Research Collection

Doctoral Thesis

Bedload transport capacity of slurry flows at steep slopes

Author(s):
Rickenmann, Dieter

Publication Date:
1990

Permanent Link:
https://doi.org/10.3929/ethz-a-000555802

Rights / License:
In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more
information please consult the Terms of use.

ETH Library
 Diss ETH No. 9065

BEDLOAD TRANSPORT CAPACITY OF

SLURRY FLOWS AT STEEP SLOPES

A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH

for the degree of

Doctor of Technical Sciences

presented by

Dieter Rickenmann

Dipl. Bailing. ETH

born October 11,1958

citizen of Zumikon (ZH)

and Horben (TG)

accepted on the recommendation of

Prof. Dr. D. Vischer, examiner

Prof. Dr. T. Dracos, co-examiner

Dr. M. Jaggi, co-examiner

1990
Leer -
Vide -

Empty
 -
3 -

Preface

This work was performed at the Laboratory of Hydraulics, Hydrology

and Glaciology of the ETH Zurich. It forms part of a joint research

project with the Swiss Federal Institute for Forest, Snow and Landscape
Research (VSL). The study was financially supported by a research grant

("Schulratsmillion") of the Board of the Swiss Federal Institutes of

Technology.

I would like to express my appreciation to the Director of the Labo¬

ratory, Prof. Dr. D. Vischer, for the support and encouragement and to

Dr. H. Jaggi, head of the river engineering section, for the supervision

and criticism of my work. I am grateful to Dr. T. Davies from New

Zealand who helped setting up the project and reviewed parts of the

manuscript. Thanks are also extended to J. Zeller from the VSL for his

interest in the study-

Several people from the workshop helped operating the flume system; I

acknowledge the effort of Rolf Virz, Stefan Ziist, Bruno Schmid and Gery

Wanner. Thanks also go to Karl Salzmann and Andy Rohrer for the drawings
and to Hare Lehmann for the revision of the english draft.

Finally I would like to mention the moral support that I received

from my wife Regula during the work.

Dieter Rickenmann
Leer -
Vide -

Empty
 -
5 -

CONTENTS

Abstract 8
Zusanenfassung 9
Resoae 10
Riassunto 11
Photos 12

1 INTRODUCTION 15

1.1 The term "debris flow" 15

1.2 Characteristics of debris flows 16
1.3 The transition between "ordinary" floods and debris flows;
Classification schemes 17
1.4 Objective of the study 20

2 LITERATURE REVIEW 22

2.1 Approaches to explain the mechanics of debris flows 22

2.1.1 General remarks 22
2.1.2 Fluid related models 23
2.1.3 Grain flow models 26
2.1.4 Other theories 29
2.2 Rheology and flow of hyperconcentrated suspensions 34
2.2.1 Rheological properties 34
Summary 39
2.2.2 Flow resistance of a Bingham fluid 39
2.2.2.1 Definitions 39
2.2.2.2 Laminar flow 40
2.2.2.3 Laminar-turbulent transition 43
2.2.2.4 Smooth and rough turbulent flow 44
2.2.2.5 Summary 48
2.2.3 Flow and fluid characteristics 48
2.3 Sediment transport 51
2.3.1 Sediment transport in hyperconcentrated flow 51
2.3.1.1 Effect of combined increase in fluid
density and viscosity 51
2.3.1.2 Effect of change only in grain-fluid
density ratio 56
2.3.1.3 Effect of increase in fluid viscosity,
laminar flow 60
2.3.1.4 Hydraulic transport of solids in pipes ... 62
2.3.1.5 Summary 65
2.3.2 Bed load transport at steep slopes and high shear
stresses 66
Summary 75
2.3.3 Initiation of Motion 75
Slope effect and^large relative roughness 76
Low values of Re ; 9 in a Bingham fluid 77
Critical flow discharge 78

3 EXPERIMENTS 80

3.1 Experimental program 80

3.2 Flume apparatus 81
3.3 Clay material and clay suspension 83
 -
6 -

3.4 Bed material 85

3.5 Measuring methods 87
3.5.1 Slope 87
3.5.2 Fluid discharge 87
3.5.3 Sediment discharge 88
3.5.4 Fluid velocity 90
3.5.5 Flow depth 97
3.5.6 Fluid density 101
3.5.7 Rheological parameters 101
3.6 Equilibrium transport conditions 103
3.7 Reproduction of some tests of Smart and Jaeggi 105

4 RESULTS AND ANALYSIS 109

4.1 Methods used in data analysis 109

4.1.1 Correction for sidewall friction 109
4.1.2 Regression analysis 110
4.2 Rheology of clay suspension Ill
4.3 Flow resistance of clay suspension on a fixed rough bed,
without sediment transport 114
4.4 Bed load transport experiments 123
4.5 Case I: Thickness of laminar sublayer is smaller than grain
size 127
4.6 Comparison and analysis with further data 135
4.6.1 Experiments of Smart and Jaeggi 135
4.6.2 Experiments of Meyer-Peter and Mueller 143
4.7 Summary of bed load transport formulae 152
4.8 Flow resistance and other aspects of thegrain-fluid mixture 157
4.9 Case II: Laminar flow around bed load grains 169

5 DISCUSSION 173

5.1 Bed Load Transport 173

Form of proposed equations 173
Comparison with other formulae 174
Inclusion of the Froude number 176
Effect of a change in the grain fluid density ratio -
177
Equations for steep flume data 178
Effect of increasing clay concentration on grain transport .. 180
General flow characteristics 181
Comparison with stream power approach 183
*_ 9 relationships for high shear stresses
-
184
5.2 Flow resistance 186
Analysis of Bingham fluid with Newtonian formulae
a 186
Steep flume data with bed load transport; comparison with
other data and flow resistance equations 188
Comparison with Takhashi's debris flow velocity equation .... 194
Analysis of clay suspension experiments 195
5.3 Effects of high fine material concentrations; implications
for field situation 196
Effect of increasing viscosity 196
Scaling considerations 198
Wide grain size distribution 199
Pulsing behaviour of debris flows 199
 -
7 -

6 SUMMARY AND CONCLUSIONS 201

6.1 201
Summary
6.2 Recommended calculation procedure 204
208
6.3 Example calculation
209
6.4 Suggestions for further research

REFERENCES 211

LIST OF SYMBOLS 223

APPENDIX :

I: Experimental results of clay suspension flows without

sediment 231
transport

II: Experimental results of clay suspension flows with

bed load transport 236
II. 1 : Case I experiments 236
II. 2 : Case II experiments 245

Curriculum vitae 249

 -
8 -

Abstract

During a flood event in a steep torrent, sediment may be transported

by relatively clear water, by a hyperconcentrated slurry or in the form
of a (pulsing) debris flow. The objective of this study was to examine
the effect of an increasing fluid density and viscosity on the bed load
transport capacity of the flow.

clay suspension was used to simulate the slurry consisting of water

A
and particles. The suspension was recirculated in a steep flume;
fine
the maximum volume concentration of clay particles in the flow was 22%.
The clay suspension showed a non-Newtonian rheologic behaviour; it was
treated as a Bingham fluid. In one test series, the flow resistance of
the clay suspension flows without any sediment transport was measured.
By using an effective viscosity for the given flow conditions, the
results could be successfully analysed with conventional Newtonian
formulae. In a second test series, the equilibrium bed load transport
rates of the clay suspension flows were determined for different flow
conditions. With increasing fluid density, transport rates were observed
to increase continuously, for the experimental conditions by as much as
a factor of 3 as compared to the corresponding clear water flows.
Viscosity effects were found to become important if the particle
Reynolds number is smaller than about 10. Then the flow around the
grains was laminar, and transport rates decreased again.

For the
majority of the bed load transport tests, density effects
were and viscous effects were negligible. These experiments
dominant
were analysed together with the steep flume data of Smart and Jaggi
(1983). By adjusting the density factor, both data sets could be
described by a bed load transport formula (based on the flow rate)
similar to the one of Smart/Jaggi. By including also the comprehensive
data set of Meyer-Peter and Muller (1948) in the analysis, it was shown
that an alternative transport equation (based on the dimensionless shear
stress) is of more general applicability; with this relationship, bed
load transport rates can be predicted both for low and steep slope
conditions.

