failure is mostly
controlled by the resistance or vulnerability of the dam to initiation of erosion, (2) to (7), and
by the effectiveness of filters and transitions to arrest erosion (8).

___________________________________________________________________________
111
9.11.2 Concentrated leak erosion-some aids to judgment
Erosion in concentrated leaks occurs when the widths of cracks or the size of openings
is large enough to allow seepage velocities to generate hydraulic shear stresses in excess of
the hydraulic shear strength (see Sections 3.3 and 3.4). Estimating crack width and opening
size and the hydraulic gradients in them is a vital part of the process of assessing the
vulnerability to erosion, as discussed in Section 3.3 and throughout this Section 9,
particularly 9.10.3 and Table 9.7.

9.12 DECISION MAKING: ARE ANTI-EROSION

MEASURES NECESSARY OR WILL
MONITORING BE SUFFICIENT?
If PFMs are identified through which the dam’s properties are insufficient to resist
erosion, it will be necessary to consider remedial works to prevent it. If no PFMs along which
erosion occurs are identified, monitoring regimes will be required focused on identification of
the onset of erosion, particularly along what seem to be the most vulnerable PFMs.
Before making decisions, the types, effectiveness and scope of anti-erosion measures
and monitoring that could be installed should be considered using the comments in Chapters
10 and 11 as a basis.

___________________________________________________________________________
112
 10. REMEDIATION AND
IMPROVEMENTS TO DAMS TO
RESIST INTERNAL EROSION
10.1 METHODS AVAILABLE
This chapter is to give a broad overview of the methods available to remediate or
improve dams to enable them to resist internal erosion. It does not deal with ‘intervention’ in
an emergency, which is dealt with briefly in Section 2.7 above.
More information on the repairs and improvements are included in Volume 2 of the
bulletin. The information here is intended to provide sufficient information to give a
preliminary understanding of the alternatives and scope of works likely to be necessary to
make a dam able to resist erosion, and to assist in making the decision whether or not to ‘fix’
a dam which the analyses carried out above have shown to be only marginally able to resist
erosion when subject to critical loads in its present condition.
There are two basic approaches to making dams capable of resisting internal erosion:
 Barriers, to prevent seepage from passing through the dam or foundation
 Filters to trap (filter) any eroded particles from seeping water while allowing the
water to drain away downstream
Engemoen (2012b, 2012c) describes the criteria used by USBR to decide if remediation
is necessary and, if so, the types of remediation used. These include ‘seepage control’ (filters
and filtered drains) and ‘seepage reduction’ (barriers) measures. Their advantages and
disadvantages are listed. Other things being equal, Engemoen preferred seepage reduction
approaches, principally because they involve standard earthworks construction and uncover
the embankment and foundation, whereas seepage reduction measures, such as cut-off
trenches, typically work ‘blind’.

10.2 BARRIERS
Barriers can be provided in many different forms including sheet piles, diaphragm
walls, secant piles, and grouting. Some forms can be installed to substantial depths; there are
examples of cut-off walls greater than 100 m deep. It is vital that the barrier is complete. Any
openings or gaps in barriers may concentrate flow, increase seepage velocities and initiate
erosion. In extreme circumstances, locally strong seepage flows can erode concrete or grout
from the barriers and enlarge gaps in them. Providing barriers only locally along part of a
dam drives seepage towards the ends of the barrier, where the extra seepage may initiate
erosion. Ideally the barrier should make a complete cut-off into impermeable strata at its
base. Providing barriers to only part depth in permeable fill or foundation concentrates
seepage below the barrier and may initiate erosion. Barriers must make a seal into
impermeable strata at the abutments; in some circumstances considerable lengths of abutment
cut-offs are needed to prevent seepage and loss of water through permeable strata above
impermeable foundation strata. Rice (2007) reports on the long-term performance of seepage
barriers.

___________________________________________________________________________
113
10.3 FILTERS
10.3.1 Filter blankets and filtered berms
Filters, granular materials with gradings designed by methods listed in Section 7.5 to
filter the smallest particles expected to be present in the dam fill or foundations, can also be
used to protect dams against internal erosion. FEMA (2011) provides much information on
design and construction of filters.
Filtered berms, filter blankets on the downstream face loaded by new fill in a berm
downstream, are commonly used to retrofit filters to dams without filters or with inadequate
filters. Filter blankets should also completely cover the entire seepage surface. This is
because eroded particles could escape in water seeping out through any areas not protected by
the filter and allow continuation and progression of erosion.
It is usually necessary to extend filter blankets at the downstream toe of embankments
to intercept and filter seepage water from the foundation. Sometimes it is necessary to
provide filters in trenches below the toe filter blankets to intercept and filter seepage which
would otherwise pass below the filter. Filtered relief wells are also used to intercept and filter
seepage flowing below the surface. It is also usually necessary to extend downstream filter
blankets, including toe blankets, along the valley sides downstream of the dam to intercept
and filter the surfaces where seepage flowing through the abutments will emerge.

10.3.2 Filter collars on conduits and spillway channels

Many internal erosion failures and incidents occur alongside conduits and spillways
passing through dams. Filter collars should be installed around the downstream end of
conduits, culverts and pipes and spillway channels through dams to prevent the continuation
of erosion through fill or foundation materials around them. Provision of filter collars at the
upstream end and if possible at intervals along the conduits, provides crack filler to reduce
seepage velocities and thereby inhibit the initiation and continuation of erosion.

10.3. 3 Filter trenches

Filters can be installed in trenches to deal with shallow seepage routes near the surface
of embankments and at the toe, as mentioned above. Bio-degradable slurry is available (e.g.
Jairaj and Wesley, 1995) to support and form deep filter trenches, if necessary. The slurry is
degraded by biological action, leaving the sand filter in-situ. If installed in the crest, they
would cross transverse cracks through which concentrated leak erosion may occur.

10.3.4 Filtered relief wells

Filtered relief wells can be used to relieve high pore pressures at depth or under
impermeable layers and to draw seepage waters upward into foundation filters. Filtered relief
wells are simailr to water wells, with the filter pack graded as a filter to the soils through
which the well passes, with screens designed to retain the filter.

10.3.5 Multiple filters to provide drainage capacity

As mentioned above, it is usually necessary to provide both a filter layer and a drainage
layer in a filter blanket. This is because the permeability of filters capable of filtering the
small particles that may be present in the dam, and therefore eroded from it, is low. The
drainage capacity of such filters is also low and not likely to be able to drain away the
quantities of seepage that may arise if a leak occurs. In extreme circumstances a third ‘filter’
coarser than the drainage filter, perhaps in the form of ‘finger drains’, may be needed to
___________________________________________________________________________
114
provide additional drainage capacity. The filter and drains must all comply with filter rules so
that there is no erosion from one to another. Filter rules must also be applied between filters
and adjoining fill or foundation materials to prevent erosion at the interfaces. This
requirement sometimes makes it necessary to provide a filter ‘sandwich’ with say fine and
coarse filter layers on both sides of a very coarse drain filter.

10.4 CONDITION OF CONDUITS

Many internal erosion failures initiate through deterioration or shortcomings of
conduits and spillways through dams. For example, if holes form in conduits, seepage
velocities towards the hole increase and may be sufficient to erode the fill and erosion may
continue and progress in the ‘pipe’ provided by the conduit. New linings to conduits, and
filter collars in the surround to the spillways and conduits are precautions to prevent erosion.

___________________________________________________________________________
115
 11. SURVEILLANCE AND
MONITORING
11.1 OBJECTIVES AND CHALLENGES OF
SURVEILLANCE AND MONITORING FOR
TIMELY DETECTION OF INTERNAL EROSION
This chapter gives an overview of the needs and methods for surveillance and
monitoring of the dam after its vulnerability to internal erosion has been reviewed and it has
been repaired or improved if the review has shown this to be necessary. More information on
details of monitoring methods and equipment are given in Volume 2.
Following the review and improvements, the dam should have a low likelihood of
failure by internal erosion, but it is not possible to be absolutely certain that it is so, and its
condition may change over time. Therefore the dam must be kept under surveillance
(particularly visual inspection) and monitored (using instrumentation) to detect potential
vulnerabilities. The intensity and frequency of surveillance and monitoring should be
designed to detect the onset of failure by internal erosion early enough in the failure process
to give sufficient time to warn and evacuate people from the dambreak floodway downstream
of the dam or to empty or lower water level in the reservoir before breach occurs.
Before designing a monitoring and surveillance system to detect the onset of internal
erosion early in the failure process, the following points about internal erosion failures and
their detection should be noted:
(a) For internal erosion in embankment failures, initiation is most common in cracks
or hydraulic fractures in the embankment or around a conduit. These mechanisms
could be expected to initiate very rapidly or rapidly once the reservoir level reaches
the critical level at which erosion begins in cracks or the critical level at which
hydraulic fracture initiates. However this depends on the rate of erosion of the soil.
(b) Initiation of erosion by suffusion (internal instability) is likely to be a more slowly
developing process, accompanied by more gradual increases in seepage and
changes in pore pressure with time.
(c) ‘Blow-out’ or ‘heave’ in dam foundations subject to backward erosion occurs when
seepage forces create a zero effective stress condition. This should be readily
detected by carefully positioned and well monitored piezometers. Often these low
effective stresses occur below lower permeability layers which act to confine the
seepage flow and it is important that piezometers are installed to measure pore
pressures in these areas. These pressures are usually directly related to reservoir
levels and it is important to monitor the relationship between the pore pressures and
reservoir levels. Such pressures can often be released in controlled circumstances
through filtered relief wells.
(d) Failures from internal erosion in the foundation and from the embankment to the
foundation are mostly from backward erosion, following ‘blow-out’ or ‘heave’. Or
erosion into a drain or ditch which penetrates into the erodible layer. These would
not necessarily be expected to be preceded by large increases in seepage during the
time the erosion is gradually working back from the downstream exit point. When
the erosion has progressed to within a short distance of the reservoir/foundation
interface, it breaks through rapidly or very rapidly.

___________________________________________________________________________
116
 (e) The ease of detection of internal erosion in the foundation by visual and seepage
means is more readily achieved if the area downstream of the dam is not vegetated,
or is at least cleared of larger vegetation. Detection is difficult if the area
downstream is densely vegetated or if the dam toe area is covered by water.
(f) Detection by pore pressure measurement is more likely if there is extensive
instrumentation read regularly and if the erosion is widespread. However it is
unlikely the piezometers will detect concentrated leak erosion unless the
concentrated leak is associated with installation of the piezometers.
(g) The frequency at which seepage is measured can be related to the likelihood that a
dam may fail or experience an accident by internal erosion and piping, the
consequences of failure and the likely time for internal erosion to progress to form
a pipe and the dam breach. It might also be related to the reservoir level, with
enhanced measurements at times of high reservoir level. There is strong evidence
that failures and accidents from internal erosion occur above or near historic high
reservoir levels. At such times the dam is also at greatest risk from a slope
instability point of view. For dams with known deficiencies in internal erosion,
there should be a significantly increased level of inspection and monitoring under
these reservoir conditions.
(h) The seepage monitoring system needs to be capable of being calibrated to separate
out the effects of rainfall or snowmelt e.g. by prior observation and monitoring and
coupling seepage measurements to rainfall and snowmelt measurements at the dam.
It should however be recognized that there are many situations where it will be
unlikely that seepage will be detected, e.g. at night when visual surveillance is
being relied upon; if the toe of the dam is submerged; if seepage occurs high on the
dam abutments and bypass measuring weirs or in winter when the dam is covered
by snow.
(i) There are many dams which may have progression and breach times of the order of
only hours. These dams particularly include some older dams which have no filter
or transitions, dams without downstream rockfill zones to stop or slow the erosion
process, and/or those dams on erodible foundations without well designed and
constructed cut-offs or filters to intercept the foundation seepage. For these dams,
an effective seepage monitoring program would require virtually continuous
monitoring. Daily, or even twice daily inspections, or measurements may be
inadequate. The installation of anti-erosion measures should be seriously
considered at such dams (Section 9.14 and Chapter 10).

Dams analyzed and shown, after improvement, if necessary, to resist internal erosion
following the methods in this bulletin, should no longer be vulnerable to the more obvious
internal erosion failure paths identified in the examples above. However, because conditions
may change over time, and the possibility of the development of internal erosion by some
unforeseen route cannot be ruled out, a surveillance and monitoring regime should be
provided at an intensity and frequency proportionate to the consequences of failure, the rate
of progression to breach and the rate at which the reservoir can be emptied.

