Versuchsanstalt für Wasserbau

Hydrologie und Glaziologie

der Eidgenössischen
Technischen Hochschule Zürich

Mitteilungen 218

Hydraulics of dike breaching

Lukas Schmocker

Zürich, 2011

Herausgeber: Prof. Dr. Robert Boes

Zitiervorschlag für VAW-Mitteilungen:

Schmocker, L. (2011).
Hydraulics of dike breaching.
Mitteilungen 218, Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie (VAW),
R. Boes, Hrsg., ETH Zürich.

Im Eigenverlag der
Versuchsanstalt für Wasserbau,
Hydrologie und Glaziologie
ETH Zürich
CH-8092 Zürich

Tel.: +41 - 44 - 632 4091

Fax: +41 - 44 - 632 1192
e-mail: [email protected]

Zürich, 2011

ISSN 0374-0056
Preface I

Preface

Most dikes and dams are not designed to withstand overtopping by flowing water.
Overtopping therefore constitutes the primary failure mode of earthen embankments
besides internal erosion. The resulting breach flow is particularly dangerous and damag-
ing to the surrounding infrastructure and the environment as it usually occurs unpredict-
ably both on a temporal and a spatial scale. In flood defense projects, dike breach sce-
narios should be accounted for in the assessment of the residual risk for the overload
case, determining the emergency action plans and the associated preparedness measures.
This, in turn, requires a thorough understanding of the physical processes involved in
dike breaching due to overtopping.
In his Doctoral Dissertation Thesis, Mr. Schmocker has systematically analyzed the-
se dike breaching mechanisms by experimental investigations. In particular, he made
use of optical methods to determine dike profiles during the breaching process account-
ing for both the temporal and the spatial scale. Although most experiments relate to
sector dike models, i.e. to the two-dimensional failure, Mr. Schmocker also examined
the more important lateral breach widening in additional 3D experiments using a sophis-
ticated stereoscopic videometry system to non-intrusively acquire the breach profiles.
The results are an important step forward in the understanding of the overtopping
failure mechanisms of earthen embankments. They further represent a valuable data
source for the development of numerical simulation tools, currently undertaken in an-
other Doctoral Dissertation Thesis at VAW. These tools will allow to forecast the im-
pact of dike breaching events and to improve the management of extreme floods.
I wish to express my gratitude to the Swiss National Science Foundation, who finan-
cially supported Mr. Schmocker’s work (Grant-No. 200020-116680 and 200020-
129523). My thanks also go to Prof. Dr. W.H. Hager for supervising and reviewing the
current thesis and to Prof. Y. Zech, Université Catholique de Louvain, Louvain-la-
Neuve, Belgium, for taking over another co-review.

Zurich, October 2011 Prof. Dr. Robert M. Boes

II Preface
Acknowledgments III

Acknowledgments

This doctoral thesis was written during my work at the Laboratory of Hydraulics, Hy-
drology, and Glaciology (VAW) at the Swiss Federal Institute of Technology (ETH)
Zurich. I would like to thank all persons and organisations contributing to this study,
especially:

 My supervisor and co-examiner Prof. Dr. Willi H. Hager, who initiated this doc-
toral thesis and gave me the opportunity to work at VAW. His continued support
and advice was invaluable for the successful completion of this thesis.

 VAW Director and examiner Prof. Dr. Robert M. Boes for the continued support
and the detailed review of the thesis.

 Second co-examiner and external expert Prof Yves Zech (Université Catholique
de Louvain) for the review of this thesis and the valuable comments.

 All from VAW workshop and electronics workshop, the VAW photographer and
the graphic designers for their continuous support.

 All my working colleagues and friends for their helpful input.

 The research project was financially supported by the Swiss National Science
Foundation, grant number 200020-116680 and 200020-129523.

 Finally I would like to thank my parents Therese and Peter who enabled my ed-
ucation at ETH: Merci viu mau!

Zurich, October 2011 Lukas Schmocker

IV Acknowledgments
Table of contents V

Table of contents

Preface ............................................................................................................ I 

Acknowledgments ....................................................................................... III 

Table of contents .......................................................................................... V 

Abstract ....................................................................................................... IX 

Kurzfassung ................................................................................................ XI 

1  Introduction ............................................................................................ 1 

1.1  Initiation ............................................................................................................ 1 

1.2  Goals ................................................................................................................. 2 
1.3  Survey ............................................................................................................... 3 

2  Historical background ............................................................................ 5 

2.1  Dike - Terminology .......................................................................................... 5 

2.2  History of dike construction ............................................................................. 5 
2.3  Modern dike guidelines .................................................................................... 8 

3  Literature review .................................................................................. 11 

3.1  Introduction..................................................................................................... 11 

3.2  Hydraulic model tests in the past .................................................................... 11 
3.3  Dike failure modes .......................................................................................... 15 
3.3.1  Overview..................................................................................................... 15 
3.3.2  Overtopping failure ..................................................................................... 16 
3.3.3  Internal erosion (piping) failure and seepage ............................................. 17 
3.3.4  Geotechnical failure .................................................................................... 20 
3.4  Overtopping erosion ....................................................................................... 20 
3.4.1  Overview..................................................................................................... 20 
3.4.2  Flow regimes and erosion zones ................................................................. 21 
3.4.3  Sediment transport ...................................................................................... 22 
Flow hydraulics .................................................................................................. 22 
VI Table of contents

Incipient motion of single grain .......................................................................... 25 

Shields criterion .................................................................................................. 28 
Effect of bed slope on incipient motion ............................................................... 30 
Bed-load sediment transport formulae ................................................................ 31 
3.5  Breach process ................................................................................................. 35 
3.5.1  Overview ..................................................................................................... 35 
3.5.2  Plane breach process ................................................................................... 35 
3.5.3  Spatial breach process ................................................................................. 37 
3.5.4  Real-case data and field tests....................................................................... 41 
3.6  Breach outflow ................................................................................................ 42 
3.7  Instrumentation and measuring devices .......................................................... 44 
3.8  Scale effects ..................................................................................................... 45 
3.8.1  Model similarity .......................................................................................... 45 
3.8.2  Scaling sediment transport .......................................................................... 47 
3.9  Research gaps and purpose of present study ................................................... 48 

4  Hydraulic scale models ......................................................................... 51 

4.1  Introduction ..................................................................................................... 51 

4.2  Plane dike breach tests .................................................................................... 51 
4.2.1  Overview ..................................................................................................... 51 
4.2.2  Model channel I ........................................................................................... 51 
4.2.3  Test setup..................................................................................................... 52 
4.2.4  Bottom drainage .......................................................................................... 53 
4.2.5  Sediment ...................................................................................................... 54 
4.2.6  Discharge measurement .............................................................................. 55 
4.2.7  Measuring camera ....................................................................................... 55 
4.2.8  Experimental procedure .............................................................................. 55 
4.2.9  Test program................................................................................................ 56 
Preliminary Tests ................................................................................................ 56 
Scale effects Tests ................................................................................................ 56 
Breach process Tests ........................................................................................... 57 
4.2.10  Data analysis procedure........................................................................... 58 
4.3  Spatial dike breach tests .................................................................................. 64 
4.3.1  Overview ..................................................................................................... 64 
Table of contents VII

4.3.2  Model channel II ......................................................................................... 64 

4.3.3  Test setup .................................................................................................... 65 
4.3.4  Sediment ..................................................................................................... 66 
4.3.5  Discharge measurement .............................................................................. 66 
4.3.6  Measurement system .................................................................................. 66 
4.3.7  Experimental procedure .............................................................................. 68 
4.3.8  Test program ............................................................................................... 69 
4.3.9  Data analysis procedure .............................................................................. 70 
4.3.10  Measuring accuracy ................................................................................ 70 

5  Results .................................................................................................. 75 

5.1  Introduction..................................................................................................... 75 

5.2  Model limitations and scale effects ................................................................ 75 
5.2.1  Effects of seepage ....................................................................................... 75 
5.2.2  Test repeatability ........................................................................................ 78 
5.2.3  Sidewall effects ........................................................................................... 82 
5.2.4  Scale families .............................................................................................. 84 
5.2.5  Hydraulic conditions ................................................................................... 86 
5.3  Plane dike breach process – general observations .......................................... 93 
5.3.1  Introduction................................................................................................. 93 
5.3.2  Breach process and breach profiles ............................................................ 93 
5.3.3  Breach discharge ......................................................................................... 96 
5.3.4  Flow depths, pressure head, velocities and energy head ............................ 98 
5.3.5  Bed shear stress......................................................................................... 105 
5.3.6  Erosion volume and transport rate ............................................................ 109 
5.3.7  Bed-load sediment transport formulae...................................................... 110 
5.4  Effect of selected test parameters on dike breach process ............................ 114 
5.4.1  Introduction............................................................................................... 114 
5.4.2  Effect of dike slope So............................................................................... 114 
5.4.3  Effect of inflow discharge Qo ................................................................... 116 
5.4.4  Effect of sediment diameter d for non-cohesive material ......................... 118 
5.4.5  Effect of cohesion ..................................................................................... 122 
5.4.6  Effect of dike height w .............................................................................. 125 
5.4.7  Further parameters not tested.................................................................... 127 
VIII Table of contents

5.5  Normalized results......................................................................................... 128 

5.5.1  Dimensional analysis................................................................................. 128 
5.5.2  Evaluated tests ........................................................................................... 130 
5.5.3  Maximum breach discharge QM ................................................................ 130 
5.5.4  Maximum reservoir level hRM ................................................................... 133 
5.5.5  Maximum dike height zM ........................................................................... 134 
5.5.6  Dike volume VD ......................................................................................... 137 
5.6  Case studies ................................................................................................... 140 
5.7  Spatial dike breach process ........................................................................... 141 
5.7.1  Preliminary test.......................................................................................... 141 
5.7.2  Spatial dike breach tests ............................................................................ 144 

6  Conclusions and outlook .................................................................... 149 

6.1  Scale effects ................................................................................................... 149 

6.2  General breach process .................................................................................. 150 
6.3  Effect of test parameters on breach process .................................................. 150 
6.4  Normalized results......................................................................................... 151 
6.5  Spatial dike breach tests ................................................................................ 152 
6.6  Outlook .......................................................................................................... 152 

Notation ..................................................................................................... 155 

References ................................................................................................. 159 

Abstract IX

Abstract

Dikes along rivers protect a valley and its people and property from floods. However,
due to increased flood discharges during the past years and a lack of maintenance, over-
topping has caused large damages in the past. Recent dike failures due to overtopping
highlight the need to understand the damage processes in detail. The hydraulics of dikes
subjected to overtopping are still poorly understood despite some recent advances,
including the dike performance in terms of hazard scenarios, damages to be expected
and the risk protection which are only available mainly from standard flood risk evalua-
tion procedures. The accurate prediction of a dike breach process is essential to develop
effective emergency action plans and to design early warning systems and hazard maps.
This research project particularly adds to the understanding of the dike breach pro-
cesses due to overtopping. Hydraulic model tests were conducted to investigate the
main features of both plane and spatial dike breaching due to overtopping. Laboratory
tests of breach processes are related to scaling issues currently not understood. There-
fore, a series of plane dike breach tests due to overtopping were conducted to examine
the model limitations. The temporal dike breach progress was optically recorded to
allow for a detailed analysis of the sediment and water surfaces. A systematic variation
of both the dike dimension and the sediment diameter resulted in basic findings relative
to: (1) Test repeatability; (2) Side wall effect; and (3) Scale effects. The results indicate
definite minimum dimensions for both dike height and width, grain size and overtop-
ping discharge. Systematic plane dike breach tests further resulted in findings regarding
the (1) Breach process; (2) Effect of seepage; (3) Breach discharge; and (4) Bed shear
stresses including sediment transport. The governing parameters for the dike breach
process were identified as the dike height, the grain size and the critical flow depth.
Finally, basic spatial dike breach tests were conducted to apply a new 3D-
videometric recording system, developed by AICON 3D Systems Ltd., Germany. This
system allows for a continuous, non-intrusive recording of the sediment surface through
the water surface during a dike breach using a rectangular grid projection and a set of
four cameras. Preliminary tests provide motivating results and a feasible, non-intrusive
method seems therefore to be available for determining the sediment topography.
X Abstract
Kurzfassung XI

Kurzfassung

Deiche entlang von Flüssen schützen Siedlungen und Landschaften vor Überflutungen.
Überschwemmungen infolge Überströmung eines Deichs haben in den letzten Jahren
aufgrund der steigenden Hochwasserintensität zugenommen. Obwohl diese Überflutun-
gen immense Infrastrukturschäden und finanzielle Konsequenzen zur Folge haben, ist
die Hydraulik von Deichbrüchen nur wenig erforscht. Um Gefahrenkarten und Überflu-
tungsszenarien zu definieren und entsprechende Evakuationsszenarien zu planen, müs-
sen hydraulische Angaben über den zeitlichen und räumlichen Ablauf eines Deich-
bruchs bekannt sein. Dieses Forschungsprojekt soll mittels hydraulischen Modellversu-
chen zum Verständnis des ebenen sowie des räumlichen Deichbruchs beitragen und
Ansätze zur zeitlichen und örtlichen Entwicklung eines Deichbruchs liefern.
In einem ersten Schritt wurden ebene Deichbruchversuche infolge Überströmen
durchgeführt, um die Reproduzierbarkeit der Versuche, Wandeffekte sowie Massstabs-
effekte zu untersuchen. Bei allen Versuchen wurde der Deichbruch mittels Kamera
seitlich durch die Kanalscheibe aufgezeichnet, um die Entwicklung der Wasser- und der
Sedimentoberfläche zu analysieren. Mit der Einhaltung minimaler Deichdimensionen,
Korndurchmesser sowie Breschen-Durchflüsse können Massstabseffekte im hydrauli-
schen Modell weitgehend ausgeschlossen werden.
Weiterhin resultierten systematische, ebene Deichbruchversuche und deren Auswer-
tung in Ergebnissen hinsichtlich: (1) Bruchprozess; (2) Einfluss der Durchsickerung; (3)
Breschen-Durchfluss; sowie (4) Schubspannungen und Sedimenttransport. Die wich-
tigsten Parameter, die den Deichbruch im vorliegenden Fall beeinflussen, sind dabei die
Deichhöhe, der Korndurchmesser sowie die kritische Abflusstiefe.
Schliesslich wurden Prinzipversuche zum räumlichen Deichbruch infolge Überströ-
men durchgeführt. Dabei wurde ein neues videometrisches Messsystem der Firma
AICON 3D Systems GmbH verwendet. Das System erlaubt die berührungslose Auf-
nahme von Sohlenstrukturen durch die Wasseroberfläche mittels einer Gitterprojektion
und vier Kameras. Die kontinuierliche Aufnahme dreidimensionaler Bruchprofile wäh-
rend des Deichbruchs wurde in ersten Testversuchen erfolgreich durchgeführt.
XII Kurzfassung
Introduction 1

1 Introduction

1.1 Initiation

Dikes along rivers are designed to protect a valley including its population and property
from floods. However, flooding of river valleys is an old concern and still a number of
events occur yearly causing both monetary and human losses (Figure 1.1). Recent major
dike failures include those of e.g. the Yangtze flood in China 1998, the Elbe flood in
Germany 2002, the New Orleans flood in 2005, the Mississippi Flood in 2008, the
Pakistan flood 2010 or the Queensland flood in Australia in 2011. The main causes of
dike failures are increased flood discharges during the past years and a lack of both
maintenance and repair. Each dike is designed to manage a certain discharge and water
level without damage. Hydrological data are however only derived from a limited
amount of records during a number of years. Therefore, insufficient data, wrong calcu-
lations, readiness to take risk and design failure may result in dike overtopping. The
effect of the climate change and the resulting variations regarding discharge and water
levels can further lead to a change in magnitude, duration and frequency of flood events.
A thorough maintenance and inspection of dikes is a challenging task given their
quantity of e.g. 7,500 km of dikes throughout Germany (LAWA 1995) or 5,500 km
simply along the Mississippi River. Many dikes have not been maintained properly
during decades or were improved by simply dumping additional material on the existing
construction. In addition, especially older dikes were not constructed under strict design
criteria as they exist today and their structural basis is unknown.
Although the risks of dike breaches are generally known from historical records,
relatively few steps were undertaken so far from the hydraulic engineering community.
One of the main reasons was identified in the complexity of the dike breach phenomena.
A dike breach is affected by various processes regarding both hydraulic and geotech-
nical aspects. The breach process is influenced by seepage, overtopping erosion, sliding
failures and the various geotechnical characteristics of the dike material. The erosion
process itself is 3-dimensional and the hydraulic conditions vary constantly. Problems
in modelling dike erosion exist especially regarding: (1) Transition from sub- to super-
critical flow and vice versa; (2) Changes from wetted to dry surfaces and vice versa; and
(3) Erosion and deposition both in the streamwise and transverse directions. Item 3 is
2 Introduction

particularly complex because the current sediment transport formulae apply essentially
for uniform flow conditions only in an almost horizontal sediment bed. Dikes have
slopes of the order 1:5 (V:H) and steeper, so that the slope effect may not be neglected.
This information is important from the very beginning of the dike breach as the breach
geometry and consequently the breach discharge depend on it.
Given the enormous damage potential regarding dike breaching, there is a need to
understand the damage process of dikes in detail. As the majority of dikes breach due to
overtopping, this failure mode is investigated in detail in the current research project.

Figure 1.1 Dike breach at (a) River Elbe in 1825 (Lithograph, Friedrich Thöming), (b) Missis-
sippi near Meyer IL during 2008 flood event (Steve Bohnstedt / Quincy Herald
Whig via AP)

1.2 Goals

The goals of the present study are to investigate the dike breach processes due to over-
topping using hydraulic model tests. The main focus is on the following topics:

 Establish a simple model setup for testing the pure overtopping failure mode
 Investigate model limitations and scale effects regarding hydraulic dike breach
scale modelling
 Investigate the plane (2D) dike breach process due to overtopping in detail
 Establish novel test setup for spatial (3D) breach tests
 Preliminary 3-dimensional breach tests
To achieve these goals, two hydraulic models were built; one for plane and the other for
spatial dike breach tests. A systematic test program accounted for various dike dimen-
sions, grain sizes and hydraulic conditions. All data were systematically analysed and
normalized including results regarding scale effects, dike breach profiles, breach dis-
Introduction 3

charges, flow velocities and sediment transport. Geotechnical aspects including the
sediment characteristics, seepage effects and pore pressures were of inferior interest in
the current research. Although seepage influences the breach process, a combined dike
failure due to seepage and overtopping inhibits the detailed analyses of the overtopping
erosion process. Further on, measurements of the pore pressure during the breach tests
are complex and would increase the laboratory effort considerably.

1.3 Survey

The present study is organised as follows. Chapter 2 presents facts on the historical
background of dike construction. The literature review and the relevant state-of-the-art
are summarized in Chapter 3. The focus lies on both 2- and 3-dimensional laboratory
modelling of dike breaching due to overtopping. The identified research gaps are sum-
marized. Chapter 4 presents the hydraulic model, the experimental setup and the meas-
uring devices. The experimental results are presented in Chapter 5 starting with the
effects of seepage and scale effects in dike breach modelling. Further, general results on
a typically 2-dimensional dike breach tests are presented. The breach profiles, the
breach discharge and the sediment transport are discussed. Further, the effect of all
varied test parameters on the overall dike breach process is presented and dimensionless
parameters are introduced. The results of the 2-dimensional dike breach tests are pre-
sented in dimensionless form. Finally, the results regarding 3-dimensional dike breach
tests, the experimental setup and the general adaptability of the AICON measurement
system are discussed. The main results are summarized in Chapter 6.
4 Introduction
Historical background 5

2 Historical background

2.1 Dike - Terminology

The modern word dike is most probably derived from the Dutch word ‘dijk’, by which
the construction of dikes is well attested since the 12th century. In the old English lan-
guage, the word dik already existed and was pronounced with a hard c in northern Eng-
land and as ditch in the south. In the Netherlands as well as in England, the origin of the
word lies in digging a trench and forming the excavated soil into a bank alongside it.
‘Levee’ or ‘Embankment dam’ are similar terms used in the literature.
A dike is a man-made structure (usually an earthen embankment of earth and (or)
rock) designed and constructed in accordance with sound engineering practice to con-
tain, control, or divert water flow to provide protection from flooding for flows up to a
certain return period. Dikes are mainly used in river engineering to prevent floods or in
costal engineering as a protection from high-tide. In contrast, a dam is a barrier that
serves the primary purpose of retaining water for various purposes such as energy pro-
duction. Normally dikes are made up of loosely placed sediments like gravel, sand, silt
and clay. As they are only temporarily exposed to high water pressure, they are mostly
built without a surface seal or a core. Compared to dams, dikes are small in height but
can have lengths of several 1,000 kilometres. Further differences compared to a dam
are: (1) A dike may become saturated for only a short time (hours to a few days) beyond
the limit of capillary saturation; (2) Dike construction is dictated primarily by flood
protection requirements resulting often in constructions on poor foundation; and (3)
Dike material is often heterogeneous and not ideal as it is generally taken from the
vicinity of the watercourse subjected to floods.

2.2 History of dike construction

The first settlements next to rivers or the sea were mostly built on small hills to protect
the people from floods and high tide. When people started using the fertile soil next to
rivers and the sea, they required additional protection for the harvest and the cattle.
Therefore simple earth walls were built which can be regarded as the first dikes. The
poor construction technique and an absence of knowledge of hydrology, hydraulics and
6 Historical background

geotechnics resulted in numerous dike failures. Often the design flood was estimated on
the basis of the highest known flood or the last flood that occurred. With ongoing time
and industrialisation, river engineering gained more and more in significance. Both
technical and economical developments would not have been possible without river
works as dikes, ensuring secure navigation routes as well as bridge crossings and pro-
tecting both settlements and agricultural land from flooding (Figure 2.1).
The first dike constructions in Europe date from the early Middle Ages. Mostly ring-
shaped dikes to protect individual buildings were built in the 9th century at the Rhine
River and in the 12th century at the Elbe River (Schmidt 2000). The first dikes of the
Netherlands date from the 13th century. In the first decades, different large dike con-
structions were mainly built in Germany and the Netherlands. The quality and building
techniques were rather inferior as there were no general guidelines available. The first
dikes were set up of two walls made of simple fences then backfilled with loose soil. As
the soil mostly came from local borrow areas, dikes from the Middle Ages consist of
varied materials. The technical dike development was mainly a consequence of the
various dike failures and the resulting restorations.

Figure 2.1 Early dike construction (ca. 1928) along Mississippi River (Courtesy U.S. Army
Corps of Engineers)
Historical background 7

Theoretical considerations started in the 18th century, and new survey methods al-
lowed for the advance of new dike profiles. Albert Brahms (1692-1758) published the
first German book describing construction requirements for dikes of which some are
still valid to date (Figure 2.2). Of essential relevance was the realization that a flattened
upstream slope reduces wave impact significantly (Brahms 1757). An overview on
River Engineering in Switzerland along with the first hydraulic structures is found in
Minor and Hager (2004).

Figure 2.2 Dike chart of Friesland, Germany, including all dike cross section (Albert Brahms,
1730)
8 Historical background

Due to the simple tools and ‘machinery’ available, dike construction was physically
hard work and a tedious process. One major river engineering project was the Linth
works in Switzerland starting in 1784. The main part of the project was the diversion of
the River Linth into Lake Walen and its canalization between Lake Walen and Lake
Zurich. As there were no excavators or transport equipment available, the power of
water was used to transport the excavated materials downstream. The Linth diversion
protected the nearby towns from flood hazards and additional agricultural land was
reclaimed. Further, the Linth channel enabled a safe shipping navigation course from
Zurich to Walenstadt (Minor and Hager 2004).
A change in dike construction occurred with the steam engine era revolutionizing
machine work in the 19th century. Given the various vehicles and construction equip-
ment, larger dike systems within a shorter time were built. In Switzerland, the construc-
tion of the Jura Waterways improvement started in the late 18th century. The project
included the construction of four channels to divert and channelize the rivers around the
Jura border lakes. The excavation of e.g. the Aarberg-Hagneck canal of around
3.8 million m3 was to 45% made by hand. The remainder was washed out in the follow-
ing years. Other major river engineering projects were carried out around the world and
include e.g. the construction of the Suez Canal, opened in 1869, or the Panama Canal,
opened in 1914.

2.3 Modern dike guidelines

Dike construction has been improved continuously through the years up to the detailed
current set of guidelines (e.g. DIN 19712 1997, TAW 1991, USACE 2000a). All these
present basic principles used in the design and construction of dikes. However, the
general dike cross-section has not significantly changed over the years. Figure 2.3
shows typical dike cross-sections of (a) Kreuter (1921) and (b) DWA (2011). The gen-
eral components like the dike body, the dike crest and the berms on both the water and
the landside are noted in both plots. As old guidelines mainly specify information on the
dike cross-section or material types, modern dike guidelines include a detailed design
procedure. This mainly starts with the definition of the design water level and the de-
termination of the freeboard and consequently the required dike crest elevation. Then,
the dike cross-section is defined depending on the crest width, the dike slope and possi-
ble berms and maintenance tracks. One main aspect is the control of seepage. Either the
Historical background 9

dike body itself exhibits the sealing or additional elements as a surface sealing, a core or
a foot drainage are applied. Once the dike cross-section and the design parameters are
determined, field investigations are conducted to provide geotechnical information on
the foundation and the soil characteristics in the construction area. Then the entire anal-
ysis regarding dike stability, hydraulic safety, structural safety and usability has to be
conducted. Once construction starts, strict guidelines control the material specification,
the construction procedure, material compaction and quality control. Finally, detailed
guidelines are available regarding dike inspection, maintenance, repair and defence.

Figure 2.3 Typical dike cross-section of (a) Kreuter (1921) and (b) DWA (2011) (adapted)
10 Historical background
Literature review 11

3 Literature review

3.1 Introduction

This chapter summarizes the relevant scientific research on dike breach modelling. Only
the theories that are essential for this work are discussed in detail and alternatively the
relevant literature is indicated. The main topics discussed are: (3.2) Hydraulic model
tests in the past; (3.3) Dike failure modes; (3.4) Overtopping erosion; (3.5) Breach
process; (3.6) Breach outflow; (3.7) Measuring devices; and (3.8) Scale effects. The
research gaps and the resulting purpose of the present study are summarized in 3.9. No
literature review on the historical analyses of dike breaches, guidelines to dike construc-
tion and the numerical simulation of dike breaching are presented, however. A survey
on historical breaches is found in e.g. MacDonald and Langridge-Monopolis (1984),
Powledge et al. (1989b), Sing (1996), Foster et al. (2000), Chinnarasri et al. (2004), and
Froehlich (2008). Design and construction guidelines for dikes and levees were pub-
lished by authorities including e.g. DIN 19712 (1997), TAW (1991), or USACE
(2000a). Recent research on numerical modelling are listed in Zhu (2006), D'Eliso
(2007), Kahawita (2007), Morris (2009) and Pontillo et al. (2010).

3.2 Hydraulic model tests in the past

Table 3.1a and b give an overview on past hydraulic dike breach modelling due to
overtopping, whereas model investigations involving e.g. cohesive materials, piping
failure, seepage or surface protection measures are not considered herein. The parame-
ters are w = dike height, LK = crest length, b = dike width, So = upstream (subscript o)
dike slope, Sd = downstream (subscript d) dike slope and d = mean grain diameter
(Figure 4.2). The listed investigations differ particularly regarding the erosion process
(2D or 3D), constant or falling reservoir levels and the presence or absence of a surface
or core layer. The dike dimensions and the sediment diameters are comparable with the
present research. All investigations mainly determined the breach process, the breach
profiles and the breach discharge. An additional survey on available laboratory breach
tests is provided by Morris (2009).
12 Literature review

Research on dike erosion started once the basics of sediment transport were availa-
ble. Tinney and Hsu (1961) conducted laboratory and field experiments on the washout
of an erodible fuse plug and described the side erosion characteristics. They explained
the washout mechanisms based on the laws of sediment transport. Hydraulic model tests
on the protection of rock and earthfill dams with armouring against overtopping flow
erosion were investigated by Sarkaria and Dworsky (1968). Martins (1981) analysed the
stability of small overflow rockfill dams, thereby concentrating on downstream slope
failure. A large study on the mechanism of overflow erosion on embankments was
conducted by Powledge et al. (1989 a, b). Part I presents model and research activities
on dams and levees subjected by overtopping flow, discussing the effectiveness of
various protection systems, whereas Part II focuses on the hydraulics of water flowing
over a dike levee including distinctive erosion zones.
A general work on dam breach modelling techniques, including major recorded dam
breaches in past history and several empirical and mathematical models on dam breach-
ing, was presented by Singh (1996). Bechteler and Broich (1991) dealt with the mathe-
matical modelling of the dike breach process. The sediment transport was described
with the standard Meyer-Peter-Müller (MPM) and Engelund-Hansen (EH) equations.
The main purpose of their studies was to define the dike outflow hydrograph. Coleman
et al. (2002) conducted flume experiments on overtopping breaching of noncohesive
homogenous embankments. Their results include information on the breach erosion
process, the breach geometry and the breach discharge, allowing for flood prediction
due to dike failure. Chinnarasri et al. (2003) investigated flow patterns and the progres-
sive damage of dike overtopping, distinguishing various stages of dike damage and
accounting for the degradation rate of the dike crest.
A unique series of field tests to study the stability of embankment dams of various
materials was presented by Höeg et al. (2004). Their tests included failures due to over-
topping and piping of up to 6 m high dikes and the corresponding laboratory scale tests.
Cao et al. (2004) presented one of the first computational models on breach hydraulics
of erodible dikes, based on sediment transport and the morphological breach evolution.
Rozov (2003) conducted laboratory experiments to investigate the process of dam
breach erosion. He described the physical mechanism of dam washout and developed a
mathematical model for simulating the dam breach process.
Literature review 13
14 Literature review
Literature review 15

Between 2001 and 2004, the European project ‘Investigation of extreme flood pro-
cesses and uncertainty’ (IMPACT) provided valuable field and laboratory data to assess
and develop breach models. The project highlights are summarized in a special issue of
the Journal of Hydraulic Research (Garcia and Zech 2007). Recent works on dike
breaching include those of Hager and Unger (2006) who presented a preliminary study
aiming to understand the essential processes of both plane and spatial dike breaches.
Dupont et al. (2007) described an experimental and numerical study of dike breach due
to overtopping. The project ‘Integrated flood risk analysis and management methodolo-
gies’ (FLOODsite) was initiated with the objective to better assess and manage flood
risks in Europe, including a state-of-the-art review on breach modelling (Morris 2009).
The present knowledge on dike failures due to overtopping both by experimental and
theoretical breach modelling was summarised by Jandora and Jaromír (2008). Pickert et
al. (2011) presented a series of 3D dike breaching tests and described the breach pro-
files, breach outflow and soil conditions during the breach. Jovanović (2010) discussed
the extrapolation of laboratory results on dike breaching to prototype dikes. Upscaling is
in general valid if the length scale is not smaller then 1:15.