At steeper slopes, the transported grains occupy a considerable part

of the (mixture) flow depth, thus influencing the flow resistance
analysis. Depending on the known or assumed parameters, two different
calculation procedures are proposed to determine the unknown parameters
of the flow of the grain-fluid mixture.
 -
9 -

Zusammenfassung

Im Verlaufe eines Hochwasserereignisses kann in einem Wildbach Ge-

schiebe auf verschiedene Arten talwarts verfrachtet werden: Als Geschie-
betrieb durch relativ klares Wasser, als murgangartiger Transport durch
eine schlammartige Flussigkeit sowie in Form eines Murgangs (mit einer
Front). Ziel der vorliegenden Arbeit war es, den Einfluss von zunehmen-
der Dichte und Zahigkeit der Flussigkeit auf das Geschiebetransport-
vermogen zu untersuchen.

Mit einer Ton-Suspension wurde die schlammartige Flussigkeit nach-

gebildet, die in Natur aus Wasser und Feinmaterial besteht. Die Suspen¬
sion wurde in einer steilen Versuchsrinne rezirkuliert, mit einer
maximalen Volumenkonzentration an Tonpartikeln von 22%. Die Ton-Suspen¬
sion verhalt sich wie eine nicht-Newton'sche Flilssigkeit; sie wurde als
Bingham'sche Flussigkeit behandelt. In einer ersten Versuchsserie wurde
der Fliesswiderstand von tonhaltigen Abfliissen ohne Geschiebetransport
gemessen. Eine Auswertung mit bekannten Newton'schen Formeln ist mog-
lich, falls eine effektive Zahigkeit des Abflusses eingefiihrt wird. In
einer zweiten Versuchsserie wurde dieGeschiebetransportkapazitat fur
veschiedene Abflussbedingungen der Ton-Suspension bestimmt. Es zeigte
sich, dass die Transportraten mit steigender Fliissigkeitsdichte zuneh-
men. In den Versuchen betrug die Zunahme bis zu einem Faktor 3, im
Vergleich zu den entsprechenden Reinwasserabfliissen. Ein Einfluss der
Zahigkeit wurde fur Versuche mit einer Korn-Reynolds-Zahl kleiner als
etwa 10 festgestellt. In diesem Fall werden die Korner laminar umstromt,
und die Transportraten nehmen wieder ab.

Fur Mehrheit der Geschiebetransportversuche waren die Dichte-

die
effekte massgebend, wahrend der Einfluss der Zahigkeitsanderung vernach-
lassigt werden konnte. Diese Versuche wurden zusammen mit denjenigen von
Smart und Jaggi (1983) ausgewertet. Es resultierte eine Geschiebetrans-
portformel (in Funktion des Abflusses) ahnlich der von Smart/Jaggi
vorgeschlagenen, wobei der Dichtefaktor angepasst werden musste. In
einem weiteren Auswertungsschritt wurden auch die umfangreichen Ver-
suchsresultate von Meyer-Peter und Mtiller (1948) einbezogen. Dabei
zeigte sich, dass eine Transportgleichung, die in Abhangigkeit der
dimensionslosen Sohlenschubspannung formuliert wird, eine allgemeinere
Gultigkeit aufweist; mit dieser Formel konnen Geschiebetransportraten
sowohl fUr flache als auch fiir steile Gefalle berechnet werden.

Bei steileren Gefallen wird ein betrachtlicher Teil der (Gemisch-)

Ablusstiefe von transportierten Kornern besetzt; dies muss bei der
Bestimmung des Fliesswiderstandes berucksichtigt werden. Je nach Frage-
stellung werden zwei verschiedene Berechnungsverfahren vorgeschlagen,
womit die unbekannten Abflussgrossen des Korn-Fliissigkeits-Gemisches
bestimmt werden konnen.
 -
10 -

Resume

Lors d'une crue dans un torrent, le sediment peutetre transports par

de l'eau relativement claire, par un melange eau-suspensions a tres
haute concentration ou encore sous forme d'une lave torrentielle (avec
front de propagation). Le but de la presente etude etait d'examiner
1'influence d'un accroissement de la densite et de la viscosite du
fluide sur la capacite de transport solide.

Le fluide d'aspect boueux, qui en nature est compose d'eau et de ma-

teriaux fins, a ete simule en laboratoire par une suspension argileuse.
Cette suspension a ete mise en circulation dans un canal a forte pente,
la concentration maximaleen volume des particules d'argile etant de

22%. suspension argileuse presente les caractSristiques d'un fluide

La
non-newtonien; elle a ete traitee comme un fluide de Bingham. Dans une
premiere serie d'essais, la resistance a l'ecoulement de la suspension
argileuse a ete mesuree, en l'absence de tout transport solide. Une ana¬
lyse des resultats est possible sur la base des formules de Newton, si
une viscosite effective de l'ecoulement est introduite. Dans une deux-
ieme serie d'essais, la capacite de transport solide de la suspension
argileuse a ete determinee pour differentes conditions d'ecoulement. II
a ete constate que les quantites de sediment transports augmentent avec
la densite du fluide. Dans les essais realises, cette augmentation a
atteint facteur 3 par rapport au transport solide observe en eau
un

claire. influence significative de la viscosite n'est apparue que

Une
pour des nombres de Reynolds des particules (rapportes au diametre du
grain) inferieur a environ 10. L'ecoulement autour des grains est alors
laminaire et les quantites transporters decroissent a nouveau.

Pour la essais de transport solide, les effets dus a la

majorite des
densite se determinants, de telle sorte que 1'influence de
sont averes

variations de la viscosite a ete negligee. Ces essais ont ete analyses

en regard des essais de Smart et Jaggi (1983). Une formule de transport
solide (en fonction de l'ecoulement) semblable a celle de Smart et Jaggi
a pu etre proposee, apres une adaptation du terme decrivant I'effet de
densite. Une seconde analyse, incluant les nombreux resultats d'essai de
Meyer-Peter et MUller (1948), a montre qu'une equation de transport, ex-
primee en fonction de la force d'entrainement adimensionelle, presentait
un caractere plus general. A I'aide de cette relation, les quantites de
sediments transportes peuvent etre calculees dans le cas d'un canal a

faible ou aforte pente.

Aux fortes pentes, les grains du materiau transports occupent une

grande partie de la profondeur de l'ecoulement (du melange). Ce fait
doit etre pris en consideration lors de la determination de la resis¬
tance a l'ecoulement. Selon la nature du probleme a resoudre, deux pro¬
cessus de calcul differents permettant de determiner les parametres in-
connus de l'ecoulement du melange fluide-materiau solide sont proposes.
 -
11 -

Riassunto

Durante una piene in un torrente i sedimenti possono essere traspor-

tati a valle in diverse maniere: come trasporto solido di fondo con
l'acqua che resta relativamente limpida, come liquido fangoso, oppure
come propria colata di fango (con un fronte). Scopo del presente
vera e

lavoro era di indagare 1'influenza di una crescente densita e viscosita

del liquido sulla capacita di trasporto solido.

II liquido fangoso e stato riprodotto

sospensione argillosa con una

che in natura consiste di acqua e materiale

La sospensione, che fino.
raggiungeva una concentrazione volumetrica massima di argilla del 22%,
veniva recircolata in un ripido canale. Tale sospensione si comporta
come un liquido non newtoniano ed e stata trattata come un liquido
definito secondo Bingham. In una prima serie di esperienze su modello si
e misurata la resistenza idraulica della sospensione senza trasporto
solido. L'analisi con le consuete formule secondo Newton e qui possibile
se si tiene conto della viscosita effettiva del deflusso. In una seconda
fase si e capacita di trasporto solido per diverse condi-
stabilita la
zioni di deflusso della sospensione. Si e potuto constatare come la
capacita di trasporto aumenti con il crescere della densita del liquido
che nelle esperienze raggiungeva un valore massimo di 3 volte superiore
ad un corrispondente deflusso con acqua limpida. Una influenza della
viscosita e stata riscontrata nelle esperienze dove il numero di Reynold
della grana era minore di circa 10. In questo caso la corrente attorno
ai grani e laminare e le quote di trasporto diminuiscono.