11.2 GENERAL PRINCIPLES

Internal erosion is difficult to detect. Internal erosion and piping could suddenly occur
without significant warning beforehand. For instance, a homogeneous long embankment can
be breached in less than one hour by internal erosion triggered by animal holes or cracks. The
frequency of visual inspections may be ineffective. Piezometers in the lower portion of a long
embankment often indicate low water levels, whilst leakage is apparent at higher elevations.
Piezometers spaced at 200m intervals will not be sufficient to guarantee detection of leakage.

___________________________________________________________________________
117
Dirty water on the surface of a dike is a sign of ongoing internal erosion; however turbid
water flowing in the foundation may not be seen.

11.2.1 The likelihood of detection and intervention

A method for assigning likelihoods to detection and successful intervention is given in
Table 2.3 based on Fell et al (2008).
It is known that most internal erosion failures in the embankment occur at reservoir
levels close to or above historic high, and the physical processes are driven by the reservoir
water. Hence a good monitoring and inspection program will have a greatly increased
frequency of inspections and reading of critical instruments under such reservoir conditions.
There should also be increased inspection and monitoring following any significant seismic
event.
Detection may be possible in the continuation or early progression phase, or more
likely, in the advanced stages of progression and breach formation. Detection is likely to be
by:
 Observation of increased seepage or turbid seepage out of the downstream face of the
embankment or in the foundation. This may be by visual observation, or by seepage
measurement, or more sophisticated methods such as thermal monitoring of the foundation
or the downstream slope.

 Measured higher pore pressures in the foundation and/or embankment.

 Settlements, deformation and cracking in the embankment or area downstream of the dam.

Whether detection is likely depends on:

 The rate at which the internal erosion and associated processes, such as instability of
the downstream face, occurs.
 The frequency of inspections, and measurement of monitoring equipment.
 The dam zoning and the location of the concentrated leak and whether the leak will
be visible to those doing the inspection.
 Design, location and sensitivity of monitoring equipment
For example if a process goes from initiation, or first presence of a concentrated leak, to
breach in say 6 hours, and the dam is only inspected or monitored weekly, it is very unlikely
that a piping incident will be detected before breach occurs. However if the dam is visible by
the general population, there is some chance the leak may be noticed nonetheless.
Detection early in the internal erosion process is usually difficult, particularly for
erosion initiating along a crack, or by backwards erosion because the amount of leakage is
very small at the start. Fell et al (2001, 2003) record that most piping incidents are first
identified as a concentrated leak in the progression phase. Suffusion is more likely to be
detected by piezometers because the process is slower to develop. The presence of conditions
potentially leading to heave and backward erosion in the foundation may also be detected by
piezometers provided they are correctly positioned and read as reservoir levels rise.
Visual inspection is a vital tool in detecting internal erosion, whether it is successful is
dependent on the factors discussed above, but also on such practical issues as:
 Inspections are seldom practical at night, so there is 30% to 50% of the time (varying
throughout the year) when detection will not be effective, particularly for rapidly developing

___________________________________________________________________________
118
 piping mechanism. Many dams are not inspected on weekends, further reducing the
likelihood of detection.

 Dense vegetation, runoff from rainfall, snow cover can all hide the presence of a
concentrated leak. However it can be the case that melted snow is a good indicator of areas
affected by seepage.

 For very long embankments, it is not practical to walk to inspect, so it is less likely small
leaks are detected.

11.2.2 Some Information on the Rate of Internal Erosion

The likelihood of detection leading to successful intervention and action to avert the ill-
effects of the failure of the dam depends on the time from when the internal erosion process
may be detected to when breach begins.
Fell et al (2001, 2003) studied case histories of failures and accidents for piping in the
embankment, foundation, and embankment to foundation. Based on the case histories and an
understanding of the physical processes they provided guidance on the time for progression
beyond when a concentrated leak is first observed, and development of a breach.
Most of the cases studied were for breach by gross enlargement, so the method is
applicable to cases where the mechanism is gross enlargement. It is considered to be
reasonably applicable to cases where the final breach is by slope instability, following
development of a pipe. It will probably underestimate the time for breach by sloughing.
Sloughing is a slowly developing breach mode which should take days or weeks to lead to
breach.
Breach by sinkhole development is potentially a rapid process in the final stages when
the sinkhole emerges into the reservoir. The breach would be expected to occur in a small
number of hours but there is not case data to support a more refined estimate.
Note that the dispersivity of the soil does not significantly affect the rate of erosion.
Fell et al (2001, 2003) show that the method gives a reasonable estimate of the time for
progression beyond where a concentrated leak is observed and breach, and the times are
acceptably accurate for the purpose here which is to assess the likelihood of detection,
intervention and repair. Fell et al (2001, 2003) caution however, against over-reliance of
these figures for life loss estimates where the estimates are sensitive to the assumed warning
times.

11.2.3 A review of methods of monitoring and detection and

their applicability to internal erosion modes
Tables 11.1, 11.2, and 11.3 summarize the usual rate of development, based on Fell et
al (2001, 2003) as described above, ease of detection, and methods of detection in generic
terms for internal erosion in the embankment, foundation and from the embankment to
foundation. In these tables the terms are defined as follows:

Rate of development: Slow = weeks or months, even years; Medium = days or weeks;
Rapid = hours (>12 hours) or days; Very Rapid = <3 hours.

Ease of detection: Difficult = unlikely to be detected in most cases; Moderate = may

be detected in some cases; Readily = readily detected in most cases.

___________________________________________________________________________
119
 Methods of detection: Visual (inspection), seepage (measurement either visual or by
instruments), pore pressure (measurement); survey (survey of surface markers, to determine
horizontal and vertical deformation); thermal (thermal measurements within or on the
downstream surface of the embankment including use of fiber optics); geophysical
(resistivity and self potential).

The classifications for ease of detection assume the dam is well instrumented to
measure seepage and pore pressure, readings are frequent and the dam is subject to regular
inspection by experienced personnel. These classifications will be too optimistic if the dam
and the area downstream are covered in vegetation, or snow covered for part of the year or if
the toe of the dam is drowned by the river or another reservoir downstream.

Table 11.1 Rate of development, ease of detection and methods of detection of internal
erosion in the embankment (adapted from Fell et al, 2005)
Phase of Mechanism Usual Rate of Ease of Method of Detection
Develop- Development Detection
ment
Initiation Concentrated Rapid or very Difficult Seepage, visual.
leak in cracks or rapid
hydraulic
fracture,
associated with
conduits or walls
Concentrated Rapid or very Difficult Seepage, visual if the crack
leak in a crack rapid emerges on the
or in a hydraulic downstream face
fracture
Global Medium to Difficult Pore pressure, seepage,
backward slow, (unless thermal, geophysical
erosion in the potentially observed at Visual if sinkhole emerges
core rapid to very downstream on the downstream face.
rapid at the face)
end
Contact Normally Monitoring of Detection of cavities in the
slow, rapid contact erosion dam by appropriate
failure can can be methods such as
occur only if achieved by geophysics may also
another visual indicate the existence of
process like observation of contact erosion. When
piping or fines particles contact erosion is
sliding is into leakage or identified, a decrease of
initiated by by the flow velocities where it
contact identification takes place, can reduce or
erosion of sinkholes. stop the phenomenon
Suffusion Slow Moderate to Pore pressure, seepage,
/internal difficult thermal, geophysical,
instability visual if emerging on the
downstream face
Contin- Filters satisfying Erosion will

___________________________________________________________________________
120
 Phase of Mechanism Usual Rate of Ease of Method of Detection
Develop- Development Detection
ment
uation no-erosion cease
criterion
Filters satisfying Rapid to very Moderate to Seepage, pore pressure
excessive rapid difficult
erosion criterion
Filters not Rapid to very Moderate to Seepage, pore pressure
satisfying rapid difficult
continuing
erosion
criterion, or no
filters
Progression Assess time
to form a using method
pipe, and a in Section
breach 11.2.2.

Gross Very rapid or Moderate. to Seepage, visual

enlargement; and rapid readily
slope instability
linked to Medium or Readily
development of a slow
pipe.

Crest settlement Usually slow Readily Visual, survey, seepage

or/and sinkhole in to medium.
embankment

Slope instability, Usually slow Readily to Visual, survey, seepage

unraveling or to medium Moderate
sloughing unless linked
to rapid
development
of pipe.

It should be noted that monitoring of seepage, either by visual surveillance or

measurement, is the most common means of identifying whether internal erosion is
occurring.
It is not common to have sufficient change in the seepage, or in other factors such as
pore pressure changes or settlement, to identify conclusively that internal erosion has initiated
and is continuing. It is more common to recognize when the erosion has progressed to the
stage that a pipe has developed, or recognize that changes in pore pressures, seepage or
settlement have occurred, the cause of which is not clear. Such changes may be a pre-cursor
to internal erosion and piping, so it is important they are observed and, when they occur,
investigated.

___________________________________________________________________________
121
 Table 11.2 Rate of development, ease of detection and methods of detection of internal
erosion in the foundation and from embankment to foundation (adapted from Fell et al, 2005)
Phase of Mechanism Usual Rate of Ease of Method of
Development Development Detection Detection
Initiation - Backward erosion Slow Difficult Visual, seepage, pore
Foundation pressure, thermal,
geophysical
Backward erosion Rapid to very Readily to Visual, seepage, pore
following “blow-out” rapid difficult pressure, thermal,
geophysical
Concentrated leak Rapid or very Moderate to Visual, seepage
erosion rapid difficult
Suffusion Slow Moderate to Visual, seepage, pore
/internal instability difficult pressure, thermal,
geophysical
Initiation - Backward erosion Slow, Difficult Pore pressure,
embankment potentially seepage
to foundation rapid to very
rapid at the
end
Initiation – Contact erosion Slow Difficult Pore pressure,
Embank- seepage, survey,
ment to thermal, geophysical
foundation
Continuation Filters satisfying no Erosion will
erosion criterion cease if all
seepage is
intercepted
Filtered exit, Rapid to very Moderate to Seepage, pore
satisfying excessive rapid difficult pressure
erosion, or incomplete
seepage interception
Filtered exit, not Rapid to very Moderate to Seepage, pore
satisfying continuing rapid difficult pressure
erosion criterion, or
unfiltered exit
Progression Assess time
to form a using method
pipe, and a in Section
breach 11.2.2.

Gross enlargement in Very rapid or Moderate to Seepage, visual

the foundation or in the rapid readily
embankment; and slope
instability linked to Medium or Readily Visual, survey,
development of a pipe. slow seepage

Crest settlement or/and

sinkhole in Usually slow Readily Visual, survey,
embankment to medium. seepage

___________________________________________________________________________
122
 Phase of Mechanism Usual Rate of Ease of Method of
Development Development Detection Detection
Slope instability, and Usually slow Moderate to Visual, survey,
unraveling or to medium readily seepage
sloughing, for internal unless linked
erosion embankment to to rapid
foundation development
of pipe.

Table 11.3 Rate of development, ease of detection and methods of detection of internal
erosion into and along conduits or adjacent to walls (adapted from Fell et al, 2005)
Phase of Mechanism Usual Rate of Ease of Method of
Development Development Detection Detection
Initiation Concentrated leak Rapid or very Difficult Visual, seepage
in crack or rapid
hydraulic fracture
Erosion into open Slow Moderate Visual, seepage
crack or joint in
the conduit or wall
Continuation: Filters (A) Erosion will
Concentrated satisfying no cease
leak in crack erosion criterion
or hydraulic
fracture Filtered exit, Rapid to very Moderate to Seepage, pore
satisfying rapid difficult pressure
excessive erosion
criterion
Rapid to very Moderate to Seepage, pore
Filtered exit, not rapid difficult pressure
satisfying
continuing erosion
criterion, or
unfiltered exit
Continuation: Crack or joint Slow (C) Moderate to Seepage, pore
Erosion into width satisfies difficult (B) pressure
open crack or excessive erosion
joint in the criterion
conduit or Slow to medium Moderate to Seepage, pore
wall Crack or joint (C) difficult (B) pressure
width does not
satisfy continuing
erosion criterion
Progression to Assess time using
form a pipe method in Section
and a breach 11.2.2.