3.3 Dike failure modes

3.3.1 Overview

A dike failure can be either natural or man-made. Natural failures occur due to extreme
flood events, earthquakes, soil settlements, piping, seepage, overtopping, or animal
burrowing. Man-made causes include e.g. poor construction, incorrect design, improper
location, or sabotage. In many cases, a combination of several failure modes leads to the
actual dike failure. In the particular case of earth dikes, the most common failure causes
and modes are (Singh 1996):

(1) Overtopping caused by extreme floods

(2) Structural failure due to internal erosion (piping)

(3) Structural failure due to shear slide

(4) Structural failure due to foundation problems

(5) Failure due to natural or induced seismicity

16 Literature review

Causes (1) and (2) are related to hydraulic failure, whereas causes (3) to (5) are mainly
dominated by geotechnical processes and therefore correspond to 'geotechnical failures'.
All failure mechanisms are discussed below. The governing mechanisms of dike failures
were also described by the University of California in 2006 (Figure 3.1). The majority
of dikes and earthen embankments breach due to overtopping or a combination of over-
topping and internal erosion (Singh 1996). Therefore, this work focuses exclusively on
the overtopping failure mode. Most earth dikes are not designed to resist overtopping
and their resistance to surface erosion is limited. Figure 3.2 shows a prototype breach
test where the dike was destroyed due to overtopping erosion along with geotechnical
slope failures and piping.

Figure 3.1 Dike failure mechanisms (© Zina Deretsky, The National Science Foundation)

3.3.2 Overtopping failure

Dikes are designed for a defined discharge or a certain reservoir water level, for which
no overtopping occurs, including a freeboard due to wind waves, ice flow, or discharge
fluctuations. Following freeboard and discharge uncertainties it is not surprising that
overtopping is still an issue for dike safety.
Dike overtopping starts normally at the lowest crest elevation or at a local dike dis-
continuity, e.g. at a bridge abutment. The shear stresses induced by the water flow over
Literature review 17

the downstream dike slope initiate the erosion process. Erosion starts if the induced
shear stress exceeds the critical sheer stress of the dike material. The soil particles are
then entrained and transported downstream. Any initial small breach or a local soil
settlement represent a crest weakness and may develop into a larger breach resulting in
dike failure. During the 2005 Hurricane Katrina in New Orleans (LA), fifty dike
breaches occurred, of which forty breaches were due to overtopping (Daniel 2007).
Figure 3.3a shows the flooding of the Alpine Rhine River upstream of Lake Constance
in 1927 with a bridge failure caused by dike breaching (Minor and Hager 2004). The
dike failed by the excessive water discharge of 3,000 m3/s. Figure 3.3b shows dike
breaches due to overtopping at the Aare River during the 2005 flood event in Switzer-
land. The process of overtopping erosion is discussed in Chapter 3.4.

Figure 3.2 Dike failure mechanisms during prototype breach test (Photo:VAW)

3.3.3 Internal erosion (piping) failure and seepage

Internal erosion occurs if particles of the dike are entrained downstream by seepage or
groundwater. It only develops if two conditions are met: particles must be entrained and
18 Literature review

then be transported by the flow. Various processes are distinguished (Hagerty, 1991a):
entrainment, backward erosion, blowout, buoyancy, dissolution, suffusion (washout of
small soil particles from an inhomogeneous soil) and piping. Internal erosion initiated
by seepage that removes solid particles and builds tubular cavities is called piping. The
water seeping through a dike creates an open path for flow, carrying particles with it and
washes them out, resulting in a loss of material essential for dam safety. The instability
of the undercut layer located above the zone of soil loss may result in a collapse of the
dike.

Figure 3.3 (a) Dike breach at Alpine Rhine River (Minor and Hager, 2004) and (b) Aare River
(Photo: Swiss Air Force 2005)

The necessary conditions for piping include an open exfiltration face, a source of wa-
ter, concentration of flow, removal of failed and displaced soils and a sufficient hydrau-
lic gradient at the exfiltration face. Unlike overtopping, piping is often detected in its
early stage and remedied before it becomes a serious issue. For new dikes, the potential
for internal erosion and piping can be controlled by a good geotechnical design and
construction of a dike core and additional provision of filters to intercept seepage. A
discussion of piping/seeping erosion in stream banks is presented by Hagerty (1991a, b).
A catastrophe caused by piping was the failure of the St. Francis Dam in California
in 1928. The dam failed suddenly after a highly-erodible sediment rock formation was
washed out, undermining portions of the concrete dam (Figure 3.4a). Different piping
events occurred during the flood events in Switzerland 2005 along the dikes of the Linth
River channel (Figure 3.4b), fortunately not resulting in any dike breach.
Literature review 19

Figure 3.4 (a) St. Francis Dam after its failure in 1928 (Cedergren 1989), (b) Seepage at Linth
River dike during 2005 floods in Switzerland (Photo: M. Jud, Linth Administra-
tion)

Further, uncontrolled seepage through dikes can led to severe dike damage. Detailed
descriptions of seepage and drains are found e.g. in Cedergren (1989). Before the twen-
tieth century, the engineering works made of earth material had no theoretical back-
ground. The problem was the lack of knowledge on how water seeps through earth and
the lack of information on permeability. The resulting designs were often uneconomical
offering no safety against seepage failure. Darcy’s fundamental experiments with seep-
age phenomena furnished a simple explanation for the seepage process trough homoge-
nous soil and drainage systems. Further improvement for scientific and experimental
earthwork design was achieved by Terzaghi (1925). The explanation of the seepage
theory for practical aspects by Casagrande (1937, 1961) represents a major step forward
in the design of earth dams and dikes.
The most common methods to control seepage are sealing and drainage methods.
Drainage is the process by which percolating water is removed from soils and rocks by
natural or artificial means. Dikes often contain no internal drainage system but a foot
drainage or a sealing. These reduce seepage trough a dike and, depending on the dike
location, are referred to as surface sealing or internal sealing. A surface sealing mostly
consists of mineral materials and synthetic sealing lanes. The internal sealing is primari-
ly made of steel sheet piles, slotted or slurry walls, MIP (mixed in place) walls or soil
stabilisation fills.
However, all dikes will experience seepage, even if they are made relatively imper-
meable by using selected, possibly artificial materials or watertight membranes. There-
fore it is necessary to know the flow net to predict the seepage rate. The seepage de-
pends mainly on the dike geometry, the dike and foundation materials, and the hydraulic
20 Literature review

characteristics of the dike and its foundation. There are different methods available to
predict the seepage rate of dikes. For homogenous dikes or central core dikes the seep-
age can be estimated by the method of Kozeny (1927) or Casagrande (1937). The seep-
age rate is predicted by either graphical techniques (flow net) or numerical methods.
The Casagrande method and most numerical models do not consider unsaturated flow.
Chapuis and Aubertin (2001) presented a numerical model to estimate saturated and
unsaturated seepage through dikes under steady state conditions. An overview and
comparison of different methods to predict steady state or transient seepage through
dikes is given by Haselsteiner (2007).

3.3.4 Geotechnical failure

All geotechnical failures have in common that the effective driving forces on the dike or
foundation exceed the resistance, so that the original dike profile cannot be obtained.
The effective forces are the dead load and the traffic loads, horizontal forces (water
pressure, ice pressure, waves, or wind) and the dynamic forces (seismic, machine and
ram works). The effective resistances are conditional characteristics of a dike and its
foundation materials (friction, cohesion, coefficient of elasticity). The failure can affect
the dike itself (sliding) or just part of it (slope failure), as well as the entire foundation
(settlements) or part of it (bearing capacity failure, shear failure).

3.4 Overtopping erosion

3.4.1 Overview

Erosion initiates if the induced shear stress by the water overtopping a dike exceeds the
critical shear stress of the dike material or its protection system. The relevant infor-
mation regarding erosion and sediment transport for this work are summarized below.
However, the discussion of all features regarding sediment transport would go beyond
the scope of this work. Extensive information on sediment transport are found in e.g.
Shields (1936), Einstein (1942), Meyer-Peter and Müller (1948), Yalin (1972), Zanke
(1982), Smart and Jaeggi (1983), van Rijn (1984a, b), Parker (1990), Hunziker (1995), ,
Gyr and Hoyer (2006),Wong and Parker (2006), and Julien (2010).
Literature review 21

3.4.2 Flow regimes and erosion zones

The overtopping flow of a dike is similar to the flow over a broad-crested weir with low
or no tailwater. Powledge et al. (1989a, 1989b) presented model and prototype research
studies on overtopping flow over embankments, dams, dikes, roadways and stated three
flow regimes and three corresponding erosion zones. The three regimes shown in Figure
3.5 are described below.

Erosion Zone 1 – Subcritical flow over dike crest

A subcritical velocity state dominates the change from the reservoir to the upstream
portion of the dike crest. The flow proceeds from a static energy head in the reservoir to
a combination of static and dynamic heads along the upstream dike slope. Due to the
small energy gradient and minute flow velocities (F < 1) the hydraulic forces are small.
Even though the flow depth may be quite large, the small energy gradient imposes a
small tractive stress. Erosion will occur only if the dike crest consists of highly erodible
material.

Erosion Zone 2 – Supercritical flow over dike crest

In this zone transitional flow occurs, characterised by critical flow and the critical flow
depth hc. The energy level is about the same as in Zone 1, but the energy gradient can be
large resulting in a significant increase of the tractive stress. However the distance over
which they occur is limited, depending on the dike crest geometry, so that erosion is
mainly expected at the downstream crest portion. Erosion starts if the critical shear
stress of the dike material is exceeded.

Erosion Zone 3 – Supercritical flow on dike downstream slope

In erosion Zone 3, the energy head progressively increases as the water flows down the
dike face. The energy gradient is steep resulting in a significant velocity increase until
energy loss equilibrium and uniform flow are attained. This will however only occur for
high embankments. The tractive stresses are large, leading to a high erosion potential.

Based on observations of embankments slopes during overtopping, the erosion pro-

cess initiates most commonly at a point of slope discontinuity, such as the toe or base of
the dike. However, depending on the particular condition and configuration of the em-
22 Literature review

bankment, the initial erosion may start anywhere. In many cases, after the erosion was
initiated, surface discontinuities occur resulting in a cascading flow. The tractive stress-
es on the downstream dike surface are therefore not consistent.

Figure 3.5 Hydraulic flow regime and erosion zones (modified after Powledge et al. 1989b)

For embankment dams, Powledge et al. (1989b) listed the following physical factors
having a strong effect on ignition and erosion rate: (1) Embankment configuration,
types of material and fill densities; (2) Maximum flow velocity; (3) Discontinuities
(berms, roads, outlets, ditches), cracks, or voids in the slope; (4) Presence and height of
tailwater on downstream slope; (5) Flow concentration at low points along the em-
bankment or at the abutment groins; and (6) Toe drains, or blanket drains. The breach
process is discussed in Chapter 3.5.

3.4.3 Sediment transport

Flow hydraulics
Consider pure dike overtopping, for which the erosion process is exclusively governed
by surface sediment transport. Further, the main transport is in the form of bed load,
given the grain sizes and the flow conditions. Regarding sediment transport, the interac-
tion between the water flow and the sediment bed is essential. The flow is either laminar
or turbulent and characterized by the Reynolds number R describing the ratio between
inertia and viscous forces as
Literature review 23

Vm L
R (3.1)


with Vm = depth-averaged velocity, L = characteristic length and ν = kinematic fluid

viscosity. For open channel flow, the characteristic length corresponds to the hydraulic
radius Rh = A/U with A = cross-sectional area and U = wetted perimeter.
The main force acting on the sediment bed is the bed shear stress τb. First, it is pre-
sented for a small sloping bed, using the common assumption sinα = tanα. The bed
shear stress on steep slopes is presented afterwards. Figure 3.6a shows a sketch of the
hydraulics on a bed sloping at a small angle α, flow depth h, width b and the gravity
force G acting on an element of length Δx. So denotes the bed slope and S the energy
slope. The main parameter responsible for sediment transport is the bed shear stress τb
acting on the sediment bed (subscript b). Assuming uniform parallel-streamline flow
given by S = So and Rh = h (b/h >15), the shear stress along the sediment bed is
 b  ghS o (3.2)

with ρ = fluid density and g = acceleration of gravity. The shear velocity U* follows
from

b
U  (3.3)


U* is especially used when dealing with flow resistance as it characterizes the bounda-
ries between turbulent flow over hydraulic smooth beds and turbulent flow over hydrau-
lic rough beds. The bed shear stress and the shear velocity resulting from the water flow
are the main forces responsible for incipient sediment motion and bed load transport.
For uniform, hydraulically rough flow and no viscous sublayer, the logarithmic velocity
profile can be written as

V 1 z 
 ln n   8.5 (3.4)
U 
  ks 

with κ = 0.41 = von Karman’s constant, zn = normal distance to channel bed and
ks = equivalent sand grain roughness. Using the depth-averaged velocity Vm and inte-
grating Eq. (3.4), the resistance law results as (Keulegan 1938)
24 Literature review

Vm 1  h 
 ln11  (3.5)
U    k s 

The resistance relation between Eq. (3.3) and Eq. (3.5) is

 b  C f Vm2 (3.6)

where the flow resistance factor Cf is given by

2
1  h 
C f   ln11  (3.7)
   k s 

In uniform hydraulically rough open channel flow, a local point estimation of the bed
shear stress can therefore be made using the mean flow velocity and the friction coeffi-
cient.
For the current dike breach tests under laboratory conditions, the assumptions of uni-
form flow S = So, Rh = h and sinα = tanα are invalid (Figure 3.6b). The bed shear stress
for non-uniform flow involves the energy line slope S resulting in a reach-averaged
value for the bed shear stress

 b  gRh S (3.8)

The energy line slope can be derived based on the energy equation for non-uniform flow

V2
H  e  z  h cos  (3.9)
2g

with H = energy head, αe = energy correction factor (Julien 2010), z = distance of bed
above the datum plane. Given the sloping bed, the pressure head differs from the flow
depth h. For curvilinear flow, the angle of the water surface must additionally be con-
sidered when calculating the pressure head (see Chapter 4.2.10). Assuming S = tanγ,
with γ = angle of energy line, the energy slope S and τb between two points of spacing
Δx follow as

H1  H 2
S (3.10)
x

H1  H 2
 b  gRh (3.11)
x
Literature review 25

It is therefore assumed that all lost energy along the dike breach is transformed into the
bead shear stress and consequently may cause sediment transport.

Figure 3.6 Definition of bed shear stress for (a) uniform flow and (b) steep slope flow

Incipient motion of single grain

In rivers and channels of moderate gradients, mainly two transport modes occur, e.g.
bed load and suspended load. Usually three modes of particle motion are distinguished:
(1) Rolling and sliding motion, (2) saltation motion, and (3) suspended particle motion.
If the bed shear velocity exceeds the critical value for initiation of motion, particles will
be rolling, sliding or both. As the bed shear velocity increases, the particles will be
moving along the bed by saltation. The suspended particles consist of the finder sedi-
ment maintained in suspension due to turbulence. The boundary between suspended
sediment and bed-load transport is not precise and may vary with the flow strength.
Usually bed-load includes the transport of rolling, sliding and saltating particles. Bag-
nold (1973) defines bed-load transport as successive contact of the particle with the bed
due to the effect of gravity. Einstein (1950) defines bed-load transport as the movement
of sediment particles in a thin layer of two particle diameters sizes above the bed, there-
fore excluding saltating particles.
The incipient motion criteria can be simplified when assuming a single spherical ex-
posed sediment particle on a horizontal or an inclined streamwise slope (Figure 3.7a, b)
using the so-called Ikeda-Coleman-Iwagaki analysis (Iwagaki 1956, Coleman 1967,
Ikeda 1982, ASCE 2008). The flow is assumed to follow the logarithmic law of wall as
26 Literature review

V 1  z 
 ln  (3.12)
U 
  zo 

with V = time averaged flow velocity at a distance z above the bed, κ = von Karman’s
constant ≈ 0.41 and zo = bed roughness length. In theory, the bed roughness length is the
distance above the bed where the flow velocity tends to zero. In practice, the equivalent
sand grain roughness ks is used, as proposed by Nikuradse (1933). For open channel
flow, Nikuradse’s equivalent sand grain roughness is commonly taken to be proportion-
al to a representative sediment diameter d. The origin of the z-coordinate is located at
the base of the exposed particle and the particle centre is therefore at z = d/2. Drag and
lift forces act through the particle centre, turbulent forces are neglected, and the rough-
ness height ks is equal to the grain diameter d.

Figure 3.7 Forces acting on single grain on (a) horizontal gravel bed and (b) inclined gravel
bed (adapted from ASCE 2008)

The forces acting on the particle are

2
1 d 
FD   π  CDV f2 Fluid drag force (3.13)
2 2

2
1 d 
FL   π  C LV f2 Fluid lift force (3.14)
2 2

3
4 d 
Fg   ( s  1) g    submerged gravitational force (3.15)
3 2

Fr   Fg  FL  Coulomb frictional force (3.16)

with s = ρs/ρ = specific density and μ = coefficient of Coulomb friction = tanθs with
θs = submerged angle of repose. The critical condition for incipient motion of the parti-
Literature review 27

cle occurs as the drag force is just balanced by the Coulomb frictional force and follows
for a horizontal bed as

FD   Fg  FL  (3.17)

For critical fluid velocity Vf at z = d/2 follows

V f2 4 
 (3.18)
( s  1) gd 3 C D  C L

The effective fluid velocity Vf acting on the particle depends on the existence of a
viscous sub-layer and follows as

V f  F U  (3.19)

Where

1 U *d U *d
F for  13.5 (viscous sub-layer) (3.20)
2  

and

U *d
F  6.77 for  13.5 (no viscous sub-layer) (3.21)


Equation (3.18) can be rearranged to a relation in terms of the critical shear stress
τc* = τb/(g(ρs‒ρ)d) with τb = ρUc*2 and Uc* = Vf/F = critical shear velocity resulting in the
Ikeda-Coleman-Iwagaki relation for the critical shear stress as

4  1
 c*  (3.22)
3 C D  C L  F 2

For a channel bed of angle α, the downslope effect of gravity has to be included in the
force balance. Balancing forces results in

FD  Fg sin    Fg cos   FL  (3.23)

and the critical shear stress follows then as

28 Literature review

4 cos  sin  1
 c*  (3.24)
3 C D  C L  F 2

The essence of this approach on the initiation of motion relates to the possibility of
obtaining an explicit formulation of the relation explored by Shields with dimensional
analyses and experiments (see below). Equation (3.22) can be directly evaluated using
realistic assumptions for the internal friction angle μ and the drag CD and lift CL coeffi-
cients. A comparison of the Ikeda-Coleman-Iwagaki model for initiation of motion with
the Shields data is provided by ASCE (2008). The best fit results for ks = 2d,
CL = 0.85CD with CD as a function of Vfd/ν according to the standard drag curve for
spheres and μ = tan60°. Such a friction angle is relatively high but the exact value of
this parameter for the sediment used by Shields is unknown.
The incipient motion of a single grain depends mainly on the location of the grain,
the decisive velocity acting on the grain and turbulent forces. The Ikeda-Coleman-
Iwagaki model considers an exposed particle with the volume of a sphere, and the flow
is taken to follow the logarithmic law. For a random particle, the volume can be written
as V = k1d3 with k1 as a proportional factor to the grain volume. For an embedded or
hidden grain, the drag and lift forces have to be reduced, as not the entire grain is sub-
jected to the flow forces. The relation between the flow velocity at the particle Vf and
Uc* can further be generally written as Vf =f Uc* with f = a measure of friction.

Shields criterion
The most significant work regarding incipient sediment motion was conducted by
Shields (1936). Experiments with uniform sand-sized particles at bottom slopes < 0.5%
and under high relative submergence led to the critical dimensionless shear stress

 bc RS S o  U *d 
 c*    f  c  (3.25)
g  s   d s  1d   

with τc* = critical shear stress at incipient motion, Uc* = critical shear velocity and
RS = hydraulic radius of bed cross-section. Shields stated that τc* is only a function of
the grain Reynolds number R* = Uc*d/ν (Figure 3.8). Turbulent smooth flow occurs for
R* ≤ 2, transitional flow if 2 ≤ R* ≤ 500 and rough flow for R* ≥ 500. For R* ≥ 500, the
dimensionless critical shear stress remains constant at τc* ≈ 0.06, independent of viscous
Literature review 29

effects. Numerous studies were conducted on incipient motion, summarized e.g. by

Buffington and Montgomery (1997).
Up to date, the Shields diagram is the most commonly used basis to investigate incip-
ient motion of grains or to evaluate the stability of uniform sediment beds. However,
several limitations have to be respected. The bed material has to be of relatively uniform
size, i.e. the geometrical standard deviation of the grain size distribution
σg = (d84/d16)0.5 < 1.5 and relative submergence is limited to Rs / d > (8 to 10). Further,
the diagram is limited to comparatively flat bed slopes of up to So < 0.005.
Critical Shields values τc* are commonly used to denote conditions under which bed
sediment particles are stable but on the verge of being entrained into the flow. It is
known that the values obtained from the Shields diagram for initiation of motion are
indeed larger than observed by other researchers, in particular for coarse material. For
example, Neill (1968) gives τc* = 0.03 instead of 0.06 for the dimensionless critical
shear stress if R* > 500, while Gessler (1971) suggests using τc* = 0.046.

c [-]
*

R* [-]

Figure 3.8 Shields diagram τc*(R*) for incipient motion of uniform, incohesive material (Buff-
ington 1999)

The design value of τc* depends on the particular case at hand. If entrained grains are
replaced by others from upstream, ‘some motion’ can be tolerated, and the values from
the Shields curve may be used. However, if grains removed are not replaced, as on a
stream bank, τc* according to Shields is too large and should be reduced (ASCE 2008).
Neill and Yalin (1969) and Gessler (1970) stated that Shields' original values for initia-
tion of motion of coarse material are too high and should be divided by a factor of 2.
30 Literature review

The Shields diagram is not practical in the form of Figure 3.8, because the critical
shear velocity is required to find the critical shear stress for incipient motion, which
depends likewise on the critical shear stress. The fit relation in Figure 3.8 can be written
in explicit form as (Brownlie 1981, ASCE 2008)

 c*  0.22 Rep0.6  0.06exp 17.77 Rep0.6  (3.26)

with Rep = (g (s‒1) d)0.5d/ν. The value of τc* can be directly computed if the water and
sediment properties are given. Again, to be on the safe side for engineering purposes,
the value resulting from Eq. (3.26) should be divided by 2,

 c* 
1
2
 
0.22 Rep0.6  0.06exp  17.77 Rep0.6  (3.27)

Further empirical threshold curves for the Shields diagram were presented by Hager and
Oliveto (2002), Cheng (2004) or Cao et al. (2006).
Even under perfect laboratory conditions, a detailed determination of the incipient
motion of a specific grain is difficult. The incipient motion under a high channel gradi-
ent is especially complicated, due to high turbulence and small relative submergence.
Therefore, most experimental research regarding the critical shear stress was carried out
at comparatively flat beds.

Effect of bed slope on incipient motion

The work of Shields and others on the initiation of motion is restricted to nearly hori-
zontal bed slope. Especially mountain streams may have steep gradients, and the effect
of the downslope gravity component has to be accounted for. Both the buoyancy and the
gravity destabilize the sediment grain additionally. The dimensionless critical shear
stress is therefore decreased compared to a flat slope. According to Graf and Suszka
(1987) or Rosport (1997), the steeper the channel bed, the smaller the critical dimen-
sionless shear stresses. Modified Shields approaches exist in the literature to account for
large channel slopes and small relative submergence (Gregoretti 2008, Recking et al.
2008 or Weichert et al. 2009).
Chiew and Parker (1994) analysed the effects of lift, drag, buoyancy, and gravity on
a bed particle consisting of similar particles and concluded that
Literature review 31

 c*  tan 
 cos 1   (3.28)
c *
  

where ταc*= critical shear stress for sediment in a river with a longitudinal bed angle α.
Equation (3.28) applies to downward sloping beds. The critical shear stress τc* can be
derived from the Shields diagram or with Eq. (3.26). The same equation is obtained
when dividing Eq. 3.24 by Eq. 3.22, obtained from the Ikeda-Coleman-Iwagaki model.
Lau and Engel (1999) used dimensionless analyses to obtain an equivalent mathematical
equation of the form

 c* sin    
 (3.29)
 c* sin  

with θ = angle of repose. Their condition for incipient motion depends both on the
slope angle and on the flow Reynolds number. They recommended to use Eq. (3.29) for
slope angels up to the angle of repose of the sediment.

Bed-load sediment transport formulae

The basic problem regarding transport formulae is the characterisation of bed-load
transport depending on discharge, channel geometry and grain size. Several approaches
exist to determine the sediment transport capacity, describing the potential volume of
bed-load per time transported in a river without additional erosion or sedimentation. The
sediment transport capacity corresponds to an equilibrium transport capacity for a given
channel and discharge.
Sediment transport is driven by the bed shear stress and mainly depends on the grain
size. A grain is entrained only if the bed shear stress exceeds a certain threshold. There-
fore, sediment transport formulae for uniform sediment are commonly based on the
excess shear stress above the critical value. Most developed formulae concern uniform,
steady flow conditions with movable uniform sediment particles. Modifications are
required if the formulae are used for non-uniform flow conditions, non-uniform sedi-
ment, armouring or paving of the bed surface.
The development of sediment transport formulae is mostly based on flume experi-
ments, field measurements or a combination of both. Thus, most formulae are based on
a comparatively restricted data basis and only valid for certain boundary conditions. The
most widely quoted formulae are those of Schoklitsch (1934), Meyer-Peter et al. (1934),
32 Literature review

Einstein (1942), Meyer-Peter and Müller (1948), Yalin (1963), Ackers and White
(1973), Bagnold (1980), Parker et al. (1982), Daido (1983), Misri et al. (1984), Smart
and Jaeggi (1983), Graf and Suzka (1987) or Rickenmann (1990). Some of these as well
as futher formulae are shortly described below. Additional overviews on sediment
transport formulae are found e.g. in Gomez and Church (1989), Sieben (1993), ASCE
(2008), or Julien (2010).
A large number of bed-load relations can be expressed using the dimensionless quan-
tities:

qb
q  dimensionless bed-load transport rate (3.30)
s  1gd 3

b
  dimensionless shear stress (3.31)
g  s   d

 s  1g 
1/ 3

d d   dimensionless grain size (3.32)
 
2

with qb = volumetric bed-load rate. The following bed-load equations consider only the
grain shear stress and uniform sediment size. Bed-load transport by fraction is not dis-
cussed.

Meyer-Peter and Müller (1948)


q   8     c 
3/ 2
where τc* = 0.047 (3.33)

This formula is empirical and was verified with data for uniform coarse sand and gravel
if So ≤ 0.5%. Wong and Parker (2006) reanalysed the data used by Meyer-Peter and
Müller and found that a better fit is provided using one of the two alternative forms


q   4.93    0.047 
1.6
(3.34)


q   3.97    0.0495 
3/ 2
(3.35)

This was further analysed by Hager and Schmocker (2008) assuming that the hydrau-
lic condition are much above the sediment entrainment condition, i.e. τ*>>τc*. They
Literature review 33

presented the dimensionless bed shear stress with basic parameters and b = channel
width as

HS  H  Rh 
1 / 12

  1  1.2   d *1 / 4  (3.36)
Rh d  b  S  

Yalin (1963)

 ln 1  ar  
q   0.635r   1  (3.37)
 ar 

where

    c
r and a  2.45s 0.4  c (3.38)
 c

The dimensionless critical shear stress τc* is evaluated from the Shields curve.

Parker (1979)

q 
 11.2
 
 0.03  4.5

(3.39)
3


Eq. (3.40) has been developed as a simplified fit to the relation of Einstein (1950) for
the range of Shields numbers likely for gravel-bed rivers.

van Rijn (1984a)

2.1
0.053    
q  
  1 (3.40)
 0.3   
d  c 

Eq. (3.41) can be used for particles with mean sediment sizes between 0.2 and 2.0 mm.

Nielsen (1992)


q   12     c    (3.41)

Eq. (3.42) can be used for uniform sand and gravel bed-load transport.
34 Literature review

Regarding the dike breach process, bed-load transport formulae for steep slopes S
are of special interest. However, only a small number of studies were carried out for
So > 1% and the present knowledge is therefore limited. Cao (1985) conducted experi-
ments for 0.5%<So<9%, Graf and Suszka (1987) for 0.5%<So<2.5%, Smart and Jaeggi
(1983) and Rickenmann (1990) produced data essentially for very steep slopes
3%<So<20%, and Recking et al. (2008) for 1.0%<So<9%.

Smart and Jäggi (1983)

0.2
 d 90  1.5   c 
q  4

 S o C 1   
0.6
(3.42)
 d 30    

with C = flow resistance factor. For grain Reynolds numbers R*>103, τc* = 0.05 is rec-
ommended, which has to be corrected for steep slopes as

 So 
 c   0.05cosarctanS o 1   (3.43)
 tan 

Rickenmann (1990)

The formula of Rickenmann (1990) is based on laboratory tests on slopes between 7%

and 20% using gravel material with d = 0.01 m and fluid densities of
1,000 kg/m3 ≤ ρ ≤1,400 kg/m3. The formula is further limited to a relative submergence
of Rs / d < 200, reading

0.2
d  0.5    c
 
q  3.1 90   

F1.1 (3.44)
d
 30   s  10.5

with F = V/(gh)0.5 = Froude number.

Sediment transport formulae are especially used when numerically simulating dike
breaches. Physically-based dike breach models use the principles of hydraulics, sedi-
ment erosion and soil mechanics to develop temporal solutions of the actual dike breach
and its outflow. The most common formulaa used are the Schoklitsch-, the MPM-, the
Einstein- and the Bagnold-Visser formulae. An overview on sediment transport formu-
lae used in breach models is given in Wahl (1997), Lecointe (1998), Broich (1999), Zhu
et al. (2004), or Hager and Schmocker (2008). However, all of the applied formulae are
Literature review 35

actually limited to uniform, steady flow conditions on small slopes and do not account
for the non-uniform flow features during a dike breach. Most numerical models are
calibrated using physical laboratory tests and the bed-load transport formulae are
adapted to satisfy the dike erosion process.

3.5 Breach process

3.5.1 Overview

Dike breaching represents a complex interaction between hydraulic, geotechnical and

structural processes. The breaching process varies with material type and state, hydrau-
lic load and dike condition. Several distinct stages have been observed and differ espe-
cially for cohesive or non-cohesive materials. As the current project is limited to
breaches of incohesive dikes due to overtopping, only this process is discussed below.
Various model studies in the past investigated both plane (2D) and spatial (3D) breach
processes due to overtopping using hydraulic modeling. Further, the evaluation of real
case data resulted in additional information on the breach process.