Per la maggior parte delle esperienze con trasporto solido gli

effetti rilevanti erano dati dalla variazione delle densita, mentre
1'influenza della variazione della viscosita era trascurabile. Queste
esperienze sono state analizzate tenendo anche conto dei risultati delle

precedenti ricerche svolte da Smart e Jaggi (1983). II risultato e una

formula che permette il calcolo del trasporto solido (in funzione del
deflusso) simile a quella proposta da Smart e JSggi, dove per6 il ter-
mine concernente la densita e stato adattato. In una fase successiva
sono state considerate anche le ricerche piu estese svolte da Meyer-
Peter e Muller (1948), grazie alle quali si
potuto dimostrare che una e

formula per il calcolo del trasporto solido espressa in funzione della

forza di trascinamento resa senza dimensioni pud trovare applicazione
generale indipendentemente dalla pendenza longitudinale.

Quando le pendenze longitudinali diventano ripide una porzione impor-

tante della profondita di deflusso viene occupata da grani comportando
un cambiamento della resistenza idraulica. A seconda delle costellazioni

richieste vengono proposti due diversi procedimenti con i quali e possi¬

bile stabilire le condizioni idrauliche di deflusso della miscela acqua-
grani mancanti.
 12

Photo 1: Example of clay suspension flow in the experimental flume

(at slope of 5 %, flow rate of 10 L/s and 15 vol.% clay con¬

centration). Note similarity to flow conditions in Photo 3.

 -
13 -

Photo 2: Front of debris wave moving down channel (about 3 m wide) at

Wrightwood, California, 20 May 1969. There is fluid mud in

the foreground, bouldery front in centre. (Photo reproduced

from Johnson and Rodine, 1984, "Debris Flow", in Slope

Instability [eds. D. Brunsen and D. Prior], reprinted by

permission of Jon Wiley & Sons, Ltd.)

 -
14 -

;y\j' sM

Photo 3: Steadily flowing debris about 10 m behind bouldery front

shown in Photo 2. (Photo reproduced from Johnson and Rodine,

1984, "Debris Flow", in Slope Instability [eds. D. Brunsen

and D. Prior], reprinted by permission of Jon Wiley & Sons,

Ltd.) This situation represents similar flow conditions as

simulated in the laboratory experiments.

 -
15 -

1 INTRODUCTION

1.1 The term "debris £lov"

Johnson's (1970) definition of a debris flow as a "gravity induced

mass movement intermediate between landslide and waterflooding, with

mechanical characteristics different from either of these processes" may

be characteristic for the state of knowledge in this field. In a posi¬
tive formulation, one could state that a "debris flow is the downslope
movement of granular solids dispersed in a clay-water fluid and is known
to be a significant sediment-transport process on land" (Hampton, 1975).

Other terms used for the same or a similar phenomenon include: mud(-
rock)-flow, mud slide, earth flow, lahar, debris slide, debris ava¬

lanche, debris torrent, alpine mudflow and mountain debris flow. To some

extent their use depends on the physical characteristics of a particular

flow but sometimes they are also used synonynomously. Often the term

"mudflow" refers more to flows containing mainly fine-grained earth

material. A comprehensive collection of "mudflow" definitions is given
in Naik (1983).

VanDine (1984) defined a debris torrent as "a mass movement that

involves water-charged, predominantly coarse-grained inorganic and or¬

ganic material flowing rapidly down a steep, confined, pre-existing

channel" and suggested that this term should be preferred since "debris
flow" can also refer to laterally unconfined flows.

Besides the fact that the mechanics of a debris flow are still poorly

understood the variety of terms used for more or less the same pheno¬
menon illustrates that there is also no generally accepted terminology
and classification. It seems, however, that the term "debris flow" has

been mostly used in the literature to refer to the phenomenon charac¬

terised by the above definitions and that it is probably the most widely
accepted common expression.
 -
16 -

1.2 Characteristics of debris flows

Probably the first comprehensive work on the topic dates back to 1910

when the Austrian engineer Stiny compiled a book entitled "Die Muren",

which refers mainly to debris flows in the Alps. Excellent descriptions

of the conditions of occurence, the flow characteristics, the physical

properties and geomorphological aspects of debris flows are given in

Costa (1984), Johnson and Rodine (1984), and Innes (1984). These reports
also contain extensive literature reviews and present a good introduc¬

tion into the subject.

Debris flows are often associated with heavy rainfall, saturated

slopes, availability of loose material, steep channels and potentially

disastrous effects in the fan area. As compared to "ordinary" floods in
steep channels the flow behaviour and the physical properties of debris

flows are distinctly different. While floods with minor sediment con¬

centrations are in most cases treated as stationary and uniform flows,

it is often observed that the front of a debris flow consists of a wave-

shaped front part containing a lot of coarse boulders, possibly wood and

comparativly little water while the preceding flow is more fluidic and

less deep. It is also possible that a series of additional pulses follow

the initial surge wave. Debris flow data suggests that such flows

usually start at slopes steeper than 21% and that deposition occurs at

valley or fan slopes steeper than 5% (Takahashi, 1981).

The flow depth of the front or a pulse can be up to 10 or 15 m, and

the corresponding velocities are often in the range between 5 and

15 m/s. The bulk density of the mixture is usually larger than 1.4 T/m3

and can reach values of about 2.4 T/m3, which is possible due to the

generally very wide grain size distribution of the transported sediment.

Such a dense mixture can have viscosities several orders of magnitude
higher than that of pure water. Viscometric measurements of debris flow

materials show a non-Newtonian rheologic behaviour. It is not surpri¬

sing, therefore, that debris flows are often observed to flow laminarly

and that they have been said to show a similar flow behaviour to wet

concrete.
 -
17 -

1.3 The transition between "ordinary" floods and debris flows;

Classification schemes

A rather comprehensive classification scheme considering various

aspects of debris flows has been proposed by Kurdin (1973). According to

this scheme the four following criteria are used: Nature of debris flow

formation (mechanism of water and solids supply), rheological and flow

characteristics, composition (relative proportion of water, fine and

coarse material) and size of a debris flow event (eroded volumes, de-

structiveness). With regard to the flow behaviour the second and third

items are mainly important. A similar classification is applied in China

(Du et al., 1986).

Beverage and Culbertson (1964) made one of the first attempts to

distinguish between different debris flow types according to the con¬

centration of solids in the flow and to determine these limiting condi¬

tions. Since then a number of researchers have introduced similar clas¬

sifications. A summary of the most common classification schemes based

on sediment concentration is given by Bradley (1986) and is reproduced

in Table 1.1. As is evident from the table there is not only some dis¬

agreement on the appropriate terms to be used but also on the delinea¬

tion of the different flow types. According to O'Brien and Julien (1984)

the maximum packing density of non-uniform silts, sands and gravels

ranges from 67 to 88 percent by volume. This upper limit of approxi¬

mately 90 percent solids by volume corresponds to a maximum mixture

density of about 2.4 T/m3.

Thus it seems that the use of the sediment concentration alone is not

sufficient for a proper classification of different flow types. Pierson

and Costa (1984, as indicated by Bradley, 1986) distinguish between

streamflow and slurry flow according to whether the flow is Newtonian or

non-Newtonian, and between slurry flow and granular flow, the limit

beeing a function of particle size and gradation. According to Davies

(1986, 1988), who reviewed numerous debris flow descriptions, basically

three different types can be distinguished: Low density, steadily moving
turbulent flows, carrying coarse particles as bed load only, the fluid

consisting of a slurry of fine material (type 1); high density, laminar

flows, coarse and fine particles being distributed uniformly over the

depth, of unsteady nature with several pulses (type 2) or with only a

 Concentration Percent WT =
by Weight (100% by 1,000,000 ppm)

23 40 52 63 72 80 87 93 97 100

Concentration Percent Volume =

by (S.G. 2.65)

Source 10 30 40 50 60 70 80 90

Beverage and
:ulbertson (1964) High Extreme Hyperconcentrated Mud Flow

:osta (1984) Water Flood Hyperconcentrated Debris Flow

3'Brien and Julien Water Flood Mud Flood Mud Landslide

(1985) using National Flow
Research Council (1982)

Fluid Flow Debris or Grain Flow Fall, Landslide,

Takahashi (1981) Creep, Sturzstrom,
Pyroclastic Flow

< Debris or Mud Flow

Chinese Investigators
(Fan and Dou, 1980) Hyperconcentrated Flow

Sediment Laden

STREAMFLOW SLURRY FLOW GRANULAR FLOW Fast

Pierson & Costa Normal. Hyperconcentrated (Debris Torrent), Sturzstrom, Debris
(1984) Debris & Mud Flow, Avalanche, Earthlow, Slow
Solifluction Soil Creep
 -
19 -

single pulse (type 3). His classification scheme is reproduced in Table

1.2. The last two types have a higher viscosity than the first one, and

a selective deposition of the coarser particles does not seem possible;

because of greater flow depths and velocities their destructive power is

considerable. As limiting density to distinguish between the steady and

the unsteady type debris flow Davies proposed a density of about 1.6 to

1.8 T/m3, which corresponds to a solids volume concentration of approxi¬

mately 36 to 492. He pointed out, however, that the transition seems to

be rather abrupt if a particular flow changes from one type to the

other. This abrupt transition is also reflected in deposits of either a

"water flood" or a debris flow (Costa, 1984).