Gross enlargement; Very rapid or Moderate to Seepage, visual

and slope rapid readily
instability linked to Readily
development of a
pipe
___________________________________________________________________________
123
Phase of Mechanism Usual Rate of Ease of Method of
Development Development Detection Detection

Crest settlement Medium or slow Readily Visual, survey,

or/and sinkhole in Usually slow to seepage
the embankment medium.

Slope instability, Usually slow to Moderate to Visual, survey,

unraveling or medium unless readily seepage
sloughing linked to rapid
development of
pipe.
Conduits and Erosion into a conduit or crack in a wall, usually only progresses towards
walls breach by initiation of erosion along the conduit or wall, in which case
factors are as for gross enlargement and instability. For cases which
continue towards development of a sinkhole in a crest mode of breach,
ability to hold a roof/or sinkhole is critical and the rate of development is
usually slow, readily detectable by visual, survey or seepage.
Notes:
(A) Filters are those controlling internal erosion around the conduit, or adjacent
walls.
(B) Often the crack and seepage is visible on inspection, but erosion may be
intermittent, or only when conduit is flowing with water, so not readily observed.
(C) The evidence seems to be that even for conduits surrounded by erodible soils, e.g.
fine sand, the rate of erosion into the conduit is slow.

___________________________________________________________________________
124
 12. METHODS FOR ASSESSMENT OF
THE LIKELIHOOD OF FAILURE OF A
DAM BY INTERNAL EROSION
12.1 INTRODUCTION
As discussed in Section 9.2 there are a number of approaches for assessing a dam for
internal erosion. These are:
(1) Deterministic: This can be regarded as a load versus resistance approach using
the Spillway Design Flood and Maximum Design Earthquake.
(2) Engineering Judgment: This can also be regarded as a load versus resistance
approach, but applied not just to the Spillway Design Flood and Maximum Design
Earthquake, so engineering judgment or qualitative risk assessments can be made.
(3) Quantitative Risk Assessment, which requires more formalized estimates of
probabilities of failure and the consequences.
This section provides information on how the likelihood of failure of an existing dam
by internal erosion can be determined. This Bulletin does not set out to be a guideline on risk
assessment. It provides information which may be used in such assessments. For detailed
information on risk management see ICOLD Bulletin 130, Risk Assessment in Dam Safety
Management (ICOLD, 2005); ANCOLD Guidelines on Risk Assessment (ANCOLD, 2003),
Hartford and Baecher (2004), Risk and Uncertainty in Dam Safety, and UK Risk Assessment
and Reservoir Safety (Environment Agency, 2013). These publications also give guidance on
methods for estimating loss of life and financial and economic losses.

12.2 GENERAL PROCEDURE

The general procedure for estimation of the likelihood of failure a dam by internal
erosion in the embankment, the foundations, and from embankment to foundations is very
similar to that for the engineering analysis described in Sections 9.10 and 9.11, but ultimately
requires that the likelihood (probability) of failure be assessed:
1. Assemble all the relevant information about the dam and its foundations as
described in Section 9.8.
2. Determine the loading conditions for the dam resulting from the reservoir and
earthquakes as described in Sections 9.3, 9.4, 9.5 and 9.6 and Section 3.4 in
relation to concentrated leaks.
3. Carry out a detailed Potential Failure Modes Analysis (PFMA) as described in
Section 9.9.
4. Screen out PFM which are not applicable to the dam or which will contribute a
negligible amount to the likelihood of failure. Some guidance on how this may
be done is given in Section 9.10. Screening may be done on the zoning of the
dam and the foundation conditions, or a more detailed understanding of the
factors influencing the initiation and continuation of erosion.
5. For the remaining PFM use engineering judgment, qualitative or quantitative
risk analysis methods to estimate the likelihood of failure taking account of the
frequency of the applied reservoir and seismic loads. Some guidance on how
this may be done is given in Section 12.4.
___________________________________________________________________________
125
12.3 ESTIMATION OF LIKELIHOODS OF FAILURE
FOR THE POTENTIAL FAILURE MODES
APPLICABLE TO THE DAM
12.3.1 Some general principles
There are some general principles which should be applied when estimating the
likelihood of a PFM leading to failure. They are similar to those in Section 9.11.1, except that
a range of loads applied at different frequencies (e.g. varying water levels) must be
considered.

12.3.2 Concentrated leak erosion-some aids to judgment

Fell et al (2008) use tables which for each PFM of concentrated leak erosion list the factors
which most affect the likelihood of a crack and hydraulic fracture. These are assigned a
relative importance factor (RF) and likelihood factor (LF) and the product is summed to give
an overall  (RF x LF), between 6 and 24. In Fell et al (2008) these are then linked to
estimate conditional probabilities of cracking or hydraulic fracture. The tables are included in
the “Piping Toolbox” described in Section 12.4.2, which also gives sources of copies of the
document.

12.4 QUANTITATIVE AND SEMI QUANTITATIVE

RISK ASSESSMENT METHODS
12.4.1 Methods based on historic databases of failures and
incidents
12.4.1.1 Interim Guide to Quantitative Risk Assessment for UK
Reservoirs
The Interim Guide to Quantitative Risk Assessment for Reservoirs in the United
Kingdom (Brown & Gosden, 2004), is a quantitative method, quantified on the historic
performance of the large population of similar dams in the United Kingdom, for which
records of incidents of poor performance have been kept for many years (Brown & Tedd,
2003). The guide has a comprehensive text and includes probability assessments estimated
on spreadsheets for the major threats, overtopping, stability and particularly, internal erosion;
a simple dambreak assessment to estimate the consequences of failure, and guidance on
unacceptable and acceptable risk, with emphasis on following the ALARP principle to
maintain the probability of failure as low as practicable at proportionate cost.
The method was developed from an earlier scoping project by KBR (2002) to compare
the probability of failure due to floods with internal erosion, the project being managed under
a Steering Group appointed by Defra and including trials of the prototype risk assessment
methodology on ten UK dams.
The methodology is shown on Figure 4.1 of the Guide and comprises devising Current
and Intrinsic Condition scores in ranges of 0 to 10, and then mapping these onto a probability
distribution based on historic performance of UK dams. Default values of annual probability
of failure are given in Table 4.2 of the Guide (reproduced here in Table 12.1), with
adjustments for dam type included in Sheet 4.3 of the workbook. Experience in application of
the Guide (Brown & Gosden, 2006, Brown et al, 2008) included that some panel engineers
___________________________________________________________________________
126
found difficulty in use of Excel workbooks with multiple sheets, and that guidance on
scoring, particularly where there was no data on dam construction, should be improved to
obtain better consistency.
This method is only applicable to UK dams which are mostly puddle clay core dams, a
design not used widely elsewhere. There are also many ‘homogeneous’ dams. Both types do
not have filters. The probability elements of the method are derived from the performance of
these dams in UK and cannot strictly be applied elsewhere, unless users can identify
substantial similarities to their dams.
Table 12.1 Suggested default values of probability of failure for anchor points for
internal erosion for UK dams
Current Description Dams regulated under Small dams which do not
Condition Reservoirs Act 1975 (Table come under legislation
score 4.2 of Brown & Gosden, (Table 4.5 of Brown &
2004) Gosden, 2004)
Internal Internal Internal Internal
stability stability stability stability
(embankment) (appurtenant (embankment) (appurtenant
works) works)
10 Emergency 1.4 x 10-2 1.0 x 10-2 3.8 x 10-2 2.6 x 10-2
drawdown
considered
necessary to avert
failure
8 Concern over 3.8 x 10-4 2.7 x 10-4 3.3 x 10-3 2.3 x 10-3
behavior leading
to works outside
periodic safety
review
0 Best condition 4.7 x 10-8 3.3. x 10-8 As dams under the Act
dam

An updated UK guide will soon be published (Environment Agency, 2013). It updates

the 2004 approach but remains focussed on UK dams. The updated UK guide also
recommends that this Bulletin be used to examine the mechanics of internal erosion in
conjunction with the quantitative risk assessment.

12.4.1.2 UNSW Method

The UNSW method is described by Foster et al (2000(a), (b)). The database of dam
incidents was compiled primarily from the three ICOLD studies (ICOLD 1974, 1983 and
1995) and supplemented with additional incident cases from the other existing compilations,
from the literature and dam engineering organizations.
The method allows estimation of the annual probability of a dam failing within the
embankment, in the foundation and from embankment to foundation based on the dam
zoning, the material in the core, the filters, foundation geology and cut-off details.
The method allows a broad categorization of the dam types into those less and more
likely to fail. However the authors of the method have found that it does not allow for
important details of the dam to be allowed for, and it cannot model the frequency of reservoir
___________________________________________________________________________
127
loading which is an important factor in estimating the likelihood of failure. They prefer to
use event tree methods such as the “piping toolbox” described below.

12.4.2 Event tree methods

Event tree methods have been used by a number of organizations including USBR,
USACE, and Australian Dam organizations since the late 1990’s. The methods have been
refined with time and are still being refined as the understanding of the mechanics of internal
erosion is improved.
These methods mostly rely on “expert opinion elicitation” (e.g. Brown & Aspinall,
2004) to estimate conditional probabilities within the event trees. They are successful
provided those doing the analysis are familiar with the mechanics of internal erosion and
quantitative risk analysis. Because of this their use has been largely restricted to USBR,
USACE, and Australian, USA and UK Consultants.
The “Piping toolbox” or “Risk analysis for Dam Safety, A Unified Method for
Estimating Probabilities of Failure of Embankment Dams by Internal Erosion and Piping” is
an event tree system for estimating the annual probability of breach of embankment dams by
internal erosion and piping in the embankment, the foundation and from the embankment into
the foundation.
It was developed as a unified approach for carrying out risk analyses by the United
States Bureau of Reclamation (USBR), United States Army Corps of Engineers (USACE),
Consulting Engineers URS and the University of New South Wales.
The document (Fell, Foster, Davidson, Cyganiewicz, Sills and Vroman (2008), see
references for details) is in two parts, a Guidance Document, and a Supporting Information
Document. It is available as a University of New South Wales UNICIV Report in electronic
form by emailing Emeritus Professor Robin Fell, UNSW, [email protected], or Dr Mark
Foster, URS Australia, [email protected]. Note that it includes the tables referred to in
12.3.2 giving guidance on cracks and concentrated leak erosion.
The “Piping toolbox” is designed for use in quantitative risk analysis. If it is to be used
for that purpose the Development Team and their organizations strongly recommend that all
risk analysts that will use this toolbox be trained in its use. The complexity of the issues and
importance of the end product demands that all analysts fully understand the methodology.
The key goals of the training would be to provide an understanding of all features and
components of the methodology; to outline for the analyst the supporting information and
background that was used in the development of the methodology; and to guide the analyst
through a detailed, real life example use of the methodology.
The authors of the Toolbox have seen a number of examples where the lack of adequate
training has resulted in incorrect application of the methods.
Some perspectives on the use of the Toolbox in the United Kingdom are given in
Eddleston et al (2009, 2010)
It is recognized that not many will wish to use the document for quantitative risk
analysis. It is however a valuable document for those managing the safety of dams for
internal erosion. The Tables of Conditional Probabilities it contains are a useful guide to the
factors which affect the likelihood of initiation, continuation, progression and breach, and of
the likelihood of successful detection and intervention.
An electronic copy of the document can be obtained by emailing Emeritus Professor
Robin Fell, UNSW, [email protected], or Dr Mark Foster, URS Australia,

___________________________________________________________________________
128
[email protected]. Note that it includes the tables referred to in 12.3.2 giving guidance
on cracks and concentrated leak erosion.