3.5.2 Plane breach process

Plane dike erosion was investigated among others by Powledge et al. (1989 a, b), Chin-
narasri et al. (2003) or Dupont et al. (2007). Powledge et al. (1989b) described the
breach development for both granular and cohesive embankments due to overtopping.
Especially for granular embankments, seepage can have a significant effect on the ero-
sion process. Once overtopping occurs, the seepage exiting on the downstream dike face
accelerates the erosion. Surface slips and sliding failures result in a fast enlargement of
the breach. Cohesive embankments are generally more resistant to overtopping breach-
ing due to the greater erosion resistance and the reduction of seepage. Erosion starts
often on the embankment toe and propagates upstream, undercutting the slope, causing
large chunks of material to be removed by soil tensile and shear failure on the over-
steepened slope. Powledge et al. (1989a) listed several factors influencing the overtop-
ping erosion, mainly: (1) Dike configuration, material types and density for earthfill, (2)
Maximum flow velocity, (3) Discontinuities, cracks in the slope and anomalies at the
36 Literature review

toe, (4) Presence and height of tailwater, (5) Flow concentration at low points along the
embankment, and (6) Toe drains, or blanket drains.
Chinnarasri et al. (2003) observed four stages in plane dike erosion (Figure 3.9),
namely: (a) Small erosion on dike crest after initial overtopping, (b) Slope sliding fail-
ure with ongoing erosion, (c) Wavelike-shaped dike profile, and (d) Large sediment
wedge deposition with small slope at erosion end.

Figure 3.9 Process of dike failure due to overtopping (Chinnarasri et al. 2003)

Similar processes were also noted by Dupont et al. (2007). They additionally ob-
served sliding of the lower part of the downstream slope just before actual overtopping.
Literature review 37

The erosion progresses from the downstream face toward the crest, with the down-
stream face rotating around a pivot point (Figure 3.10). With ongoing erosion, anti-
dunes are generated on the downstream face.

Figure 3.10 Rotation of downstream dike face around pivot point in the first stage after over-
topping (Dupont et al. 2007)

The observed dike breach profiles differ mostly due to particular test setups. Pure
plane erosion can only be obtained by limiting the dike length. Otherwise 3D erosion
patterns occur and influence the breach process. Further, the existence of a surface layer
or a core may affect the erosion process significantly. The volume of the upstream
reservoir or increasing inflow discharge mainly determine the breach duration. Despite
several extensive hydraulic model tests made in the past, a general plane dike erosion
profile is not yet available. Most of the available data are limited to the test range and
were directly used to validate numerical models.

3.5.3 Spatial breach process

An overview on spatial breach formation is provided by Singh (1996), Coleman et al.

(2002), Rozov (2003), Chinnarasri et al. (2004), Spinewine et al. (2004), Visser et al.
(2006), Morris et al. (2008), or Pickert et al. (2011). This erosion process normally
occurs in engineering applications and includes both vertical and lateral erosion. Espe-
cially the fundamentals of the lateral erosion are currently not fully understood. Cole-
man et al. (2002) presented embankment breach tests under constant reservoir level and
described the breach process in detail. Flow through a pilot channel located at the chan-
nel side wall initially erodes a small breach channel on the downstream embankment
face from the crest to the toe. The breach then expands primarily vertically to predomi-
nantly lateral erosion as the breach channel approaches the embankment foundation.
38 Literature review

Embankment material is eroded by the process of tractive shear stress and turbulence.
Undermining of the breach channel side slopes causes large volumes of material to
collapse into the channel centre and gets transported downstream. The breach channel is
of curved (hourglass) shape in plan (Figure 3.11). This shape increases in curvature with
time until the vertical erosion is inhibited by the embankment foundation. Coleman et
al. (2002) further accounted for both longitudinal breach profiles and breach cross-
sections. A similar breach process was observed by Rozov (2003), hence with the pilot
channel located in the embankment centre. After the initial vertical breach had occurred,
the lateral erosion proceeded to both sides.

t = 44 s t = 113 s

t = 217 s t = 269 s

Figure 3.11 Breach development for coarse-sand embankment (Coleman et al. 2002)

Chinnarasri et al. (2004) investigated the breach geometry in homogenous embank-

ments under falling reservoir level. They observed primarily vertical erosion in the
initial breach with subsequent lateral erosion after the breach approached the fixed
embankment foundation. The lateral erosion was predominated by a combination of
tractive shear forces and instabilities of the side slopes, causing the collapse of large
volumes of embankment material. The breach shape was rectangular at initial breaching
and developed into a trapezoidal shape with ongoing erosion and reservoir drainage.
Visser et al. (2006) distinguished five stages in the process of overtopping breaching,
for both sand and clay dikes. The breach started with the flow entering a small initial
breach located at the top in the middle of the dike. During Stages I and II, the breach
developed gradually by decreasing the height of the dike and increasing the width of the
Literature review 39

initial breach channel. During Stage III the breach growth accelerated until the dike was
completely washed out in its breach section. The breach then expanded mainly laterally
during Stage IV and decelerated in Stage V under decreasing backwater and hence
decreasing breach discharge.
Schmocker et al. (2011) presented hydraulic model tests of fuse plug embankments.
They investigated both the vertical and lateral erosion processes due to overtopping.
Figure 3.12 shows the temporal erosion process, initiating as the flow enters the pilot
channel and in a first step, proceeds vertically. With ongoing vertical erosion, the lateral
erosion increases and is mainly dominated by a collapse of large sediment chunks.

Figure 3.12 Advance of spatial fuse plug erosion at different times t (Schmocker et al. 2011)

Pickert et al. (2011) conducted a series of spatial embankment breach tests due to
overtopping. They divided the failure of homogeneous embankments into two breaching
phases (Figure 3.13). The “breach evolution” started with initial embankment overtop-
ping and ended when the erosion reached the upstream embankment shoulder. The
second phase “breaching” was dominated by a rapid increase of both erosion and breach
outflow due to vertical and lateral breach widening. Further, the erosion process de-
pended on the sediment diameter and was especially influenced by apparent cohesion.
No continuous erosion was observed for fine sand and the classification of the cross-
40 Literature review

sectional breach profile was difficult. The lateral erosion was a combination of constant
erosion and sudden collapse of the breach side slopes.
The laboratory effort for spatial breach tests is comparatively high. Therefore, sys-
tematic investigations are not available and most past research was limited to some five
individual tests. Further, recording breach profiles and breach cross-sections is complex
as the spatial breach process remains mostly invisible and cannot be simply recorded
through the channel side wall.

Figure 3.13 Spatial breach development for medium sand embankment (Pickert et al. 2011)
Literature review 41

3.5.4 Real-case data and field tests

MacDonald and Langridge-Monopolis (1984) analyzed the breach characteristics of

historical dam failures. They stated that the breach shape for both earthfill and non-
earthfill dams is either (1) triangular with 2:1 (V:H) side slopes, if the breach develops
to the embankment base, or (2) trapezoidal with 2:1 side slopes if additional material is
washed out after the breach reaches the embankment bottom.
A unique series of field tests was carried out in Norway to study the stability of em-
bankment dams (EBL 2003, Höeg 2004). Overtopping failure was initiated with a pilot
channel in the embankment centre. The first breaching phase included the slow and
gradual erosion of the downstream slope. As the scour reached the upstream dam crest,
the breach became rapid and dramatic. Similar observations were made for rockfill,
gravel and clay dams. The opening of the breach first progressed down to the dam base
before it expanded laterally. The breach sides were almost vertical for all three materi-
als. Figure 3.14 shows examples of the Norwegian field test.
Froehlich (2008) approximated 74 embankment dam failures regarding breach
height, breach width and side slope ratio with a trapezoidal breach geometry. He stated
that the breach height is either limited by the dam foundation, which is more resistant to
erosion, or the reservoir water volume at failure time. Further, he proposed regression
equations for the average breach width, the breach side slope ratio and the breach for-
mation time.

Figure 3.14 (a) Initial erosion in pilot channel and (b) breach enlargement during dam breach
field test in Norway (EBL 2003)

Morris et al. (2008) discussed the importance of soil type and state on the breach
process. They observed that the breach shape during failure is not necessarily trapezoi-
42 Literature review

dal, but a function of soil type, soil state and hydraulic loading. During many observed
failures, the breach had vertical or even undercut sides, rather than sloping faces form-
ing the ‘trapezoidal shape’. For non-cohesive material, soil suction, negative pore pres-
sures or the dynamic erosion pattern can as well lead to vertical faces. Many photos can
be misleading, as they were taken after the breach event when the breach cross-section
had collapsed after the soil had dried.

3.6 Breach outflow

Both experimental and field observations indicate that the discharges over and through a
dike can be estimated by broad-crested weir flow (Pugh 1985, Singh 1996) using the
general overflow formula


Q  C d b 2 gH o 
3 1/ 2
(3.45)

with Q = overflow discharge, Cd = discharge coefficient, b = overflow width,

g = gravitational acceleration, Ho = ho+Q2/[2gb2(ho+w)2] = approach flow energy head,
ho = approach flow overflow depth, and w = dike height. The discharge coefficient
depends on the dike profile and can be determined for broad-crested weirs as (Fritz and
Hager 1998)

Cd  0.43  0.06sinπ  0.55 (3.46)

with ξ = Ho/(Ho + LK) = relative crest length and LK = crest length. However, the breach
shape may vary during the breach process, thereby affecting the discharge coefficient.
Especially for 2D dike breaching, the breach profile is often similar to a circular-crested
weir (Schmocker and Hager 2010). The discharge coefficient follows the standard
circular-crested weir formula as (Montes 1970, Hager 1994)

2  3 k 
Cd  1   (3.47)
3 3  11  Ω k 

with ρk = Ho/R = relative crest curvature, R = crest radius and Ω = 4.5. The effect of
various weir slopes on the circular-crested weir flow was investigated by Ramamurthy
and Vo (1993) or Schmocker et al. (2011) resulting in
Literature review 43

2  3 k ' 
Cd  1   with (3.48)
3 3  11  Ω k ' 

1
H    2 d  3
k'  o  o  (3.49)
R  270 

with [(αo+2αd)/270]1/3 = weir angle ratio with αo = upstream (subscript o) and

αd = downstream (subscript d) weir angle (Figure 3.15). For the standard circular-
crested weir αo=αd=90°, this ratio equals 1, and ρk' = ρk = Ho/R. Further methods to
determine the overflow discharge involve e.g. velocity measurement, reservoir water
balance calculation or numerical modelling.
Visser et al. (2006) determined the breach discharge and the flow velocities with
broad-crested weir formulae for free breach discharge and for submerged breach dis-
charge due to backwater. Spinewine et al. (2004) recorded both water levels and surface
velocities during the breach process and determined the resulting outflow discharge.
Both Coleman et al. (2002) and Pickert et al. (2011) measured the breach discharge
with a V-notch weir at the channel end. Dupont et al. 2007 estimated the breach dis-
charge by integrating the breach profiles in a hydrodynamic model computing the dis-
charge for the given temporal topography.

Figure 3.15 Definition of flow over circular-crested weir with various up- and downstream weir
angles

Equations for the peak breach discharge derived from past embankment dam failures
were also determined (Froehlich 1995, Chinnarasri et al. 2004, Wahl 2004). The main
parameters affecting peak discharge are the reservoir storage volume, reservoir depth at
failure time measured from the bottom of the final breach to the reservoir surface, and
the embankment cross-section at the breach location. The peak outflow was however
difficult to estimate especially for earth embankment dams and dikes as they fail gradu-
ally. The discharge depends therefore significantly on the embankment soil properties,
the embankment geometry and the location of impervious elements.
44 Literature review

3.7 Instrumentation and measuring devices

Several laboratory methods were used in the past to determine the breach and free sur-
face profiles during a dike breach. The plane dike erosion process is comparatively
simple to record, as no 3D flow patterns occur. Both the temporal water and sediment
surfaces during an erosion process can be recorded across a glass channel sidewall
(Figure 3.16a). Most researches simply recorded the breach process using a standard
CCD camera. The profiles can be directly determined from the images using standard
graphical programs.
For a spatial breach topography, an observation of the breach profiles across the
channel sidewall is impossible, given the 3D erosion patterns. If the pilot channel is
located at the channel side wall, the dike breach may be recorded but only the breach
profile directly at the channel side wall is visible (Figure 3.16b). Therefore, other meas-
uring techniques have to be applied to the spatial dike breach problem. Intrusive instru-
mentation including point gages are not applicable as they disturb the erosion process.
Coleman et al. (2002) drained the reservoir at several stages of the erosion process to
record the ‘dried’ 3D surface topography using a surface profile measurement system.
However, the effort to determine these profiles is comparatively high, and the degree of
disturbance the embankment during discharge turn-off considerable. The embankment
had to be entirely rebuilt once the reservoir was drained and the test was repeated to
measure the breach profile at an advanced stage. One experiment only resulted in one
temporal breach topography.

(a)

(b)

Figure 3.16 Typical (a) plane and (b) spatial dike breach profiles recorded through channel side
wall
Literature review 45

Spinewine et al. (2004) used a range of measuring systems to determine both dis-
charge parameters and breach geometry during a dike breach test. They introduced a
non-intrusive technique involving a sweeping laser light sheet system combined with
digital imaging algorithms. The principle involves a laser light sheet to illuminate a
particular breach cross-section capturing the line imprinted on the dike onto digital
images. The technique can be used to obtain full digital terrain models of an evolving
breach topography. However, as the laser passes through the water surface it undergoes
refraction effects and therefore the measured bed level has an error proportional to the
water depth and the water surface slope.
Pickert et al. (2004, 2011) used a so-called Fringe Projection to continuously meas-
ure 3D breach profiles. An alternating black/transparent stripe pattern was produced on
the dike surface using a slide projector and recorded with a video camera through the
channel sidewall. The spatial breach profile at various times was determined from the
camera images. The breach process could only accurately be observed during the initial
phase, however, as the observation through the sidewall failed for large breach widths.
Rozov (2003) measured the breach process using washout indicators installed inside
the model dam. These indicators work on the principle of circuit breaking if they are
reached by the moving sediment. A detailed analysis of the breach profiles was not
provided, however.

3.8 Scale effects

3.8.1 Model similarity

To achieve perfect similitude between a hydraulic model and its prototype, geometric,
kinematic and dynamic similitude must be satisfied (Novak 1984, Heller 2007). Geo-
metric similarity exists if the ratio of all linear dimensions between a model and the
prototype are identical. The geometric scale factor can be defined as: (1) Ratio of a
variable in a model to the corresponding variable in its prototype (Yalin 1971); or (2)
Ratio of a variable in a prototype to the corresponding variable in its model (Heller
2011). Here, the scale ratio is defined as λ = Lm/Lp where subscripts m and p refer to
model and prototype, respectively. Kinematic similarity implies that the ratios of veloci-
ty and acceleration between model and prototype are equal. Dynamic similarity requires
46 Literature review

that the ratio of all hydraulic forces between the model and prototype are identical. The
governing hydraulic forces include gravity, viscosity, pressure, surface tension and
elasticity.
For hydraulic modelling, the important parameters are dimensionless ratios formed
with respect to inertia. The ratio of inertia and gravity forces is represented by the
Froude number F as

V
F (3.50)
gL 0.5

with V = flow velocity, g = gravitational acceleration, and L = scaling length.

The Reynolds number R accounts for viscosity and is the ratio of inertia and viscous
forces as

VL
R (3.51)


with ν = kinematic fluid viscosity (ν = 1.01·10-6 m2/s for water of 20°C). The Weber
number W accounts for surface tension and is the ratio of inertia and surface tension
forces as

V 2 L
W (3.52)


with σ = surface tension (0.073 N/m for water of 20°C).

If the model is different in scale from the prototype, no fluid may satisfy all force ra-
tio requirements. An open-channel model, such as a dike breach due to overtopping,
generates normally a free surface process dominated by gravity and therefore operates
according to Froude similitude. To respect the Froude similitude, λF = 1 is required,
where λF is the scale factor for the Froude number. The scale relationships based on
Froude similitude are shown in Table 3.1.
Regarding modelling erosion and sediment transport in small scale, too small Reyn-
olds numbers lead to reduced turbulence and therefore decreased sediment transport in
the model. A too small Weber number leads to decreased discharge and flow velocity
and therefore as well to reduced sediment transport. To ensure that both viscous and
surface tension forces be sufficiently small in a Froude-scale model, several limitations
Literature review 47

have to be respected. Flow in the scale model has to be fully turbulent link in the proto-
type. In analogy to the turbulence limit of R = 2,320 in pipe flow (L = pipe diameter),
the Reynolds number has to be higher than 2,000 to 5,000 in a scale model (Kobus
1984, Hentschel 2007, Recking 2008). For R ≥ 5·104 the turbulence structure is general-
ly independent of geometrical length (Kobus 1984).

Table 3.1 Scale factors λ = Lm/Lp according to Froude similitude

Characteristics Scale factor

Length [m] λ = Lm/Lp
2
Area [m ] λ2
Volume [m3] λ3
Time [s] λ1/2
Velocity [m/s] λ1/2
Acceleration [m/s2] 1
Force [N] λ3
Discharge [m3/s] λ5/2

The effects of viscosity and surface tension on embankment weir flow were studied
by Ranga Raju et al. (1991). They identified the characteristic parameter R0.2W0.6. For
R0.2W0.6 > 103 scale effects due to surface tension and viscosity are absent, almost
independent of embankment height and discharge. Fritz and Hager (1998) found that
scale effects due to viscosity are absent for flow depths h ≥ 0.05 m. Martin and Pohl
(2000) stated that the minimum flow depth has to be h > 5ks, with ks = equivalent sand
grain roughness, and h > 5 mm to minimize effects of surface tension. To avoid effects
of cohesion, the mean sediment diameter has to be d > 0.20 mm (Zanke 1982), whereas
Hager and Oliveto (2002) postulated that d > 0.80 mm.

3.8.2 Scaling sediment transport

Sediment transport is the governing mechanism in the dike breaching process. There-
fore, the model must simulate the tractive shear stress τ0 accurately, because it causes
the drag force required to overcome the gravity forces resisting particle entrainment.
The shear stress on a sediment particle fluctuates due to turbulence and both the drag
force and turbulence are functions of viscous forces and consequently the Reynolds
number. Therefore, a Froude similitude model does not necessarily simulate the tractive
48 Literature review

forces and sediment transport accurately, as the Reynolds number may be too small in
the model.
The critical shear stress τc* for incipient motion depends on the grain Reynolds num-
ber R* (Shields 1936). In general, τc* is constant for R* > 500. However, if the sediment
is scaled according to geometrical scaling, R* may fall below 200. The critical shear
stress of the model flow is therefore smaller than in the prototype and grains are already
entrained for comparatively smaller discharges. For a model with 5 < R* < 200, the unit
sediment discharge rate for the model would be higher than for the prototype. There-
fore, the sediment size must be adjusted in these models to compensate for a grain
Reynolds number that is too small to properly simulate sediment transport. Pugh (1985)
presented the settling velocity adjustment. For grains with d > 1 mm, the settling veloci-
ty is a function of d1/2. Smaller particles settle, however, at progressively smaller veloci-
ties. By increasing the size of all model grain sediment for 2 ≤ R* ≤ 100, the settling
velocity can be corrected to the proper value for Froude scaling in the model. Zarn
(1992) presented a scaling method for grain size distributions. First, the grain size dis-
tribution is scaled down according to the model length scale. To maintain the scaling
law according to Froude for the critical shear velocity in the model and nature, all grains
that exhibit a grain Reynolds number R* below 200 must then be coarsened. The criti-
cal shear stress τc* in the model are therefore properly scaled for all grain sizes.
As flow depth, flow velocity and slope are given by Froude similitude, the sediment
transport initiation similitude can be achieved by using a suitable alternative material.
By modifying the sediment density (artificial sediment) and the grain size distribution,
both incipient motion and bed-load transport are properly modelled (Hentschel 2007). In
particular cases it is impossible to adjust the scaled sediment. To ensure that viscous
effects remain insignificant as the same length scale is adopted for the model dimension
and sediment size (λd = λL), the particle Reynolds number R* should be kept higher than
the critical threshold of 70 (Dupont et al. 2007, Recking et al. 2008).

3.9 Research gaps and purpose of present study

The literature review reveals that a number of comprehensive hydraulic model studies
on dike breaching due to overtopping are available. These studies cover a wide range of
parameters regarding hydraulic conditions, dike designs and sediment properties. De-
spite this knowledge, the breach process is still poorly understood because only limited
Literature review 49

test data are available due to a significant laboratory effort. Systematic model tests are
scarce and most research is limited to one specific test setup or one specific failure
mode. Most results may therefore only be applied to particular problems, e.g. to special
dike configurations or to a particular grain size. Further, new measuring devices to
simultaneously determine water surface and sediment profiles were rarely used and are
not yet completely established regarding reliability and measuring errors. Most systems
were only applied by one specific research team and the overall experience is therefore
limited.
In addition, scaling problems related to material type (cohesive, non-cohesive) and
breach formation (seepage, sediment transport) are not yet completely understood. The
conclusions from small-scale experimentation and their applicability to prototype condi-
tions must therefore be handled with caution. Most laboratory research must be consid-
ered principally qualitative. Regarding the breach process, sediment transport on steep
slopes and the lateral erosion process are currently not fully understood. Most sediment
transport equations are only valid for comparatively flat slopes under almost uniform
flow conditions. While the plane dike breach process is only governed by surface sedi-
ment erosion, the lateral erosion process is additionally affected by various geotechnical
failure mechanisms.
Accurate and thorough laboratory tests are essential for validating numerical models.
There is still a lack of laboratory data on the breach process (breach profiles, free sur-
face profiles, or breach discharge) to test and further improve existing numerical breach
models. The main purposes of the present study are therefore summarized as follows:
 Establish a simple hydraulic test setup that allows for investigating plane dike
breaching due to overtopping under a wide range of dike dimensions and hy-
draulic conditions with a comparatively small laboratory effort
 Investigate possible scale effects and define model limitations for dike breach
experimentation
 Investigate the plane erosion processes with a systematic test procedure
 Obtain information regarding dike breach profiles, breach discharge, flow ve-
locities, shear stresses, and erosion rates
 Develop a new measurement system to investigate the spatial dike breach
process due to overtopping
 Provide laboratory data for numerical modelling purposes
50 Literature review

Two new model channels were installed at VAW to achieve these requirements. Sys-
tematic plane dike breach tests due to overtopping have been carried out and analysed in
detail. A novel optical measurement system was applied to the dike breach phenomenon
to record the spatial breach process.
Hydraulic scale models 51

4 Hydraulic scale models

4.1 Introduction

Two sets of flume experiments were conducted using two different hydraulic models.
The first channel was 0.40 m wide and suited for plane dike breach tests. The spatial
dike breach tests were conducted in a 1.0 m wide channel. Both channels are hereafter
described in detail, including the sediments and measuring devices used, the test proce-
dure, the test program and the data analysis procedure.

4.2 Plane dike breach tests

4.2.1 Overview

The main boundary conditions for the plane hydraulic tests are summarized as follows:

 Trapezoidal model dike

 Homogenous, non-cohesive sediments
 Neither surface layer nor core, i.e. homogenous embankment
 Seepage controlled using bottom drainage
 Steady inflow, falling reservoir level
 Optical recording using camera

Preliminary tests were carried out to establish the most favourable test setup. A general
test series was conducted regarding model and scale effects. Systematic tests were then
conducted regarding the hydraulics of the dike breach process.

4.2.2 Model channel I

The plane dike breach tests were conducted in the glass-sided VAW dike-breach chan-
nel (Figure 4.1). It is equipped with a pump of 70 l/s as maximum discharge capacity.
The horizontal channel is 0.40 (0.002) m wide, 0.70 m high and 8 m long. The channel
52 Hydraulic scale models

width was adjusted with PVC insets to reduced widths of 0.20 m and 0.10 m, respec-
tively. The intake is 0.66 m long and a flow straightener generated undisturbed inflow.
Dikes were inserted at various distances xD from the intake. The eroded sediment was
collected in a basket at the channel end. The seepage through the dike was reduced with
a bottom drainage (Chapter 4.2.4). The seepage flow through the bottom drainage was
separately collected at the end of the channel.

0.66 m
5
xD
8.0 m
4

0.4 m
3
1 6 0.7 m
Q 8

2
1 IDM 5 Flow straightener
2 Computer 6 Dike
Q 7
3 Valve 7 Sediment basket
4 Intake 8 Bottom drainage

Figure 4.1 Model channel for plane dike breach tests

4.2.3 Test setup

Figure 4.2 shows the test setup. Trapezoidal dikes (subscript D) of 0.40 m maximum
height were inserted at various distances xD from the intake. The origin of the coordi-
nate system (x, y, z) was located at the upstream dike toe. The main parameters investi-
gated herein were: Approach flow (subscript o) discharge Qo = 1 to 64 l/s, dike height
w = 0.10, 0.15, 0.20, 0.30, 0.40 m, dike width b = 0.10, 0.20, 0.40 m, dike crest (sub-
script K) length LK = 0.05, 0.10, 0.20 m and uniform sediment size of d = 0.31, 1.0, 2.0,
4.0, 5.5, 8.0 mm with an uniformity coefficient σg = (d84/d16)0.5 < 1.2. Both the up- and
downstream dike slopes So = Sd = 1:m were always identical and varied between 1:2
(V:H), 1:2.5 and 1:3 resulting in a total dike length of LD = 2·m·w + LK. During each test
both the water (subscript w) surface hw (x,t) and the sediment surface z (x,t) with
t = time were recorded using a 30 Hz CCD camera.
Hydraulic scale models 53

4.2.4 Bottom drainage

To control the seepage through the dike and prevent sliding failure (Chapter 5.2.1) a
bottom drainage was added to the channel. A second PVC bottom was inserted 20 mm
above the original invert. Drainage holes were located in the second bottom along the
upstream dike portion. Drainage holes of 0.02 m in diameter were arranged in a
0.05 m 0.035 m grid. A fine wire net of 0.5 mm opening width prevented dike sedi-
ment washout. As the upstream water level increased, a small discharge portion passed
the drain and bypassed the dike. The drainage discharge was collected in a separate tank
and measured after each Test. The drainage discharge does vary during the experiment
and increases especially towards test end, once all sediment is eroded and the water can
directly enter the drainage holes. However, this was not recorded due to the test setup
and the very small drainage discharges. To compare the drainage discharge with the
constant inflow discharge, the accumulated water volume in the separate tank was con-
verted into a constant drainage discharge using the test duration. The drainage discharge
was typically 3% of Qo, reaching a maximum of 8%.

Flow straightener
LK

Qo 1 1
w
(a) So Sd
x z d

xD LD
Bottom drainage
Intake

(b) Qo Dike b
y
x

0.66 m xD LD

Camera

Figure 4.2 VAW plane dike breach model (a) streamwise section, (b) plan view, and notation
54 Hydraulic scale models

4.2.5 Sediment

Uniform sediment of various mean sediment sizes d = 1.0, 2.0, 4.0, 5.5, 8.0 mm was
used for the standard breach tests (Table 4.1, Figure 4.3). The sediment was homoge-
nous and non-cohesive (sand or gravel) with a density of ρs ≈ 2,650 kg/m3. To addition-
ally test the limit of application and to analyze cohesive effects, a cohesive sediment
with d = 0.31 mm was used. The sediment showed cohesive characteristics but still
eroded gradually without sticking together or forming clumps. All sediments were
directly screened at VAW from conventional sand and gravel bought at local gravel pits.
The standard VAW screens resulted in geometrical standard deviation of the grain size
distribution σg = (d84/d16)0.5 < 1.2. The sediment was completely dry for all tests.

Table 4.1 Grain size characteristics of used sediments

Sediment VAW Screen d ρs

[mm] [mm] [kg/m3]
1 0.25-0.37 0.31 2,650
2 0.75-1.02 1.0 2,650
3 1.49-2.00 2.0 2,650
4 3.48-4.10 4.0 2,650
5 5.10-6.00 5.5 2,650
6 7.50-8.30 8.0 2,650

Figure 4.3 Granular sediments for plane dike breach tests

Hydraulic scale models 55

4.2.6 Discharge measurement

The channel was equipped with a pump of maximum discharge capacity up to 70 l/s.
Water passed from the pump through a supply pipe located below the channel into the
intake section (Figure 4.1). The water discharge was measured in the supply pipe
( = 150 mm) by an electromagnetic flow meter (IDM) to ±0.5% of the measured
value. The constant inflow discharge was regulated with a valve manually for
Qo ≤ 20 l/s and remote-controlled by a computer for Qo > 20 l/s. For discharges up to
20 l/s, the inflow discharge was manually set to the test value, as fluctuations were too
high when using the remote-controlled valve. The accuracy was then about ± 0.1 l/s.
Discharges higher than 20 l/s were directly adjusted by the computer resulting in an
accuracy of ± 1 l/s.

4.2.7 Measuring camera

All breach processes were recorded using a 30 Hz FireWire CCD camera (AVT Guppy,
Allied Vision Technologies) of a resolution of 1,034(H)  778(V) pixels, resulting in
an accuracy of ±2 mm or higher, depending on the image frame. The camera with a
wide angle lens (focal length = 5mm) was mounted up to 3 m from the channel axis,
resulting in image distortion, which was ortho-rectified with an image correction pro-
gram using a calibration panel. The camera recorded single frame images via FireWire
directly on the PC.

4.2.8 Experimental procedure

All experiments were conducted using a constant approach flow discharge Qo to obtain
simple boundary conditions and to additionally reduce the laboratory effort. The dry
dike material inserted was neither compacted nor were surface or core layers added. A
dike cross-section made out of cardboard was attached to the outer channel glass wall
and used as scheme for dike construction. The dike setting was accurate to one grain
size. Steady discharge Qo was added without any tailwater submergence, such that the
dike overflow was always free and the tailwater did not affect the basic erosion process.
The valve was opened as fast as possible to attain the steady discharge before overtop-
ping started. Therefore, reservoir filling was relatively fast, typically within 30 s. The
56 Hydraulic scale models

rapid approach flow generated a small bore-like wave propagating on the dry approach
flow channel floor. The wave was reflected at the dike, travelled upstream to the chan-
nel inlet and returned with a much smoother surface due to the flow straightener. Once
dike overtopping had started the water surface was almost horizontal. The duration from
erosion start to test end was between 500 s and 1000 s, depending on sediment size and
inflow discharge. The tests were stopped when the equilibrium state was reached and
the dike remained stable, or when the entire dike was eroded. Given the short breach
duration and the relatively high overflow discharges, no constant reservoir level was
attained. The results are therefore limited to dike breach processes with a falling reser-
voir level. The entire breach process was recorded with the CCD camera through the
channel side wall. The images were analysed using a standard graphical program (Chap-
ter 4.3.9).

4.2.9 Test program

Preliminary Tests
Several preliminary tests were carried out to establish a simple test procedure. The main
aspect of these tests was to guarantee a dike failure process that is limited to overtop-
ping erosion. The preliminary tests indicated the decisive effect of seepage on the dike
breach process. A bottom drainage was subsequently added to the test setup to achieve a
‘pure’ overtopping breach process. The results of the preliminary tests are presented in
Chapter 5.2.1 but not further used for data analyses.