Flow Tvce 1 2 3

Characteristic

Flow Steady Pulsing Single pulse

Appearance Turbulent Laminar Laminar

Sizes present

above bed Fine Fine + coarse Coarse + fine

Coarse load At bed Throughout Throughout depth

depth

Density < 1.6 T/b3 > 1.8 T/b3 i 1.8 T/m3

Viscosity odO-100 x water >1000 x water >1000 x water

Velocity Low, o* 2 m/s High; a 3-5b/s High

Effect on bed Depositlonal Very erosive Very erosive

Table 1.2 : Debris flow characteristics, after Davies (1988).

Chinese researchers (Du et al., 1986) use a similar classification

based on fluid and kinematic properties. They make a distinction between

low-viscous and viscous debris flows. Their transition region lies in

the density range of 1.8 to 2.0 T/m3, and it is associated with rapidly
increasing viscosity parameters in this range. Possibly due to the fact

that Chinese debris flows consist mainly of fine material, their limi¬

ting values are higher than Davies' (1988) values. A similar separation
into rapid turbulent and quasilaminar (structural) mud-streams is also

proposed by Soviet researchers (Syanozhetsky et al., 1973).

 -
20 -

In Japan Takahashi's debris flow models are widely accepted. Mainly

based on flow characteristics and composition, Takahashi (1987) sug¬

gested a fourfold subdivision of the general term debris or mud flow:

Stony debris flows only have a minor content of water and of fine

material. Muddy debris flows or turbulent mudflows consist predominantly

of smaller particles, and large scale turbulence is present due to ran¬

domly moving clusters of particles. A third case is a mixture of a stony

debris flow in the near-bed layers with a turbulent mud flow prevailing
in the upper layers. In an immature debris flow particles move basically
as an intense bed load while the upper layer contains only very few

grains.

1.4 Objective of the study

Debris flows primarily occur on steep slopes and can obviously have

an enormous erosive power and transport capacity. In the past, studies

on sediment transport in rivers and flumes were limited to bedslopes up

to a few percent. It is only recently that interest on sediment trans¬

port processes in steep channels has grown. However, it is very dif¬

ficult to measure sediment transport rates in torrents in the field, and

only a few experimental studies have been carried out in steep flumes.

Such studies, including flume slopes in excess of 10%, were performed by

Mizuyama (1977), Mizuyama and Shimohigashi (1985), and Smart and Jaeggi
(1983).

The transporting fluid used in these flume tests was clear water. It

is likely that in torrents, already under "ordinary" flood conditions,

considerable amounts of fine material are transported in suspension if

the geology and the flood history allow so. However, it is difficult to

predict how the increasing density and viscosity of the fluid influence

the flow behaviour and the sediment transport rates if the flow changes
from an "ordinary" flood to a low density debris flow (type 1 of Davies'

classification). Therefore it was decided to carry out a series of steep

flume tests using a clay suspension to represent the fine slurry of a

debris flow, and to measure the flow resistance and the sediment trans¬

port capacity for various clay concentrations. Since the experiments of

Smart and Jaeggi (1983) were performed at the same hydraulic laboratory

they could serve as a reference condition for the clear water case.
 - 21 -

Based on the various proposed classification schemes it can be con¬

cluded that the transition from the steady (type 1) to the unsteady

(type 2 or 3) debris flow should occur at a mixture density in the range

of 1.6 to 2.0 T/m3, probably depending also on flow conditions. Calcu¬

lated bed load concentrations (by volume) for the Smart/Jaeggi tests are

as high as 33% for the same gravel material as was to be used in the new

experiments. In preliminary tests clay suspensions with (volume) con¬

centrations up to 20% and more were successfully recirculated in a small

flume. From this information it could be estimated that bulk densities

in the critical transition range should be reached with the planned

experimental arrangement. Thus it seemed possible that unsteady flow
would develop as a result of a natural flow instability in this critical

density range.

To study the flow behaviour of debris flow surges an experimental

investigation was performed by Davies (1988) at the same hydraulic labo¬

ratory using a conveyor belt flume where the bed moves upstream and the
-
in reality unsteady -
front remains stationary with respect to the

side walls and the observer.

In this study, the term "debris flow" is restricted to the unsteady,

pulsing flows (type 2 and 3 of Davies' (1988) classification), while the

steady flows (type 1) are referred to as slurry flows or hyperconcen¬
trated flows with sediment transport.

Photo 2 (p. 13) shows a typical front of a debris flow. The flow

conditions examined in the present study are illustrated by Photos 1 (p.

12) and 3 (p. 14).
 -
22 -

2 LITERATURE REVIEW

2.1 Approaches to explain the mechanics of debris flows

2.1.1 General remarks

There are many countries that have to face problems due to the oc¬

curence of debris flows which can impose a constant threat to certain

regions. Debris flows have long been known in mountainous regions in the

Alps, in the Soviet Union, in Japan, in the Rocky Mountain areas of

North America and in other places of the world.

Research on debris flows and related problems started in the 1960-

ties, especially in Japan and in the United States. It began probably

around the same period in China although many of their studies were only
published in English towards 1980. Papers from Russian scientists ap¬

peared from time to time but it seems difficult to get a comprehensive

picture of their research activities. In Austria it is primarily en¬

gineers that have been concerned with this topic, and their interest was

mainly focussed on technical and practical aspects of planning counter-

measures against debris flow events. Recently some Canadian researchers

have also summarized approaches to determine the flow characteristics,

with special regard to mitigation of debris flow hazard.

Topics of debris flow research cover various aspects such as: Mecha¬

nisms of formation and initiation; mechanics and rheological cha¬

racteristics of motion, including fluid related and grain flow models;

geologic analysis of deposits and determination of catchment charac¬

teristics and geomorphological aspects; and development of models for

prediction of debris flow events and of (rough) rules for the design of
torrent control projects.

A very concise review on the state of the art in modelling the motion
of a debris flow was made by Iverson and Denlinger (1987). As a first

approximation, the debris flow mass can be separated into a fluid

matrix, consisting of water and the fine material, and the coarser par¬

ticles. It is then assumed that the slurry with the fines behaves like a
 -
23 -

pseudo-homogeneous fluid that carries along the coarse fraction. There

are two main problems connected with this approach: Which is the limi¬

ting particle size that delinates the fine from the coarse particles,
and what is the relative importance of grain-grain contacts and of

fluid-grain interactions with respect to the total energy dissipation ?

In the past, two principally different debris flow theories have been

developped. The viscoplastic model was applied independently by Yano and

Daido (1965) and by Johnson (1965) to flows containing a lot of fine

material in a viscous slurry, and is based on the approach by Bingham

(1922) to describe a particular type of non-Newtonian flow. To treat the

granular-type debris flows that consist mainly of coarser particles and

only have little water, Takahashi (1978, 1980) proposed a set of equa¬

tions based on the concept of dispersive grain stresses introduced by

Bagnold (1954).

More recently attempts have been made to incorporate both fluid

effects and grain-grain interactions in one single model (Ward and
O'Brien, 1981; O'Brien and Julien, 1984; Chen, 1983,1986a,1988a,b).
However, these theories have not yet been tested sufficiently with ex¬

perimental data because it is very difficult also in the laboratory to

examine the grain and fluid processes in detail.

2.1.2 Fluid related models

Suspensions of fine particles show a trend towards non-Newtonian

rheological behaviour, the effect becoming more pronounced with in¬

creasing conentration. The fluid viscosity is no longer a constant but

depends on the velocity gradient dv/dy (shear rate) in the flow. The

rheological behaviour can be determined for example with a viscometer

where the shear stress T is measured as a function of the applied shear

rate, for laminar flow conditions.