___________________________________________________________________________
129
 13. TERMINOLOGY – INTERNAL
EROSION PROCESSES –
INTERNATIONAL GLOSSARY
13.1 TERMINOLOGY USED IN RELATION TO
INTERNAL EROSION
The following terminology is used in relation to internal erosion and piping.
It is important that all those working in the area adopt these terms so as to use a
common language to describe the processes and avoid confusion. The terminology is based
on that adopted at the International Workshop on Internal Erosion of Dams and their
Foundations, held in Aussois, France in April 2005, (Fell and Fry, 2007).
Backward erosion: Backward erosion involves the detachment of soils particles when
the seepage exits to a free unfiltered surface, such as the ground surface downstream of a soil
foundation or the downstream face of a homogeneous embankment or a coarse rock fill zone
immediately downstream from the fine grained core. The detached particles are carried away
by the seepage flow and the process gradually works its way towards the upstream side of the
embankment or its foundation until a continuous pipe is formed.
There are two forms of backward erosion, “backward erosion piping” in which the roof
of the pipe is formed by a cohesive soil layer of embankment and the pipe is essentially
horizontal (Figure 2.5); and “global backward erosion” where near-vertical pipes form within
broadly graded silt sand gravel (non-plastic) cores of embankments (Figure 2.7).
Breach initiation: The initiation of breach is the final phase of internal erosion. It may
occur by one of the following five phenomena (listed below in order of their observed
frequency of occurrence):
 Gross enlargement of the pipe.
 Overtopping (e.g. due to settlement of the crest from suffusion and/or due to
the formation of a sinkhole from a pipe in the embankment).
 Slope instability of the downstream slope
 Unraveling of the downstream face.
 Static liquefaction: A form of slope instability in loose non-plastic soils which
involves a large increase of pore pressure and loss of undrained strength.
Breach process: There are three phases in the breach process, initiation, formation and
development. The breach initiation is where there is an uncontrolled release of water from the
reservoir but the crest does not collapse, the breach formation occurs after the crest collapses
up to the erosion of the whole section of the dam and the development where there is finally
the enlargement of the breach width by side erosion.
Concentrated leak erosion: Where there is an opening, through which concentrated
leakage occurs, the walls of the opening may be eroded by the leaking water. Such
concentrated leaks may occur through a crack caused by settlement or hydraulic fracture in a
cohesive clay core for example, or through desiccation and tension cracks at high levels in the
fill, or through cracks resulting from settlement of fill. In some circumstances, these openings
___________________________________________________________________________
130
may be sustained by the presence of structural elements such as spillways and pipes, or by the
presence of cohesive materials able to ‘hold a roof’, as it is described, below which an
opening is sustained, the periphery of which is eroded. It may occur in a continuous zone
containing coarse and/or poorly compacted materials which form an interconnecting voids
system. The concentration of flow causes erosion (some organizations call this scour) of the
walls of the crack or interconnected voids.
Contact erosion: (also known as parallel contact erosion): Soil contact erosion is a
form of internal erosion which involves selective erosion of fine particles from the contact
with a coarser layer, caused by the flow passing through the coarser layer. For instance along
the contact between silt and gravel sized particles. It relates only to conditions where the flow
in the coarser layer is parallel to the interface between the coarse and fine layer (Figure 13.1).
Selective erosion of fine particles from the contact with a coarser layer caused by the flow
through the fine layer with flow normal to the contact surface is considered by another phase
of the erosion process: filtration or continuation.

Figure 13.1 Diagram of soil contact erosion.

Continuation (filtration): Continuation is the phase where the relationship of the
particle size distribution between the base (core) material and the filter controls whether or
not erosion will continue. Foster and Fell (2001) and Foster (1999) define four levels of
severity of continuation; no erosion, some erosion, excessive erosion and continuing erosion.
These are shown conceptually in Figure 13.2 and defined individually below:

Continuing Erosion
CONTINUING Boundary
EROSION

DF15 EXCESSIVE
EROSION Excessive Erosion
Boundary
SOME
EROSION

No Erosion
NO EROSION Boundary

Other factor e.g. DB85

Figure 13.2 Conceptual filter erosion boundaries (Foster, 1999), Foster & Fell (2001)
___________________________________________________________________________
131
  Continuing erosion: The filter is coarser than the continuing erosion criteria
and is too coarse to allow the eroded base materials to seal the filter allowing
unrestricted erosion of the base soil.
 Excessive erosion defines conditions where erosion of the base soil will be
excessive before it seals. Filters between the excessive erosion boundary and the
continuing erosion boundary will eventually seal but only after significant
erosion of the base soil. In dams there may be large leakage flows before the
filter seals by clogging of the surface of the filter by eroded base soil.
 Some erosion: The filter is between the no erosion and excessive erosion
boundaries. The filter quickly seals after particles of the base material clog the
surface of the filter.
 No erosion: The filter is finer than the no erosion criterion and seals with no or
practically no erosion of the base material. Filters designed and constructed
according to modern filter design criteria will satisfy no erosion criteria.
Detachment: Detachment is the first stage of the erosion process. Particle detachment
occurs by the hydraulic shear forces developed by the seepage flow velocity. The mechanics
are determined by whether the soil is cohesive or cohesionless.
Detection: Detection is the ability to detect the internal erosion process during one of
the erosion phases. Currently, internal erosion is most easily detected in the progression or
breach phases.
Dispersion (also known as deflocculation) is a physico-chemical phenomenon
contributing to concentrated leak erosion.
Soils in which the clay particles will detach from each other and from the soil structure
without a flow of water, and go into suspension are termed dispersive clays.
The dispersivity of a soil is directly related to its clay mineralogy. In particular soils with a
high exchangeable sodium percentage such as montmorillonite present, tend to be dispersive,
while kaolinite and related minerals (e.g. halloysite) are non dispersive. Soils with illite
present tend to be moderately dispersive.
The dispersivity depends also on the pore water chemistry. Low pore water salt
concentrations lead to greater dispersivity and high salt concentrations can suppress
dispersion in susceptible soils.
Erosion of a soil particle is the detachment and transport of that particle by forces
caused by seepage flow (drag and lift forces, which are related to viscosity and turbulent
shear stresses). Detachment is a first condition expressing that drag and lift forces are larger
than resisting forces. Transport is the second condition expressing first a hydro-mechanical
criterion that the drag forces are larger than the rolling or sliding resistance of the buoyancy
weight of the particle and secondly a geometric criterion: voids or cracks exist which are
large enough for eroded particles to pass through. This last criterion makes the difference
between internal and external erosions.
Failure process: Failure process describes how the dam could fail by internal erosion
for the particular loading condition. It includes the definition of loading, the internal erosion
path and internal erosion phases (initiation of erosion, continuation, progression, failure to
intervene, and breach). In event tree analysis, each internal erosion failure process is
represented by one complete event tree.
Failure mode: Failure mode describes how the dam could fail for the particular loading
condition. The three principle failure modes for embankment dams are overtopping due to
inadequate spillway capacity, internal erosion and slope instability. For internal erosion, it
___________________________________________________________________________
132
includes the definition of loading, the internal erosion path and internal erosion phases
(initiation of erosion, continuation, progression and initiation of breach). In event tree
analysis, each failure mode is represented by one complete event tree.
Flaw: A flaw is a physical condition that allows seepage concentration Examples
include continuous crack; or poorly compacted layer in which a concentrated leak may form.
Heave (also known as ‘blow out’) occurs in cohesionless soils which are confined by
an overlying lower permeability stratum when seepage pore pressures are such that the
effective stress becomes zero (pore pressure equals total stress). ‘Heave’ may often be
followed by backward erosion of the cohesionless soil if the seepage gradients remain high at
the surface.
Hydraulic fracture occurs in the core or foundation of embankment dams when the
pore pressure exceeds the minor principal stress and the effective stress in the core or
foundation becomes zero (or even slightly negative if the soil can withstand tensile stresses).
The pressure of the water seeping through the core from the reservoir exceeds the remaining
compressive stress and forms a crack or further opens an existing crack in which internal
erosion by concentrated leak may initiate.
Initiation: Initiation is the first phase of internal erosion, when one of the phenomena
of detachment of particles occurs. Four initiation phenomena are defined: concentrated leak,
backward erosion, suffusion and soil contact erosion (each of which are defined separately in
this section).
Internal erosion occurs when soil particles within an embankment dam or its
foundation, are carried downstream by seepage flow. Internal erosion can initiate by
concentrated leak erosion, backward erosion, suffusion and soil contact erosion.
Internal erosion path is the path of moving eroded particles inside the dam and/or its
foundation.
Internal erosion phases (or internal erosion mechanisms) are the mechanisms during
an internal erosion process leading to failure. Internal erosion of embankment dams and their
foundations can be represented by four phases (or mechanisms):
1. Initiation of erosion.
2. Continuation of erosion (i.e. whether there are filters capable of stopping the
erosion process).
3. Progression to form and sustain a pipe and/or to increase seepage and pore
pressures in the downstream part of the embankment or its foundation.
4. Breach initiation resulting in uncontrolled release of the water from the
reservoir.
Internal Instability: In soils subject to internal instability finer particles in the soil are
able to move within the soil mass under the forces imposed on the particles by seepage flow.
The phenomenon does not require a crack within the soil in which erosion may occur as is
required in concentrated leak erosion, or a free surface from which particles detach and a roof
of cohesive soil as is required for backward erosion.
Coarse graded and gap graded soils, such as those shown schematically in Figure 13.3
are internally unstable. In these circumstances, the internal instability is known as suffusion.
The term internal instability has sometimes been used in the literature to describe internal
erosion of soils which do not self filter, and internal erosion of broadly graded soils which do
not satisfy the rules for suffusive soils but which erode under very high gradients.

___________________________________________________________________________
133
 FINER FRACTION
Finer fraction

Figure 13.3 Soil gradation types which are potentially internally unstable and
susceptible to suffusion

Intervention: Intervention is the ability to stop the internal erosion process during one
of the first three phases and prior to the initiation of a breach.
Load: The main loads exerted on a dam are hydrostatic loads (which are related to the
reservoir level) and seismic loads (which may be expressed in terms of earthquake magnitude
and peak ground acceleration [PGA]). Other potential loads include acts of terrorism or an
aero plane crashing into the dam. The likelihood of a given reservoir level load is expressed
as the annual probability or frequency that the maximum reservoir level is within a certain
range. The likelihood of a given seismic load is expressed as the annual probability or
frequency of a loading parameter (e.g. peak ground acceleration) being in a certain range.
Load Condition: A load condition is a theoretical combination of different loads
exerted on a dam at a given time for analysis purposes. The loading condition is the first
node of the internal erosion assessment. The loading condition controls whether certain
failure processes are possible, and affects the likelihood of each of the internal erosion
phases.
Location of initiation of erosion: This is the physical initiating point of internal
erosion in the structure such as in the embankment, foundation, along a conduit or along a
spillway interface.
Piping: Piping a potential progression phase of internal erosion which initiates by
backward erosion, or erosion in a crack or high permeability zone, and results in the
formation of a continuous tunnel called a ‘pipe’ between the upstream and the downstream
side of the embankment or its foundation. Internal erosion is commonly described as
‘internal erosion and piping’ but piping is actually the culmination of a process of erosion in
which a number of phases must occur and be sustained in order that a ‘pipe’ develops through
the dam or its foundation and allows the passage of considerable quantities of water which
may lead to a breach.
Progression: Progression is the phase of internal erosion where hydraulic shear stresses
within the eroding soil may or may not lead to the erosion process being on-going, and in the
case of backward and concentrated leak erosion to formation of a pipe. The main issues are
whether the pipe will collapse, or whether upstream zones may control the erosion process by
flow limitation.

___________________________________________________________________________
134
 Self-filtering: In soils which self-filter, the coarse particles prevent the internal erosion
of the medium particles, which in turn prevent erosion of the fine particles. Soils which
potentially will not self-filter include clayey sandy gravels, soils which are susceptible to
suffusion, and very broadly graded soils such as those which fall into the grading envelope
shown in Figure 13.4.
Movement of the particles in such broadly graded soils may be in concentrated leaks in
cracks or openings caused by hydraulic fracture as proposed by Sherard (1979).

Figure 13.4 Examples of grading envelopes of some broadly graded soils which did not
self filter (Sherard, 1979)
Sinkhole: A sinkhole is a cavity formed in the core of a dam or in a dike which may
form as a result of erosion initiated by a concentrated leak, contact erosion or suffusion, or
erosion into open joints in a foundation or conduit. The sinkhole cavity is usually near
vertical and forms above the zone or point where erosion initiates. The progression of
development of sinkholes is often a slow process with soil falling from the top of the sinkhole
being carried away by the initiating erosion mechanism. Sinkholes may take years to manifest
themselves at the crest of the dam or dike
Static Liquefaction: Static liquefaction occurs in loose saturated contractive
cohesionless soils. As shear strains develop under the static loads in the slope, the pore
pressures in the soil increase significantly and effective stresses approach zero, resulting in a
large reduction in shear strength in undrained loading. The soil may exist in a contractive
state in its natural condition or may become contractive as a result of suffusion.
Suffusion is a form of internal erosion of internally unstable soils which involves
selective erosion of finer particles from the matrix of coarser particles, in such a manner that
the finer particles are removed through the voids between the larger particles by seepage
flow, leaving behind a soil skeleton formed by the coarser particles. The volume of finer
particles is such that they fit within the voids formed by the coarser particles, that is the voids
are under-filled. Suffusion usually involves little or no change in volume of the soil mass.
Suffusion occurs at vertically upward seepage gradients less than the Terzaghi critical
gradient and the effective stresses are carried largely by the coarser particles. This
phenomenon is sometimes referred to as suffosion in the literature.