Scale effects Tests

A total of 39 experiments were conducted (Table 4.2) to: (1) Verify the test repeatabil-
ity, (2) Determine the side wall effect, and (3) Detect scale effects by means of scale
families. The repeatability was tested using five dikes of various scales and sediment
sizes (Tests 1 to 15). The side wall effect was determined using three dikes of identical
cross-section and sediment size with widths of b = 0.10, 0.20 and 0.40 m. To account
for the effects of sediment size and dike dimension, the side wall effect was investigated
with five different dikes (Tests 16 to 30). All tests were conducted with So = Sd = 1:2.
Scale effects were analysed using scale families. The scale factor λ = Lm/Lp is the ra-
tio of a variable in the model to the corresponding variable in its prototype (Novak
1984). Three model dikes of different scaling factors were considered (Figure 4.4), in
Hydraulic scale models 57

which the largest scale model was taken as reference. The largest tested dike had a scale
factor of λ1 = 1 (w = 0.40 m, b = 0.40 m), the medium dike of λ2 = 0.5 (w = 0.20 m,
b = 0.20 m) and the smallest dike of λ3 = 0.25 (w = 0.10 m, b = 0.10 m), respectively
(Figure 4.4). Potential scale effects exist if the results of the smallest scale model devi-
ate from the two larger models. The uncertainty of this approach is obviously the rela-
tion of the largest scale model to the prototype.
A dike breach due to overtopping is a free surface process dominated by gravity,
therefore Froude similitude was applied (Chapter 3.8.1). To respect the Froude (sub-
script F) similitude, λF = 1 is required, where λF is the scale factor of the Froude number
F. The geometric scale factor is given by λ = Lm/Lp. The scale factors for velocity V,
time t and discharge Q follow from Froude similitude as λV = λ0.5, λt = λ0.5, λQ = λ2.5,
respectively. Herein, the dike dimensions for λ = 1 were limited by the test facility,
allowing for a maximum dike height of 0.40 m and dike width of 0.40 m. The two other
dike dimensions resulted from λ = 0.5 and λ = 0.25. Further parameters linked to the
model scale are the discharge Qo, the water storage volume VW upstream from the dike
before overtopping, related to the toe distance xD in Figure 4.2, and the sediment diame-
ter d. The toe distance for λ3 = 0.25 was set to xD = 0.17 m resulting in a storage volume
(including the volume of the intake) of

VW  bx D  w  w 2  0.66  w = 0.0093 m

3
(4.1)

The storage volume for the other two scales result from the Froude similitude as
VW = 0.074 m3 for λ2 = 0.5 and VW = 0.595 m3 for λ1 = 1. This in turn results in the toe
distances xD = 1.0 m for λ2 = 0.5 and xD = 2.66 m for λ1 = 1. These toe distances guaran-
teed that the deposition zone for λ1 = 1 was large enough.
The sediment diameter was modelled with the length scale and the erosion threshold
expressed with the Shields number. Therefore, the tractive shear stresses of the different
sediments are not properly scaled. Further, the percolation was not properly scaled by
the Froude similitude as other laws dominate this process. However, the percolation
effect only controlled the overtopping feature. Three scale families were analysed, each
consisting of three individual tests (Test 31 to 39).

Breach process Tests

A total of 29 Tests were conducted to systematically investigate the breach process and
the breach outflow (Table 4.3). The main parameters varied were the constant inflow
58 Hydraulic scale models

discharge Qo, the mean sediment diameter d and the dike height w. Both the up- and
downstream dike slopes were varied in Test 66 (So = Sd = 1:2.5) and Test 67
(So = Sd = 1:3). Test 68 was carried out with the cohesive sediment of d = 0. 31 mm. The
following parameters were kept constant during all tests: Dike width b = 0.20 m, crest
length LK = 0.10 m and toe distance xD = 1.0 m. Several tests of the scale effects test
series (Table 4.2) are as well presented along with the breach process tests to cover a
wider parameter range.

LK

Qo λ=1 w
λ = 0.5 1 1
λ = 0.25 So Sd
d

xD LD

Figure 4.4 Model dikes for scale family tests, notation

4.2.10 Data analysis procedure

All required data were directly determined from the recorded CCD images. Prior to each
breach test, a calibration panel was recorded. All images were then corrected for distor-
tion with the program DaVis©, LaVision (Göttingen, Germany). The pixels were as-
signed to the real length and width of the images and the origin of the coordinate system
(x, y, z) was fixed at the upstream dike toe. All dimensions could consequently be meas-
ured directly from the camera images. The following data were obtained at various
times t:
 Water surface profile hw (x,t)
 Sediment surface profile z (x,t)
 Overflow depth ho
 Upstream reservoir water level hR
 Crest Length LK or crest radius R
 Up- and downstream weir face angles αo and αd

Both the water and sediment surface profiles were normally determined with a spacing
of Δx = 20 mm.
Hydraulic scale models 59

Table 4.2 Test program and basic test parameters for scale effects

Test λ xD w LK b d Qo
[-] [m] [m] [m] [m] [mm] [l/s]
1,2,3 1.00 0.10 0.05 0.10 1.0 2.0
Repeatability 4,5,6 1.00 2.0
7,8,9 1.00 4.0
10,11,12 2.66 0.20 0.10 0.20 2.0 11.3
13,14,15 2.66 4.0
16 1.00 0.10 0.05 0.10 1.0 2.0
17 0.20 4.0
18 0.40 8.0
19 1.00 0.10 0.05 0.10 2.0 2.0
20 0.20 4.0
21 0.40 8.0
Side wall effect

22 1.00 0.10 0.05 0.10 4.0 2.0

23 0.20 4.0
24 0.40 8.0
25 2.66 0.20 0.10 0.10 2.0 5.7
26 0.20 11.3
27 0.40 22.6
28 2.66 0.20 0.10 0.10 4.0 8.0
29 0.20 16.0
30 0.40 32.0
31 0.25 0.17 0.10 0.05 0.10 1.0 2.0
32 0.5 1.00 0.20 0.10 0.20 2.0 11.3
33 1 2.66 0.40 0.20 0.40 4.0 64.00
Scale families

34 0.25 0.17 0.10 0.05 0.10 1.0 1.00

35 0.5 1.00 0.20 0.10 0.20 2.0 5.7
36 1 2.66 0.40 0.20 0.40 4.0 32.0
37 0.25 0.17 0.10 0.05 0.10 2.0 2.0
38 0.5 1.00 0.20 0.10 0.20 4.0 11.3
39 1 2.66 0.40 0.20 0.40 8.0 64.0
60 Hydraulic scale models

Table 4.3 Test program for breach process tests

Test d [mm] w [m] So, Sd [-] Qo [l/s]

40 2.0
41 1.0 0.20 4.0
42 6.0
43 2.0 0.10 8.0
44 2.0
45 4.0
2.0 0.15
46 8.0
47 12.0
48 1.0
49 2.0
50 4.0
51 6.0
52 2.0 0.20 8.0
1:2
53 10.0
54 12.0
55 16.0
56 20.0
57 2.0 0.30 8.0
58 6.0
59 8.0
60 4.0 0.20 10.0
61 12.0
62 16.0
63 6.0
64 5.5 0.20 8.0
65 12.0
66 2.0 0.20 1:2.5 8.0
67 2.0 0.20 1:3 8.0
68 0.31 0.20 1:2 6.0
Hydraulic scale models 61

Figure 4.5 shows a definition sketch including the measured parameters. The measur-
ing accuracy was ±2.0 mm or higher, depending on the image frame. To compare the
overall erosion process for different breach tests, the sediment surface profiles z(x) at
various times t were normalized as X = x / LD and Z = z / w. The starting point t = 0 s
was always set when the upstream water level reached the dike crest.
The crest radius was determined by measuring three (x, z)-coordinates of the circular-
crested dike profile. The breach discharge Q was calculated using the broad-crested
weir formula (Eq. 3.46) with Cd according to Eq. (3.47) for a trapezoidal breach profile
and with Eq. (3.49) for a circular-crested breach profile.

Figure 4.5 Measured data from camera images

To determine the energy head H, curvature and slope effects have to be considered,
as the pressure head differs from the flow depth. The slope of both the dike breach
profile and the water surface profile vary constantly along x. Three quantities are im-
portant, namely the flow depth hw determined from the camera pictures (which is the
vertical distance from the bottom to the surface), the length of the orthogonal trajectory
of the streamlines N, and the pressure head h (which equals the vertical projection of the
normal N). Figure 4.6 shows a definition sketch of the flow geometry where θu and θo
are the local inclination angles of the bottom (dike surface) and top (water surface)
streamlines, respectively.
The two boundary streamlines are connected by the orthogonal trajectory of length
N, which is assumed to form a circular arc. According to Hager and Hutter (1984),
geometric relations imply the following relations

h sin u  sin o
 (4.2)
N u  o

hw cos u tan u  tan o 

 (4.3)
N u  o
62 Hydraulic scale models

With hw given from the camera pictures, both N and h can then be determined. For
θu = θo, i.e. parallel streamlines, it follows that

N  hwcos u for θu = θo (4.4)

h  hwcos 2 u for θu = θo (4.5)

For small slopes of the streamlines, hw = N = h, with h as the flow depth.

Figure 4.6 Definition of flow geometry at position x. The pressure head h differs from the
flow depth hw (Hager and Hutter 1984, adapted)

To determine the local angles θu and θo, two points spaced by Δx of the water surface
profile and the sediment surface profiles were used, respectively. The angles follow as
tanθu = Δz/Δx and tanθo = Δ(z+hw)/(Δx). All further parameters were calculated for each
cell of length Δx and applied in the cell center.
The cross-sectional average velocity along the dike breach is V = Q/(bN). The local
Froude number along the dike is F = V/(gh)0.5 and the local Reynolds number
R = V·4Rh/ν, respectively, with the hydraulic radius Rh = A/U = bN / (b+2N). The ener-
gy head follows from H = αe V2/2g + h + z (with αe = 1). Additionally, the bed shear
stress τb (Eq. 3.2), the shear velocity U* (Eq. 3.3), and the grain Reynolds number R*
can be computed along the dike breach profile for each cell of length Δx. The critical
shear stress τc* for each material follows from Eq. 3.28.
Hydraulic scale models 63

Once the bed shear stress along the dike breach profile is given, the dimensionless
bed shear stress τ* (Eq. 3.31) can be calculated. The dimensionless bed-load transport
rate q* follows then from one of the existing transport formulae and the volumetric
transport rate qb [m3/ms] for each cell is calculated with Eq. (3.30).
As sediment was deposited almost along the entire channel and therefore not entire-
ly recorded, the bed-load transport rate was determined at the original downstream dike
toe x = LD. The volumetric transport rate qb [m3/s] follows as

qb  qb b (4.6)

The calculated qb is only valid for the actual time step as it varies constantly. For
t < 20 s, qb was determined in an interval of 1 s. For t ≥ 20 s, the interval was increased
to 10 s or more to decrease the evaluation effort. The bed-load transport rate qb during
the breach process can then be displayed for each calculated time step.
The calculated bed-load transport rate qb can be compared with the actual transport
rate during a dike breach obtained directly from the camera images. With the sediment
surface z (x,t) and the constant dike width b, the temporal dike volume VD(t) is found.
For a dike slope of So = 1:m, the initial dike volume is


VDo  b LK  w  m  w 2  (4.7)

at t = 0 s before overtopping starts. The eroded dike material ΔVD(t) up to the original
dike length LD follows from the difference of the dike volumes at two subsequent times
ΔVD = VD(t1)–VD(t2) and the bed-load transport rate per second follows as

VD (t1 )  VD (t 2 )
qb  (4.8)
t 2  t1

Various bed-load sediment transport formulae can therefore be applied and compared to
the actual bed-load transport.
Note that all results depend essentially on the determination of both the water surface
profile and the sediment surface profile. The spacing Δx between two points essentially
defines the accuracy of e.g. the dike volume, dike slope and water surface slope. How-
ever, especially the slopes of the dike and water surface are sensitive to small variations
and have a significant effect on the energy head and the bed shear stress. If Δx is too
small, the fluctuations of both the dike and water surface slopes may increase consider-
64 Hydraulic scale models

ably. If Δx is too large, changes in the dike erosion profile and curvatures of the water
surface may be ignored. For the Test 52 evaluated in Chapter 5.3, the spacing of
Δx = 20 mm provided the most consistent results.

4.3 Spatial dike breach tests

4.3.1 Overview

The main boundary conditions for the spatial hydraulic tests are summarized as follows:
 Trapezoidal model dike
 Homogenous, non-cohesive sediments
 Neither surface layer nor core
 Pilot channel at channel side wall
 Constant reservoir level
 Optical recording using cameras
The purpose of the spatial dike breach tests was to apply and test a new non-intrusive
measurement system, developed by AICON 3D Systems Ltd., Braunschweig, Germa-
ny. A preliminary spatial dike breach test was carried out in the hydraulic model chan-
nel I (Chapter 4.2.2) to obtain general information about breach physics. Only the longi-
tudinal breach profiles along the breach centreline and the corresponding flow depths
were recorded through the channel side wall. Compared to the fully spatial breach char-
acteristics, this approach is considerably less complex both from the model set-up and
the data analysis. However, 3-dimensional erosion pattern can not be recorded in chan-
nel I.

4.3.2 Model channel II

The 3D-experiments were conducted in another glass-sided model channel at VAW.

The channel was described in detail by Kohli (1998). Its discharge capacity is
Q = 130 l/s, it is 1.0 m wide and 9 m long, with a working section of roughly 5 m. The
inflow discharge can be regulated with a valve both manually and remote-controlled by
a PC. A flow straightener generated undisturbed inflow. Test dikes of up to 0.50 m
height can be erected at various distances xD from the intake section. The tailwater was
Hydraulic scale models 65

remote-controlled with a flap gate at the channel end and the flow features were optical-
ly recorded either through the large channel windows or from the top.

4.3.3 Test setup

The notation is equal to that of plane dike breach tests (Figure 4.2), except that the
erosion was initiated using a pilot channel. The test setup is shown in Figure 4.7. The
trapezoidal dike consisted of the same material as the sediment bed (d = 1.14 mm) and
was erected on top of it. No surface layer or core were added. A 0.04 m deep and
0.07 m wide triangular pilot channel was cut in the embankment crest along the channel
sidewall, to initiate the breach process, with the channel wall simulating the breach
centreline. This allowed recording the longitudinal breach profiles and the correspond-
ing flow depths along the breach centreline. The reservoir was filled comparatively fast
by opening the valve. Once the water entered the pilot channel, the discharge was throt-
tled to prevent entire dike overtopping. During the breach process, the discharge was
manually adjusted to keep the reservoir level at a more or less constant level. This was
achieved in the beginning, but failed for large times as the erosion process was too fast.
The eroded material was deposited behind the dike in the model channel. Sediment
washout was prevented with a sill.

Flow straightener

b = 1.0 m
Qo Dike Sill
(a)
Sediment bed

Intake

Dike
(b) Qo b

Pilot channel

Figure 4.7 Test setup for spatial dike breach tests with (a) streamwise section, (b) plan view
66 Hydraulic scale models

4.3.4 Sediment

A horizontally graded washed river sand of mean diameter d = 1.14 mm was used for
both sediment bed and dike construction (Figure 4.8). The sand was homogenous and
non-cohesive with a density of ρ ≈ 2,650 kg/m3.

Figure 4.8 Granular sediment for spatial dike breach tests

4.3.5 Discharge measurement

The channel was equipped with a pump of maximum capacity of 130 l/s. The water
discharge was measured in the supply pipe ( = 150 mm) by an electromagnetic flow
meter (IDM) of ±0.5% measuring accuracy of the measured value. The variable inflow
discharge Qo was continuously measured and recorded during a dike breach test.

4.3.6 Measurement system

A stereoscopic-videometric AICON measurement system was used to determine the

breach profiles. It was originally developed for Bundesanstalt für Wasserbau (BAW),
Karlsruhe, Germany, to continuously measure bed elevations and flow depths in a phys-
ical model of Elbe River (Godding at al. 2003). For the present research, the system was
further developed and adapted by AICON 3D Systems Ltd.
The so-called AICON-method refers to videometry involving a rectangular grid pro-
jected onto a sediment surface using a high-power slide projector. The test setup is
shown in Figure 4.9. The projector was mounted above the hydraulic channel. The
sediment surface above and the bed forms below the water surface reflect a non-
rectangular distorted ‘grid’ structure. The horizontally-projected grid spacing on the
Hydraulic scale models 67

sediment surface was about 30 mm in the present test set-up. The grid was recorded
using three synchronised CCD cameras located above the channel. The cameras have a
resolution of about 2 Mio. pixels and a recording frequency of 30 Hz. The digital imag-
es are directly transferred to the measuring computer.
A fourth camera close to the channel was used to record the flow depth along the
channel sidewall. This allowed for the determination of the projected grid distortion
below the water surface for all flow depths that are equal to the flow depth along the
channel side wall. As the flow depth varies along the breach width, it was impossible to
properly correct the distortion effect at some distance from the sidewall. For the first
tests, the flow depth along the channel sidewall was approximately used to correct the
entire projected grid below the water surface.
Before each test, the relative camera orientation and the camera parameters (inner
orientation) were determined using a calibration panel. It consisted of a number of
measuring marks whose 3D-coordinates were known. Each measuring mark had a
distinctive code number. The calibration panel was recorded several times at different
positions with all four cameras and all required parameters were calculated using the
software AICON ProSurf (Godding 1993) to define the 3D-position of each camera.
During a test, the projected grid was measured with the cameras and 3D-coordinates
were calculated for each grid point.
To transform all calculated coordinates in an overall, superior coordinate system, the
hydraulic channel was equipped with coded measuring marks. The coordinates of these
marks were determined prior to a test. The coded marks on the hydraulic channel were
recorded along with the projected grid and assigned to 3D-coordinates, to allow for a
3D-Helmert-transformation of the measured grid points into the channel system. During
a test, the channel was only lit by the slide projector. Figure 4.10 shows all four camera
images before reservoir filling. During the entire breach process, the projected grid was
recorded with the cameras. All images were then processed using the AICON ProSurf
software. A set of four images was available for each time step, with the 3D-coordinates
of the breach profile determined as follows: (1) Water elevation was manually measured
from the side camera image to determine the distortion of the grid projection on the
sediment surface below the water; (2) A starting grid point was manually located in two
of the three top camera images; (3) Based on the starting point, the corresponding grid
points were automatically searched and determined using the three associated camera
68 Hydraulic scale models

images; (4) Using camera calibration and channel orientation data, each grid point was
assigned to 3D-coordinates.

Figure 4.9 Test setup for spatial dike breach tests with AICON measurement system

4.3.7 Experimental procedure

Before each experiment, the sediment bed was flattened using a grader according to
Unger (2006). The dike was then erected on top of the sediment bed. The dike cross-
section was marked on both side walls and used as pattern for dike construction. The
dike setting was accurate to ± 2d. The 0.04 m deep and 0.07 m wide triangular pilot
Hydraulic scale models 69

channel was then cut by removing the crest material along the channel side wall. The
slide projector was started and the dry dike was recorded with the AICON system to
obtain the original dike geometry. The reservoir was filled until the water entered the
pilot channel. The inflow discharge was kept low in the beginning to avoid overtopping
of the entire dike and increased with ongoing erosion to attain a constant reservoir level.
During the entire test, the projected grid was recorded with a frequency of 1 Hz. The
breach test was stopped when the erosion reached the original sediment bed or the entire
dike was eroded. The water was drained and the dry breach profile again recorded with
the AICON system.

Figure 4.10 Camera images before reservoir filling with projected grid on dike surface and
measuring marks on channel side wall, () principal flow direction

4.3.8 Test program

All tests were conducted using the same dike dimension, as the main goal of these tests
was to apply and test the AICON system. The dike dimensions of all tests are given in
Table 4.4.
70 Hydraulic scale models

Table 4.4 Test program for spatial breach tests

Test xD w LK b So = Sd d Qo
[m] [m] [m] [m] (V:H) [mm] [l/s]
1 2.66 0.20 0.10 1.0 1:3 1.14 varied

4.3.9 Data analysis procedure

The detected grid points (Figure 4.11) from the Pro Surf software were further analysed
using standard graphical programs. As each grid point was assigned to 3D-coordinates,
3D breach profiles were directly obtained from the data. The adaptability of the AICON
system to the dike breach problem is discussed in detail and evaluated regarding appli-
cation limitations as well as advantages and disadvantages. Systematic spatial dike
breach tests will be part of the follow-up research project.

Figure 4.11 (a) Recorded camera image during dike breach test, (b) detected grid points by
AICON ProSurf software

4.3.10 Measuring accuracy

The measuring accuracy of the AICON system depends mainly on the test setup regard-
ing illumination, water quality and sediment visibility. A series of plane dike breach
tests were conducted to identify errors resulting from the AICON instrumentation. The
measured dike breach profiles were compared with those resulting from the standard
VAW instrumentation using the side camera. The grid spacing of the AICON instru-
Hydraulic scale models 71

mentation was 0.01 m and the channel width b = 0.10 m was kept small to allow for a
plane flow structure.
In Test 1 a rigid dike body was subjected to a constant discharge Qo. Figure 4.12
compares zeff measured using the side camera with zmeas measured with the AICON
instrumentation, plus the relative error δ. The overall agreement between the VAW and
AICON data is good. Larger deviations are seen at the dike crest due to flow separation
at the upstream edge of the dike crest. Further, the water surface in the crest zone is
rather wavy due to the transition from sub- to supercritical flow. This complicates the
surface detection with the AICON system. The relative error is δ < 2% for x < 750 mm
and increases up to δ = 10-15% for large x and small dike heights. An accuracy of 10%
results in an error of some 2 mm. The same order of measuring accuracy results from
the standard VAW instrumentation using the side camera.

300 10
 [%]
z [mm]

200 0

100 -10

0 -20
0 500 x [mm] 1000

Figure 4.12 Comparison between effective and measured dike surfaces z(x) from recorded
AICON instrumentation and flow depths h(x) for rigid dike body and Qo = 4 l/s

In Test 2, a loose dike body made up of uniform sediment of median grain size
d = 1.5 mm was subjected to plane erosion with Qo = 3 l/s. Figure 4.13 compares the
VAW and AICON-measured sediment surfaces at various times t. The general agree-
ment is remarkably good. Some wavy pattern is noted in the AICON measurement due
to local determination of the dike surface profile given the fixed grid spacing of 0.01 m.
72 Hydraulic scale models

z [mm]
t = 12 s
200

100

0
(a) 0 500 x [mm] 1000

z [mm]
t = 15 s
200

100

0
(b) 0 500 x [mm] 1000

z [mm]
t = 30 s
200

100

0
(c) 0 500 x [mm] 1000

Figure 4.13 Effective and AICON-measured dike surfaces z(x) at various times t
Hydraulic scale models 73

Note that the accuracy of the AICON instrumentation was so far only determined for
the plane dike breach test. The 3D breach profiles cannot be compared with the standard
VAW instrumentation as the dike surface remains invisible from the side. To verify the
AICON accuracy, the 3D breach process has to be stopped at a certain time and the
breach profile has to be measured manually with a point gauge or a laser. This process
is challenging, as the breach profile changes rapidly and the reservoir cannot be emptied
instantly. A test series to determine the AICON accuracy during a 3D breach will be
part of the follow-up research project.
74 Hydraulic scale models
Results 75

5 Results

5.1 Introduction

The results obtained with the physical model investigation are presented here. First,
general results regarding the test setup, the model limitations and the scale effects for
plane dike breach tests are summarized in chapter 5.2. In chapter 5.3, the general ero-
sion process is exemplarily illustrated for a representative dike breach test. The effect of
the various test parameters on the dike breach process is presented in chapter 5.4. Chap-
ter 5.5 refers to normalized results regarding breach discharge, breach profiles and the
erosion process. Finally general results for spatial dike breach tests are summarized in
chapter 5.6.
All results are presented in real-time, with the time t retained in seconds, as chapter
5.5 includes the dimensionless time T.

5.2 Model limitations and scale effects

5.2.1 Effects of seepage

A challenge of the hydraulic model was to limit the failure mechanism to ‘pure’ over-
flow erosion thereby excluding any other failure mode. Preliminary tests indicated that
seepage through the dike has a decisive effect on the dike breach process. As the dike is
not protected against seepage, its infiltration and saturation was initiated as soon as the
reservoir filling started. Once the seepage flow line reaches the downstream dike face,
several failure modes occur: (1) Sliding of the downstream dike toe including backward
erosion, (2) Sliding of the downstream dike face, and (3) Shear failure.
Failure modes 1 and 2 occur if the seepage line reaches the downstream dike face be-
fore overtopping starts. Particular problems result for comparatively long reservoir
filling time. The water flow through the dike decreases its stability and initiates toe
erosion and sliding failure. The water is drained from the dike body and erodes the
downstream dike toe similar to a piping failure. For fine sediment diameters (d = 1 to
2 mm), backward erosion of the downstream dike face results (Figure 5.1). The eroded
76 Results

dike face is still stable, yet with a smaller slope Sd close to the submerged angle of
repose of about 25°. For coarse material (d = 8 mm) the toe erosion triggers additional
sliding of the downstream dike face (Figure 5.2), again without control of the precise
breach features.

Figure 5.1 Backward erosion due to seepage for Qo = 1.0 l/s, b = 0.20 m, w = 0.20 m and
d = 2.0 mm

Figure 5.2 Erosion of downstream dike toe for Qo = 3.88 l/s, b = 0.10 m, w = 0.20 m and
d = 8.0 mm

Even though dike overtopping and erosion have started, shear failure of the down-
stream dike slope still occurs if the seepage line reaches the downstream dike face. As
soon as the seepage line reaches the downstream dike surface, sliding failure occurs and
a major part of the dike collapses instantly, with a significant reduction of test reproduc-
ibility. Figure 5.3 shows the typical sliding process. At t = 8 s overflow erosion has
started and the seepage line is about to reach the downstream dike face. The sliding
process occurred at t = 8.8 s with the sliding mass highlighted. The dike breach process
was therefore not limited to ‘pure’ overflow erosion. Similar slope sliding failures were
noted by Wu and Wang (2008).
All these failure modes due to seepage occur randomly with a significant effect on
the test reproducibility. Although geotechnical failure has to be included in the under-
standing of the breaching process, it prevents both the control of surface erosion and test
comparison, given the stochastic character of this breach mode. To allow for an investi-
gation of dike failure due to pure overflow erosion, seepage was reduced in the past by:
Results 77

(1) covering the upstream dike surface (Dupont et al. 2007, Wu and Wang 2008), or (2)
addition of a toe drain (Coleman et al. 2002, Pickert et al. 2004). Both methods signifi-
cantly complicate the test procedure, as the sediment has to be reconditioned after each
test. Further, as the water quality in terms of turbidity should be excellent to assure the
optical recording, seepage was reduced in this research with a dike bottom drainage.

Figure 5.3 Sliding of downstream dike slope due to combined erosion and seepage at various
times t for Qo = 5 l/s, b = 0.10 m, w = 0.20 m and d = 1.50 mm

A second bottom made of PVC was added to the original channel and drainage holes
were located in the upstream dike portion. The holes had a diameter of 0.02 m and were
arranged in a 0.05  0.035 m grid. A fine wire net of 0.5 mm opening width prevented
dike sediment washout. As the upstream water level increases, a small discharge portion
passes the drain. The drained discharge is measured at the channel end. It was typically
78 Results

3% of the approach flow discharge Qo, reaching a maximum of 8%. The bottom drain-
age results in an almost vertical seepage line which does not reach the downstream dike
toe prior to overtopping. Given the reduced seepage, sliding of the downstream dike
face does not occur if d ≤ 5.5 mm, whereas it cannot be completely avoided for
d = 8.0 mm due to high sediment porosity. Dikes with d  8.0 mm were excluded to
assure the test repeatability and the sidewall effects (Chapter 5.2.3). However, to expand
the test range of the scale families, a test with d = 8.0 mm was carried out in the scale
family analysis. In general, sediment diameters of d > 5.5 mm are inappropriate for the
present dike breach test setup using a bottom drainage.
The bottom drainage reduces the seepage but does not completely prevent it. De-
pending on the approach flow discharge and the reservoir volume, dike toe erosion due
to seepage cannot be completely prevented using the bottom drainage. Due to the long
reservoir filling time for Qo < 4 l/s, the seepage line may reach the downstream dike toe
prior to overtopping.

5.2.2 Test repeatability

Given the various factors affecting the dike breach process, it is necessary to accurately
reproduce a certain Test. The repeatability was therefore analyzed with five tests involv-
ing various dike dimensions, discharges and sediment sizes (Chapter 4.3.8). A Test
conducted by the same person was repeated three times using identical test parameters.
Figure 5.4 shows the dike surface profiles Z(X) at various times t for Tests 10, 11 and
12, with X = x / LD and Z = z / w. The ordinates are stretched to allow for a better com-
parison. At all time steps, the dike surface profiles Z(X) generally correlate well. At
t = 0 s, the upstream water level reaches the dike crest and overtopping is about to start.
Small differences in Z(X) at t = 2 s result from small overflow depths and minute differ-
ences in the erosive start. As time passes, these small differences reduce, so that all 3
tests tend to a single dike surface profile. Maximum deviations in Z(X) of always less
than 5% of the original dike height w are seen to occur at the instantaneous dike crest
region (X0.5) and at the original dike end zone (X1).
The repeatability of Tests 13, 14 and 15 was less perfect, mainly due to the larger
sediment diameter of d = 4.0 mm (Figure 5.5). After initial overtopping, the scatter was
particularly large along the downstream dike surface involving a wavy surface pattern.
Results 79

1 1
Z t=0s Z t = 10 s

0.5 0.5

Test 10
Test 11
(a) Test 12 (e)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=4s Z t = 40 s

0.5 0.5

(c) (g)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 1 X 2 0 1 X 2

Figure 5.4 Test repeatability: Dike surface profiles Z(X) at various times t for Tests 10, 11 and
12 (Qo = 11.3 l/s, w = 0.20 m, b = 0.20 m, d = 2.0 mm)
80 Results

1 1
Z t=0s Z t = 10 s

0.5 0.5

Test_13
Test_14
(a) Test_15 (e)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=4s Z t = 40 s

0.5 0.5

(c) (g)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 1 X 2 0 1 X 2

Figure 5.5 Test repeatability: Dike surface profiles Z(X) at various times t for Tests 13, 14 and
15 (Qo = 11.3 l/s, w = 0.20 m, b = 0.20 m, d = 4.0 mm)
Results 81

1 1
Z t=0s Z t = 10 s

0.5 0.5

Test_1
Test_2
(a) Test_3 (e)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=4s Z t = 40 s

0.5 0.5

(c) (g)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 1 X 2 0 1 X 2

Figure 5.6 Test repeatability: Dike surface profiles Z(X) at various times t for Tests 1, 2 and 3
(Qo = 2.0 l/s, w = 0.10 m, b = 0.10 m, d = 1.0 mm)
82 Results

A surface slip after initial overtopping resulted in an increased dike toe erosion for
Test 14 (Figure 5.5b). These effects decrease with increasing discharge and as time
progresses, so that the dike profiles nearly collapse for t > 10 s.
For the dike of height w = 0.10 m (Tests 1 to 9), the test repeatability was less accu-
rate as for w = 0.20 m (Figure 5.6). This is mainly due to the small discharge of only
2 l/s and possible effects of viscosity. The initial erosion could not always be deter-
mined accurately from the camera images given the small overtopping breach discharge.
After initial overtopping, the small breach flow transported sediment downward similar
to a debris flow, resulting in small sediment accumulations (Figure 5.6b). These effects
decrease again as time progresses and the dike profiles nearly collapse for t > 10 s.
The test repeatability reaches its limit for large sediment diameters. Tests with a sed-
iment diameter of d  8.0 mm were excluded, therefore, because sliding failure due to
seepage was the dominant dike breach process (Chapter 5.2.1).