According to Brauer (1971), the most commonly used approaches to

describe a non-Newtonian viscous fluid are the power law model of

Ostwald and De Waele, the Prandtl-Eyring model and the Bingham model.
The power law relation is given as:

- ' A
<!>" (2.1)
 -
24 -

where A is an empirical constant, v is the local fluid velocity, y is

the coordinate perpendicular to the flow direction, and n an empirical

constant called the fluid index. The Bingham model is written as:

x =
xB yg) (2.2)

where T_ is the Bingham (yield) stress and f\- the Bingham viscosity. The

apparent viscosity, being defined analogous to the viscosity of a New¬

tonian fluid, is given for a Bingham fluid as:

na " +
"B <2-3>
dv7d7

The reological behaviour of fluids that can be characterised with

these approaches is depicted in Fig. 2.1. In the treatment of hyper-

concetrated flows or of the viscous slurry of debris flows it is

primarily the Bingham model and sometimes the power law relation that

are commonly used. Debris flows have been modelled as Bingham fluids,
dilatant fluids, or as viscoplastic fluids.

Viscous non-Newtonian fluids are usually divided into three cate¬

gories: Pseudoplastic fluids, described by the power law model having an

exponent n<l, and Bingham fluids both show a shear-thinning effect,

which means that the viscosity decreases with increasing shear rate.

Dilatant fluids, having an exponent n>l in the power law model, on the

other hand show a shear-thickening effect, the viscosity increasing with

increasing applied shear. For n=l the power law model is equal to the

Newtonian relation, with K being equal to the Newtonian viscosity. It is

interesting that both for power law fluids with n<l and for Bingham
fluids a certain minimum shear stress is required for the flow to start.

It should be noted that these models refer to ideal fluids. In re¬

ality many fluids show a behaviour that is described by more than only
one model, depending on the shear range of interest. And even within a

limited range of shear rates, the use of a particular model is often an

approximation to the true rheologic behaviour.

Yano and Daido (1965) made rheological measurements and determined

velocity profiles of a clay suspension of various concentrations flowing

in an open channel. They showed that both the pseudoplastic and the
 -
25 -

s^Y*

>

Shear velocity ^
dy

Fig. 2.1 : Rheological models for non-Newtonian viscous fluids.

Bingham model can be used to analyse the flow behaviour. D.G. Thomas

(1963a) performed pipe flow experiments with different fine material

suspensions including a kaoline slurry. In the analysis he compared

various rheological models and concluded that the use of a more complex
model than the Bingham approach is not justified. Johnson (1965), based

on laboratory tests and observations of natural debris flows, proposed a

modified Bingham model, the so called Coulomb-viscous model:

+ p
rn
tana
s
+
v£> (2.4)

where c' is the cohesion, p is the normal stress and o is the static

angle of internal friction. In the case of dilute suspensions or if

there are only smaller grain sizes present in the flow, the first two

terms of equ. (2.4) may be replaced by a single shear strength para¬

meter, becoming equal to the Bingham yield stress T-. The Bingham model
or a modified form thereof have been extensively used in the United
 -
26 -

States (Johnson and Rodine, 1984; Naik, 1983) and in China (Cheng,
1986b) to explain many aspects of debris flows.

Since the Bingham model appeared adequate to describe the clay sus¬

pension used in the experimental work the rheology and the flow re¬

sistance of a Bingham fluid is treated in more detail in section 2.2 .

2.1.3 Grain flow models

In a fundamental work Bagnold (1954) studied the dynamics of grain

dispersions. In his experiments he sheared neutrally buoyant wax spheres
in a Newtonian fluid in an annular space between two concentric

cylinders. By measuring the torque necessary to shear the grain

dispersion, he could determine the solids shear stress T as a function

of the applied shear rate for various grain concentrations. Since the

flow depth normal to the shear plane was confined, the dilation of the

moving grain mass caused a normal dispersive pressure P, which was also

recorded. He found that this dispersive pressure P is proportional to

the shear stress T acting between different grain layers. The proportio¬
nality factor can be termed dynamic friction angle a, and the relation

is written as:

T = P tana (2.5)

Depending on grain size, grain concentration, fluid viscosity and

flow conditions, Bagnold (1954) distinguished two different flow regimes

where a is approximately constant. If grain inertia dominates the flow
behaviour tana was found to be about 0.75, and about 0.32 if the fluid

viscosity is predominant. In the first case the resistive shear stress

is mainly due to grain to grain contacts while in the second case it is

primarily the interaction between the grains and the fluid that deter¬

mines the flow behaviour. To separate the two flow regimes a dimension-

less grain flow parameter G is given by Bagnold (1956) as:

G>=5g (2.6)

where a is the grain density, d the grain diameter, r\ the dynamic fluid
viscosity, and X the linear grain concentration which can be expressed
 - 27 -

in terms of the solids volume concentration C and the maximum possible

packing concentration C^ as:

X =
(2-7)
(CVC )"3-l
v
* s'

Grain shearing will be inertial if G2>1500 and macroviscous if G*<100;

in the transition region both grain to grain and grain to fluid inter¬

actions are important. Bagnold (1954) showed that the dimensionless

grain number G is analogous to the particle Reynolds number Re defined

Re* = d(T^g)"' for the inertial case (2.8)

and Re = ,
for the macroviscous case (2.9)

it
where p is the fluid density. In terms of Re the transition region lies
between 55 and 10. For volume grain concentrations C <57% (X<12) in his

experiments, Bagnold (1954) determined two empirical relations to ex¬

press the solid transmitted shear stress T:

T =
a.c(Xd)2(du/dy)2sina for the inertial case (2.10)

3/2
and T = a X (du/dy)sina for the macroviscous case (2.11)

where a. and a are empirical constants, u denotes the local grain

velocity, and du/dy is the velocity gradient normal to the shear plane.

Bagnold (1956) noted that in his experiments for X < 22 grain

shearing became possible and then the dispersion behaved like a granular
paste. He found that a residual shear stress at zero shear rate only
existed for X > 14. He termed this concentration value the "fluid" limit

below which the grain dispersion would behave like a Newtonian fluid.

For natural, reasonably rounded uniform bed material these X-values cor¬

respond to volume concentrations of 57% and 52%. At the "fluid" limit

(X=14, C =52%) Bagnold found that the overall shear stress should be at

least 100 times greater than the fluid shear due to viscous forces only.
He could not determine the residual fluid shear stress due to turbulence

(as modified by the presence of grains) but he stated that this effect
 -
28 -

seems to become less and less important with increasing grain concentra¬

tion.

Based on Bagnolds (1954) concept of dispersive pressure, Takahashi

(1978) developed a set of equations to describe the flow of a concen¬

trated grain-water mixture in a wide open channel. In his theoretical

analysis he considered two different cases which he verified by experi¬

ments. In a steep flume he simulated the formation of a bore front by

adding an abrupt water supply over a nearly saturated, erodible gravel

bed. In one case the movement of the front asymptotically reached a

quasi-steady state because the scouring depth of the bed remained limi¬

ted. In the other case, at very steep flume slopes, he observed a pro¬

gressive erosion of the bed and a continuous growth of the bore front,
while the front velocity remained almost constant. Having determined the

experimental constants, the grain concentration, the propagation velo¬

city, and the mixture depth could be calculated by the theoretical

relationships. For the given experimental conditions the formation of a

quasi-steady state front which reaches an equilibrium grain concentra¬

tion should theoretically occur in the slope range between 26% and 42%,

while the bore front is supposed to continuously grow on slopes steeper

than 42%.

Takahashi (1980) extended his analysis to describe the experimentally

observed longitudinal profile of the snout for the quasy steady-state
case where no bed erosion occurred beneath the snout. Additional tests

with a non-uniform bed material showed a tendency for the velocity

profile to become more uniform with the wider grain size distribution.
He observed an accumulation of the coarser particles at the front which

he attributed to the larger dispersive pressure associated with the

coarser particles, driving them to the flow surface that moves faster

than the lower layers.