___________________________________________________________________________
135
13.2 DEFINITIONS RELATED TO THE ANALYSIS OF
THE PROBABILITY OF FAILURE
The following definitions relate to the analysis of the probability of failure, more details
are given in ICOLD Bulletin 130:
Annual Exceedance Probability (AEP): The estimated probability that an event of
specified magnitude will be exceeded in any year.
Frequency: A measure of likelihood expressed as the number of occurrences of an
event in a given time or in a given number of trials (see also likelihood and probability).
Likelihood: Conditional probability of an outcome given a set of data, assumptions and
information. Also used as a qualitative description of probability and frequency.
Probability: A measure of the degree of certainty. This measure has a value between
zero (impossibility) and 1.0 (certainty). It is an estimate of the likelihood of the magnitude of
the uncertain quantity, or the likelihood of the occurrence of the uncertain future event.
There are two main interpretations from ICOLD:
Statistical – frequency or fraction – The outcome of a repetitive experiment of some
kind like flipping coins. It includes also the idea of population variability. Such a number is
called an “objective” probability because it exists in the real world and is in principle
measurable by doing the experiment.
Subjective probability (degree of belief) – Quantified measure of belief, judgment, or
confidence in the likelihood of an outcome, obtained by considering all available information
honestly, fairly, and with a minimum of bias. Subjective probability is affected by the state of
understanding of a process, judgment regarding an evaluation, or the quality and quantity of
information. It may change over time as the state of knowledge changes.
Risk assessment: The process of reaching a decision recommendation on whether
existing risks are tolerable and present risk control measures are adequate, and if not, whether
alternative risk control measures are justified or will be implemented.
Screening: An initial assessment of the risk of internal erosion at a dam, used to inform
the need for further investigations or studies, or whether the risk is tolerable

13.3 ICOLD DEFINITIONS OF LARGE DAM,

FAILURE, ACCIDENT AND INCIDENT
Large dam: ICOLD define a large dam as a dam which is more than 15 meters in
height (measured from the lowest point in the general foundations to the crest of the dam), or
any dam between 10 and 15 meters in height which meets one of the following conditions:
 the crest length is not less than 500 meters
 the capacity of the reservoir formed by the dam is not less than one million
cubic meters
 the maximum flood discharge dealt with by the dam is not less than 2000 cubic
meters per second
 the dam is of unusual design.
Accident: Three categories of accidents are listed by ICOLD (1974):

___________________________________________________________________________
136
  Accident Type 1 (A1): An accident to a dam which has been in use for some
time but which has been prevented from becoming a failure by immediate
remedial measures, including possibly drawing down the water.
 Accident Type 2 (A2): An accident to a dam which has been observed during
the initial filling of the reservoir and which has been prevented from becoming a
failure by immediate remedial measures, including possibly drawing down the
water.
 Accident Type 3 (A3): An accident to a dam during construction, i.e. settlement
of foundations, slumping of side slopes etc., which have been noted before any
water was impounded and where the essential remedial measures have been
carried out, and the reservoir safely filled thereafter.
Failure: Collapse or movement of part of a dam or its foundation, so that the dam
cannot retain water. In general, a failure results in the release of large quantities of water,
imposing risks on the people or property downstream (ICOLD, 1995). Incidents to dams
during construction are only considered as failures when large amounts of water were
involuntarily released downstream. Two categories of failures are listed by ICOLD (1974):
 Failure Type 1 (F1): A major failure involving the complete abandonment of
the dam.
 Failure Type 2 (F2): A failure which at the time may have been severe, but yet
has permitted the extent of damage to be successfully repaired, and the dam
brought into use again.
Incident: Either a failure or accident, requiring major repair.

___________________________________________________________________________
137
 REFERENCES
An electronic copy of the Bulletin and some other internal erosion references can be
downloaded from the ‘sharespace’ of the Technical Committee on Embankment Dams on the
ICOLD website (www.icold-cigb.net). It is accessible to all national committee members. A
password (available from your national committees) is required to access the members’ area,
and another password (available from ICOLD) to access the Technical Committee
sharespace.
ANCOLD (2003). Guidelines on risk assessment. Australian National Committee on Large
Dams.
Arulanandan, K. and Perry, E.B. (1983). Erosion in relation to filter design criteria in earth
dams. Journal of Geotechnical Engineering Division, ASCE, 109 (GT5), 682-698.
Bakker, K.J. (1987). Hydraulic filter criteria for banks. Proc 9th European Conference on Soil
Mechanics and Foundation Engineering, Dublin.
Beguin, R., Guidoux, C., Faure, Y.H. and Ho, C.C. (2009). Stability conditions and erosion
rates at the contact between two soils, subjected to a water flow parallel to the interface.
Presentation at ICOLD European Working Group on Internal Erosion St Petersburg April
Beguin, R. (2011). Etude multi-échelle de l'érosion de contact au sein des ouvrages
hydrauliques en terre. PhD Thesis. Université Joseph Fourier, Grenoble, France.
Belkova I. N., Glagovsky V.B., and Radchenko, V.G. (2011). Assessment of Seepage
Strength of earth dams. Presentation at Annual meeting ICOLD European Working Group
on Internal Erosion Brno April
Benahmed, N. and Bonelli, S. (2012). Investigating piping erosional behavior of cohesive
soils. Accepted for publication in European Journal of Environmental and Civil
Engineering.
Bendahmane, F., Marot, D., Rosquoët, F. and Alexis. A. (2008). Experimental parametric
study of suffusion and backward erosion. International Journal of Geotechnical and
Geoenvironmental Engineering, 134(1):57-67.
Bezuijen, A., Klein-Breteller, and Bakker, K.J. (1987) Design criteria for placed block
revetments and granular filters. Proceedings of the 2nd international conference on coastal
& port engineering in developing countries, Beijing, China.
Bonelli, S. and Brivois. O. (2008). The scaling law in the hole erosion test with a constant
pressure drop. Int. J. Numerical Methods in Engineering, 1573-1595.
Bonelli, S, Olivier, B. and Damien, L (2007). The scaling law of piping erosion. Proceedings
18th Congress Français Mechanique, Grenoble.
Bowles, D.S., Anderson, L.R., Glover, T.F. and Chauhan, S.S. (2003). Dam safety decision-
making: Combining engineering assessments with risk information. Proceedings of the
2003 USSD Annual Meeting, Charleston, SC.
Brauns, J. (1985), Erosionsverhalten geschichteten Bodens bei horizontaler Durchstromung,
Wasserwirtschaft, vol 75, p. 448-453.
Brauns, J. and Witt, K. J. (1987). Proposal for an advanced concept of filter design, Proc. 9th
European Conf. on Soil Mechanics and Foundation Engineering, Dublin.

___________________________________________________________________________
138
Bridle, R. (2007). Assessing the vulnerability of dams to internal erosion. Internal Erosion of
Dams and their Foundations. Editors R.Fell and J-J Fry. Taylor and Francis, London. 25-
34.
Bridle, R. (2008). Assessing the vulnerability of a typical British embankment dam to internal
erosion. 15th Conference British Dam Society, University of Warwick, Thomas Telford
London 13-28.
Bridle, R., Delgado, F. and Huber, N.P. (2007). Internal erosion: continuation and filtration:
current approaches illustrated by a case history. Assessment of the Risk of Internal Erosion
of Water Retaining Structures: Dams, Dykes and Levees. Intermediate Report of the
Working Group of ICOLD Deutsches Talsperren Komitee, Technical University of
Munich, Nr.114 ISBN 1437-3513, ISBN 978-3-9404476-04-3 70-84
Brown, A.J. and Aspinall, W.P. (2004). Use of expert opinion elicitation to quantify the
internal erosion process in dams. Proc 13th British Dam Society Conference, Canterbury.
Thomas Telford, 282-297
Brown, A.J. and Gosden, J.D. (2004) Interim guide to quantitative risk assessment for UK
reservoirs. Thomas Telford London ISBN 0 7277 3267 6
Brown A.J. and Gosden J.D. (2006) Preliminary feedback on the Interim Guide to
Quantitative Risk Assessment for UK reservoirs. Proc 14th Conference British Dam
Society. University of Durham, Improvements in reservoir construction, operation and
maintenance. Thomas Telford, 239-250
Brown, A.J. and Gosden, J.D. (2008). Defra research into internal erosion. 15th Conference
British Dam Society, University of Warwick, Thomas Telford London 39-50.
Brown, A.J. and Tedd, P. (2003). Probability of a safety incident at an embankment dam
based on historical data. J Hydropower and Dams. Issue 2, 122-126
Brown, A.J., Yarwood, G., King, S. and Gosden, J.D. (2008). Application of the Interim
Guide to Quantitative Risk Assessment across multiple dam owners by multiple Jacobs
offices. British Dam Society 15th Biennial Conference, Warwick, Thomas Telford, 65-79
Bui, H., Fell, R. and Song, C. (2004). Two and three dimensional numerical modeling of the
potential for cracking of embankment dams during construction. UNICIV Report No. 426,
The School of Civil and Environmental Engineering, The University of New South Wales,
Sydney.
Bui, H., Tandrijana, V., Fell, R., Song, C. and Khalili, N. (2005). Two and three dimensional
numerical modeling of the potential for cracking of embankment dams-supplementary
report UNICIV Report No. 438, The School of Civil and Environmental Engineering, The
University of New South Wales, Sydney.
Burenkova, V.V. (1993). Assessment of suffusion in non-cohesive and graded soils.
Proceedings, the First International Conference “Geo-Filters”, Karlsruhe, Germany, 20–
22 October 1992, Filters in Geotechnical and Hydraulic Engineering, Brauns, Heibaum &
Schuler (eds), Balkema, Rotterdam, 357-360.
CFGB, (1997). Internal erosion: typology, detection and repairs. Barrages & Réservoirs N° 6
- French Committee on Large Dams
Chang, D.S., Zhang, L. and Xu, T.H. (2012) Laboratory investigation of initiation and
development of internal erosion in soils under complex stress states. Proc ICSE-6 (6th
International Conference on Scour and Erosion) Paris, August 27-31, Société
Hydrotechnique de France, Paris, Paper 58, 895-902.