5.2.3 Sidewall effects

Systematic dike breach experiments demand a considerable modeling effort. Various

breach tests in the past were therefore limited in number. To reduce these efforts, the
effect of dike width was investigated, resulting in a model dike of minimum volume and
setup time. The dike width required to avoid sidewall effects was tested using three
dikes of equal streamwise shape but of widths b = 0.10, 0.20, 0.40 m and discharges
according to the Froude similitude (Table 3.1). Figure 5.7 compares the dike surface
profiles Z(X) of Tests 28, 29 and 30, where the largest sidewall effect was observed. For
0 < t < 10 s, the erosion of the dike of width b = 0.10 m advanced slower than for the
two wider dikes. This was attributed to a sediment stratification process along both
sidewalls, and a resulting dike stabilization for small dike width. For 10  t  100 s, the
dike erosion advance was reversed, with the largest erosion rate for the dike of width
b = 0.10 m. Note the agreement of Z(X) for the two wider dikes. Tests 16 to 27 with
different sediment sizes, discharges and dike dimensions indicate a similar trend. It can
therefore be stated that the dike width has no significant effect on the plane dike erosion
process provided b  0.20 m. Note also that a small dike width complicates the exact
dike setting, so that this criterion satisfies also practical modeling requirements.
Results 83

1 1
Z t=0s Z t = 10 s

Test 28
0.5 b = 0.10 m 0.5
Test 29
b = 0.20 m
Test 30
(a) b = 0.40 m (e)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=4s Z t = 40 s

0.5 0.5

(c) (g)
0 0
0 1 X 2 0 1 X 2
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 1 X 2 0 1 X 2

Figure 5.7 Sidewall effect: Dike surface profiles Z(X) at various times t for Tests 28, 29 and
30
84 Results

5.2.4 Scale families

Potential scale effects were identified using three scale families, each consisting of three
individual tests of different scale factors λ. The temporal dike surface profiles Z(X) were
compared based on Froude similitude. In Tests 31 to 33 (Table 4.2) of scale family I,
the dike dimensions with the scale factor of λ = 1 were w = 0.40 m, b = 0.40 m,
LK = 0.20 m, d = 4.0 mm and Qo = 64 l/s. The dike toe was located at xD = 2.66 m so
that the 3.5 m long tailwater reach for sediment deposition was sufficiently long to
avoid effects of the channel outlet. It was observed that dikes located too close to the
channel end generate a different erosion pattern than considered herein. The dimensions
of the two other dikes resulted from the scale factors λ = 0.5 and λ = 0.25 (Table 4.2).
The recorded images at various times for scale family I are shown in Figure 5.8.

Figure 5.8 Scale Family I: Recorded images at various times t for Tests 31, 32 and 33
Results 85

The three profiles Z(X) at various times t are shown in Figure 5.9. Deviations in Z(X)
are particularly large in Figure 5.9c, d, with the maximum dike height for λ = 1 being
lower than for the two other. This effect was attributed to the high turbulence level at
the downstream dike toe, an effect that may be observed also in Figure 5.9b, because
deposition of the eroded sediment was partially prevented, resulting in a reduced stabili-
zation of the downstream dike face. As the bottom drainage was less efficient for λ = 1
than for the other λ, seepage for λ = 1 proceeded faster, resulting in a higher dike water
content than for the smaller dikes. A slight settlement was further observed once the
overflow had reached the downstream dike toe. At the overtopping start, the dike with
λ = 0.25 was more resistant to erosion than the other two dikes. Small overflow depths
at the erosion start result in a reduced relative sediment transport, particularly for the
small dike due to viscous effects (Ranga Raju et al. 1991). The three dike surface pro-
files Z(X) converge shortly after the initial overtopping phase and from then tend to a
common profile (Figure 5.9f, g). For large times (Figure 5.9h), the dike with λ = 1 has
the minimum elevation again because of larger turbulence.
In scale family II, the discharge was half of the first test series, resulting in Qo = 1 l/s
for λ = 0.25, whereas the sediment sizes and the dike geometries were conserved. Figure
5.10 shows the dike surface profiles Z(X) at various times t for Tests 34, 35 and 36.
Overall, the plots are similar to Figure 5.9, with major deviations of Z(X) particularly
again shortly after initial overtopping (Figure 5.10c), yet without a toe deformation as in
Figure 5.9 (b). This temporal shift results from the temporal scaling law. Once a certain
discharge is attained, the erosion profiles converge (Figure 5.10g) with a similar profile
spread as time passes, as observed in Figure 5.9h. A small discharge of Qo = 1 l/s has no
major effect on the general erosion process, but there remains an effect on the hydraulic
overflow features (Ranga Raju et al. 1991).
In scale family III, the sediment diameter of the reference model with λ = 1 was in-
creased to d = 8 mm, with Qo = 64 l/s, yet with a dike geometry as in scale family I.
Figure 5.11 shows the dike surface profiles at various times t for Tests 37, 38 and 39. A
major deviation of Z(X) can be seen in Figure 5.11b, in which the dike with d = 8 mm
has undergone a larger tailwater slope erosion than the two smaller dikes. This results
again from the bottom drainage, whose effect is obviously too small for the large sedi-
ment diameter (Chapter 5.2.1). As overtopping starts, the dike with d = 8 mm was al-
ready saturated, resulting in a sliding failure of the downstream dike slope, as described
above. The dike breach due to surface erosion started only after sliding failure had
86 Results

occurred. Note also a similar yet much smaller effect for λ = 0.5 at the tailwater dike toe
(Figure 5.11b). A sediment diameter of d = 8 mm along with the remaining test charac-
teristics is therefore not suitable for plane dike erosion. As previously, the dike surface
profiles converge surprisingly well at later times (Figure 5.11e, f) with similar devia-
tions toward test end as in Figure 5.9 and Figure 5.10.
The scale families I to III indicate that no major scale effects are present if sliding
failure is avoided. Note that the previous indications focused on dike erosion, whereas
the hydraulic conditions are investigated below. Surprisingly, neither a small dike
height of w = 0.10 m, nor a small discharge of Qo = 1 l/s have a significant effect on the
general dike erosion process. Further experimentation may suitably be made using a
minimum dike height of 0.20 m and dike width of 0.20 m, for ease in laboratory dike
setting and to avoid errors due to too small dimensions.

5.2.5 Hydraulic conditions

The hydraulic features were investigated using Froude similitude for the tests of scale
family I. The instantaneous breach discharge Q (t) during dike breach was estimated
with the circular-crested weir formula Q = Cdb(2gHo3)1/2, with Ho as approach overflow
energy head and Cd as discharge coefficient. Ho was in a first step assumed to be equal
to the approach overflow depth ho and then determined iteratively with the resulting
approach flow velocity V = Q/(bhR). The cross-sectional average velocity along the dike
breach profile is V = Q/(bN), the local Froude number is F = V/(gh)0.5 and the local
Reynolds number R = V·4Rh/ν with ν = 1.12·10-6 m2/s (Figure 4.6).
Figure 5.12 shows the Froude number profile F(X) at various times t for Tests 31, 32
and 33. In Figure 5.12a, the overflow has just started, transition occurs close to the dike
crest (X ≈ 0.5) and F increases as the water flows over the downstream dike face. The
Froude number at the flow front, marked with () in Figure 5.12a, tends to F. In
Figure 5.12b, c, F along the downstream dike face is smaller for λ = 1 than for the other
two tests. This results from the reduced dike slope for λ = 1 at this stage, as compared to
the other two tests (Figure 5.9c, d). As shown in Figure 5.12b, the deposition profile for
λ = 0.25 (Figure 5.9c) results in a reduction of F(X) between 1 < X < 1.5. Further devia-
tions of F(X) are seen in Figure 5.12d to f, mainly because of the small absolute flow
depths, resulting in smaller Froude numbers along the downstream dike face for
λ = 0.25 than for the two other.
Results 87

1 1
Z t=0s Z t = 10 / 14.1 / 20 s

Test_31

0.5 Test_32 0.5
 = 0.5
Test_33
=1
(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 2 / 2.8 / 4 s Z t = 20 / 28.3 / 40 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 4 / 5.7 / 8 s Z t = 40 / 56.6 / 80 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 6 / 8.5 / 12 s Z t = 100 / 141.4 / 200 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.9 Scale Family I: Dike surface profiles Z(X) at various times t for Tests 31, 32 and 33
88 Results

1 1
Z t=0 Z t = 10 / 14.1 / 20 s

Test_34
 = 0.25
0.5 Test_35 0.5
 = 0.5
Test_36
=1
(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 2 / 2.8 / 4 s Z t = 20 / 28.3 / 40 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 4 / 5.7 / 8 s Z t = 40 / 56.6 / 80 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 6 / 8.5 / 12 s Z t = 100 / 141.4 / 200 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.10 Scale Family II: Dike surface profiles Z(X) at various times t for Tests 34, 35 and
36
Results 89

1 1
Z t=0s Z t = 10 / 14.1 / 20 s

Test_37
 = 0.25
0.5 Test_38 0.5
 = 0.5
Test_39
=1
(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 2 / 2.8 / 4 s Z t = 20 / 28.3 / 40 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 4 / 5.7 / 8 s Z t = 40 / 56.6 / 80 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 6 / 8.5 / 12 s Z t = 100 / 141.4 / 200 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.11 Scale Family III: Dike surface profiles Z(X) at various times t for Tests 37, 38 and
39
90 Results

For λ = 0.25 the overflow depths ho are generally below the limit of 50 mm stated by
Fritz and Hager (1998) to inhibit viscous effects. Note from Figure 5.12g, h that the
flows for λ = 0.25 and λ = 0.50 remain in the subcritical regime. For large erosion times,
the dike is almost entirely eroded and the remaining dike body becomes almost horizon-
tal (Figure 5.9h). The flow is not accelerated anymore and no transition from sub- to
supercritical flow occurs. In Figure 5.12h the transition from F<1 to F>1 occurs for
λ = 1 due to a major sediment deposition at X = 0.6.
Figure 5.13 shows the Reynolds number profile R(X) at various times t for Tests 31,
32 and 33. In Figure 5.13a, the overflow has just started and R increases as the water
flows over the downstream dike face. The profiles are then equal for all evaluated time
steps with R increasing along X due to flow acceleration. The Reynolds numbers for
λ = 0.25 are always the lowest and below 105, i.e. below fully turbulent flow. The
Reynolds numbers for λ = 0.5 are below 105 along the upstream dike face but increase
above 105 along the dike crest. For λ = 1 the flow is always fully turbulent with R > 105.
Although no significant scale effect on the general dike erosion process was ob-
served from the dike erosion profiles Z(X), the hydraulic conditions indicate scale ef-
fects for λ = 0.25. Aside from minimum dike dimensions, a unit discharge of approxi-
mately qo > 20 l/(sm) should be maintained to avoid scale effects. This condition assures
that the required hydraulic features are satisfied, and scale effects mainly due to fluid
viscosity are avoided. The results on model limitations and scale effects in plane dike
breach modelling are summarized by Schmocker and Hager (2009).
Results 91

4 3
t = 2 / 2.8 / 4 s Test_31 t = 20 / 28.3 / 40 s
3  = 0.25
Test_32 F
F 2
 = 0.5
2 Test_33
=1
1
1

(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
4 3
t = 4 / 5.7 / 8 s t = 40 / 56.6 / 80 s
3
F 2 F

1
1

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
4 3
t = 6 / 8.5 / 12 s t = 100 / 141.4 / 200 s
3
F 2 F

1
1

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
4 3
t = 10 / 14.1 / 20 s t = 200 / 282.8 / 400 s
3
F 2 F

1
1

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.12 Scale Family I: Froude number F(X) at various times t for Tests 31, 32 and 33
92 Results

5 5
5 ·10 5 ·10
t = 2 / 2.8 / 4 s Test_31 t = 20 / 28.3 / 40 s
5 5
4 ·10 = 0.25 4 ·10
R Test_32 R
5 5
3 ·10  = 0.5 3 ·10
Test_33
5 5
2 ·10 =1 2 ·10
5 5
10 10
(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
5 5
5 ·10 5 ·10
t = 4 / 5.7 / 8 s t = 40 / 56.6 / 80 s
5 5
4 ·10 4 ·10
R R
5 5
3 ·10 3 ·10
5 5
2 ·10 2 ·10
5 5
10 10
(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
5 5
5 ·10 5 ·10
t = 6 / 8.5 / 12 s t = 100 / 141.4 / 200 s
5 5
4 ·10 4 ·10
R R
5 5
3 ·10 3 ·10
5 5
2 ·10 2 ·10
5 5
10 10
(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
5 5
5 ·10 5 ·10
t = 10 / 14.1 / 20 s t = 200 / 282.8 / 400 s
5 5
4 ·10 4 ·10
R R
5 5
3 ·10 3 ·10
5 5
2 ·10 2 ·10
5 5
10 10
(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.13 Scale Family I: Reynolds number R(X) at various times t for Tests 31, 32 and 33
Results 93

5.3 Plane dike breach process – general observations

5.3.1 Introduction

The following relates to general observations of the dike breach process due to overtop-
ping erosion. All observations and results are first exemplarily presented for Test 52
(w = 0.20 m, b = 0.20 m, LK = 0.10 m, So = Sd = 1:2 (V:H), d = 2.0 mm and Qo = 8 l/s).
The representative results include: (1) Breach process and breach profiles (Chapter
5.3.2); (2) Breach discharge (Chapter 5.3.3); (3) Flow depths, velocities, energy heads
(Chapter 5.3.4); (4) Bed shear stresses (Chapter 5.3.5); and (5) Erosion rate (Chapter
5.3.6 and 5.3.7).

5.3.2 Breach process and breach profiles

Figure 5.14 shows a typical temporal advance of a dike breach test and Figure 5.15
indicates the corresponding temporal dike breach profiles z(x). The constant approach
flow discharge Qo = 8 l/s results in dike overtopping at t = 0 s. At t = 2 s, erosion starts
at the downstream dike crest initiating sediment transport. The minor sediment accumu-
lation at t = 2 s results from the small overflow discharge transporting sediment down-
ward. The water front consisted of a sediment-water mixture, similar to a debris flow.
At t = 4 s the flow front reaches the downstream dike toe and impacts the channel bot-
tom leading to minor turbulence increase and air entrainment in this region (Figure 5.14
at t = 4 s). Sediment is then eroded and transported along the downstream channel bot-
tom. The dike crest has already developed to a round-shaped form. At t = 6 s, erosion
advances and the entire visible channel bottom is covered with eroded material. A wavy
deposition pattern is observed beyond the original dike toe. At t = 20 s almost half of
the dike is eroded. Up to this point, the thickness of the eroded material downstream of
the original dike still increases. From then erosion advances less fast until the dike
surface is nearly horizontal at t = 100 s. The eroded material forms a large tailwater
wedge and the breach profile remains almost constant. A slight 3-dimensional erosion
pattern is visible at t = 200 s and t = 400 s but with no effect on the overall erosion
process. A similar erosion process was described by Chinnarasri et al. (2004) and
Dupont et al. (2007).
94 Results
Results 95

Figure 5.14 Typical advance of dike erosion at various times t for Test 52

200
t=0s
z [mm] t=2s
t=4s
t=6s
150 t=8s
t = 10 s
t = 15 s
t = 20 s
100 t = 30 s
t = 50 s
t = 100 s
t = 200 s
50 t = 400 s

0
0 500 x [mm] 1'000 1'500

Figure 5.15 Dike breach profiles z(x) at various times t for Test 52
96 Results

The dike overflow features are related to the hydraulic flow regimes during overtop-
ping, as indicated by Powledge et al. (1989b). Shortly after test initiation, both the
breach discharge and the erosion potential are small. Then the breach discharge over the
dike increases rapidly exceeding the constant inflow discharge. The flow condition
upstream of the dike is subcritical. Transitional flow occurs on the dike crest whereas
the flow on the downstream slope is supercritical. The hydraulic control point is located
at the highest instantaneous dike elevation. As the highest dike elevation moves up-
stream with ongoing erosion (Figure 5.15), the hydraulic control point also moves up-
stream. The velocity increases significantly over the downstream dike slope so that the
tractive shear stresses result in high erosion.
With a reduced dike crest height, the energy head decreases resulting toward test end
in a smaller erosion potential. Further, with ongoing erosion the breach discharge de-
creases until a constant breach discharge equal to the constant inflow discharge is at-
tained. Depending on the dike and hydraulic characteristics, an equilibrium sediment
deposit is asymptotically generated. For a small sediment size and a large inflow dis-
charge, the dike may be completely eroded, whereas an equilibrium dike breach profile
develops otherwise.

5.3.3 Breach discharge

The breach discharge can be determined for each time step of the dike erosion using the
overflow formula for circular-crested weirs of various up- and downstream slopes
(Schmocker et al. 2011).


Q  C d b 2 gH o 
3 1/ 2
(5.1)

2  3 k ' 
Cd  1   (5.2)
3 3  11  Ω k ' 

1
H    2 d  3
k'  o  o  (5.3)
R  270 

To compute the breach discharge, the energy head Ho is in a first step assumed to be
equal to the overflow depth ho and then determined iteratively with the resulting ap-
proach flow velocity. As overtopping starts, the dike breach crest profile is not yet
Results 97

round-shaped. For t < 2 s, the discharge coefficient was therefore determined with the
standard formula for broad-crested weirs (Eq. 3.47). The breach discharge is very small
for t < 2 s and the water flow has not yet reached the downstream dike toe.
The crest radius R, both the up- and downstream dike face angles αo, αd and the over-
flow depth ho were directly measured from the camera images. Two problems occurr,
namely (1) when overtopping starts, it is hard to distinguish the dike and water profiles,
as the flow depths are very small (Figure 5.14, t = 2 s), and (2) for large times, the crest
radius becomes very large and almost impossible to determine with the downstream
dike face angle αd tending to zero (Figure 5.14, t ≥ 50 s). For large erosion times, the
breach discharge was assumed to be equal to the constant inflow discharge.
Figure 5.16 shows (a) the overflow depth ho(t) along with the reservoir water level
hR(t) and the maximum (subscript M) dike height zM(t), (b) the temporal development of
the crest radius R(t), (c) both the up- and downstream dike face angles αo(t), αd(t), and
(d) the breach discharge Q(t) during the dike breach for 0 ≤ t ≤ 50 s.

300 3'000
z, h [mm] hR R [mm]
zM
ho
200 2'000

100 1'000

0 0
(a) 0 25 t [s] 50 (b) 0 25 t [s] 50
30 12
o , d [°] Q [l/s]

20 o 8
d

10 4
Q [l/s]
Qo [l/s]
0 0
(c) 0 25 t [s] 50 (d) 0 25 t [s] 50

Figure 5.16 Temporal development of (a) overflow depth ho(t), reservoir water level hR(t) and
maximum dike height zM(t), (b) crest radius R(t), (c) up- and downstream dike face
angles αo(t), αd(t), and (d) breach discharge Q(t) for Test 52
98 Results

As soon as overtopping starts, Q increases fast and reaches its maximum of

QM ≈ 10.4 l/s at t  10 s (Figure 5.16d). The breach discharge Q therefore exceeds Qo as
the upstream reservoir volume is emptied. After reaching QM, the breach discharge
decreases slowly until Q = Qo for t = 50 s and then remains constant. Accordingly, the
reservoir level hR first increases fast and reaches a maximum of about hRM = 259 mm
after t = 5 s. The overflow depth and hence the discharge depend mainly on how fast the
dike erodes and on the resulting reservoir drawdown. After t = 10 s, ho remains almost
constant as both hR and zM decrease equally (Figure 5.16a). As previously stated, no
constant reservoir level was attained for the plane dike breach tests (Figure 5.16a).
The dike crest remains initially plane until erosion starts at the downstream dike
crest. Then the crest radius increases fast in the first 10 seconds and then almost linearly
up to t = 50 s (Figure 5.16b). No definite crest radius can be determined for t > 50 s. The
upstream dike face angle remains almost constant at αo = 26.56° (1:2) during the entire
breach test (Figure 5.16c). The downstream dike face angle decreases instantly within
the first two seconds, as the eroded material from the dike crest is transported down the
downstream weir face, and then increases again. With ongoing erosion, αd decreases
gradually and reaches a constant value of αd ≈ 3° (Figure 5.16c).

5.3.4 Flow depths, pressure head, velocities and energy head

Once the breach discharge Q(t) is determined, the velocity V along the dike breach
profile can be computed using the continuity equation V = Q/A with A = bN. The energy
head follows as

V2
H hz (5.4)
2g

with V2/2g = local velocity head, h = local pressure head and z = local dike height. Due
to the curvilinear flow and the high flow velocities, both N and h differ from the flow
depth hw. The vertical flow depth hw was measured from the camera images (Figure 4.6)
and then N and h are determined with Eq. (4.1) and Eq. (4.2). Both the water surface
profile hw (x) and the dike breach profile z (x) were determined from the camera images
with a spacing of Δx = 20 mm for −160 mm ≤ x ≤ 1'800 mm. The local inclination
angles of the water surface follows from θo = arctan(Δ(z+hw)/Δx) and the dike surface
from θu = arctan(Δz/Δx), respectively. Figure 5.17 and Figure 5.18 show the dike breach
Results 99

profiles z(x) along with the flow depth z(x) + hw(x), the calculated pressure head
z(x) + h(x) and the deviation δ between the flow depth and the pressure head for Test 52
at various times t. The gap at x = 1'060 mm results from the steel truss of the channel
avoiding the visual determination of the flow depth at this point.

300 30
z [mm] t=4s
 [%]

200 20

100 10

0 0
(a) 0 500 x [mm] 1000 1500

300 30
z [mm] t=6s
 [%]

200 20

100 10

0 0
(b) 0 500 x [mm] 1000 1500

Figure 5.17 Dike breach profile z(x), flow depth z(x) + hw(x), calculated pressure head
z(x) + h(x) and deviation δ between flow depth and pressure head for Test 52 at
t = (a) 4 s, (b) 6 s
100 Results

Deviations between the flow depth and the pressure head at the inflow section are
small due to the small velocities and the low curvature of the water surface. Deviations
increase as the water flows over the rounded dike crest and are especially large on the
downstream dike face as the flow accelerates and the velocity increases. Due to the
large velocities, the pressure head is always equal or smaller than the flow depth, there-
fore, resulting in negative pressure along the dike breach profile.

300 30
z [mm] t = 10 s
 [%]

200 20

100 10

0 0
(a) 0 500 x [mm] 1000 1500

300 30
z [mm] t = 20 s
 [%]

200 20

100 10

0 0
(b) 0 500 x [mm] 1000 1500

Figure 5.18 Dike breach profile z(x), flow depth z(x) + hw(x), calculated pressure head
z(x) + h(x) and deviation δ between flow depth and pressure head for Test 52 at
t = (a) 10 s, (b) 20 s
Results 101

After t = 4 s (Figure 5.17a), the maximum deviations are δ  30%, mainly due to the
small flow depths at the initial stage of overtopping. With ongoing time, the deviations
reduce as the crest radius increases, the downstream dike slope flattens and the flow
depths increase. At t = 20 s (Figure 5.18b), the curvature radii of both the dike breach
profile and the water surface decrease considerably so that δ < 5%. For large erosion
times, the dike and water surfaces become almost parallel and the downstream inclina-
tion angle tends to zero, resulting in hw = h. However, to determine the energy head
during the dike breach, it is essential to account for the curvature of both the sediment
and water surface.
The temporal development of the velocity V(x) and the energy head H(x) along the
plane dike breach profile is hereafter described for Test 52. Two difficulties arise for the
energy head. For the initial overtopping stage, the flow depths are small and the velocity
tends to zero at the flow front. The wavy depositions downstream of the original dike
result in a rapid local change of cross-section A and therefore both the velocity and the
energy head fluctuate (Figure 5.19). This leads to an impossible local increase of the
energy head and the energy slope regarding energy conservation. The energy head shall
in fact decrease along the entire dike breach profile, as energy is dissipated due to the
bed shear stress that initiates the erosion and due to friction along the channel sidewalls.
Therefore, the computed energy head was fitted with a 3rd order polynomial. For
small times, the energy head upstream of the dike was manually expanded to attain a
constant energy head at the inflow section. Figure 5.20 shows A(x), V(x) and H(x) along
with the fit at various times t for Test 52. For x < 0 m, the cross-section A is constant,
then decreases almost linearly along the upstream dike face, tends to a constant value
along the downstream dike face and remains more or less constant in the deposition
zone (x > 1 m). Velocity V(x) is small at the inflow section for x < 0 m, increases as the
flow accelerates along the dike breach profile and tends to an almost constant value for
large x as an energy loss equilibrium is reached. For x > 1.5 m, the velocity decreases
again because of energy loss due to induced bed shear stress and friction. The fluctua-
tions in the velocity profile are observed in Figure 5.20a, b, but decrease with ongoing
time as the sediment deposition is then smooth and plane.
The energy head H equals the reservoir water level hR for x < 0 m as the velocity is
comparatively low and then decreases for x > 0 m. Again, due to the velocity fluctua-
tions, the energy head shows the same fluctuations, especially after initial overtopping
102 Results

(Figure 5.20a, b). With ongoing time, the energy head decreases almost linearly along
the downstream dike face. Both V(x) and H(x) indicate the zones of high erosion poten-
tial on the dam crest and especially on the downstream dike slope as observed also by
Powledge et al. (1989a).
The fit equations and the resulting coefficients of determination R2 in Figure 5.20 are
as follows:

t = 6 s: H  0.0008 x 3  0.022 x 2  0.0148 x  0.259 R2 = 0.92

t = 10 s: H  0.0064 x 3  0.0355 x 2  0.0019 x  0.239 R2 = 0.91

t = 20 s: H  0.0128 x 3  0.0481x 2  0.0083x  0.194 R2 = 0.99

t = 50 s: H  0.0075 x 3  0.0247 x 2  0.0077 x  0.161 R2 = 0.99

0.25
t = 10 s
z [m]
0.2

0.15
z [m]
0.1 z + h w [m]
H [m]

0.05

0
0 0.5 1 1.5 x [m] 2

Figure 5.19 Dike breach profile z(x), flow depth z(x) + hw(x) and energy head H(x) for Test 52

For the current Test 52, no energy head could be determined for t < 5 s as both the
flow depth and velocity fluctuate strongly due to the small overtopping flow and the
sediment-water mixture flowing down the downstream dike face (Chapter 5.3.2).
Results 103

0.05 2.5 0.3

2 t=6s t=6s
A [m ] V [m/s] z [m]
0.04 2
0.2
0.03 1.5
A [m2]
0.02 1 z [m]
V [m/s] 0.1 H [m]
fit
0.01 0.5
x [m] x [m]
0 0 0
0 0.5 1 1.5 2 (b) 0 0.5 1 1.5
(a)
0.05 2.5 0.3
2 t = 10 s t = 10 s
A [m ] V [m/s] z [m]
0.04 2
0.2
0.03 1.5

0.02 1
0.1
0.01 0.5
x [m] x [m]
0 0 0
0 0.5 1 1.5 2 (d) 0 0.5 1 1.5
(c)
0.05 2.5 0.3
2 t = 20 s t = 20 s
A [m ] V [m/s] z [m]
0.04 2
0.2
0.03 1.5

0.02 1
0.1
0.01 0.5
x [m] x [m]
0 0 0
0 0.5 1 1.5 2 (f) 0 0.5 1 1.5
(e)
0.05 2.5 0.3
2 t = 50 s t = 50 s
A [m ] V [m/s] z [m]
0.04 2
0.2
0.03 1.5

0.02 1
0.1
0.01 0.5
x [m] x [m]
0 0 0
0 0.5 1 1.5 2 (h) 0 0.5 1 1.5
(g)
Figure 5.20 Cross section A(x), velocity V(x) and energy head H(x) at various times t, Test 52
104 Results

Figure 5.21 and Figure 5.22 show again the fitted energy heads H(x) and the result-
ing energy slopes S(x) for various times. The energy slope S follows directly as the
streamwise derivation of the energy head approximation. The energy head H(x) de-
creases gradually along the breach profile as material is eroded and H(t) decreases with
ongoing breach time as the dike height decreases. The curvature of the energy head is
high for small erosion times and tends to zero as the energy line gets parallel to the
breach profile for large times. For small breach times S(x) increases linearly along the
breach profile because the flow constantly accelerates along the dike breach profile
(Figure 5.22, t = 6 s, 10 s). For large breach time, S(x) first increases along the original
dike breach profile and then decreases again along the deposition zone (Figure 5.22,
t = 20 s, 50 s).