For a stony type debris flow in the inertial regime where the mo¬

mentum exchange is mainly due to grain contacts and the role of the

interstitial fluid can be neglected, Takahashi (1978) obtained a re¬

lation for the vertical velocity distribution of a quasi-steadily moving

front as:

u
Id <fiifa (V^-VS
1
)1/2 c^)1'3-1)
s
("3/2-("-y)V2) (2-^)
 -
29 -

where 0 is the slope angle, g the acceleration due to gravity, h the

(mixture) flow depth and the vertical coordinate y is measured from the

bed upwards. Takahashi (1987) later modified his approach to account for

a nonuniform grain distribution over the depth. The velocity distribu¬

tion calculated with this new model matches his experimental results
better than his first model assuming a constant concentration, and it

also agrees fairly well with experimental data of Tsubaki et al. (1982).

The effect of clay in the interstitial fluid is discussed by Taka¬

hashi (1980) to suggest that most debris flows are in Bagnold's inertial

regime. He estimated the yield strength t„ and the apparent viscosity by

formulae taking into account size and concentration of the clay par¬

ticles, and concluded that the effect of these parameters would be

negligible in real debris flows. However, as shown below, the presence

of silt and fine sand - in addition to the clay -

in the fluid matrix

can increase the apparent viscosity by several orders of magnitudes. In

fact, Davies (1985, 1988) put forward the hypothesis that the unsteady,

pulsing behaviour of debris flows is associated with the macroviscous

and not the inertial flow regime.

According to Iverson and Denlinger (1987) the main deficiencies of

Takahashi's theory are the neglection of dynamic effects of the inter¬

stitial fluid matrix and the assumption (in Takahashi's first model)
that the solid particles are uniformly dispersed throughout the flow

depth. They claim that this latter assumption leads to a mathematical

overdeterminacy in that his theory basicly requires four equations

having only three unknowns. Other restrictions are the assumption of an

average grain size to represent an often very nonuniform material and

the fact that a yield strength cannot be predicted by the dilatant fluid
model (Naik, 1983). In spite of these limitations Takahashi's model has
been applied in Japan with some success to treat granular type debris
flows.

2.1.4 Other theories

Davies (1988) suggested that his type 1 flows (s. section 1.3) pro¬

bably do not differ fundamentally from ordinary streamflow. Conventional

flow resistance and sediment transport formulae may still apply although
they might have to be modified to account for the changed density and
 -
30 -

viscosity of the slurry. He expected that the single-pulse, unsteady

(type 3) flows can be at least partly explained by Takahashi's (1978,

1980) theory.

Based on the analysis of existing debris flow descriptions Davies

(1988) offered a new hypothesis concerning the occurrence of the

multiple-pulse, unsteady (type 2) flows. He proposed that macroviscous

flow (Bagnold's grain flow number G2 being less than 100) is a necessary

condition for the more or less uniform distribution of the coarser

grains over the flow depth and for the onset of several pulses in large
debris flow. According to Bagnold (1955) the increase of the shear

stress in a macroviscous flow of high grain concentration leads to a

decrease of the transport capacity of the flow. Since a selective depo¬

sition of larger grains is not possible in a macroviscous flow the

decrease in the transport rate must be achieved by a decrease in the

mean flow velocity. This, in turn, can bring about an instability of the
flow, in that slower moving, deeper reaches begin to form and develop
into surges. The initial increase in shear stress may be due to a change
in flow depth, slope or grain concentration.

Assuming typical values for the slurry density, the flow depth, the

grain size, the grain concentration, and the bed slope of a hypothetical
debris flow, Davies (1988) calculated the necessary apparent viscosity
t| of the slurry for macroviscous flow with equ. (2.6) to be about 5000

times that of water. In view of the measured rheological parameters of

real debris flow slurries by Chinese researchers, this requirement does
not seem to be unrealistic.

Ward and O'Brien (1981) also suggested that debris flows can occur in

Bagnold's macroviscous, transitional or inertial regime, and simulated

flows in the three regimes in flume experiments. To separate between the

different regimes, they proposed to use the particle Reynolds number as

defined by equs. (2.8) and (2.9), with limiting values as given by

Bagnold. They presume that in actual debris flows mainly three types of

resistive stresses are important. While in the macroviscous regime vis¬

cous stresses dominate and the flow appears to be laminar, turbulent and

dispersive stresses should dominate in the inertial regime. Defining the

turbulent shear stress according to Prandtl's mixing length concept,

they expressed the total shear stress t' in the inertial regime as:
 -
31 -

T' =

aio-(Xd)2sina(^)2 +
p(.K'y)>(^)' (2.13)

where K' is the von Karman constant as modified by the presence of

grains. Assuming that the particles and the fluid move with the same

velocity, they integrated equ. (2.13) to obtain the vertical velocity

distribution.

A similar approach was proposed by Takahashi (1987) to analyse his

muddy debris flows (or turbulent mud flows, s. section 1.3), but he

suggested to use the mixture density instead of the fluid density in the

turbulent shear stress term. A comparison between the obtained velocity

distribution with equ. (2.12) and experimental data showed that the

effects of turbulent mixing become more important for lower solids con¬

centrations C and higher relative depths h/d.

O'Brien and Julien (1984) stated that the energy dissipation in a

debris flow is mainly caused by viscous, turbulent, dispersive and yield

shear stresses. Their relative importance is a function of fluid and

grain properties. To account for all these effects a general equation

for the total shear stress is proposed as:

x' =

h +
V& ci02 + + "- <2-u>

where C. is a variable that depends on depth, concentration, sediment

size and boundary roughness. The first and second term of equ. (2.14)
contain yield and viscous stress and represent the Bingham model; the

effect of turbulent and dispersive stresses are combined in the third

term because they both depend on the shear gradient squared and because
it is difficult to separate them in debris flows. By using a rotational

viscometer O'Brien and Julien measured the rheological properties of the

fluid matrix of actual debris flow deposits for various sediment con¬

centrations, and determined values for x_, n_ and C.. However, in con¬

ventional viscometers only sediment-water mixtures with relatively small

grain sizes can be sheared, and thus dispersive stresses caused by coar¬

ser grains are not included in the analysis. As indicated by the dots in

equ. (2.14) and pointed out by O'Brien and Julien this relation is still

a simplification of the complex processes in a debris flow, and they

noted that more experimental research is especially required to deter¬
mine possible values of the paramter C..
 -
32 -

Chen (1986a, 1988a,b) noted that Bagnold's model and also Takahashi's
extension thereof have major constraints in modelling real debris flows.
On the one hand Bagnold's formulation is strictly valid only for the
fully dynamic state because in the transition to a quasy-static state

the grain stress relation should not only contain a rate-dependent but
also a rate-independent part, including cohesion. On the other hand the

assumption of constant concentration and pressure over the flow depth is

also too restrictive. Releasing the latter constraint, Chen (1988b)

proposed that a normal stress relation should be solved simultanously
with the conventionally used shear stress versus shear rate relation to

obtain the velocity and pressure distribution for uniform debris flow.

He termed his model the generalized viscoplastic fluid model, and the

constitutive equations are given as (Chen, 1988b):

Tyx = c' cosa' + p sina' +

V^g)" (2.15)

and
Tyy =
-p +
y2(g)n (2.16)

where T ,T denote the total shear and normal stresses, respectively,

x is the coordinate in flow direction, p the pressure, a' can be the

dynamic or the "static" angle of internal friction, and p. and u, are

the consistency and cross-consistency indices, respectively. By expres¬

sing the relative viscosity X] =t| /h of the suspension with a theore¬

tical relation given by Krieger and Dougherty (1959):

rir = (1 -

KCs)"B/K (2.17)

Chen (1988b) came to the following equation for u. and y„:

y. = a. o-n-1d2<n-1V-nC -B'/K(1-KC )"B'/K *

(2.18)
J J w s

where j=l,2, a. are numerical constants, ft is the dynamic viscosity of

water, B' the "intrinsic" viscosity (shown to be close to Einstein's

value of 2.5 in his theoretical equation for the relative viscosity of

dilute suspensions, VI =1+B'C ), and K is a factor accounting for the

interaction of the colliding particles (usually taken as
K=l/C^).