___________________________________________________________________________
139
Chapius, R.P. (1986a). Use of rotational device on cohesive soils. Transportation Research
record 1089.
Chapius, R.P. (1986b). Quantitative measurement of the scour resistance of natural soil clays.
Canadian Geotechnical Journal, 23, 132-141.
Charles, J.A. (1997). General Report: Special problems associated with earthfill dams.
Nineteenth International Congress on Large Dams, Florence. GR Q73, Vol II, 1083-1198.
International Commission on Large Dams, Paris.
de Graauw A., van der Meulen T. and van der Does de Bye M. (1983). Design criteria for
granular filters. Deflt Hydr. 287 (January).
de Wit, G.N., Sellmeijer, J.B. and Penning, A. (1981). Laboratory tests on piping. Proc. 10th
Int. Conf. Soil Mechanics and Foundation Engineering, Stockholm, June 1981. Balkema,
Rotterdam, 517-520.
Delgado, F., Huber, N.P., Escuder, I. and De Membrillera, M.G. (2006). Revised criteria for
evaluating granular filters in earth and rockfill dams. Transactions of the 22nd Congress
International Commission on Large Dams Barcelona Q86 R29 Vol III, 445-456.
Den Adel, H., Koender, M.A. and Bakker, K.J. (1994). The analysis of relaxed criteria for
erosion-control filters. Can. Geotech. J., vol 31(6), 829-840.
Dyer, Mark, Utili, Stefano, and Zielinski, Marcin, (2007). The influence of desiccation fine
fissuring on the stability of flood embankments. FRMRC Research Report UR11 (available
from: www.floodrisk.org.uk)
EBL Kompetanse AS (2005). Stability and breaching of embankment dams; Report on sub-
project 2: Stability of shell and downstream toe during large through-flow. Publikasjon
186-2005 ISBN 82-436-0521-5
Eddleston, M., Gardiner, K.D. and Cyganiewicz, J. (2010). The use of the Unified Method of
Risk Analysis on UU dams. British Dam Society Conference, University of Stathclyde.
Managing Dams Challenges in a time of Change, The British Dam Society 241-253.
Eddleston, M., Rigby P., Margret R., Mason P. J., Gardiner K. D. and Cyganiewicz J.,
(2011). Application of ‘A Unified Method for Estimating Probabilities of Failure of
Embankment Dams by internal erosion and piping’ – A UK Perspective. United States
Society on Dams, 2011 Annual Conference. San Diego, California, 1033-1043.
Engemoen, W.O. (2011). Bureau of Reclamation experiences with internal erosion incidents.
Proc. of Workshop on Internal Erosion in embankment dams and their foundations, Brno,
Czech Republic, April 2011. Institute of Water Structures, FCE, BUT, Brno, 11-18.
Engemoen, W. O. (2012a). Personal Communication
Engemoen, W. O. (2012b) Determining when to remediate an embankment for potential
internal erosion concerns. Proc ICSE-6 (6th International Conference on Scour and
Erosion) Paris, August 27-31, Societe Hydrotechnique de France, Paris, Paper 78, 1481-
1488.
Engemoen, W. O. (2012b) Methods to mitigate internal erosion risks in embankment dams.
Proc ICSE-6 (6th International Conference on Scour and Erosion) Paris, August 27-31,
Societe Hydrotechnique de France, Paris, Paper 77, 1567-1575.
Engemoen, W.O. and Redlinger, C.G. (2009). Internal erosion incidents, Bureau of
Reclamation Dams, Proc. USSD Annual Conference, Nashville Tennessee.

___________________________________________________________________________
140
Environment Agency (2013). UK Risk Assessment and Reservoir Safety. In press,
http://evidence.environment-agency.gov.uk/FCERM
Fell, R. and Fry, J-J (2007) Editors Internal Erosion of Dams and their Foundations. Taylor
and Francis, London, ISBN: 978-0-415-43724-0.
Fell, R. and Fry, J-J. (2007). The state of the art of assessing the likelihood of internal erosion
of embankment dams, water retaining structures and their foundations. In Internal Erosion
of Dams and their Foundations. Editors R.Fell and J-J Fry. Taylor and Francis, London,
1-24.
Fell, R., MacGregor, P., Stapledon, D. and Bell, G. (2005) Geotechnical Engineering of
Dams. Balkema, Leiden ISBN 04 1536 440 x
Fell, R., Foster, M., Davidson, R., Cyganiewicz, J., Sills, G. and Vroman, N. (2008). A
Unified Method for Estimating Probabilities of Failure of Embankment Dams by Internal
Erosion and Piping. UNICIV Report R 446, The School of Civil and Environmental
Engineering, University of New South Wales, Sydney, Australia 2052.ISBN:85841 413 9
Fell, R. and Wan, C.F. (2005). Methods for estimating the probability of failure of
embankment dams by internal erosion and piping in the foundation and from embankment
to foundation. UNICIV Report No R-436, The University of New South Wales, Sydney,
Australia 2052.ISBN: 85841 403 1.
Fell, R., Wan, C.F., Cyganiewicz, J. and Foster, M.A. (2001). The time for development and
detectability of internal erosion and piping on embankment dams and their foundations.
UNICIV Report No. R-399, ISBN: 84841 366 3. School of Civil and Environmental
Engineering, University of New South Wales.
Fell, R., Wan, C.F., Cyganiewicz, J. and Foster, M.A. (2003). Time for development of
internal erosion and piping in embankment dams. ASCE Journal of Geotechnical and
GeoEnvironmental Engineering, Vol. 129, No.4, 307-314.
Fell, R., Wan, C.F. and Foster M. A. (2007). Assessment of the likelihood of initiation of
erosion in embankment dams. In Internal Erosion of Dams and their Foundations. Editors
R.Fell and J-J Fry. Taylor and Francis, London, 71-102.
FEMA (2005a). Impacts of animals on Earthen Dams, US Federal Emergency Agency
Report 473
FEMA (2005b). Technical Manual for Dam Owners, Impacts of Plants on Earthen dams, US
Federal Emergency Management Agency Report 534
FEMA (2006). Conduits through Embankment Dams. Best practices for design, construction,
problem identification and evaluation, inspection, maintenance, renovation and repair. US
Federal Emergency Management Agency.
FEMA (2011). Filters for embankment dams, best practices for design and construction. US
Federal Emergency Management Agency (http://www.fema.gov)
Fong, F.C. and Bennett, W.J. (1995). Transverse cracking on embankment dams due to
earthquakes. Presented at the 1995 ASDSO Western Regional Conference, Red Lodge,
Montana.
Forster, I.R. and MacDonald, R.B. (1998). Post-Earthquake Response Procedures for
Embankment Dams - Lessons from the Loma Prieta Earthquake. ANCOLD Bulletin No.
109, 46-64.

___________________________________________________________________________
141
Foster, M.A. (1999). The probability of failure of embankment dams by internal erosion and
piping. PhD thesis, School of Civil and Environmental Engineering, The University of
New South Wales.
Foster, M.A. (2007). Application of no, excessive and continuing erosion criteria to existing
dams. In Internal Erosion of Dams and their Foundations, Editors R Fell and J-J Fry,
Taylor Francis, 103-114.
Foster, M.A. and Fell, R. (1999a). Assessing Embankment Dam Filters which do not Satisfy
Design Criteria. UNICIV Report No. R376. School of Civil and Environmental
Engineering. The University of New South Wales. ISBN 85841 343 4.
Foster, M.A. and Fell, R. (1999b). A Framework for Estimating the Probability of Failure of
Embankment Dams by Piping Using Event Tree Methods. UNICIV Report No. R377.
School of Civil and Environmental Engineering, The University of New South Wales.
ISBN 85841 343 4.
Foster, M. and Fell, R. (2001). Assessing embankment dams filters which do not satisfy
design criteria. J. Geotechnical and Geoenvironmental Engineering, ASCE, Vol.127, No.4,
May 2001, 398-407.
Foster, M., Fell, R. and Spannagle, M. (2000(a)). The statistics of embankment dam failures
and accidents, Canadian Geotechnical Journal, Vol.37, No.5, National Research Council
Canada, Ottawa, 1000-1024, ISSN: 0008-3674.
Foster, M., Fell, R. and Spannagle, M. (2000(b)). A method for assessing the relative
likelihood of failure of embankment dams by piping. Canadian Geotechnical Journal,
Vol.37, No.5, National Research Council Canada, Ottawa, pp.1025-1062, ISSN: 0008-
3674.
Foster, M., Spannagle, M. and Fell, R. (1998). Report on the analysis of embankment dam
incidents. UNICIV Report No.R374, School of Civil and Environmental Engineering,
University of New South Wales ISBN: 85841 349 3; ISSN 0077-880X.
Fry, J-J (2007). Current practice and new view on granular filter. Intermediate Report of the
European Working Group of ICOLD. Freising, Germany, 60-69.
Garner, S.J. and Fannin, R.J. (2010). Understanding internal erosion: a decade of research
following a sinkhole event, International Journal on Hydropower and Dams, 17(3), 93-98.
Gillon, M.D. (2007). Re-evaluation of internal erosion incidents at Matahina Dam, New
Zealand. In Internal erosion of dams and their foundations, Editors R Fell and J-J Fry,
Taylor Francis, 115-132
Giroud, J-P (2003). Filter criteria. Anniversary of K. Terzaghi "Erdbaumechanik" ("Soil
Mechanics"), H. Brandl (ed.), Reports of the Institute for Soil Mechanics and Geotechnical
Engineering, TU Vienna, Austria, 221-229.
Giroud, J-P (2010). Development of criteria for geotextiles and granular filters. Terzaghi
Lecture. Proceedings 9th International Conference on Geosynthetics, Guaruja, Brazil,
May. (http://www.igs-uk.org/downloads.php)
Glynn, M.E. and Kuszmaul, J. (2004). Prediction of piping erosion along middle Mississippi
River levees - an empirical model. US Army Corps of Engineers, Engineering Research
Center, Geotechnical and Structural Laboratory, ERDC/GSL TR-04-12.
Glynn, E., Quinn, M. and Kuszmaul, J. (2012). Predicting piping potential in levees. Proc
ICSE-6 (6th International Conference on Scour and Erosion) Paris, August 27-31, Societe
Hydrotechnique de France, Paris, Paper 71, 1473-1780.

___________________________________________________________________________
142
Goltz, M., Etzer T., Aufleger, S. and Muckenthaler, P. (2009). Assessing the critical seepage
velocity causing transport of fine particles in embankment dams and their foundations.
Long Term Behavior of Dams Proc. 2nd Int. Conf. 12-13 October 2009 Graz Austria 479-
484
Guidoux, C., Faure, Y.H., Beguin, R. and Ho, C.C. (2010). Contact erosion at the interface
between granular filter and various base-soils with tangential flow, Journal of
Geotechnical and Geoenvironmental Engineering, Vol. 136(5), 655-775.
Hanson, G.J. (1990). Surface erodibility of earthen channels at high stresses. Part II,
developing an in-situ device. Transactions American Society of Agricultural Engineers,
33(1), 132-137.
Hanson, G.J. (1991). Development of a jet index to characterize erosion resistance of soils in
earthen spillways. Transactions American Society of Agricultural Engineers, 36(5), 2015-
2352
Hanson, G.J. and Cook, K.R. (2004). Apparatus test procedure and analytical methods to
measure soil erodibility in-situ. Applied Engineering in Agriculture, 20(4), 455-462
Harder, L.F. Jr., Bray, J.D., Volpe, R.L., Rodda, K.V. (1991). Performance of Earth Dams
during the Loma Prieta Earthquake. Proceedings, Second International Conference on
Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, Paper No.
LP05.
Hartford D.N.D. and Baecher G.B. (2004). Risk and Uncertainty in Dam Safety. Thomas
Telford, London.
Høeg, K., Johansen, P.M., Kjærnsli, B. and Torblaa, I. (1998). Internal erosion in
embankment dams. Research project sponsored by Norwegian Electricity Federation
(EnFO). (In Norwegian with English summary).
Hoffmans, G.J.C.M. (2013). The influence of laminar flow on piping. Deltares, Netherlands
(to be published).
Holtz, R.D. and Kovacs, W.D. (1981). Introduction to Geotechnical Engineering. Prentice
Hall.
Hunter, G. (2003). The deformation behavior of embankment dams and landslides in natural
and constructed soil slopes. PhD Thesis. School of Civil and Environmental Engineering,
The University of New South Wales.
Hunter, G. and Fell, R. (2003). The deformation behavior of embankment dams. UNICIV
Report No. R-416. ISBN: 0077-880X, School of Civil and Environmental Engineering, The
University of New South Wales.
ICOLD (1974). Lessons from Dam Incidents. International Commission on Large Dams,
Paris.
ICOLD (1983). Deterioration of Dams and Reservoirs. International Commission on Large
Dams, Paris.
ICOLD (1994). Embankment dams; granular filters and drains. Bulletin 95, International
Commission on Large Dams, Paris.
ICOLD (1995). Dam Failures Statistical Analysis. Bulletin 99, International Commission on
Large Dams, Paris.