0.3
H [m]

0.2

0.1 t=6s
t = 10 s
t = 20 s
t = 50 s
x [m]
0
0 0.5 1 1.5

Figure 5.21 Energy heads H(x) at various times t, Test 52

The fitted energy heads at several time steps result in a constant energy head if
x < 0 mm in the upstream reservoir (Figure 5.21) and therefore in an energy slope of
S = 0. This impossible result originates from small velocities and minute changes in the
water level not detected due to insufficient image resolution. Therefore, the first meas-
ured energy slope S > 0 was used as constant energy slope for the upstream reservoir.
Results 105

0.15
S [-]
t=6s
t = 10 s
t = 20 s
0.1 t = 50 s

0.05

x [m]
0
0 0.5 1 1.5

Figure 5.22 Energy slopes S(x) at various times t, Test 52

5.3.5 Bed shear stress

The bed shear stresses τb along the breach profile was determined as

 b  gRh S (5.5)

with the energy line slope S from the derivation of the energy head fit equation. Figure
5.23 and Figure 5.24 show the dike breach profiles z(x), the flow depths z(x) + hw(x), the
fits of the energy heads H(x) and the bed shear stresses τb(x) at various times t for Test
52. Depending on the advance of the dike breach process, the bed shear stresses have
different characteristics. For all times, τb increases fast and almost linearly along the
upstream dike slope. For t = 6 s (Figure 5.23a), τb decreases between 0.3 m ≤ x ≤ 0.5 m
due to a major decrease of the hydraulic radius along the strongly curved dike crest. On
the downstream dike face and along the deposition zone, τb increases again due to the
high energy line slope. After t = 10 s, τb increases along the entire breach profile. The
increase is gradual over the dike crest as Rh decreases but S increases. Once Rh remains
almost constant for x > 0.5 m, τb increases linearly.
106 Results

z [m] z + hw [m]
H [m] b
[m]

0.3 30
2
z [m] t=6s b [N/m ]

0.2 20

0.1 10

x [m]
0 0
(a) 0 0.5 1 1.5

0.3 30
z [m] t = 10 s 2
b [N/m ]

0.2 20

0.1 10

x [m]
0 0
(b) 0 0.5 1 1.5

Figure 5.23 Dike breach profiles z(x), flow depths z(x) + hw(x), fits of energy heads H(x) and
bed shear stresses τb(x) for Test 52 at t = (a) 6 s, and (b) 10 s

For t = 20 s and t = 50 s the dike surfaces and the energy heads are almost parallel
with τb increasing gradually up to x ≈ 1.2 m. Both the energy line slopes and the bed
shear stresses reduce then again as the sediment depositions result in a decreased slope
and the flow converges to an energy loss equilibrium.
Results 107

z [m] z + hw [m]
H [m] b
[m]

0.3 30
2
z [m] t = 20 s b [N/m ]

0.2 20

0.1 10

x [m]
0 0
(a) 0 0.5 1 1.5

0.3 30
2
z [m] t = 50 s b [N/m ]

0.2 20

0.1 10

x [m]
0 0
(b) 0 0.5 1 1.5

Figure 5.24 Dike breach profiles z(x), flow depths z(x) + hw(x), fits of energy heads H(x) and
bed shear stresses τb(x) for Test 52 at t = (a) 20 s, and (b) 50 s

For the given bed shear stress τb, both the grain Reynolds number R* and the dimen-
sionless shear stress τ* (Eq. 3.31) along the dike breach profile can be determined. Fig-
ure 5.25 shows R*(x) and Figure 5.26 τ*(x) at various times t for Test 52. The character-
istic of these is similar and again depends on the advance of the dike breach process.
108 Results

300
*
R [-]
t = 10 s

t=6s
200

t = 50 s
100

x [m]
0
0 0.5 1 1.5

Figure 5.25 Grain Reynolds number R*(x) at various times t for Test 52

0.75
* [-]

0.5
t=6s

0.25
t = 50 s

t = 10 s
x [m]
0
0 0.5 1

Figure 5.26 Dimensionless shear stress τ*(x) at various times t for Test 52

Along the dike breach profile, R*(x) is always above the critical threshold of 70
(Dupont et al. 2007, Recking et al. 2008). Both incipient motion and bed-load transport
are therefore modeled properly for Test 52. If the erosion threshold is expressed with
the Shields number (Chapter 3.4.3), a constant value for incipient motion of τc* = 0.03
can be assumed if 100 < R* < 300. This value is considerably exceeded during the entire
dike breach process (Figure 5.26).
Results 109

5.3.6 Erosion volume and transport rate

With the sediment surface z (x,t) and the constant dike width b, the temporal dike vol-
ume VD(t) between 0 ≤ x ≤ LD can be determined. The dike has a maximum volume
VDM = 20 dm3 at t = 0 s before overtopping. The bed-load transport rate per second
follows as the difference of the dike volumes at two subsequent times 1 and 2 as

VD (t1 )  VD (t 2 )
qb  (5.6)
t 2  t1

Figure 5.27a shows the temporal advance of the dimensionless dike volume
VD/VDM(t) and Figure 5.27b the bed-load transport rate qb(t). VD/VDM exponentially
decreases with half of the dike eroded after t ≈ 50 s. With ongoing time, VD/VDM de-
creases slowly, as the energy head and the bed shear stresses decrease. Depending on
the inflow discharge, the dike erodes completely and VD/VDM → 0, or constant value
remnants VD/VDM tend to the equilibrium (subscript e) VDe.

1 1 3
VD qb [dm /s]
VDM
0.75 0.75

0.5 0.5

0.25 0.25

0 0
(a) 0 100 t [s] 200 (b) 0 50 t [s] 100

Figure 5.27 (a) Dimensionless dike volume VD/VDM(t) and (b) bed-load transport rate qb(t) for
Test 52

The overall bed-load transport rate increases fast within the first seconds after overtop-
ping due to the high energy head. The maximum transport rate of qb = 0.9 dm3/s is
attained at t = 6 s. Variations in qb are seen between 10 s < t < 30 s, as the dike breach
profile develops from a round-crested weir profile to a large wedge. With ongoing time,
qb decreases and converges to qb → 0 for large erosion time. The oscillations in qb are
not physically explainable but a result of the inaccurate measurement of the sediment
surface
 110 Results

5.3.7 Bed-load sediment transport formulae

Most bed-load transport formulae assume uniform flow over an almost flat bed and are
therefore not applicable to the present dike breach problem. However, several bed-load
transport formulae are used below to determine the bed-load transport rate qb during a
dike breach and to compare it with the actual sediment transport rate. The bed shear
stresses follow from chapter 5.3.5 and the bed-load transport rate qb is calculated ac-
cording to chapter 4.2.10. The actual bed-load transport rate follows directly from the
camera images (Chapter 5.3.6). Both bed-load transport rates are only determined up to
the original dike length LD.
The calculated qb depends mainly on the bed shear stresses and therefore on the en-
ergy line slope S from the derivation of the energy head fit equations. No qb was calcu-
lated for t < 5 s as no definite energy line could be determined due to the small overflow
depth and the sediment-water mixture on the downstream dike face (Figure 5.14). The
transport formulae in Table 5.1 were used to calculate the bed-load sediment transport.

Table 5.1 Bed-load sediment transport formulae used to determine qb

MPM (1948) q   8    c   c  0.047 (5.7)
3/ 2

MPM (1948) q   8    c   c*  tan  (5.8)

3/ 2
 cos 1  
c *
 tan 
with   30 ,  c  0.047
Wong&Parker q   3.97   0.0495  c  0.047 (5.9)
3/ 2

(2006)
Hager&Schmocker q   3.97   0.047  HS   (5.10)
3/ 2 1 / 12
 H  R 
(2008)  

1  1.2  h  d *1 / 4 
Rh D   B  S  
Nielsen (1992) q   12    c     c  0.05 (5.11)

Smart&Jäggi d 
0.2
1.5     c  0.05 , C = V/(gRhS)0.5 (5.12)
(1983) q  4 90

 S 0.6 C  1  c 
 d 30    
Rickenmann d
0.2
  0.5     c 1.1  c  0.05 (5.13)
(1990) q  3.1 90

  F
 d 30  s  10.5

Figure 5.28 compares various qb(t) calculated with the bed-load transport formulae
based on the original MPM formula up to t = 50 s. The transport rates vary over time
due to the various characteristics of the bed shear stress τb (Figure 5.23, Figure 5.24)
and the rapid change in both the hydraulic radius and the energy line affecting τb. Figure
5.29 shows again qb(t), but up to t = 200 s to demonstrate the overall trend. In agree-
Results 111

ment with the dike breach process, qb initially increases fast, reaches the maximum,
decreases gradually with time and finally tends to qb→0.
The smallest transport rate results from the Wong & Parker formula due to the re-
duced factor of 3.97 in Eq. (5.9). Consequently, about twice the bed-load results from
the original MPM-formula (Eq. 5.7). In Eq. (5.8), the critical shear stress τc* was adjust-
ed to account for the bed slope. The bed-load transport does however not increase sig-
nificantly when using τcα*. This results from the fact that the hydraulic conditions during
a dike breach are much above the sediment entrainment condition e.g. τ* >> τc* (Figure
5.26). The critical shear stress τc* has therefore almost no effect on the bed-load
transport formulae. The bed-load transport for Eq. (5.10) using the original MPM-
formula and the dimensionless bed shear stress τ* calculated according to Hager and
Schmocker (2008) lies between the original MPM and the Wong & Parker formulae.

0.25 3
q [dm /s]
b

0.2

0.15

0.1

0.05

0
0 10 20 30 40 t [s] 50

Figure 5.28 Bed-load transport rate qb(t) calculated with various transport formulae for Test 52

Figure 5.30 compares various qb(t) calculated with the bed-load transport formulae of
Wong & Parker (2006), Nielsen (1992), Smart & Jäggi (1993) and Rickenmann (1990).
The critical bed shear stress τc* was not adjusted and kept constant at 0.05 for Nielsen,
Smart & Jäggi and Rickenmann. The maximum qb(t) results from Nielsen (Eq. 5.11).
Both the formulae of Smart & Jäggi (1993) and of Rickenmann (1990) that were de-
rived for steep slopes result in higher qb(t) as compared to Wong & Parker (2006). The
112 Results

overall trend of all formulae is similar with high qb(t) shortly after initial overtopping
and then a continuous decrease of the bed-load transport.

0.25 3
q [dm /s]
b

0.2

0.15

0.1

0.05

0
0 50 100 150 t [s] 200

Figure 5.29 Bed-load transport rate qb(t) calculated with various transport formulae for Test 52

0.35
3
q [dm /s]
b
0.3

0.25

0.2

0.15

0.1

0.05

0
0 10 20 30 40 t [s] 50

Figure 5.30 Bed-load transport rate qb(t) calculated with various transport formulae for Test 52

Figure 5.31compares the calculated qb(t) with the actual qb(t) derived from the cam-
era images. All transport formulae underestimate the actual bed-load sediment transport,
especially for t < 10 s. Surface slips or sliding of large sediment conglomerates occur on
the downstream dike face within the first instants after overtopping and increase the
Results 113

erosion rate significantly. The transport formulae do not account for this transport
mechanism so that the bed-load transport is considerably underestimated at e.g. t =6 s.
The bed-load transport calculated with the formula of Nielsen (1992) shows the best
results especially for t > 10 s, but still underestimates the actual qb. The formula of
Wong & Parker (2006) results in the lowest qb. However, this had to be expected given
the non-uniform, transitional flow on a steep slope during the dike breach. An adjust-
ment of τc* does not result in a better agreement, as the hydraulic conditions are much
above the sediment entrainment condition. Even when considering τc* = 0, qb(t) does not
increase considerably. For large times, the transport formulae may reproduce the actual
bed-load transport, as almost uniform flow conditions are attained.
Note that the calculation of qb(t) is primarily based on the bed shear stresses obtained
with the energy line slope. The reach-averaged value for the bed shear stress may un-
derestimate the actual shear stress at the sediment bed. Further, the energy line was
fitted as otherwise no representative trend of the bed shear stress could be determined.
The breach process may further be too fast in the beginning to actually calculate qb(t) at
a certain time step. The evaluation of more tests is necessary to finally assess the direct
adaptability of uniform bed-load transport formula to the dike breach problem.

1 3
q [dm /s]
b

0.8

0.6

0.4

0.2

0
0 10 20 30 40 t [s] 50
Figure 5.31 Calculated bed-load transport rates qb(t) and actual qb(t) for Test 52
114 Results

5.4 Effect of selected test parameters on dike breach process

5.4.1 Introduction

In the following, the effects of the varied hydraulic and dimensional parameters on the
dike breach process are discussed. For each parameter, the temporal breach profiles z(x)
or Z(X) with X = x / LD and Z = z / w are analyzed. The main parameters are: (1) Dike
slope; (2) Inflow discharge; (3) Sediment size; and (4) Dike height. The effects of fur-
ther parameter that were not systematically varied in the present research are discussed
also. The main goal of this chapter is to identify the governing parameters controlling
the dike breach process.

5.4.2 Effect of dike slope So

The effect of the dike slope on the breach process was investigated using three Tests
with different slopes So = 1:2, 1:2.5 and 1:3 (Table 4.3: Tests 52, 66, 67). The distance
from the intake to the upstream dike crest was kept constant at xK = 1.4 m. This results
in a slightly reduced water storage volume VW for So = 1:2.5 and 1:3. However, its effect
on the breach process can be neglected. All other parameters were kept constant.
The difference in the erosion process is observed when comparing the breach dis-
charge Q (Figure 5.32). The maximum breach discharges were: QM = 10.4 l/s for
So = 1:2, QM = 10.0 l/s for So = 1:2.5, QM = 9.8 l/s for So = 1:3. As the erosion proceeds
faster for a steep slope, QM is higher and attained faster compared to a flat dike slope.
The maximum breach discharge for So = 1:2 is attained after t = 10 s and for So = 1:3
after t = 15 s. Due to the rapid decrease of the dike height for So = 1:2, a constant breach
discharge is faster attained compared to So = 1:2.5 and So = 1:3.
Figure 5.33 shows the dike surface profiles Z(X) at various times t for Tests 52, 66
and 67. The ordinates are stretched to allow for a better comparison. Note that the ero-
sion advances faster for a steep dike slope because of increased flow velocities and thus
increased energy slope for So = 1:2 as compared to So = 1:3. The difference is especially
noted when comparing the maximum dike elevation, e.g. at t = 10 s (Figure 5.33d). The
deposition patterns are similar for all tested dike slopes. For large erosion times when an
almost equilibrium erosion stage is attained, the remaining dikes exhibit the same max-
Results 115

imum dike height and show a similar dike surface profile for all dike slopes. Both the
erosion process and the breach discharge therefore reduce with decreasing dike slope.

12
Q [l/s]
10

6
So = 1:2
4 So = 1:2.5
So = 1:3
2
t [s]
0
0 50 100

Figure 5.32 Effect of dike slope So: Breach discharge Q(t) for So = 1:2 (Test 52), So = 1:2.5
(Test 66) and So = 1:3 (Test 67)

1
1 Z t = 20 s
Z t=2s

So = 1:2
So = 1:2.5
0.5 So = 1:3 0.5

(a) (c)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 10 s Z t = 100 s

0.5 0.5

(b) (d)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.33 Effect of dike slope So: Dike surface profiles Z(X) at various times t for So = 1:2
(Test 52), So = 1:2.5 (Test 66) and So = 1:3 (Test 67)
116 Results

5.4.3 Effect of inflow discharge Qo

The effect of the constant inflow discharge on the breach process is investigated below
for Qo = 4 l/s, Qo = 8 l/s and Qo = 16 l/s with all other parameters kept constant (Table
4.3: Tests 50, 52, 55). Depending on the inflow discharge, different breach shapes were
experimentally observed. For Qo < 4 l/s, an almost triangular breach shape results in
comparison with a round-crested breach shape for high Qo (Figure 5.34). For small Qo,
the eroded material deposits directly below the original dike, as the stream power is
small. Therefore, the dike slope downstream of the highest dike point has an almost
constant gradient. For high Qo, the eroded material is transported further downstream,
resulting in a steeper downstream dike face angle and the typical rounded breach shape.
Whereas the crest radius remains almost constant for Qo = 4 l/s with ongoing breach
time, it increases continuously for Qo = 8 l/s and Qo = 16 l/s.

Figure 5.34 Effect of inflow discharge Qo: Triangular breach profile for Qo = 2 l/s (Test 50),
round-crested breach profile for Qo = 8 l/s (Test 52) and Qo = 16 l/s (Test 55)

Figure 5.35 shows the dike surface profiles Z(X) at various times t for Tests 50, 52
and 55. The ordinates are stretched to allow for a better comparison. Note that the high-
er the inflow discharge, the faster the dike breach. This is obvious given the increased
energy head for high inflows and the resulting bed shear stresses.
Results 117

1 1
Z t=0s Z t = 20 s
Qo = 4 l/s
Qo = 8 l/s
Qo = 16 l/s
0.5 0.5

(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=2s Z t = 50 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=6s Z t = 100 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t = 10 s Z t = 200 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.35 Effect of inflow discharge Qo: Dike surface profiles Z(X) at various times t for
Qo = 4 l/s (Test 50), Qo = 8 l/s (Test 52) and Qo = 16 l/s (Test 55)
118 Results

For Qo = 4 l/s, only the downstream dike crest is slightly eroded at t = 2 s as the wa-
ter is just overtopping. For Qo = 16 l/s, the eroded material is already deposited down-
stream of the original dike. At t = 100 s, an almost equilibrium stage with a large depo-
sition wedge forms for Qo = 4 l/s as compared to an almost complete erosion for
Qo = 16 l/s.
The constant inflow discharge scenario was chosen to achieve a simple test setup and
a basic upstream boundary condition. However, the relevance of Qo in practice is diffi-
cult to quantify. In prototypes, a dike will get overtopped once the water level in the
river exceeds the dike elevation. The overtopping flow depends on the breach progress
and especially on the current discharge in the main river. The breach discharge and the
main discharge in the river influence each other and there is hardly a constant inflow
discharge to the breach section. A precise upstream boundary condition, similar to Qo in
the hydraulic experiments, will therefore not be present in nature. However, the qualita-
tive effect of various inflow discharges on the breach shape and the breach time can be
applied to the practical use.

5.4.4 Effect of sediment diameter d for non-cohesive material

An effect of the sediment diameter on the dike breach process is obvious for all tested
sediments. First, the breach profiles and the breach outflows for d = 2 mm, 4 mm, and
5.5 mm are compared. The breach process for d = 1 mm is individually discussed as
effects of apparent cohesion were observed. Finally, the effect of cohesive material on
the breach process is presented for d = 0.31 mm.
Figure 5.36 shows the dike surface profiles Z(X) at various times t for d = 2 mm
(Test 52), d = 4 mm (Test 59) and d = 5.5 mm (Test 64) with all other parameters kept
constant. The ordinates are stretched to allow for a better comparison. The breach pro-
cess has to be divided into the initial phase up to t = 10 s and the second breach phase
for t > 10 s. After t = 2 s, the breach profiles still collapse as overtopping starts and the
erosion of the downstream dike crest is initiated. At t = 4 s and t = 6 s (Figure 5.36c, d),
the breach profiles differ considerably, with the lowest dike elevation for d = 5.5 mm
and the highest for d = 2 mm. During the initial breach phase, the dike erodes faster
with increasing sediment diameter d. For d = 2 mm, the erosion proceeds uniformly as
the grains are entrained and transported mostly by rolling. The downstream dike slope is
still stable despite the water flow over the dike.
Results 119

1 1
Z t=0s Z t = 10 s
d = 2 mm
d = 4 mm
d = 5.5 mm
0.5 0.5

(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=4s Z t = 50 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.36 Effect of sediment diameter d: Dike surface profiles Z(X) at various times t for
d = 2 mm (Test 52), d = 4 mm (Test 59) and d = 5.5 mm (Test 64)
120 Results

For d = 5.5 mm, the downstream dike slope is unstable as water flows down the dike
face. Small sediment portions are sliding down the downstream dike face but the coarser
particles are transported by rolling and sliding. Further, the bottom drainage prevents
sliding failure prior to overtopping but does not prevent seepage. When overtopping
starts, the saturation of the dike is therefore higher for coarser material. Once overtop-
ping starts, the downstream dike body gets further seeped through as water enters from
above. After t = 4 s, the downstream dike slope is therefore already considerably re-
duced for d = 5.5 mm compared to d = 2 mm (Figure 5.37). For the sediment tested
herein, the erosion during the first ten seconds is therefore faster for the coarser materi-
al. A comparison of Tests 51, 58, 63 and Tests 54, 61, 65 with different inflow dis-
charges Qo indicate the same trend. Pickert et al. (2011) also observed faster erosion for
coarse sand compared to fine sand.
At t = 10 s and t = 20 s (Figure 5.36e, f), the breach profiles for the tested sediment
diameters converge. In this breach phase, the dike erosion is primarily governed by
rolling sediment transport as surface sliding no longer occurs. All dikes are fully satu-
rated. Later, the erosion proceeds slower for the coarse material given its higher erosion
resistance (τc* = 0.051 for d = 5.5 mm and τc* = 0.041 for d = 2 mm, Eq. 3.26). At
t = 50 s and t = 100 s (Figure 5.36g, h), the breach profile for d = 5.5 mm has the high-
est elevation, whereas the breach profile for d = 2 mm has the lowest. All breach pro-
files remain almost stable and reach an equilibrium stage after t ≈ 300 s. The final
breach profiles again differ depending on the sediment diameter. Whereas the profiles
for d = 2 mm and d = 4 mm are smooth with a plane sediment surface, the profile for
d = 5.5 mm has a wavy deposition pattern (Figure 5.38).
The effect of the sediment diameter on the erosion process can further be seen when
comparing the breach discharge Q (Figure 5.39). As the erosion for finer sediment
proceeds slower, the maximum reservoir level is higher and consequently the overflow
depth and the maximum breach discharge are higher for fine material. The maximum
breach discharges were QM = 10.4 l/s for d = 2 mm, QM = 9.5 l/s for d = 4 mm and
QM = 8.7 l/s for d = 5.5 mm. Sametz (1981) tested sediments of 2.25 ≤ d ≤ 12.5 mm and
also observed the maximum breach discharge for d = 2.25 mm.
The breach discharge for d = 2 mm and d = 4 mm increases fast after initial overtop-
ping until QM is attained, thereby exceeding the constant inflow discharge Qo. After
reaching the maximum outflow, the breach discharge decreases slowly until Q = Qo and
then remains constant. In contrast, the breach outflow for d = 5.5 mm first increases up
Results 121

to QM but then falls back below the constant inflow Qo, due to the fast erosion and
reservoir drainage. With ongoing time, the breach discharge increases again up to
Q = Qo. This trend was observed for all Tests with d = 5.5 mm.

Figure 5.37 Effect of sediment diameter d: Images of dike breach profiles at various times t for
d = 2 mm (Test 52) and d = 5.5 mm (Test 64)

Figure 5.38 Effect of sediment diameter d: Breach profiles after t = 300 s for d = 2 mm (Test
52), d = 4 mm (Test 59) and d = 5.5 mm (Test 64)

12
Q [l/s]

d = 2 mm
4 d = 4 mm
d = 5.5 mm

t [s]
0
0 50 100

Figure 5.39 Effect of sediment diameter d: Breach discharge Q(t) for d = 2 mm (Test 52),
d = 4 mm (Test 59) and d = 5.5 mm (Test 64)
122 Results

5.4.5 Effect of cohesion

The main goal here is to clearly distinguish the cohesive from the non-cohesive dike
breach. A sediment with d = 1 mm is not cohesive but indicated different breach pro-
files for Qo > 2 l/s (Test 41, 42) as compared to d ≥ 2 mm. Figure 5.40 shows the tem-
poral dike breach profiles z(x) for (a) d = 2 mm (Test 51), (b) d = 1 mm (Test 42) and
(c) d = 0.31 mm (Test 68) with all other parameters kept constant. The breach profiles
for d = 2 mm show the typical round-shaped dike crest and the deposited material di-
rectly below the downstream dike toe. The downstream dike slope decreases with ongo-
ing time. For d = 0.31 mm, no eroded material was deposited downstream of the dike
(Figure 5.40c) as the fine material is transported away from the dike toe. Therefore, the
downstream dike slope does not decrease and remains almost constant during the entire
breach process. The dike breaches due to a backward erosion starting at the downstream
dike toe. Figure 5.41 shows two images during the dike breach with cohesive material.
Similar breach profiles were reported by Sametz (1981) and Pickert et al. (2011). How-
ever, no abrupt failure and no spontaneous collapse of the dike body were observed. The
erosion was always gradual with smooth erosion profiles. For highly-cohesive material
e.g. clay, a partial collapse of the dike must be expected.
The breach process for d = 1 mm can be classified into a non-cohesive and a cohe-
sive breach (Figure 5.40b). The eroded material is not deposited directly downstream of
the original dike toe but only further downstream (e.g. t = 15 s). The downstream dike
slope flattens therefore with ongoing time as compared to d = 2 mm but does not form a
continuous tailwater deposition. Although soil suction was not measured in the experi-
ments, the dike breach for d = 1 mm seems to be affected by apparent cohesion. The
negative pore water pressure results in steeper breach slopes compared to Tests with
d ≥ 2 mm. During Test 42, the sediment deposition between 1000 < x < 1500 resulted
further in a decrease of the flow velocities and finally in a hydraulic jump at t = 22 s
(Figure 5.42). The hydraulic jump remains stable until the deposited sediment is trans-
ported further downstream. A large tailwater wedge is again formed for t > 50 s.
In scale family I and II, two tests were carried out with d = 1 mm but comparatively
small inflow discharges of Qo = 1 l/s and Qo = 2 l/s. Due to the small stream power, the
sediment was deposited directly downstream of the original dike and resulted therefore
in comparable breach profiles as for d ≥ 2 mm.
Results 123

200
t=0s t = 20 s

z [mm]
t=2s t = 25 s
t=4s t = 30 s
t=6s t = 40 s
150 t=8s t = 50 s
t = 10 s t = 100 s
t = 15 s

100 d = 2 mm

50

0
(a) 0 500 x [mm] 1000 1500

200
z [mm]

150

100 d = 1 mm

50

0
(b) 0 500 x [mm] 1000 1500

200
z [mm]

150

100 d = 0.31 mm

50

0
(c) 0 500 x [mm] 1000 1500

Figure 5.40 Effect of sediment diameter d: Breach profiles at various times t for (a) d = 2 mm
(Test 51), (b) d = 1 mm (Test 42) and (c) d = 0.31 mm (Test 68)
124 Results

Figure 5.41 Round-shaped erosion profile with no deposition downstream of original dike for
dike erosion with d = 0.31 mm (Test 68)

Figure 5.42 Images of dike erosion for d = 1 mm (Test 42). Deposition downstream of original
dike resulted in a hydraulic jump at t = 22 s

The erosion process of a cohesive dike is slower compared to the non-cohesive dike.
The maximum dike height after e.g. t = 20 s is zM = 110 mm for d = 2 mm compared to
zM = 160 mm for d = 0.31 mm. This results in a smaller maximum breach discharge for
the cohesive dike Q(t). Figure 5.43 shows Q(t) for d = 2 mm and d = 0.31 mm. Whereas
Q increases fast and QM exceeds Qo for d = 2 mm due to the fast erosion and the reser-
voir drainage, Q increases slowly for d = 0.31 mm and tends to Qo without exceeding it.
The maximum breach discharge for the tested cohesive dike breach is therefore equal to
the constant inflow discharge.
Results 125

8
Q [l/s]

4
d = 2 mm
d = 0.31 mm
2

t [s]
0
0 50 100

Figure 5.43 Effect of cohesive sediment: Breach discharges Q(t) for d = 2 mm (Test 51) and
d = 0.31 mm (Test 68) with Qo = 6 l/s

5.4.6 Effect of dike height w

The effect of the initial dike height w on the breach process with all other parameters
kept constant is evident. Due to the decreased dike volume, the erosion process for
small dike heights is considerably faster. Figure 5.44 shows Z(X) for w = 0.15 m (Test
46), w = 0.20 m (Test 52) and w = 0.30 m (Test 57) with all other tests parameters iden-
tical. At t = 4 s, the erosion for w = 0.15 m and w = 0.20 m already advanced. The dike
with w = 0.30 m still exhibits its original shape as the overtopping flow has not yet
reached the downstream dike toe.
All dikes then show continuous erosion with ongoing time. The breach profile for
w = 0.30 m was evaluated up to X = 1.27 only due to the limited image frame. Although
the dikes with w = 0.15 m and w = 0.20 m are eroded faster than the dike with
w = 0.30 m, the height and slope of the deposited material is similar for all Tests. At
t = 100 s, all three dikes have a large tailwater wedge with almost equal maximum dike
height. The breach profiles therefore converge with ongoing time as only a flat bed of
equal sediment diameter exists in the end stage of the breach process.
126 Results

1 1
Z t=0s Z t = 10 s
w = 0.15 m
w = 0.20 m
w = 0.30 m
0.5 0.5

(a) (e)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=2s Z t = 20 s

0.5 0.5

(b) (f)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=4s Z t = 50 s

0.5 0.5

(c) (g)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5
1 1
Z t=6s Z t = 100 s

0.5 0.5

(d) (h)
0 0
0 0.5 1 X 1.5 0 0.5 1 X 1.5

Figure 5.44 Effect of dike height w: Dike surface profiles Z(X) at various times t for w = 0.15 m
(Test 46), w = 0.20 m (Test 52) and w = 0.30 m (Test 57)
Results 127

5.4.7 Further parameters not tested

Several parameters were not systematically varied and analyzed in the present study but
affect the dike breach process. These are shortly discussed qualitatively hereafter.

 Reservoir volume VW
The reservoir volume can be varied by changing the toe distance xD. Herein, it was
only varied for the scale family tests to satisfy the scaling laws. All further breach
tests were carried out with xD = 1.0 m. A systematic variation of VW is impractical,
given the comparatively small available reservoir volume for the present model
channel (Chapter 4.2). The reservoir volume can have a decisive effect on the dike
breach, as it determines the overflow depth and therefore the breach discharge. For
a large reservoir volume, the reservoir level will remain almost constant during the
dike breach resulting in an increased erosion potential. Therefore, the dike breach
proceeds faster for larger reservoir volume.
 Dike crest length LK
A constant dike crest length of LK = 0.10 m was adopted for all systematic dike
breach tests. As the crest length simply determines the dike volume, a smaller
crest length would accelerate the erosion process whereas a larger crest length
would decelerate it.
 Sediment density ρs
Sediment with a constant density of ρs = 2,650 kg/m3 was used for all tests. Ac-
cording to Eq. (3.25), the sediment density affects the critical shear stress τc*. A
decreased sediment density would therefore accelerate the erosion process due to a
decreased grain stability.
 Sediment compaction
The dike material was not actively compacted and not inserted in layers. An active
compaction would decrease the porosity and consequently increase the relative
sediment density and would therefore decelerate the dike breach process. Howev-
er, the relative sediment density can only slightly be increased, given the homoge-
nous sediment.
 Grain size distribution
Only homogenous sediment was used of geometrical standard deviation
σg = (d84/d16)0.5 < 1.2. A well-graded material would allow for an increase of the
128 Results

relative sediment density and a good compaction. The erosion resistance of well-
graded material is generally higher than of uniform material. The breach process
would therefore be decelerated when using a well-graded material.
 Foundation slope
The model channel was horizontal for all breach tests. An inclined foundation
would accelerate the breach process given the increased energy slope and the in-
creased dike slope, resulting in less resistance against sliding (Eq. 3.24).
 Downstream water level
The Tests were conducted under steady flow condition without any tailwater sub-
mergence, so that the dike overflow was always free and the tailwater did not af-
fect the erosion process. A tailwater reduces the erosion as the energy down the
slope is partially dissipated into the tailwater (Powledge et al. 1989a).

5.5 Normalized results

5.5.1 Dimensional analysis

To generalize the results obtained with the plane dike breach tests, the governing pa-
rameters are discussed hereafter. The dike breach process can be described using the
following basic parameters:

Dike height w [m]

Dike width b [m]
Dike crest length LK [m]
Dike slope So [-]
Dike length LD [m]
Toe distance xD [m]
Inflow discharge Qo [m3/s]

Mean sediment diameter d [m]

Sediment density ρs [kg/m3]
Grain size distribution σg [-]
Submerged angle of repose θs [°]
Results 129

Water density ρ [kg/m3]

Surface tension of fluid σ [kg/s2]
Viscosity of fluid υ [m2/s]

Acceleration of gravity g [m/s2]

The dike length LD follows directly from w, LK and So. The acceleration of gravity and
both the sediment and water densities can be replaced by the submerged specific gravity
of the sediment as

 
g '   s  g (5.14)
  

The constant inflow discharge Qo can be replaced by the critical flow depth

1/ 3
 Q2 
hc   o2  (5.15)
 gb 

The breach process and the erosion rate can be described as a functional relationship
with parameters that accelerate the breach process in the numerator and parameters that
decelerate the breach process in the denominator, i.e.

hc , x D
Breach process ~ (5.16)
w, g ' , LK , S o , d ,  g , s

The viscosity and surface tension are assumed to be constant for all tests and therefore
not taken into account here. Lk, g' and xD were not systematically varied and are there-
fore dropped for the dimensionless analysis. Too few experiments with various dike
slopes So were conducted and So is not considered therefore. As only uniform sediment
was tested and the submerged angle of repose was similar for all tested sediments, both
σg and θs are dropped as well. From chapter 5.4.4, the effect of d on the breach process
depends on the breach stage. For large sediment diameters, the saturation of the dike
material is higher due to increased seepage. Sliding sediment transport and small sur-
face slips are therefore increased for coarse material within the first seconds after over-
topping. For large breach times, the erosion proceeds slower for coarse than for fine
130 Results

material given its higher erosion resistance. For the dimensional analysis, it is assumed
that an increase of d generally decelerates the erosion process.
The governing parameter for the dike breach process is therefore

hc
Breach process ~ (5.17)
w, d

With w as the global length scale and g and ρ as the two other dominant parameters, all
governing parameters can be expressed in dimensionless terms. The main results of the
systematic dike breach tests are presented in dimensionless form below.