Chen (1988a) noted that the first two terms of equ. (2.15) represent
a soil yield stress which might become negligible if a flow is in the

fully dynamic range. The value of the flow behaviour index n can vary
 - 33 -

from 1 (macroviscous regime) to 2 (inertial regime) or possibly assume a

higher value. For the assumption of a constant grain concentration over

the depth, Chen (1988a) showed that Bagnold's (1954) model is a par¬

ticular case of the generalized viscoplastic fluid model (for the fully

dynamic state and with B'/K =

0). Chen (1988b) obtained analytical solu¬
tions for the velocity distribution in the macroviscous flow regime

(n=l), with which he could illustrate the influence of the parameter

B'/K. Chinese researchers have found (Chen, 1986b) that the value of
B'/K should increase with increasing fine sediment concentration,

thereby, according to Chen's model, decreasing the velocity gradient

near the bed. Numerical solutions (Chen, 1988b) for the inertial flow

regime (n=2) with B'/K # 0 show good agreement with experimental

velocity profiles of Takahashi (1980) and Tsubaki et al. (1982). While

the theoretical determination of the parameter B'/K is relatively easy,

assumptions are usually required for the value of the flow behaviour

index n and for a relation between density and pressure.

Recent studies in the field of granular flow experiments and model¬

ling are summarised by Iverson and Denlinger (1987). Extensive work has

been done to study the behaviour of dry granular flows. It is found that

generally grain collisions are inelastic which means that energy is

dissipated during this process. Both translational and rotational energy

can be exchanged during the grain contacts. In experiments performed by

Drake and Shreve (1986, as described by Iverson and Denlinger, 1987) it
could be observed that the particle fluctuation energy mainly originated
from direct interactions of grains with the bed and that both the par¬

ticle concentration and the frequency of collisions decreased with

height above the bed. At the higher flow depths particles followed more

irregular paths while near the bed they tended to flow parallel to the

bed and occasionally formed densely packed clusters. Iverson and Den¬

linger presume that the random interactions of a relatively small number

of grains (as compared to the flow depth) possibly results in a random

macroscopic flow behaviour. Dissipation of fluctuation energy from the

boundary into the flow is expected to occur over a shorter distance in

the case of inelastic collisions such as for clay aggregates as compared

to flows with hard quartz grains. This statement is somewhat contra¬

dictory to Chen's (1988b) model which predicts that the near bed shear

zone should increase with increasing fine sediment concentration.

 -
34 -

As to the role of an intergranular fluid matrix, Iverson and Den¬

linger (1987) report that during a collision the fluid will be pres¬

surized and squeezed out of the gap between the grains, thus forming a

cushion that dampens the collision of the solids. If grain collisions

are mainly inelastic this fluid "cushioning" may lead to a reduced

energy dissipation and thus increase the mobility of the flow. There are

mixture theories that allow for large deformations and varying concen¬

trations and contacts of particles and that include viscous drag effects

arising from the relative movement of an interstitial fluid. However,

according to Iverson and Denlinger, effects of inertial solid-fluid

interactions, of pore-scale pressure fluctuations and of nonuniform

particle size distributions have not been treated so far in a compre¬

hensive theory.

2.2 Rheology and flow of hyperconcentrated suspensions

2.2.1 Rheological properties

In debris flow modelling it is often assumed that the finer particles

(e.g. smaller than 1 mm) and the water form a fluid matrix which carries

along the coarser particles or interact with them (Iverson and Den¬

linger, 1987; Davies, 1988). Many experimental studies have been made to

determine the changed fluid properties of hyperconcentrated suspensions

and the effect on the flow behaviour.

Naik (1983) summarised the factors that affect the rheological pro¬

perties of a debris flow :

-
concentration of solids

clay content

-
type of clay

-
absolute size of the solid material

-
size distribution of clay, silt, sand, and gravel fractions
-
characteristics of clastic materials, such as shape, size
and density

packing arrangement
-

proportions of fine-grained material to coarse clastic

material such as rocks and boulders

 -
35 -

-
electro-chemical characteristics of the liquid phase
This list illustrates that it is difficult to develop theoretical
equations to predict the rheological properties of natural grain
materials. Therefore the main approaches to treat hyperconcentrated
flows are based on an experimental determination of the rheological
behaviour. The term hyperconcentrated flow or suspension is used here in

the sense of water-sediment mixtures containing predominantly fine

grains that are more or less uniformly dispersed within the flow.

A comprehensive review of theoretical relations to predict the change

in viscosity as a function of increasing particle concentration and
other factors is given by Naik (1983). To determine the viscosity of a

suspension, n ,
at low concentrations of spherical grains A. Einstein

(1906, cited by Naik, 1983) proposed the equation:

ns = *i(l +
2-5Cs) (2.19)

where h, is the viscosity of the suspending medium. According to Naik

this formula should only be used up to volume concentrations C =0.01.

Above this concentration level hydrodynamic interactions between par¬

ticles become important, and empirical and theoretical equations have

been proposed to express the relative viscosity ti =
\] /r\ as a function

of the ratio C /C^, termed the reduced volume fraction. A comparison of

a theoretical equation of Ackerman and Shen (1979) with experimental

data shows good agreement over a wide range of reduced volume fractions

from 0 to about 0.95 where the relative viscosity of the dispersion

attained values of approximately 200. Ackerman and Shen's analysis
showed that the grain size distribution should also affect the relative

viscosity, and according to Moshev (1979) this is particularly important

for bimodal dispersions, which are indeed often found in deposits of

debris flows. Theoretical models that account for the effects of wide

grain size distributions would be very complex. It has been shown,

however, both by Moshev and by Ackermann and Shen that dispersions con¬

taining several discrete particle sizes which differ at least by a

factor of ten can be analyzed by successively determining the relative

viscosity of the fluid system including first only the smallest and then
the next larger particle size. Naik (1983) proposed to apply this pro¬

cedure to actual debris flow materials by dividing the wide grain size

distribution into several discrete size classes.

 -
36 -

In a debris flow the fluid matrix consisting of a suspension of fine

particles is an important factor in causing a yield strength and non-

Newtonian flow behaviour. To examine this aspect Naik (1983) measured

the rheological properties of a clay suspension and velocity profiles in

an open channel. Clay suspensions are generally considered to be purely

viscous time-independent fluids, and the Bingham model has been shown

appropriate to treat the flow behaviour at higher shear rates. Naik

proposed to use a theoretical relation developed by D.G. Thomas (1961)

to predict the Bingham parameters:

tB =
K^ (2.20)

and
ITg =

hvexp(K2Cf) (2.21)

where is the volume concentration of fine particles, and K, is

Cf a

dimensional and K. a dimensionless constant for a given particle type,

depending on grain size and shape. According to D.G. Thomas (1963a) the
above equations are valid up to fine material volume concentrations C,

between 0.2 and 0.3; for higher concentrations the flow behaviour will

change from shear-thinning (pseudoplastic) to shear-thickening (dilatant

flow). Based on the assumption that the approach of Ackerman and Shen

(1979) can also be used for grain dispersions in a Bingham fluid, Naik

(1983) developed rather complex theoretical relations to predict the

Bingham yield stress and the Bingham viscosity for such a dispersion.
His equations show good agreement with measured Bingham parameters of
Mills (1983) for uniform dispersions of spherical glass beads sheared in

a clay slurry. When applied to dispersions of crushed quartz sand and

coarse particles from Mount St. Helens mudflow deposits, the predicted
rheological parameters deviated from the measured ones, probably due to

the effect of irregular particle shape.

Wan (1982) measured the Bingham parameters for a kaoline and a ben-

tonite clay suspension and determined the following empirical relation

for the Bingham yield stress:

TB =
K3C3 (2.22)

where K, is a dimensional constant. It is interesting that this equation

shows the same dependence on concentration as the theoretical formula
 -
37 -

(equ. 2.20) given by D.G. Thomas (1961). For the kaoline suspension Wan

presented also a relation for the Bingham viscosity:

rig =

K4 +
K5C^68 (2.23)

where K,, K,- are dimensional constants. Wan made further maesurements

for a bentonite slurry containing sand particles up to 0.21 mm and found

that both x„ and n_ increase more strongly with concentration above a

total volume concentration (including clay and sand) of approximately

30%.