___________________________________________________________________________
143
ICOLD (2005). Risk Assessment in Dam Safety Management: a reconnaissance of Benefits.
Methods and Current Applications. Bulletin 130, International Commission on Large
Dams, Paris.
ICOLD (2008). Weak rocks and shales in dams. Bulletin 134, International Commission on
Large Dams, Paris.
ICOLD (2011). Reservoirs and Seismicity – State of Knowledge. Bulletin 137, International
Commission on Large Dams, Paris
Indraratna, B, and Raut, A.K. (2006). Enhanced criterion for base soil retention in
embankment, J. Geotechnical and Geoenvironmental Engineering, 132 (12) 1621-1627.
Indraratna, B., Raut, A. K. and Khabbaz, H. (2007). Constriction-based retention criterion for
granular filter design. Journal of Geotechnical and Geoenvironmental Engineering, 133,
266-276
Istomina, V.S. (1957). Filtration stability of soils (in Russian). Gostroizdat, Moscow.
Jairaj, V. and Wesley, L. D. (1995). Construction of a chimney drain using bio-polymer
slurry at Hays Creek dam. International Water Power & Dam Construction, February, pp
30-33.
Johnston, T.A., Millmore, J.P., Charles, J.A. and Tedd, P (1999). An engineering guide to the
safety of embankment dams in the United Kingdom. Second edition. BRE Construction
Research Communications Ltd, Watford
KBR (2002) Reservoir safety – Floods and Reservoirs safety integration. Research contract
for UK Government (Defra). Presented in 3 volumes. Copy available on request from
Defra
Kenney, T.C. and Lau, D. (1985). Internal stability of granular filters. Canadian Geotech. J.,
Vol. 22, 215-225.
Kenney, T.C. and Lau, D. (1986). Closure to: internal stability of granular filters. Canadian
Geotech. J., Vol. 23, 420-423.
Kenney, T.C. and Westland, J. (1993). Laboratory study of segregation of granular materials.
In filters in Geotechnical and Hydraulic Engineering. Editors Brauns, Helbaum and
Schuler. Balkema, Rotterdam, 313-319.
Kenney, T.C., Chahal, R., Chiu, E., Ofoegbu, G.I., Omange, G.N. and Ume, C.A. (1985)
Controlling construction sizes of granular filters Canadian Geotech. J., Vol. 22, 32-43
Kjaernsli, B., Valstad, T. and Hoeg, K. (1992). Rockfill dams, design and construction.
Norwegian Institute of Technology, Oslo.
Koenders, M.A. and Sellmeijer, J.B. (1992). Mathematical model for piping. Journal of
Geotech. Eng., ASCE, Vol.118, No.6, 943-946.
Kozeny, J. (1953). Hydraulik: Ihre Grundlagen und praktische anwendung. Springer-Verlag,
Vienna.
Lafleur, J., Mlynarek, J. and Rollin, A.L. (1993). Filter criteria for well graded cohesionless
soils. Proceedings, the First International Conference “Geo-Filters”, Karlsruhe,
Germany, Filters in Geotechnical and Hydraulic Engineering, Brauns, Heibaum and
Schuler (editors), 1993 Balkema, Rotterdam, 97-106.
Lawrence (2002). Summary of outcomes of “Workshop on cracking of homogeneous earth
dams in semi-arid environments” Maricopa County, USA.

___________________________________________________________________________
144
Lawton E.C., Fragaszy, R.J. and Hetherington, M.D. (1992). Review of wetting induced
collapse in compacted soil, ASCE Geotech, Vol 118, No 9, 1376 – 1394
Li, M. and Fannin, R.J. (2008). Comparison of two criteria for internal stability of granular
soil. Canadian Geotechnical Journal, Vol 45 pp 1303-1309
Li Nenghui (2012). Dam failure records in China. Report in personal communication to R
Bridle.
Lim, S.S. (2006). Experimental investigation of erosion in variably saturated clay soils. PhD
thesis. School of Civil and Environmental Engineering, University of New South Wales,
Sydney, Australia.
Lim, S.S. and Khalili, N.K. (2010). Laboratory measurement of erosion of clay soils using
rotating cylinder test. Geotechnical Testing Journal, Vol.32, No 3.
Marot, D., Alexis, A. and Bendahmane, F. (2007). A specific triaxial device for the study of
internal erosion in cohesive soils. In Internal erosion of dams and their foundations,
Editors R Fell and J-J Fry, Taylor Francis, 159-166.
Marot D., Regazzoni P.L. and Wahl, T. (2011). Energy based method for providing soil
surface erodibility rankings. Journal of Geotechnical and Geoenvironmental Engineering
(ASCE), Vol. 137 N°12, pp. 1290 – 1294.
Marot D., Le V.D., Garnier J., Thorel L. and Audrain, P. (2012). Study of scale effect in an
internal erosion mechanism. European Journal of Environmental and Civil Engineering,
Vol.16, No.1, pp.1–19.
Marsland, A. and Cooling, L. F. (1958). Tests on full scale clay flood bank to study seepage
and the effects of overtopping. Building Research Station Internal Report C562, Watford.
Messerklinger, S. and Straubhaar, R. (2011). Filter Design for a High Earth Core Rockfill
Dam made of Moraine. Proc. 79th ICOLD Annual Meeting, Lucerne, Switzerland: 229-
238.
Milligan, V. (2003). Some uncertainties in embankment dam engineering. Journal of
Geotechnical and Geoenvironmental Engineering (ASCE) Vol 129 No 9 785-797
Mitchell, J.K. (1993). Fundamentals of soil behavior. John Wiley & Sons.
Moffat, R.M., Fannin, R.J. and Garner, S.J. (2011). Spatial and temporal progression of
internal erosion in cohesionless soil. Canadian Geotechnical Journal, Vol 48 399-412
Moffat, R.M. and Fannin, R.J. (2011). A hydromechanical relation governing the internal
stability of cohesionless soil. Canadian Geotechnical Journal, Vol 48 413-424
Monnet, A. (1998). Boulance, erosion interne, renard. Les instabilities sous ecoulement.
Revue Francais de Geotechnique, Vol. 82, pp 3-10.
Muckenthaler, P. (1989). Hydraulische Sicherheit von Staudämmen. Institut für Wasserbau
und Wassermengenwirtschaft, Technische Universität München, Bericht Nr. 61.
Newman, S. and Foster, M. (2006). Lake Buffalo Dam risk reduction upgrade. ANCOLD
Bulletin No 132, 45-56. Australian National Committee on Large Dams, Brisbane.
Nilsson, A. (2007a). The susceptibility of internal erosion in the Suova dam. In Internal
Erosion of Dams and their Foundations, Editors R Fell and J-J Fry, Taylor Francis, 167-
172.
Nilsson, A. (2007b). Filters and internal erosion in Swedish dams. In Internal Erosion of
Dams and their Foundations, Editors R Fell and J-J Fry, Taylor Francis, 173-178.

___________________________________________________________________________
145
NGI (1984). Jukla Sekundærdam Lkkaasjer, Reparasjoner, Påvisning av erosjonskanaler
Gjennom Tetningskjernen (Jukla Saddle Dam, Leakages, Rehabilitations, Documentations
of Piping Through the Core). Internal Report 53600-1, Norges Geotekniske Institutt, Oslo.
(In Norwegian).
Park, Y. (2005) Investigation of the ability of filters to stop erosion through cracks in dams.
PhD dissertation, Virginia Polytechnic Institute, Blacksburg, Virginia, USA.
Pavchich (1961) Transition filters. BE Vedeneev St Petersburg in English 8 pages.
Pells, S. and Fell, R. (2002). Damage and Cracking of Embankment Dams by Earthquakes,
and the Implications for Internal Erosion and Piping. UNICIV Report No. R-406, School of
Civil and Environmental Engineering, The University of New South Wales, ISBN: 85841
375 2.
Pells, S. and Fell, R. (2003). Damage and Cracking of Embankment Dams by Earthquake and
the Implications for Internal Erosion and Piping. Proceedings 21st Congress on Large
Dams, Montreal. Q83-R17, International Commission on Large Dams, Paris.
Perzlmaier, S. (2007). Hydraulic criteria for internal erosion in cohesionless soil. In Internal
Erosion of Dams and their Foundations. Editors R. Fell and J-J Fry. Taylor and Francis,
London. 179-190.
Pravedny G.H. (1966). Design and selection of filter grain size distribution for transition
zones of dams. Energia, Moscow (in Russian).
Radchenko, V., Belkova, I. and Rumyantsev, O. (2012) The Russian practice of earth dam
seepage strength. Proc ICSE-6 (6th International Conference on Scour and Erosion) Paris,
August 27-31, Societe Hydrotechnique de France, Paris, Paper 281, 577-584.
Raut, A.K. and Indraratna, B. (2008). Further advancement in filtration criteria through
constriction-based techniques. ASCE J. Geotechnical and Geoenvironmental Engineering,
Vol 134, No 6, 883-887.
Redlinger, C.G., O’Leary, T.M., Henn, K.E. and Scott, G.A. (2011) Current methodology and
research at the Corps of Engineers for assessing internal erosion risk at dams. In Internal
Erosion in Embankment Dams and their Foundations. J-J Fry, J Riha and T Julinek,
editors, Proceedings of the Institute of Water Structures, FCE, BUT, Brno 123-130.
Regazzoni P-L and Marot D. (2011). Investigation of interface erosion rate by Jet Erosion
Test and statistical analysis. European Journal of Environmental and Civil Engineering,
Vol. 15, N°8, pp. 1167-1185.
Rice J D (2007). Study on the long-term performance of seepage barriers in dams. PhD
Thesis, Virginia Polytechnic Institute and State University.
Rönnqvist, H. (2007). Assessing potential for internal erosion in glacial moraine core
embankment dams. Dam Engineering, Vol. XVIII, Issue 2, 101-117.
Rönnqvist, H. (2008). Review of moraine core dams and internal erosion. Dam Engineering,
Vol. XIX, Issue 2, 99-121.
Rönnqvist, H. (2009). Long-term behavior of internal erosion afflicted dams comprising
broadly graded soils. Dam Engineering, Vol. XX, Issue 2, 149-197.
Rönnqvist, H. (2010). Predicting surfacing internal erosion in moraine core dams. Licentiate
thesis, KTH Land and Water Resource Engineering, Stockholm.
Rumyantsev, O. A., Belkova, I. N., Burenkova, V. V. and Radchenko, V. G. (2012). Russian
design methods. Report of working group 13pp. Available from ICOLD website.

___________________________________________________________________________
146
Salehi Sadaghiani M. R. and Witt K. J. (2012) Analysis of internal stability of widely graded
soils based on identification of mobile grains. Proc ICSE-6 (6th International Conference
on Scour and Erosion) Paris, August 27-31, Societe Hydrotechnique de France, Paris,
Paper 50, 257-264.
Salehi Sadaghiani M. R., Witt K. J. and Jentsch H (2012) A statistical approach to identify
mobile particles of widely graded soils. Proc ICSE-6 (6th International Conference on
Scour and Erosion) Paris, August 27-31, Societe Hydrotechnique de France, Paris, Paper
49, 839-845.
Sail, Y., Marot D., Sibille, L. and Alexis, A. (2011). Suffusion tests on cohesionless granular
matter. European Journal of Environmental and Civil Engineering, Vol. 15, N°5.
Schaefer, J.A., Redlinger, C.G. and O’Leary, T.M. (2011). Recent U.S. Army Corps of
Engineers Dam Case Histories of Internal Erosion. In Internal Erosion in Embankment
Dams and their Foundations, J-J Fry, J Riha and T Julinek, editors, Proceedings of the
Institute of Water Structures, FCE, BUT, Brno.
Scheuermann, A., Vardoulakis, I., Papanastasiou, P. and Stavropoulou, M. (2002). A Sand
Erosion Problem in Axial Flow Conditions on the Example of Contact Erosion due to
Horizontal Groundwater Flow, in Wolfgang Ehlers, ed., IUTAM Symposium on Theoretical
and Numerical Methods in Continuum Mechanics of Porous Materials', Springer
Netherlands, 169-175.
Schmertmann, J.H. (2000). The non-filter factor of safety against piping through sands. ASCE
Geotechnical Special Publication No. 111, Judgment and innovation. Edited by F. Silva
and E. Kavazanjian, ASCE, Reston.
Schmitz, S. (2007). Zur hydraulischen Kontakterosion bei bindigen Basiserdstoffen. PhD
thesis, Universität der Bundeswehr, Munich, Germany.
Schuler, U. and Brauns, J. (1993). Behavior of coarse and well graded filters. In Filters in
Geotechnical and Hydraulic Engineering, Editors Brauns, Helbaum and Schuler, Balkema,
Rotterdam, 3-17.
Schuler, U. and Brauns, J. (1997). The safety of geotechnical filters. Hydropower and dams,
Issue 6, 72-74
Sellmeijer, J.B. (1988). On the mechanism of piping under impervious structures. Doctorate
dissertation, Technische Universiteit Delft.
Sellmeijer, J.B. and Koenders, M.A. (1991). A mathematical model for piping. Applied
Mathematical Modelling, Vol.115, 646-661. Butterworth-Heinemann.
Sellmeijer, J.B., Lopez De La Cruz, J., Van Beek, V. M. and Knoeff, J.G. (2012). Fine tuning
of the backward erosion piping model through small scale, medium scale and IJKdijk
experiments. Accepted for publication in European Journal of Environmental and Civil
Engineering.
Shannon and Wilson, Inc. (1995). Summary Data Report-Inspection, Testing, and
Performance of Underseepage Control Systems - Mississippi River Levees, High Waters of
1973 and 1995, Volumes I and II, prepared for U.S.Army Corps of Engineers, St. Louis
District.
Sherard, J.L. (1973). Embankment Dam Cracking. In Embankment Dam Engineering Ed
Hirschfield and Poulos. John Wiley and Sons.