5.5.2 Evaluated tests

The symbols in Table 5.2 were used to present the normalized results. To maintain a
certain systematic in the Figures, not all sixty-eight tests, but only distinctive tests are
presented below. The selected tests cover the entire range of tested parameters to
demonstrate the relevance of the dimensionless parameters selected. The application of
the results is subjected to the following limitations regarding the governing parameters:

Dike height 0.10 m ≤ w ≤ 0.40 m

Sediment diameter 0.001 m ≤ d ≤ 0.0055 m
Critical flow depth 0.014 m ≤ hc ≤ 0.138 m

Test 68 with d = 0.00031 m, Test 39 with d = 0.008 m, Tests 31, 34, 43 with
w = 0.10 m and Tests 40, 48, 49 with Qo < 4 l/s are as well considered to demonstrate
the model limitations. All these mentioned Tests indicated scale effects regarding the
hydraulic conditions, cohesion or sliding failure. These Tests are plotted in the diagrams
(open symbols) but are otherwise not included in the fit equations.

5.5.3 Maximum breach discharge QM

Figure 5.45 shows the dimensionless maximum breach discharge QM/Qo as a function of
the constant inflow discharge Qo for various Tests. It can be stated that: (1) For a fixed
dike height w and sediment diameter d, QM/Qo decreases with increasing Qo. The higher
the inflow discharge, the faster the erosion initiates and the less important is the volume
Results 131

of the upstream water reservoir. For very high Qo, the relation QM/Qo  1 as the dike
would be washed away instantly and would not affect the breach discharge; (2) For a
fixed dike height w and inflow discharge Qo, QM/Qo decreases with increasing d. For
large d, the erosion initiates fast due to increased seepage and surface sliding failures.
The reservoir level does therefore not significantly increase as compared to a dike made
up of fine sediment; and (3) For a constant sediment diameter d and inflow discharge
Qo, QM/Qo increases with increasing w. The increased reservoir volume for higher dikes
results in an increased breach discharge due to the stored water upstream of the dike. QM
was always attained within the first 15 s after initial overtopping.

Table 5.2 Tests and corresponding symbols for normalized results

As a constant inflow discharge scenario was tested, the maximum breach discharge
depends on Qo and will always be QM ≥ Qo. The maximum breach discharge would be
different for a scenario with a constant reservoir level, where the breach discharge
would increase constantly, or a very small reservoir that just drains and will be empty
after a certain time.
132 Results

2.5
Q
M
Q
o
2

1.5

0.5

Q [l/s]
o
0
0 10 20 30 40 50 60 70

Figure 5.45 Dimensionless maximum breach discharge QM/Qo as function of Qo for various
dike breach Tests. Symbols in Table 5.2

All data were normalized and the governing dimensionless coordinate was identified
as hc2/3·d1/3/w. These exponents of hc and d resulted also in accurate fits for the other
normalized results. Figure 5.46 shows QM/Qo ‒ 1 as a function of hc2/3·d1/3/w for all
Tests of Table 5.2. The fit equation may be expressed with R2 = 0.91 as

QM  hc2 / 3d 1 / 3 

 1  1.7 exp  20  (5.18)
Qo  w 

When excluding the Tests with scale effects (open symbols), the limitation of Eq. (5.18)
is 0.053 ≤ hc2/3·d1/3/w ≤ 0.16.
All data collapse well with Eq. (5.18). The data for the Tests with scale effects (open
symbols) are generally too high but follow the overall trend. Test 68 with the cohesive
sediment deviates significantly from the fit equation as the breach discharge did not
exceed the constant inflow discharge.
In summary, QM/Qo decreases with decreasing w and both increasing hc and d. For
small hc2/3·d1/3/w, QM is almost 1.5-times higher than Qo and QM/Qo ‒ 1 tends to a con-
stant value of approximately 1.7. For hc2/3·d1/3/w > 0.20, QM/Qo  1, indicating that the
dike washes away instantly or that the reservoir has no effect on the breach discharge.
Results 133

2
Q
M
1
Q Eq. (5.18)
o

2
R = 0.91
1

2/3
hc d 1/3
w
0
0 0.1 0.2 0.3

Figure 5.46 Dimensionless maximum breach discharge QM/Qo ‒ 1 as function of hc2/3·d1/3/w for
various dike breach Tests along with (‒) Eq. (5.18). Symbols in Table 5.2

5.5.4 Maximum reservoir level hRM

Figure 5.47 shows the dimensionless maximum reservoir level hRM/w as a function of
hc/w for various dike breach Tests. The sediment diameter has no effect on hRM, as hRM
was always attained within the first 7 s after initial overtopping. The dike erosion has
then not fully started yet and the reservoir level mainly depends on the backwater effect
resulting from the constant inflow discharge Qo. The governing dimensionless coordi-
nate was therefore identified as hc/w and the linear fit may be expressed with R2 = 0.91
for 0.07 ≤ hc/w ≤ 0.55 as

hRM h
 1.05  c  0.97 (5.19)
w w

The maximum reservoir level increases with increasing hc and decreasing w. A maxi-
mum reservoir level of 1.5-times the dike height is reached for the breach Tests with
high inflow discharges and low dike heights. As hRM is almost independent of the dike
erosion process, the data for the Tests that exhibit scale effects collapse well with
Eq. (5.19). For hc/w = 0, e.g. no inflow discharge, the reservoir levels remains constant
at hRM ≈ w. The fit equation was however not forced through hRM/w = 1 and is therefore
limited.
134 Results

2
hRM Eq. (5.19)
w

2
R = 0.91

hc
w
0
0 0.2 0.4 0.6

Figure 5.47 Dimensionless maximum reservoir level hRM/w as a function of hc/w for various
dike breach Tests along with (‒) Eq. (5.19). Symbols in Table 5.2

5.5.5 Maximum dike height zM

Figure 5.48 shows exemplary the temporal advance of the dimensionless maximum dike
height zM/w for a constant dike dimension and sediment diameter and various constant
inflow discharges Qo. For all Qo, zM decreases fast soon after initial overtopping and
then slows down with time as both the dike height and the energy slope decrease. The
analysis of all tests indicates that the maximum dike height decreases faster with: (1)
increasing Qo due to the high erosion potential; (2) decreasing dike height due to the
reduced dike volume that has to be eroded; and (3) decreasing sediment diameter due to
the increased erodibility. For all tests, the maximum dike height reaches eventually the
maximum equilibrium (subscript e) height zMe after a certain time te. For e.g. Qo = 1 l/s,
the maximum equilibrium dike height is zMe ≈ 76 mm at te ≈ 600 s. For each Test, the
maximum dike height zM was normalized with w and zMe as

zM  z Me
ZM  (5.20)
w  z Me

All Tests start therefore at ZM = 1 and tend to ZM = 0 with ongoing time t. The govern-
ing dimensionless time was identified as
Results 135

h
T  t ( g ' )1 / 2 c
1/ 2
(5.21)
d w

Figure 5.49 shows the normalized dimensionless dike height ZM(T) for various dike
breach tests. The fit equation may be expressed with R2 = 0.72 for T ≤ 104as

Z M  exp 0.1T 0.4  (5.22)

For small Qo, ZM remains constant during the first seconds after initial overtopping as
the flow first passes along the dike crest before erosion starts at the downstream dike
crest (e.g. Qo = 1 l/s in Figure 5.48). The larger the dike crest length LK, the longer the
maximum dike height will remain constant after initial overtopping. The trend of these
Tests is therefore not exclusively exponential but more like a half Gaussian bell curve.
However, as this only occurs for small Qo given the tested LK, the overall fit equation
was selected exponential. Except for Test 68 with cohesive sediment, all data follow
well Eq. (5.22).

1
d = 2 mm Qo = 1 l/s
zM w = 0.2 m Qo = 2 l/s
w Qo = 6 l/s
Qo = 12 l/s
Qo = 20 l/s
0.5

0
0 50 100 150 t [s] 200

Figure 5.48 Temporal advance of dimensionless maximum dike height zm/w versus time t for
d = 0.002 m, w = 0.2 m and various Qo

For large times, the maximum dike height reaches the equilibrium height zMe. Figure
5.50 shows the dimensionless expression zMe/w as a function of hc/(d2/3w1/3) for various
136 Results

dike breach Tests. The fit equation may be expressed for 2.5 ≤ hc/(d2/3w1/3) < 12 with
R2 = 0.92 as

z Me   hc  
2/3

 exp  0.54 2 / 3 1 / 3  (5.23)
w   d w  


Besides the inflow discharge Qo, the sediment diameter is the governing parameter,
as the equilibrium stage is mainly controlled by sediment transport and therefore by
incipient motion. The maximum equilibrium dike height decreases with both decreasing
d and w and increasing hc. For small hc/(d2/3w1/3), the dike remains in its original shape
and zMe/w → 1. For large hc/(d2/3w1/3), the dike erodes completely as zMe/w → 0. The
data for the Tests with scale effects (open symbols) follow the overall trend. Test 68
again deviates from the fit equation due to the cohesive sediment.

1
ZM
Eq. (5.22)

2
R = 0.72
0.5

0
1.10
4
0 2500 5000 7500 T

Figure 5.49 Dimensionless maximum dike height ZM(T) with (‒) Eq. (5.22). Symbols in
Table 5.2

The equilibrium stage is subjected to a certain error, as the end stage of the dike
breach is difficult to determine. Bed forms and 3-dimensional erosion patterns for large
times complicate the exact determination of zMe. In addition, all Tests were stopped once
only minor changes in the breach profile occurred. Changes in zM could still occur after
several hours as the equilibrium stage is attained asymptotically. Further, the parameter
zMe is closely related to the present experimental setup using a constant inflow dis-
Results 137

charge. The equilibrium height is reached once sediment transport stops and depends
therefore on the constant inflow discharge. For a constant reservoir level, all dikes will
eventually be eroded completely.

1
z Me Eq. (5.23)
w

0.5

2
R = 0.92

hc
2/3 1/3
d w
0
0 5 10 15 20

Figure 5.50 Dimensionless maximum equilibrium dike height zMe/w as function of hc/(d2/3w1/3)
for various dike breach Tests along with (‒) Eq. 5.23. Symbols in Table 5.2

5.5.6 Dike volume VD

The temporal dike volume VD was determined between 0 ≤ x ≤ LD for various Tests. The
dike has an initial maximum volume VDo at t = 0 s before overtopping starts. Figure 5.51
shows the temporal advance of the dimensionless dike volume VD/VDo(t) for a constant
dike dimension and sediment diameter and various constant inflow discharges Qo. For
all Qo, VDo decreases fast soon after initial overtopping and then slows down with time
as the erosion decreases. Similar to zM(t), the dike volume decreases fast with: (1) in-
creasing Qo due to the high erosion potential; (2) decreasing dike height due to the
reduced dike volume that has to be eroded; and (3) decreasing sediment diameter due to
the increased erodibility.
For all tests, VD reaches eventually the equilibrium dike volume VDe after a certain
time te. For each Test, the dike volume VD was normalized with VDo and VDe as

VD  VDe
VD  (5.24)
VDo  VDe
138 Results

All Tests start therefore at V̅ D = 1 and tend to V̅ D = 0 as tte. The dimensionless time is
again T (Eq. 5.21). Figure 5.52 shows V̅ D(T) for various Tests. The fit equation may be
expressed with R2 = 0.89 for T ≤ 104as


V D  exp  0.01T 0.7  (5.25)

All data collapse well with the proposed fit. The data for the Tests with scale effects
(open symbols) are generally too low but follow the overall trend.

1
V /V d = 2 mm Qo= 1 l/s
D Do
w = 0.2 m Qo= 2 l/s
Qo= 6 l/s
Qo= 12 l/s
Qo= 20 l/s

0.5

0
0 50 100 150 t [s] 200

Figure 5.51 Dimensionless dike volume VD/VDo(t) for d = 0.002 m, w = 0.2 m and various Qo

For large times, the dike volume reaches the equilibrium value VDe. Figure 5.53
shows the dimensionless equilibrium dike volume VDe/VDo as a function of hc/(d2/3w1/3)
for various dike breach Tests. The fit equation may be expressed with R2 = 0.93 for
2.5 ≤ hc/(d2/3w1/3) < 12 as

VDe   h  
0.85
 exp  0.30 2 / 3 c 1 / 3   (5.26)
VDo   d w  


Besides the inflow discharge, the sediment diameter is again the governing parameter,
as the equilibrium stage is mainly controlled by sediment transport and therefore by
incipient motion. The equilibrium dike volume decreases with both decreasing d and w
Results 139

and increasing hc. The data for the Tests with scale effects (open symbols) follow the
overall trend.
Again due to bed forms and 3D erosion patterns for large times, the determination of
the equilibrium dike volume was complicated. Changes in VD could still occur after
several hours as the equilibrium stage is attained asymptotically.

1
VD
Eq. (5.25)

2
R = 0.89
0.5

0
1.10
4
0 2500 5000 7500 T

Figure 5.52 Dimensionless dike volume V̅ D(T) with (‒) Eq. (5.25) for various dike breach
Tests. Symbols in Table 5.2

1
VDe Eq. (5.26)
VDo

2
R = 0.93
0.5

hc
2/3 1/3
d w

0
0 5 10 15

Figure 5.53 Dimensionless equilibrium dike volume VDe/VDo versus hc/(d2/3w1/3) for various dike
breach Tests along with (‒) Eq. (5.26). Symbols in Table 5.2
140 Results

5.6 Case studies

A brief application of the obtained results to a real case study is presented below. The
main breach parameters obtained in Chapter 5.5 shall be calculated for a typical em-
bankment dam in the canton of Zurich. The main prototype values (subscript p) are in
the range of wp ≈ 2.0 m, dp ≈ 10 mm, Sop ≈ Sdp ≈ 1:2 and LKp ≈ 1 m. Compared with the
hydraulic model tests, these values result in a model scale of 1:5 to 1:10. The reservoir
volume is assumed to be small and the breach width is set to b = 2 m. The most difficult
parameter to predict is the constant inflow discharge Qop in the prototype. In the model
it is set to Qo = 10 l/s and up-scaled to the prototype value with a scale of 1:10. This
results in Qop = 3.2 m3/s and hcp = 0.634 m for the prototype. The basic parameters are
thus wp ≈ 2.0 m, dp ≈ 0.01 m and hcp = 0.634 m resulting in the fit parameters (Chapter
5.5) hcp2/3dp1/3/wp = 0.079, hcp/wp = 0.317, hcp/(dp1/2wp) = 3.17 and hcp/(dp2/3wp1/3) = 10.8.
The maximum breach discharge results from Eq. (5.18) in QMp = 4.3 m3/s and the
maximum reservoir level results from Eq. (5.19) in hRMp = 2.6 m. Both values are about
30% higher than the constant inflow discharge Qop and the dike height wp, respectively.
The equilibrium dike height results from Eq. (5.23) in zMep = 0.143 m. The erosion of
the 2.0 m high dike will therefore stop once a dike height of about 14 cm is reached.
With the equilibrium dike height, the development of the maximum dike height zMp can
be determined using Eqs. (5.20), (5.21) and (5.22). Figure 5.54 shows the temporal
decrease of the maximum dike height zMp(t). The dike erodes fast and 14 s after initial
overtopping, the maximum dike height has decreased already to 1 m. After t ≈ 200 s, the
dike erodes slowly and attains the equilibrium dike height of zMep = 0.14 m after
t ≈ 40 min.
The maximum breach discharge and the temporal decrease of the embankment dam
can therefore be calculated using the normalized results. However, the obtained results
are only valid for an embankment dam made up of uniform material with no surface
seal or core. The reservoir is very small and subjected to a constant inflow and the
breach develops only vertically. If the dike exhibits a grass layer, various material types
or a different cross section, the calculated results should only be considered qualitative.
Results 141

2
zMp [m]
z
Mp

1.5

0.5

0
0 100 200 300 t [s] 400

Figure 5.54 Temporal decrease of maximum dike height zMp(t) for a representative prototype
embankment dam

5.7 Spatial dike breach process

5.7.1 Preliminary test

A preliminary spatial dike breach test was carried out in the hydraulic model channel I
to obtain general information on the breach physics. This allows for recording the longi-
tudinal breach profiles along the breach centreline and the corresponding flow depths.
Compared to the fully-spatial breach characteristics, this approach is considerably less
complex both from the model set-up and the data analysis. However, 3-dimensional
erosion patterns cannot be recorded in channel I.
The dike had the following dimensions: dike height w = 0.20 m, dike width
b = 0.20 m, dike crest length LK = 0.10 m, up- and downstream dike slopes So = Sd = 1:2
(V:H) and uniform sediment size d = 1.0 mm. A triangular pilot channel was cut into
the dike crest along the channel sidewall, to initiate the breach process, with the channel
wall simulating the breach centreline. The reservoir was slowly filled to prevent over-
topping of the entire dike. No additional discharge was added to the reservoir once the
breach process initiated at time t = 0 s with water entering the pilot channel (Qo = 0 l/s).
To decrease seepage due to the long reservoir filling time, an inclined thin sand layer of
142 Results

d = 0.154 mm was added to the dike body (Figure 5.55). The sand layer did not affect
the overall erosion process, as it was continuously eroded along with the dike body.
Figure 5.55 shows the temporal advance of the spatial dike erosion recorded through
the channel side wall. The longitudinal breach profiles along the channel side wall and
the corresponding flow depths are clearly visible and can directly be determined from
the camera images. Erosion starts at t = 0 s as water enters the pilot channel. The ero-
sion first progresses slowly and only vertically along the channel side wall. Due to the
small overtopping discharge, a wavy erosion pattern is noted at t = 50 s. After t = 60 s,
the breach profiles along the channel side wall are similar to the breach profiles of the
plane dike breach tests. The lateral erosion advances both due to the tractive shear stress
exerted by the water flow and the collapse of large sediment volumes. The breach chan-
nel side slopes become unstable due to undermining, so that they slump down into the
centre of the channel from where the material is transported downstream. Due to seep-
age, the small sediment diameter and the resulting negative pore pressure, the breach
shape is not trapezoidal but has nearly vertical slopes. The breach wall is vertical with-
out the presence of undercutting and hence overhanging sides. The breach shape is
therefore mainly a function of soil type, soil saturation and hydraulic loading. As the
reservoir is drained, the erosion process slows down and finally stops after t = 200 s.

Figure 5.55 Sequences of spatial dike breach, recorded through channel side wall

Figure 5.56 shows the end stage after the entire reservoir was emptied. The dike
breach profile along the channel side wall shows the large tailwater wedge, equal to the
Results 143

plane dike breach tests. The breach crest line is curved and of hourglass shape as de-
scribed by Coleman et al. (2002). The vertical and overhanging sides are clearly visible.

Figure 5.56 End stage of spatial dike breach test (after drainage)

The longitudinal breach profiles along the breach centreline are consequently similar
to the breach profiles of the plane dike breach tests. The latter can therefore be used to
describe the vertical breach process along the breach channel centreline of a spatial dike
breach. To account for the entire spatial breach process, both the vertical and lateral
erosion has to be analyzed simultaneously during the dike breach. It is however not
possible to account for both erosion processes using only one side camera, as noted
from Figure 5.55. In prototypes, side erosion is the important mechanism describing the
temporal advance of flood plain discharge, dike erosion and sediment deposition. Espe-
cially the cross-sectional breach profiles are essential to compute the breach discharge.
Therefore, the 3D-videometric measurement system developed by AICON 3D Systems
Ltd., Germany, was used to record the temporal dike breach profiles. This system is
presented below and first results regarding the experimental setup and the general
adaptability of the AICON system are discussed.
144 Results

5.7.2 Spatial dike breach tests

The spatial dike breach tests were conducted in Channel II. The main goal was to obtain
general information on the spatial breach process and especially to test the adaptability
of the AICON system. Figure 5.57 shows the temporal advance of a spatial dike breach
test recorded with the side camera of the AICON system. A triangular pilot channel was
cut in the embankment crest along the channel sidewall to initiate the breach process.
The breach process initiated at time t = 0 s with discharge entering the pilot channel.
The erosion mainly advanced vertically in this first phase due to the pilot channel locat-
ed along the channel side wall. After 80 s, an erosion channel parallel to the down-
stream dike slope had formed along the side wall. Note that the dike remained stable at
its toe, and that the tailwater portion was slightly wetted.
In the second phase, the breach develops mainly laterally. Embankment material is
eroded by the process of tractive shear stress and turbulence. Undermining of the breach
channel side slopes causes large sediment volumes to collapse into the channel centre
from where it is transported to the tailwater. The breach side slope is almost vertical due
to the negative soil-moisture tension. Nearly half of the original dike was eroded after
t = 330 s. Similar to Coleman et al. (2002) and Pickert et al. (2004), the hourglass
breach shape is visible.
The flow condition upstream of the dike is subcritical. Transitional flow occurs on
the dike crest so that the flow on the downstream dike slope becomes supercritical. The
velocity increases significantly over the latter portion so that the tractive shear stresses
result in high erosion. As the energy head and the tractive shear stresses decrease with
reduced dike crest height, the erosion both at and beyond the original dike slows down.
The reservoir was filled fast in the beginning and then the inflow discharge was de-
creased to Qo = 2 l/s to prevent entire dike overtopping. As the water entered the pilot
channel and erosion started, the inflow discharge was increased to keep the reservoir
water level constant. For the first experiments, this was done manually by adjusting the
inflow discharge once a reservoir drawdown was noticed. This worked out fine for
small erosion times but failed once the lateral erosion accelerated as the response time
of the manual adjustment of the pump was too slow. For future tests, the reservoir level
has to be constantly measured with an ultrasonic device and automatically controlled
using a computer. However, without a large reservoir, it will be challenging to keep the
reservoir level constant for large erosion times.
Results 145

Figure 5.57 Temporal advance of spatial dike breach, recorded through channel side wall with
AICON videometric measurement system
146 Results

During the entire test, the projected grid was recorded with a frequency of 1 Hz and
then evaluated using the AICON software. The grid covered the dike crest and the entire
downstream dike slope. The upstream dike face was not recorded in the present test
setup. The derived 3D-coordinates were further processed to maintain contour plots of
the dike breach surface. Figure 5.58 shows the 3D breach profiles at four time steps
along with the original dike shape indicated with a black line. The origin of the coordi-
nate system (x, y, z) was located at the downstream dike toe of the channel side wall.
The contour plots show the downstream dike face up to the dike crest. For t = 0 s, the
entire dike surface was recorded without any problems, as there was no water flow over
the crest and the sediment surface was smooth. The AICON system detected about 850
grid points. After t = 80 s, the channel formed along the channel side wall becomes
clearly visible. With ongoing breach development and increasing breach discharge, the
optical data acquisition became challenging and not the entire breach could be recorded.
The breach crest line and the steep breach slopes in the breach channel are still visible
after t = 130 s and t = 330 s. The entire breach profile was however not detected by the
AICON software.
The quality and integrity of the breach profiles depend essentially on the detected
grid points of an image. There are several reasons reducing the visibility of the grid
points and impeding an automatic computation:
 Too large grid spacing may inhibit detection of small erosion patterns.
 Sediment colour changes from dry to wet conditions. The contrast between
the sediment and the projected gird varies therefore during a test.
 Projected grid is partly distorted due to inclined dike surfaces and the projec-
tor position.
 Light of projector is reflected on water surface, especially at large times under
high breach discharge. These reflections disturb the visibility of the sediment
surface.
 Surface waves and suspended sediment complicate the grid point detection
due to reduced visibility.
 Breach side slopes may be steep and grid detection therefore fails as the verti-
cal displacement of the grid points is too large.
Results 147

Figure 5.58 Plots of 3D breach profiles at various times t

Figure 5.59 shows the ongoing breach process recorded by camera 1. The light re-
flections are clearly seen and the projected grid is not visible below them. The dry and
wetted part of the dike body can be distinguished. The breach side slopes are especially
steep at the upstream dike crest. Due to the negative pore pressure, vertical dike faces
can result during the breach process. The vertical displacement of the grid points is
therefore too large and not all points can be detected. Further, black or white artificial
sediment would be an optimum because of its excellent contrast qualities in addition to
148 Results

the small amount of impurities. This can be demonstrated when measuring the com-
pletely dry dike before the breach test. Due to the uniform color of the dry sediment, the
protected grid has an optimal contrast. Figure 5.60 shows approx. 2000 detected grid
points from the AICON measurement.

Figure 5.59 Ongoing dike breach with projected grid and both surface waves and light reflec-
tions

However, the first evaluations are surprisingly good, given the various factors dis-
turbing a proper grid recording. A non-intrusive method is therefore available in princi-
ple to determine a sediment surface topography during a spatial dike breach. To increase
grid visibility and to improve the automatic computation, the test setup has to be opti-
mized, however.

Figure 5.60 Detected grid points of initial dry dike from AICON measurement system
Conclusions and outlook 149

6 Conclusions and outlook

6.1 Scale effects

The following model limitations and scale effects were determined. The minimum
sediment size should be limited to 1 mm to avoid cohesive effects, and to apply Froude
similitude. Dike sliding during overtopping initiation is excluded with a bottom drain-
age. Sediment sizes of d > 5.5 mm have to be excluded because seepage is then not
controlled with the bottom drainage.
The test repeatability was successfully demonstrated to be independent of dike
height, dike width and sediment size within the present test program. A minor sidewall
effect was detected for a dike width of 0.10 m, which was absent for the minimum dike
width of 0.20 m. Scale families I to III indicate no major scale effects if sliding failure is
excluded. Neither a 0.10 m model dike height nor a small model discharge of only 1 l/s
had a significant effect on the general dike erosion profiles, except for the initial over-
topping phase, and after large times during the development of the equilibrium dike
profile. The hydraulic conditions during the dike breach process are governed by instan-
taneous discharge and the resulting local Froude and Reynolds numbers along the dike.
A data analysis indicates that for the smallest dike test with a dike height of 0.10 m, the
Froude numbers are smaller than for the larger scales and the Reynolds numbers are
below the fully-turbulent flow threshold of R = 105.
Although no significant scale effect on the general dike erosion process was ob-
served from the dike erosion profiles Z(X), the hydraulic conditions indicate scale ef-
fects for the smallest dike breach tests (λ = 0.25 ) with Qo = 1 and 2 l/s.
In summary, (1) the minimum dike dimensions should be 0.20 m in height and
0.20 m in width to allow for a representative test performance, (2) the minimum grain
size be d = 1 mm due to effects of viscosity, (3) the maximum grain size be d = 5.5 mm
to prevent sliding due to seepage, and (4) the minimum unit discharge be qo = 20 l/(sm)
again due to viscous effects. If these conditions are satisfied for water flow over a gran-
ular dike with a sediment density close to 2,650 kg/m3, then physical model testing in a
wide range of discharges with a comparatively small effort is amenable.
150 Conclusions and outlook

6.2 General breach process

The constant approach flow discharge results in dike overtopping at t = 0 s. The erosion
starts at the downstream dike crest initiating downstream sediment transport. The dike
crest develops with time from a sharp-crested to a round-shaped weir body. For large
erosion times, the eroded material forms a large tailwater wedge and the breach profile
remains almost constant.
The breach discharge Q for each time step of the dike erosion can be calculated using
the overflow formula for circular-crested weirs with various up- and downstream slopes.
As soon as overtopping starts, Q increases fast, exceeds the constant inflow discharge
Qo and reaches a maximum. Then the breach discharge decreases slowly toward Q = Qo.
Flow depths, velocities and energy heads follow directly from the derived sediment
and water surface profiles. Due to the curvilinear flow and the high flow velocities, the
pressure head differs from the flow depth. Maximum deviations after initial overtopping
are about 30 % due to the small flow depths and flow acceleration.
The wavy depositions downstream of the original dike and the local determination of
the flow depth result in rapid local change of the cross-section so that both the velocity
and the energy head fluctuate. Therefore, the energy head was fitted with a third-order
polynomial. The energy head slope follows then as the derivation from the energy head
fit equations.
The bed shear stresses can be used to calculate the bed-load sediment transport using
existing transport formulae. All transport formulae underestimate the actual sediment
transport rate considerably. A direct application of transport formulae for uniform flow
to the present dike breach process is therefore not possible at this time.

6.3 Effect of test parameters on breach process

The effect of various parameters on the dike breach process were investigated and indi-
vidually presented. In summary, the dike breach process is accelerated with increasing
dike slope So and inflow discharge Qo and decreasing dike height w within the range of
tested parameters. After initial overtopping, the erosion process is accelerated with
increasing sediment diameter d due to surface slips and the increased dike saturation for
coarse sediment. For large times, the erosion process is decelerated with increasing
sediment size, as this equilibrium stage is mainly controlled by sediment transport and
Conclusions and outlook 151

therefore by the incipient motion criteria. The shape of the breach profiles is more or
less independent of So, Qo, w and d > 1 mm and differs mainly for cohesive and non-
cohesive sediments. Sediment with d = 1 mm is considered non-cohesive but shows
cohesive characteristics due to apparent cohesion.

6.4 Normalized results

For the dimensional analyses, the breach process and the erosion rate can be described
as a functional relationship with parameters that accelerate or decelerate the breach
process. The governing parameters for the dike breach process were identified as the
dike height w, the sediment diameter d and the critical flow depth hc derived from the
constant inflow discharge Qo. All governing parameters can therefore be expressed in
dimensionless terms. Table 6.1 states the dimensionless parameters used to describe the
main results of the systematic dike breach tests. As the effect of sediment diameter on
the erosion process differs in function of time e.g. after initial overtopping compared to
large times, d appears both in the numerator and the denominator of the fit parameters.