___________________________________________________________________________
147
Sherard, J.L., Decker, R.S. and Ryker, N.L. (1972). Piping in earth dams of dispersive clay.
Proceedings, Specialty Conference on Performance of Earth and Earth-Supported
Structures, ASCE, Vol.1, Part 1, pp. 589-626.
Sherard, J.L. (1979). Sinkholes in dams of coarse, broadly graded soils. 13th Int. Congress on
Large Dams, New Delhi. Q47, R2, 325-334. International Commission on Large Dams,
Paris.
Sherard, J.L. (1985). Hydraulic Fracturing in Embankment Dams. In Seepage and Leakage
from Dams and Impoundments. ASCE Geotechnical Engineering Division Conference.
Sherard, J.L. (1986). Hydraulic Fracturing in Embankment Dams. ASCE J. Geotech Eng.,
112, GT10, 905-927.
Sherard, J.L., Decker, R.S. and Ryker, N.L. (1972). Hydraulic Fracturing in Low Dams of
Dispersive Clay. Proc. Specialty Conf. on Performance of Earth and Earth Supported
Structures, ASCE, 1, 1, 653-689.
Sherard, J.L. and Decker, R.S. (1977). Summary evaluation of Symposium. ASTM STP 623,
pp 467-479
Sherard, J.L., Dunnigan, L.P. and Talbot, J. (1984). Basic properties of sand and gravel
filters. ASCE J Geotech Eng. Vol 110, No 6.
Sherard, J.L. and Dunnigan, L.P. (1985). Filters and leakage control. In Proceedings
Symposium on Seepage and Leakage from Dams and Impoundments. ASCE, R.L. Volpe
and W.E. Kelly Editors, 1-30.
Sherard, J.L. and Dunnigan, L.P. (1989). Critical filters for impervious soils. J. Geotech. Eng.
ASCE, Vol.115, No.7, 927-947.
Sherard, J.L., Dunnigan, L.P., Decker, R.S. and Steel, E.F. (1976) Pinhole test for identifying
dispersive soils. Journal of the Geotechnical Engineering Division, ASCE, 102 (GT1), 69-
85.
Sherard, J.L., Woodward, R.J., Gizienski, S.F. and Clevenger, W.A. (1963). Earth and Earth
Rock Dams. Chapter 2 Part 5. John Wiley and Sons, New York.
Shields, A. (1936). Anwendung der Ähnlichkeitsmechanik und der Turbulenzforschung auf
die Geschiebebewegung. Preusische Versuchsanstalt für Wasserbau und Schiffbau, Berlin,
Heft 26.
Sills, G.L. and Vroman, N (2007). A review of Corps of Engineers levee seepage practices in
the United States. In Internal Erosion of Dams and their Foundations, Editors R Fell and
J-J Fry, Taylor Francis, 209-218.
Silvis, F. (1991). Verificatie piping model; Proeven in de Deltagoot. Evaluatierapport.
Rapport Grondmechanica Delft, CO 317710/7
Skempton, A.W. and Brogan, J.M. (1994). Experiments on piping in sandy gravels.
Geotechnique 44, No.3, 449-460.
Solvik, O. (1991). Through flow and stability problems in rockfill dams exposed to
exceptional loads. 17th ICOLD Congress on Large Dams, Q67, R20, Vienna. ICOLD
Paris.
Solvik, O. (1995). Stenfyllningsdammars stabilitet vid genomstrőmning, VASO Dam
Committee, Report No.17.
Sutherland K J (2002). Quantifying and controlling segregation in earth dam construction
MASc Thesis, Univ of Toronto, Dept Civil Engineering, Toronto.
___________________________________________________________________________
148
Sun, B.C.B. (1989). Internal stability of clayey to silty sands. PhD Thesis, Department of
Civil Engineering, University of Michigan.
Talbot, J.R. (1994). The mechanics of cracking in embankment dams. Proc. Fracture
Mechanics applied to Geotechnical Engineering, ASCE, 118-130.
Terzaghi, K. and Peck, R.B. (1948). Soil Mechanics in Engineering Practice. First edition,
Wiley International, New York.
Technical Advisory Committee (1999). Technical Report on Sand Boils (Piping). Technical
Advisory Committee on Flood Defences, The Netherlands. Road and Hydraulic
Engineering Institute, Delft.
Townsend, F.C., Bloomquist, D., Shiau, J-M., Martinez, R. and Rubin, H. (1988). Evaluation
of Filter Criteria and Thickness for Migrating Piping in Sands. Department of Civil
Engineering, University of Florida, Report for U.S. Bureau of Reclamation.
USACE (1956). Investigation of under-seepage and its control, Lower Mississippi River
levees. USACE Waterways Experimental Station, Vicksburg, Technical Memorandum No
3-424.
USACE (1984). EM 1110-3-138. Engineering and Design – Pavement Criteria for Seasonal
Frost Conditions – Mobilization Construction. US Army Corps of Engineers.
USACE (2000). Engineering Manual EM 1110-2-1913, Design and construction of levees.
US Army Corps Engineers.
USACE (2005) Engineering Technical Letter ETL 1110-2-569. Engineering Design - Design
guidance for levee under-seepage. US Army Corps of Engineers.
USBR (1979). Laboratory procedures for determining the dispersibility of clayey soils.
Report No. REC-ERC-79-10. US Bureau of Reclamation, Denver.
USBR (2004). Design Standard No 13, Protective Filters. US Bureau of Reclamation,
Denver.
USBR (2011). Binding Agents in Embankment Dam Protective Filters. Report No. DSO-11-
04. Prepared by Robert V. Rinehart and Mark W. Pabst. Bureau of Reclamation, Denver,
Colorado.
USDA-SCS (1994). Gradation Design of Sand and Gravel Filters. United States Department
of Agriculture, Soil Conservation Service, Part 633, Chapter 26, National Engineering
handbook.
Van Beek, V.M., Bezuijen, A. and Zwanenburg, C. (2010a). Piping: Centrifuge experiments
on scaling effects and levee stability. Physical Modelling in Geotechnics, Zürich, Taylor
and Francis Group.
Van Beek, V. M., Knoeff, J.G., J. Rietdijk, J., Sellmeijer, J.B. and Lopez De La Cruz, J.
(2010b). Influence of sand and scale on the piping process - experiments and multivariate
analysis. Physical Modelling in Geotechnics, Zürich, Taylor and Francis Group.
Van Beek, V., Yau, Q., Van, M. and Barends F. (2012a) Validation of Sellmeijer’s model for
backward piping under dikes on multiple sand layers. Proc ICSE-6 (6th International
Conference on Scour and Erosion) Paris, August 27-31, Societe Hydrotechnique de
France, Paris, Paper 39, 543-550.
Van Beek, V. M., Knoeff J.G. and Sellmeijer, J. B (2012b). Observations on the process of
piping by under-seepage in cohesionless soils in small, medium and full scale experiments.
Accepted for publication in European Journal of Environmental and Civil Engineering.

___________________________________________________________________________
149
Vaughan, P.R. and Soares, H.F. (1982). Design of filters for clay cores of dams. J. Geotech.
Div., ASCE. Vol. 108, GT1, 17-32.
Vaughan, P.R. and Bridle, R.C. (2004). An update on perfect filters. Long-term benefits and
performance of dams. Proceedings 13th conference of the British Dam Society and
European Club of ICOLD. University of Canterbury, Thomas Telford, London, 516-531
Verheij, H., Hoffmans, G., Dorst, K. and Vandesande, S. (2012) Interface stability of
granular filter structures under currents. Proc ICSE-6 (6th International Conference on
Scour and Erosion) Paris, August 27-31, Societe Hydrotechnique de France, Paris, Paper
225, 241-248.
Vuola, P., Konrad, J.M. and Bartsch, M (2007). Effects of frost and thaw on dams. In
Assessment of the Risk of Internal Erosion of Water Retaining Structures: Dams, Dykes
and Levees. Intermediate Report of the European Working Group of ICOLD, Editors M
Aufleger, J-J Fry, M.Goltz and S. Perzlmaier.
Wahl, T.L., Regazzoni, P. and Erdogan, Z. (2008). Determining Erosion indices of cohesive
soils with the Hole Erosion Test and Jet Erosion Test. US Department of Interior, Bureau
of Reclamation Report DSO-08-05.
Wallace, M. (1987) How to prevent frost heave. Concrete Construction, April 1987, 369-372
Wan, C.F. (2006). Experimental investigation of piping erosion and suffusion of soils in
embankment dams and their foundations. PhD thesis, School of Civil and Environmental
Engineering, The University of New South Wales, Sydney, Australia.
Wan, C.F. and Fell, R. (2002). Investigation of internal erosion and piping of soils in
embankment dams by the slot erosion test and the hole erosion test. UNICIV Report No. R-
412, ISBN: 85841 379 5, School of Civil and Environmental Engineering, The University
of New South Wales.
Wan, C.F. and Fell, R. (2004a). Investigation of rate of erosion of soils in embankment dams.
ASCE Journal of Geotechnical and GeoEnvironmental Engineering, Vol. 130, No. 4, 373-
380.
Wan, C.F. and Fell, R. (2004b). Laboratory tests on the rate of piping erosion of soils in
embankment dams. Geotechnical Testing Journal, vol.27, No.3, 295-303.
Wan, C.F. and Fell, R. (2004c). Experimental investigation of internal instability of soils in
embankment dams and their foundations. UNICIV Report No.429, School of Civil and
Environmental Engineering, The University of New South Wales, Sydney, ISBN 85841
396 5.
Wan, C.F. and Fell, R. (2007). Investigation of internal erosion by the process of suffusion in
embankment dams and their foundations. In Internal Erosion of Dams and their
Foundations. Editors R.Fell and J-J Fry. Taylor and Francis, London. 219-234.
Wan. C.F. and Fell, R. (2008). Assessing the Potential of Internal erosion and Suffusion in
Embankment Dams and Their Foundations. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, Vol 134, No3, 410-407.
Weijers, J.B.A. and Sellmeijer, J.B. (1993). A new model to deal with the piping mechanism.
Filters in Geotechnical and Hydraulic Engineering. Brauns, Herbaum and Schuler
(editors), Balkema, Rotterdam.
Wilson, J. D. (2003). Middle Mississippi River flood performance: Assessing the occurrence
of piping through empirical modelling. M.S. thesis, University of Mississippi, Oxford, MS.

___________________________________________________________________________
150
Wolff, T.F. (1974). Performance of underseepage control measures during the 1973
Mississippi River flood, Columbia Levee District, Illinois. Master’s Thesis, Oklahoma
State University.
Wolff, T.F. (2002). Performance of levee under-seepage controls: A critical review. US
Army Corps of Engineers, Engineering Research Center, Geotechnical and Structural
Laboratory, ERDC/GSL TR-02-19.
Wörman A. and Olafsdottir R. (1992). Erosion in a granular medium interface. J. Hydraul.
Res., 30(5):639-655.
Wray, S. (2006) A guide to rabbit management. CIRIA Publication C645, 58pp. ISBN 0-
86017-645-2
Zhilenkov, V.N., Radchenko, V.G. and Rumyatsev, O.A. (2011) Proposals on text for a new
ICOLD bulletin: Internal Erosion of Dams and their Foundations. In Internal Erosion in
Embankment Dams and their Foundations, J-J Fry, J Riha and T Julinek, editors,
Proceedings of the Institute of Water Structures, FCE, BUT, Brno, 204-209.

___________________________________________________________________________
151