Table 6.1 Dimensionless fit parameters used to present the normalized results

Parameter Dimensionless form Fit parameter

QM hc
2/3
d 1/ 3
Maximum breach discharge QM
Qo w
hRM hc
Maximum reservoir level hRM
w w
z  z Me h
Maximum dike height zM ZM  M T  tg '1 / 2 1 / c2
w  z Me d w
Maximum equilibrium dike height z Me hc
zMe w d w1 / 3
2/3

V  VDe h
Dike volume VD VD  D T  tg '1 / 2 1 / c2
VDo  VDe d w
VDe hc
Equilibrium dike volume VDe
VDo d w1 / 3
2/3
152 Conclusions and outlook

6.5 Spatial dike breach tests

Spatial dike breach tests due to overtopping were conducted in a laboratory channel to
apply and test a new videometric measurement system of AICON 3D Systems Ltd.
The dike consisted of homogenous and non-cohesive sand, without the addition of a
surface protection or core to the model dike. The breach was initiated with a pilot chan-
nel located at the channel sidewall. The AICON measurement system was used to de-
termine the 3-dimensional breach profiles during a spatial dike breach. This system
allows to record a sediment surface through the water surface using a rectangular grid
projection and a set of four cameras. Preliminary tests confirmed the general adaptabil-
ity of the AICON system to the dike breach problem. The spatial breach profiles were
directly computed from the camera images. Preliminary tests provide motivating results
and a feasible, non-intrusive method seems therefore available for determining a sedi-
ment surface topography during a spatial dike breach.
Several issues regarding the test set-up have to be improved for a complete and relia-
ble application of the technique, however. Surface waves and light reflections disturb
the projected grid and complicate the optical grid detection. The accuracy depends
essentially on the water quality, the reflection pattern by the water and on the light
absorption characteristics of the sediment. An adaption of test setup and an analysis of
the derived data are necessary for further application of the AICON measurement sys-
tem.

6.6 Outlook

The present work investigates in particular the plane dike breach process of uniform
granular dikes due to overtopping. All other failure modes were excluded and seepage
was intentionally limited. These simplifications are necessary to examine the basic
physical principles of the breach process due to surface erosion. However, a dike breach
mostly results from a combination of several failure mechanisms. The tested dikes
exhibit a minimum resistance against surface erosion, as no surface protection or core
was installed. Several geotechnical and soil parameters that affect the dike breach were
not investigated. The dike was not compacted and both the pore water pressure and
seepage were not analysed in detail. All these simplifications were accepted to decrease
the laboratory effort and allow for a systematic investigation of the dike breach process.
Conclusions and outlook 153

Once the hydraulic and physical mechanisms of the simple model dike are understood,
the test setup can be adapted to account for further parameters affecting the dike breach
as e.g. surface protection, soil layers, core, compaction or soil moisture content. All
these parameters increase the laboratory effort and may further impede the optical re-
cording of the plane dike breach profiles due to e.g. increased turbidity.
The systematic parameter variation included the dike height, the sediment diameter
and the inflow discharge. Based on these results, the dike breach knowledge should be
expanded particularly in terms of the effect of sediment mixtures on the plane breach
process. With a variation of the reservoir volume, the dike crest length and the bottom
slope, the current findings can be further completed.
Finally, the new AICON measurement system was installed and successfully applied
to the spatial dike breach test. Systematic tests should be carried out to account for 3-
dimensional breach profiles and the breach discharge. The findings of the spatial dike
breach test should further be combined with the results from the plane dike breach tests
to achieve an overall knowledge of the dike breach phenomenon.
154 Conclusions and outlook
Notation 155

Notation

A Cross-sectional area
a Variable
b Width
C Flow resistance factor
CD Drag coefficient
Cd Discharge coefficient
CL Lift coefficients
d Mean grain diameter
di Average grain size of fraction i
*
d Dimensionless grain size
F Froude number
F Variable
FD Fluid drag force
Fg Submerged gravitational force
FL Fluid lift force
Fr Coulomb frictional force
f Friction coefficient
g Acceleration of gravity
g' Submerged specific gravity
H Energy head
Ho Approach flow energy head
h Flow depth, pressure head
hc Critical flow depth
ho Approach flow overflow depth
hR Reservoir water level
hRM Maximum reservoir water level
hw Water surface level
k1 Constant accounting for grain shape
ks Equivalent sand grain roughness
L Length
156 Notation

LD Dike length
LK Crest length
N Length of the orthogonal trajectory of the streamlines
Q Breach discharge
QM Maximum breach discharge
Qo Approach flow discharge
qb Volumetric bed-load transport rate
q* Dimensionless bed-load transport rate
R Reynolds number
*
R Grain Reynolds number
R Crest radius
R2 Coefficient of determination
Rh Hydraulic radius
Rep Factor
Rs Hydraulic radius of bed cross-section
r Variable
S Energy slope
Sw Water surface slope
So Bed slope, upstream dike slope
Sd Downstream dike slope
s Specific density
T Dimensionless Time
t Time
U Wetted perimeter
U* Shear velocity
Uc* Critical shear velocity
V Velocity
Vf Effective fluid velocity
Vm Depth-averaged velocity
VD Dike volume
V̅ D Normalized dike volume
VDo Initial dike volume
VDe Equilibrium dike volume
VW Water storage volume
Notation 157

W Weber number
w Dike height
X Dimensionless streamwise coordinate
x Streamwise coordinate
xD Dike toe distance
xK Dike crest distance
y Transverse coordinate
Z Dimensionless vertical coordinate
ZM Dimensionless maximum dike height
z Vertical coordinate, distance of bed above datum plane
zeff z from side camera (4.3.10)
zmeas z from AICON measurement (4.3.10)
zM Maximum dike height
zMe Maximum equilibrium dike height
zn Normal distance to channel bed
zo Bed roughness length
α Angle
αe Energy correction factor
γ Angle of energy line
δ Relative error
θ Angle of repose
θs Submerged angle of repose
θo Inclination of top streamline
θu Inclination of bottom streamline
κ von Karman’s constant
λ Scale factor
μ Angle of friction / coefficient of Coulomb friction
ν Kinematic fluid viscosity
ξ Relative crest length
ρ Fluid density
ρs Sediment density
ρk Relative crest curvature
ρk' Dimensionless parameter
σ Surface tension
158 Notation

σg Geometrical standard deviation of grain size distribution

τb Bed shear stress
τ* Dimensionless shear stress
τc* Critical dimensionless shear stress
τcα* Critical dimensionless shear stress for a bed of angle α
Ω Constant

Subscripts
b Bed
D Dike
d downstream
e Equilibrium
F Froude
K Crest
M Maximum
m Model
o upstream, approach
p Prototype
References 159

References

Ackers, P., White, W.R. (1973). Sediment transport: New approach and analysis. Jour-
nal of the Hydraulics Division ASCE 99(HY11), 2041-2060.

ASCE (2008). Sedimentation engineering: processes, management, modelling and

practise. M.H. Garcis, ed., ASCE Report no. 110, Virginia, USA.

Bagnold, R.A. (1973). The nature of saltation and of bed-load transport in water. Proc.
Royal Society London 332(1591), 473-504.

Bagnold, R.A. (1980). An empirical correlation of bedload transport rates in flumes and
natural rivers. Proc. Royal Society London, Series A 372(1751), 453-473.

Bechteler, W., Broich, K. (1991). Effects of dam-break modelling. Proc. 24th Congress
of IAHR, Madrid, Spain, 189-200.

Brahms, A. (1757). Anfangs-Gründe der Deich- und Wasserbaukunst. Schuster Verlag,

Leer, Germany (in German).

Broich, K. (1999). An overview of breach modelling. Proceedings of the Zaragoza

Meeting – CADAM, 309-325.

Brownlie, W.R. (1981). Re-examination of Nikuradse roughness data. Journal of the

Hydraulics Division 89(5), 327-323.

Buffington, J.M., Montgomery, D.R. (1997). A systematic analysis of eight decades of

incipient motion studies, with special reference to gravel-bedded rivers. Water Re-
source Research 33(8), 1993-2029.

Buffington, J.M. (1999). The legend of A.F. Shields. Journal of Hydraulic Engineering
125(4), 376 – 387.

Cao, H. H. (1985). Résistance hydraulique d’un lit à gravier mobile à pente raide; étude
expérimentale. Ph.D. Thesis, Ecole Polytechnique Federale de Lausane, Lausanne,
Switzerland, (in French).
160 References

Cao, Z., Pender, G., Wallis, S., Carling, P. (2004). Computational dam-break hydraulics
over erodible sediment bed. Journal of Hydraulic Engineering 130(7), 689-703.

Cao, Z., Pender, G., Meng, J. (2006). Explicit formulation of the Shields diagram for
incipient motion of sediment. Journal of Hydraulic Engineering 132 (10), 1097–
1099.

Casagrande, A. (1937). Seepage through dams. Journal of the New England Water
Works Association 51(2), 295-336.

Casagrande, A. (1961). Control of seepage trough foundations and abutment dams.

Géotechnique 11(3), 161-182.

Casey, H. (1935). Über Geschiebebewegung, Mitteilung 19, Preussische Versuchsan-

stalt Wasserbau Schiffbau, 86 p., (in German).

Cedergren, H.R. (1989). Seepage, drainage and flow nets. Wiley, New York, USA.

Chapuis, R.P., Aubertin, M. (2001). A simplified method to estimate saturated and

unsaturated seepage through dikes under steady-state conditions. Canadian Ge-
otechnical Journal 38(6), 321-1328.

Cheng, N.S. (2004). Analysis of bedload transport in laminar flows. Advances in Water
Resources 27(9), 937–942.

Chiew, Y., Parker, G. (1994). Incipient sediment motion on nonhorizontal slopes. Jour-
nal of Hydraulic Research 32(5), 649-660.

Chinnarasri, C., Tingsanchali, T., Weesakul, S., Wongwises, S. (2003). Flow patterns
and damage of dike overtopping. International Journal of Sediment Research 18(4),
301-309.

Chinnarasri, C., Jirakitlerd, S., Wongwises, S. (2004). Embankment dam breach and its
outflow characteristics. Civil Engineering and Environmental Systems 21(4), 247-
264.
References 161

Coleman, N.L. (1967). A theoretical and experimental study of drag and lift forces
acting on a sphere resting on a hypothetical stream bed. 12th Congress, International
Association for Hydraulic Research, Fort Collins Colorado, 185-192.

Coleman, S.E., Andrews, D.P., Webby, M.G. (2002). Overtopping breaching of nonco-
hesive homogenous embankments. Journal of Hydraulic Engineering 128(9), 829-
838.

Daido, A. (1983). Incipient motion and bed load of sediment in steep channel. Proc. of
the 20th Congress of IAHR, Moscow 7(2), 293-296.

Daniel, D.E. ed. (2007). The ERP Report: What went wrong and why? Civil Engineer-
ing 77(6), 54-76.

D'Eliso, C. (2007). Breaching of sea dikes initiated by wave overtopping - a tiered and
modular modelling approach, PhD thesis, University of Braunschweig, Braun-
schweig, Germany.

DIN 19712 (1997). Flussdeiche. Deutsches Institut für Normung e.V. (DIN) (in Ger-
man).

Dupont, E., Dewals, B.J., Archambeau, P., Erpicum, S., Pirotton, M. (2007). Experi-
mental and numerical study of the breaching of embankment dam. Proc. 32nd Con-
gress of IAHR Venice 1(178), 1-10, IAHR, Madrid.

DWA (2011). Deiche an Fliessgewässern. DWA-Regelwerk, Merkblatt DWA-M 507-1,

Deutsche Vereinigung für Wasserwirtschaft, Abwasser und Abfall, (in German).

EBL, Energibedriftenes landsforening (2003). Stability and breaching of dams. Report

123-2003, ISBN: 82-436-0475-8.

Einstein, H.A. (1942). Formulas for the transportation of bed load, Transactions ASCE
107, 561-577.

Einstein, H.A. (1950). The bed-load function for sediment transportation in open chan-
nel flow. Technical Bulletin 1026, United States Department of Agriculture, Wash-
ington D.C.
162 References

Foster, M., Fell, R., Spannagle, M. (2000). The statistics of embankment dam failures
and accidents. Canadian Geotechnical Journal 37(5), 1000-1024.

Fritz, H.M., Hager, W.H. (1998). Hydraulics of embankment weirs. Journal of Hydrau-
lic Engineering 124(9), 963-971.

Froehlich, D.C. (1995). Peak outflow from breached embankment dam. Journal of
Water Resource Planning and Management 121(1), 90-97.

Froehlich, D.C. (2008). Embankment dam breach parameters and their uncertainties.
Journal of Hydraulic Engineering 134(12), 1708-1721.

Garcia, M.H., Zech, Y. (2007). Dam-break flow experiments and real-case data: A
database from the European IMPACT research program. Journal of Hydraulic Re-
search 45(Extra Issue), 109 p.

Gessler, J. (1970). Self-stabilizing tendencies of sediment mixtures with large range of

grain size. Journal of Waterway and Harbour Division 96(2), 235-249.

Gessler, J. (1971). Beginning and ceasing of sediment motion. River Mechanics Ch. 7,
H.W. Shen ed., Fort Collins, CO.

Gilbert, G. K. (1914).The transportation of debris by running water, U.S. Geological

Survey Professional Paper 86, 263 pp.

Godding, R. (1993). Ein photogrammetrisches Verfahren zur Überprüfung und Kalibrie-

rung digitaler Bildaufnahmesysteme. Zeitschrift für Photogrammetrie und Ferner-
kundung 2, 82-90 (in German).

Godding, R., Hentschel, B., Kauppert, K. (2003). Videometrie im wasserbaulichen

Versuchswesen. Wasserwirtschaft 93(4), 36-40 (in German).

Gomez, B, Church, M. (1989). An assessment of bed load sediment transport formulae

for gravel bed rivers. Water Resource Research 25(6), 1161-1186.

Graf, W.H., Suzka, L. (1987). Sediment transport in steep channels. Journal of Hydro-
science and Hydraulic Engineering 5(1), 11-26.
References 163

Gregoretti, C. (2008). Inception sediment transport relationships at high slopes. Journal

of Hydraulic Engineering 134(11), 1620-1629.

Gregoretti, C., Maltauro, A., Lanzoni, S. (2010). Laboratory experiments on the failure
of coarse homogeneous sediment natural dams on a sloping bed. Journal of Hydrau-
lic Engineering 136(11), 868-879.

Gyr, A., Hoyer, K. (2006). Sediment transport: A geophysical phenomenon. Springer,

Dordrecht, NL.

Hager, W.H., Hutter, K. (1984). Approximate treatment of plane channel flow. Acta
Mechanica 51(1-2), 31-48.

Hager, W.H. (1994). Discussion of ‘Momentum model for flow past weir’, by Rama-
murthy, A.S., Vo., N.-D., Vera, G. Journal of Irrigation and Drainage Engineering
120(3), 684-685.

Hager, W.H., Oliveto, G. (2002). Shields' entrainment criterion in bridge hydraulics.

Journal of Hydraulic Engineering 128 (5), 538–542.

Hager, W.H., Unger, J. (2006). Dike Erosion - A preliminary outlook. Proc. Intl. Conf.
River Flow, R.M.L. Ferreira, E.C.T.L Alves, J.G.A.B Leal, A.H. Cardoso, eds.,
Lisbon, Portugal, 1511-1519.

Hager, W.H., Schmocker, L. (2008). Discussion of „Reanalysis and correction of bed-

load relation of Meyer-Peter and Müller using their own database“ by Wong, M.,
Parker, G. Journal of Hydraulic Engineering 134(12), 1780-1782.

Hagerty, D.J. (1991a). Piping/sapping erosion. I: Basic considerations. Journal of Hy-

draulic Engineering 117(8), 991-1008.

Hagerty, D.J. (1991b). Piping/sapping erosion. II: Identification – diagnosis. Journal of

Hydraulic Engineering 117(8), 1009-1025.

Haselsteiner, R. (2007). Hochwasserschutzdeiche an Fliessgewässern und ihre Durchsi-

ckerung. Bericht Lehrstuhl und Versuchsanstalt für Wasserbau und Wasserwirt-
schaft der Technischen Universität München, Band 111, Strobl, ed., Munich, Ger-
many (in German).
164 References

Heller, V. (2007). Massstabseffekt im hydraulischen Modell. Wasser, Energie, Luft

99(2), 153-159 (in German).

Heller, V. (2011). Scale effects in physical hydraulic engineering models. Journal of

Hydraulic Research 49(3), 293-306.

Hentschel, B. (2007). Hydraulische Flussbaumodelle mit beweglicher Sohle. Mitteilung

Bundesanstalt für Wasserbau 90, 25-46 (in German).

Höeg, K., Lövoll, A., Vaskinn, K.A. (2004). Stability and breaching of embankment
dams: field tests on 6 m high dams. International Journal on Hydropower & Dams
11(1), 88-92.

Hunziker, R.P. (1995). Fraktionsweiser Geschiebetransport. VAW Mitteilung 138, D.L.

Vischer, ed. VAW, ETH Zurich.

Ikeda, S. (1982). Incipient motion of sand particles on side slopes. Journal of the Hy-
draulics Division ASCE 108(1), 95-114.

Iwagaki, Y. (1956). Fundamental study on critical tractive force. Japan Society of Civil
Engineering 41, 1-21.

Jandora, J., Jaromír, R. (2008). The failure of embankment dams due to overtopping.
Brno University of Technology, Brno, Czech Republic.

Jovanović, M. (2010). Discussion of ‘Modelling dike breaching due to overtopping’, by

Schmocker, L., Hager, W.H., Journal of Hydraulic Research, 48(4), 547-549.

Julien, P.Y. (2010). Erosion and sedimentation. Cambridge University Press, Cam-
bridge UK.

Kahawita, R. (2007). Dam breach modelling - A literature review of numerical models.

CEATI Report No. T032700-0207C. CEA Technologies (CEATI) Dam Safety In-
terest Group (DSIG), Montreal, Canada.

Keulegan, G.H. (1938). Laws of turbulent flow in open channels. Journal National
Bureau of Standards 1151(21), 707-741.
References 165

Kobus, H. (1984). Wasserbauliches Versuchswesen. DVWK-Schriften 39. Parey, Ham-

burg / Berlin (in German).

Kohli, A. (1998). Kolk an Gebäuden in Überschwemmungsebenen. PhD thesis 12592.

ETH Zurich, Zürich, Switzerland (in German).

Kozeny, J. (1927). Über kapillare Leitung des Wassers im Boden. Sitzungsbericht der
Akademie der Wissenschaften in Wien 136(2a), 271-306 (in German).

Kramer, H. (1932). Modellgeschiebe und Schleppkraft. PhD thesis, Technischen Hoch-

schule Dresden, Germany (in German).

Kramer, H. (1935). Sand mixtures and sand movement in fluvial models. Transaction
ASCE 100, 798-878.

Kreuter, F. (1921). Der Flussbau. Handbuch der Ingenieurwissenschaften 3(6). Engel-

mann, Leipzig, Germany (in German).

Lau, Y.L., Engel, P. (1999). Inception of sediment transport on steep slopes. Journal of
Hydraulic Engineering 125(5), 544-547.

LAWA (1995). Leitlinien für einen zukunftsweisenden Hochwasserschutz. Hochwasser

– Ursache und Konsequenzen. Länderarbeitsgemeinschaft für Wasser (LAWA) (in
German).

Lecointe, G.D. (1998). Breaching mechanisms of embankments: An overview of previ-

ous studies and the models produced. MSc Project. The University of Birmingham,
School of Civil Engineering, Birmingham UK.

MacDonald, T.C., Langridge-Monopolis, J. (1984). Breaching characteristics of dam

failures. Journal of Hydraulic Engineering 110(5), 567-586.

Martin, H., Pohl, R. (2000). Technische Hydrodynamik 4. Huss-Medien GmbH, Berlin

(in German).

Martins, R. (1981). Hydraulics of overflow rockfill dams. LNEC Mémoria 559.

Laboratorio Nacional de Engenharia Civil, Lisbon, Portugal.
166 References

Meyer-Peter, R. Favre, H. Einstein, H.A. (1934). Neuere Versuchsresultate über den

Geschiebetrieb. Schweizerische Bauzeitung 103(13), 147-150 (in German).

Meyer-Peter, E., Müller, R. (1948). Formulas for bed-load transport. 2nd Meeting of the
International Association for Hydraulic Structures Research, Appendix 2, Stock-
holm, Sweden, 1-26.

Minor, H.-E., Hager, W.H., eds. (2004). River engineering in Switzerland. Stäubli AG,
Zürich.

Misri, R.L., Garde, R.J., Ranga Raju, K.G. (1984). Bed load transport of coarse non
uniform sediment. Journal of Hydraulic Engineering 110(3), 312-328.

Montes, S.V. (1970). Flow over round crested weirs. L’Energia Elettrica 47(3), 155-
164.

Morris, M.W., Hassan, M.A.A.M., Vaskinn, K.A. (2007). Breach formation: Field test
and laboratory experiments. Journal of Hydraulic Research 45(Extra Issue), 9-17.

Morris, M.W., Hassan, M.A.A.M., Samuels, P.G., Ghataora, G.S. (2008). Development
of the HR BREACH model for predicting breach growth through flood embank-
ments and embankment dams. Proc. Intl. Conf. River Flow, M.S. Altinakar, M.A.
Kokpinar, I. Aydin, S. Cokgor and S. Kirkgoz, eds.), Çeşme, Izmir, Turkey, 679-
687.

Morris, M.W. (2009). Breaching processes - A state of the art review. FLOODsite
Report T06-06-03. FLOODsite, www.floodsite.net.

Neill, C. R. (1968). Note on initial movement of coarse uniform material. Journal of

Hydraulic Research 6(2), 157-184.

Neill, C.R., Yalin, M.S. (1969). Qualitative definition of beginning of bed movement.
Journal of the Hydraulics Division ASCE 95(1), 585-587.

Nielsen, P. (1992). Coastal bottom boundary layers and sediment transport. World
Scientific, River Edge, N.J.
References 167

Nikuradse, J. (1933). Strömungsgesetzte in rauhen Rohren. Forsch. Arb. Ing.-Wes. 361,

1-22 (in German).

Novak, P. (1984). Scaling factors and scale effects in modelling hydraulic structures.
Symposium on scale effects in modelling hydraulic structures, H. Kobus, ed., Tech-
nische Akademie, Esslingen, 0(3), 1-5.

Parker, G. (1979). Hydraulic geometry of active gravel rivers. Journal of the Hydraulics
Division ASCE 105(9), 1185-1201.

Parker, G., Klingeman, P.C., McLean, D.G. (1982). Bedload and size distribution in
paved gravel-bed streams. Journal of the Hydraulics Division ASCE 108(HY4),
544-571.

Parker, G. (1990). Surface-based bedload transport relation for gravel rivers. Journal of
Hydraulic Research 28(4), 417-436.

Pickert, G., Jirka, G.H., Bieberstein, A., Brauns, J. (2004). Soil/water interaction during
the breaching process of overtopped embankments. Proc. Intl. Conf. River Flow, M.
Greco, A. Carravetta, and R. Della Morte, eds., Naples, Italy, 903-910.

Pickert, G., Weitbrecht, V., Bieberstein, A. (2011). Breaching of overtopped river em-
bankments controlled by apparent cohesion. Journal of Hydraulic Research 49(2),
143-156.

Pontillo, M., Schmocker, L., Greco, M., Hager, W.H. (2010). Two-dimensional numeri-
cal modelling of dike erosion due to overtopping. Journal of Hydraulic Research
48(5), 573-582.

Powledge, G.R., Ralston, D.C., Miller, P., Chen, Y.H., Cloppner, P.E., Temple, D.M.
(1989a). Mechanics of overflow erosion on embankments I: Research activities.
Journal of Hydraulic Engineering 115(8), 1040-1055.

Powledge, G.R., Ralston, D.C., Miller, P., Chen, Y.H., Cloppner, P.E., Temple, D.M.
(1989b). Mechanics of overflow erosion on embankments II: Hydraulic and design
considerations. Journal of Hydraulic Engineering 115(8), 1056-1075.
168 References

Pugh, C.A. (1985). Hydraulic model studies of a fuse plug embankment. REC-ERC-85-
7, USBR, Denver, USA.

Ramamurthy, A.S., Vo, N.-D. (1993). Characteristics of circular crested weir. Journal
of Hydraulic Engineering 119(9), 1055-1062.

Ranga Raju, K.G., Srivastava, R., Porey, P.D. (1991). Scale effects in modelling flows
over broad-crested weirs. Irrigation and Power India 47(3), 101-106.

Recking, A., Frey, P., Paquier, A., Belleudy, P., Champagne, J.Y. (2008). Bed-load
transport flume experiments on steep slopes. Journal of Hydraulic Engineering
134(9), 1302-1310.

Ribi, J.M, Pury, J., Boillat, J.-L. (2008). Breach formation in a fuse plug lateral weir.
Proc. Intl. Conf. River Flow, M.S. Altinakar, M.A. Kokpinar, I. Aydin, S. Cokgor
and S. Kirkgoz, eds., Çeşme, Izmir, Turkey, 539-545.

Rickenmann, D. (1990). Bedload transport capacity of slurry flows at steep slopes. VAW
Mitteilung 192, D.L. Vischer, ed., ETH Zurich.

Rosport, M. (1997). Fließwiderstand und Sohlstabilität steiler Fließgewässer unter

Berücksichtigung gebirgsbachtypischer Sohlstruktur. Mitteilung 196. Institut für
Wasserwirtschaft und Kulturtechnik, Universität Karlsruhe, Karlsruhe (in German).

Rouse, H. (1939). Discussion of “Laboratory investigation of flume traction and trans-

portation.” Transaction ASCE 104, 1303-1308.

Rouse, H. (1949). Fundamental principles of movement. Engineering hydraulics, 786-

804, Wiley, New York.

Rozov, A.L. (2003). Modeling of washout of dams. Journal of Hydraulic Research

41(6), 565-577.

Sametz, L. (1981). Beitrag zur Frage der Flutwellenbildung bei progressiven Dammbrü-
chen infolge Überströmung. PhD thesis. Graz University of Technology, Graz, Aus-
tria (in German).
References 169

Sarkaria, G.S., Dworsky, B.H. (1968). Model studies of an armoured rockfill overflow
dam. Water Power 20(10), 455-462.

Schmidt, M. (2000). Hochwasser und Hochwasserschutz in Deutschland vor 1850: Eine

Auswertung der Quellen und Karten. Oldenburg Industrieverlag, München, Germa-
ny (in German).

Schmocker, L., Hager, W.H. (2009). Modelling dike breaching due to overtopping.
Journal of Hydraulic Research 47(5), 585-597.

Schmocker, L., Hager, W.H. (2010). Overtopping and breaching of dikes - Breach
profile and breach flow. Proc. Intl. Conf. River Flow, A. Dittrich, K. Koll, J. Aberle
& P. Geisenhainer, eds., Braunschweig, Germany, 515-522.

Schmocker, L., Rühli, E., Mayor, P.A., Weitbrecht, V., Springman, S.M., Boes, R.M.
(2010). Design and hydraulic modelling of a fuse plug spillway. Dams and Reser-
voirs under Changing Challenges A.J. Schleisss, R.M. Boes, eds., Balkema, Leiden
NL.

Schmocker, L., Halldórsdóttir, B.R., Hager, W.H. (2011). Effect of weir face angles on
circular-crested weir flow. Journal of Hydraulic Engineering 137(6), 637-643.

Schoklitsch, A. (1934). Geschiebebewegung in Flüssen und an Stauwerken. Springer,

Berlin, Germany (in German).

Shields, A. (1936). Anwendung der Ähnlichkeitsmechanik und der Turbulenzforschung

auf die Geschiebebewegung. Mitteilung 26. Preussische Versuchsanstalt für Was-
serbau und Schiffbau, Berlin, Germany (in German).

Sieben, A. (1993). Hydraulics and morphology of mountain rivers: a literature survey.

Report No. 93-4. Faculty of Civil Engineering, TU Delft, Delft NL.

Singh, V.P. (1996). Dam breach modelling technology. Kluwer, Dordrecht NL.

Smart, G.M., Jaeggi, M.N.R. (1983). Sediment transport on steep slopes, VAW Mittei-
lung 64, D.L. Vischer, ed., ETH Zurich, Zürich.
170 References

Spinewine, B., Delobbe, A., Elslander, L., Zech, Y. (2004). Experimental investigation
of the breach growth process in sand dikes. Proc. Intl. Conf. River Flow, M. Greco,
A. Carravetta, R. Della Morte, eds., Naples, Italy, 903-910.

TAW (1991). Guidelines on design of river dikes. Technical Advisory Committee for
Flood Defence (TAW), Published by the Centre for Civil Engineering Research and
Codes (CUR), Gouda NL.

Terzaghi, K. (1925). Erdbaumechanik auf bodenphysikalischer Grundlage. Deuticke,

Wien (in German).

Tinney, E.R., Hsu, E.-Y. (1961). Mechanics of washout of an erodible fuse plug. Jour-
nal of the Hydraulics Division ASCE 87(HY3), 1-29.

Unger, J. (2006). Strömungscharakteristik um kreizylindrische Brückenpfeiler. VAW

Mitteilung 138, H.-E. Minor, ed., ETH Zurich, Zürich (in German).

USACE (2000a). Design and construction of levees. Engineer Manual - Engineering

Design, US Army Corps of Engineers, Washington DC.

U.S. Waterways Experimental Station (USWES) (1935). Study of river-bed material

and their use with special reference to the Lower Mississippi River, Paper 17, 161
pp., Vicksburg, Mississippi.

van Rijn, L.C. (1984a). Sediment transport I: Bed load transport. Journal of Hydraulic
Engineering 110(10), 1431-1456.

van Rijn, L.C. (1984b). Sediment transport II: Suspended load transport. Journal of
Hydraulic Engineering 110(11), 1613-1641.

Visser, P.J., Zhu, Y., Vrijling, J.K. (2006). Breaching of dikes. 30th Int. Conf. Coastal
Engineering, San Diego CA, 2893-2905.

Wahl, T.L. (1997). Predicting embankment dam breach parameters - A needs assess-
ment. Proc. 27th Congress of IAHR San Francisco, 48-53, IAHR, Madrid.

Wahl, T.L. (2004). Uncertainty of prediction of embankment dam breach parameters.

Journal of Hydraulic Engineering 130(5), 389-397.
References 171

Weichert, R.B., Bezzola, G.R., Minor, H.-E. (2009). Bed erosion in steep open chan-
nels. Journal of Hydraulic Research 47(3), 360-371.

Wong, M., Parker, G. (2006). Reanalysis and correction of bed-load relation of Meyer-
Peter and Müller using their own database. Journal of Hydraulic Engineering
132(11), 1159-1168.

Wu, W., Wang, S.S.Y. (2008). One-dimensional explicit finite-volume model for sedi-
ment transport with transient flows over movable beds. Journal Hydraulic Research
46(1), 87-98.
Yalin, M.S. (1963). An expression for bedload transport. Journal of the Hydraulics
Division ASCE 89(3), 221-250.

Yalin, M.S. (1971). Theory of hydraulic models. Macmillan, London UK.

Yalin, M.S. (1972). Mechanics of sediment transport. Pergamon Press, Oxford UK.

Zanke, U. (1982). Grundlagen der Sedimentbewegung. Springer, Berlin (in German).

Zarn, B. (1992). Lokale Gerinneaufweitung, eine Massnahme zur Sohlenstabilisierung

der Emme bei Utzendorf. VAW Mitteilung 118, D.L. Vischer, ed., ETH Zurich, Zü-
rich (in German).

Zhu, Y.H., Visser, P.J., Vrijling, J.K. (2004). Review on Embankment Dam Breach
Modeling. New Developments in Dam Engineering, M. Wieland, Q. Ren, and J.S.Y.
Tan, eds., Taylor & Francis, London, UK, 1189-1196.

Zhu, Y. (2006). Breach growth in clay dikes. PhD Thesis. TU Delft, Delft NL.