O'Brien and Julien (1986) noted that in previous studies in which the

Bingham model was used to analyse the flow of hyperconcentrated sus¬

pensions (e.g. Thomas, 1963a,b; Wan, 1982; Mills, 1983) the rheological
parameters were determined from measurements in the high shear rate

region (dv/dy > 100 l/s). They claim that in natural debris flows and

hyperconcentrated sediment flows representative shear rates are of the

order of 5 to 50 l/s. Therefore the effective Bingham yield stress would

be smaller than values obtained from backextrapolating high shear rate

measurements, whereas larger values result for the Bingham viscosity. To

determine the Bingham parameters for ten different natural soils they
used a special viscometer in which they could shear a fluid containing

sand particles up to 0.5 mm. Their empirical relations for T and tlD
determined in the low shear rate region are given as:

TB
=

K6exp(K?Cf) (2.24)

rtg
=

K8exp(K9Cf) (2.25)

It can be seen that the relation for the Bingham viscosity (equ. 2.25)
is similar to the theoretical equation (2.21) given by D.G. Thomas

(1961). O'Brien and Julien measured the Bingham parameters also for

sand-clay mixtures. For a clay volume concentration of 6 % and sand

particles up to 0.25 mm the Bingham viscosity increased rapidly with

concentration above a total volume concentration of about 20 %, while

for a natural soil containing sand particles up to 0.5 mm a rapid in¬

crease occurred at total volume concentrations above about 50 %.

Chen (1986b) summarised Chinese concepts of modeling hyperconcen¬

trated streamflow and noted that the Bingham approach is very widely
 -
38 -

used in China. Many empirical expressions have been developed to

estimate the Bingham yield stress. Fei (1981) proposed the following
equations:

tb
-

Kiocsfw cl'73 for

cf < °-106
cs1;462 (2'26a>

TB
"
KllCsfw3cJ'33 for
Cf > °-106
St;4" (2-26b>

where C ,
denotes the weight concentration of fine particles smaller
than 0.025 mm. According to the above relations the influence of the

total particle concentration by volume C, on the yield stress is the

bigger the higher is the amount of fine particles. For example, if C c

=

stw

0.1 then the critical total concentration Ct is 31 %, and if C ,

= 0.2
i stw

then C, is 22 % above which there is a rapid increase of the Bingham

yield stress t_ with increasing total concentration. Fei (1981, 1983)
also found that T_ should depend on a critical sediment concentration

C, at which all free water is extracted (bound by cohesive particles),

C, being mainly a function of particles finer than 0.01 mm. Tang (1981)
further introduced a cohesion and a geometry index to calculate T_ and

his relation showed good agreement with field data from the Yellow

River. Kang and Zhang (1980) analysed numerous samples of debris flow
material from a particular torrent and they found a breakpoint in the t„
D

relation at critical concentration of C=40 % and in the H-

vs.
Cf a vs.

C, relation at a value of C,=45 %. Chen (1986b) concluded that at lower

concentrations x„ varies mainly with C, and the amount of colloidal

particles while at higher concentrations the internal friction between

colliding particles becomes important.

As to the Bingham viscosity Chu (1983) proposed an equation similar

to the one given by Einstein (equ. 2.19):

rig =
yi -

2.5K12Cf) (2.27)

where the is equal the concentration of sediment plus

term
K.„Cf to

bound water. Chu further suggested another relation which can be written

^B "
V1 -

cf>~2'5 <2'28)
 -
39 -

Fei (1983) modified equ. (2.28) and developed two separate forms the use

of which depends on the presence of particles finer than 0.01 mm. Chen

(1986b) noted that equ. (2.28) is similar in form to the theoretical

equation for the relative viscosity proposed by Krieger and Dougherty

(1959), equ. (2.17), and that a similar empirical relation was found by
McTigue (1982). Again it appears that the Bingham viscosity is not only
a function of the total concentration but also of the sediment compo¬

sition and the grain size distribution.

Shen and Xie (1985) measured the Bingham properties of fine slurries

containing sand particles up to 0.5 mm in diameter. They suggested to

apply equations similar to (2.24) and (2.28), and it is interesting that

they used a two step procedure analogous to that proposed by Naik

(1983). First they calculated Bingham yield stress and viscosity for the
fine slurry having only particles smaller than 0.01 mm. Knowing the
rheological properties of this "new" fluid matrix they then computed the
Bingham parameters for the whole slurry including the coarser particles.

Summary

From the above discussion it can be concluded that no unique relation

has been found to express either the Bingham yield stress or the Bingham
viscosity as a function of concentration and material parameters.

Empirical equations have been proposed that are similar to formulas

developed on a theoretical basis. Viscometric measurements of grain

dispersions with different size classes show that the Bingham parameters
tend to increase more rapidly with concentration above a certain

critical value of the total volume concentration, which appears to lie

somewhere in the range of 20% to 50%. This distinct change may be

related to the increasing influence of internal friction at higher grain

concentrations (Chen, 1986b).

2.2.2 Flov resistance of a Bingham fluid

2.2.2.1 Definitions

Naik (1983) reviewed theoretical approaches to treat the flow of a

Bingham fluid and stated that the flow resistance should be a function

of the following dimensionless parameters:

 -
40 -

f = Fn
(ks/h, Fr, ReB, Yf) (2.29)

where k denotes the equivalent sand roughness and k /h is termed the

relative roughness, and where the following definitions are used:

8x
o
f =
-jtj- : Darcy-Weissbach friction factor (2.30)
pV2

Fr =
, _.,.,
: Froude number (2.31)
(gh)"2

4Vho
Re_ = : Bingham Reynolds number (2.32)
%

2xB
Yf =
-y£ : Yield factor (2.33)

where x =
pghj is the shear stress at the bed, J is the slope (=tang)
and V is the mean cross-sectional velocity. The yield factor can also be

replaced by the Hedstroem Number He which is given as:

P"tB(4h)2
He =
r-j : Hedstroem number (2.34)
%

and the relation between Y,, He and Re_ is:

Yf =
if (2.35)

High values of
Yf or He indicate increasing non-Newtonian characteristic
of the flow.

2.2.2.2 Laminar flow

By integrating the Bingham equation (2.2) for laminar flow in a wide

open channel the velocity distribution over the depth is obtained as:

v -
«J<h'-y"> .
^1 (2.36)
ZnB %

where y' is measured downwards from the flow surface. From this equation
the mean velocity can be calculated as:
 - 41 -

hx x. x

V = 7-2 [1 -
1.5-5 +
'
o.5(-5)3] (2.37)
'"
—x. —T.
Shg 0 o

These results are confirmed by the equations given by Howard (1963) and

Kozicki and Tiu (1967). If the last term in equ. (2.37) is neglected and
the stress ratio a' = x_/x is smaller than 0.5 this leads to an error
D 0

in the mean velocity that is smaller than 6.3 %. The simplified relation
can be expressed as:

\
'

^B +
\ <2-38>

Combining equ. (2.37) with equs. (2.30), (2.32) and (2.34) a friction

factor relationship in terms of f, Re„ and He is found as:

He
— -
-
— -
+ —-f—)3 (2^•»>39)
96
ReB 8Re| 3f2^Re|;

The use of equ. (2.39) allows the friction factor to be plotted against
the Bingham Reynolds number, with the Hedstroem number as an additional

parameter.

Kozicki and Tiu (1967) considered the steady, uniform, laminar flow

of non-Newtonian fluids in straight open channels of arbitrary cross

sections. For different non-Newtonian flow models they developed a

general method to predict the flow rate and the maximum velocity. Their

equations for the Bingham model are somewhat more elaborate than the

ones given above, in that they include two cross-sectional shape para¬

meters a and b; for a rectangular channel a and b are a function of the

aspect ratio A =W/h, where W is the width of the channel. From the

equation for the mean velocity given by Kozicki and Tiu, Naik (1983)

developed a relation between f, Re. and He similar to equ. (2.39) but

including the shape parameters a and b.

Johnson (1970) presented a numerical method to compute the velocity

and shear stress distribution for the laminar flow of a Bingham fluid in
a rectangular channel. He also showed that the existence of dead regions

at the corners becomes important for narrow channels (low aspect ratio).

D.G. Thomas (1963a) used the Bingham model to analyse pipe flow ex¬

periments with non-Newtonian suspensions. He stated that besides the use

of an
f-ReB-He relation another, equally satisfactory method to predict
 -
42 -

the laminar friction loss is the introduction of an effective visocity

u which allows the conventional Newtonian relation f = 16/Re (for
e

laminar pipe flow) to be used. This method was first proposed by

Caldwell and Babbitt (1941), and the definition of u is similar to the

apparent viscosity f| except that the former is used for mean flow

values. Considering the simplified equ. (2.38) for open channel flow,
the bed shear stress x can be plotted as a function of the shear rate

term 3V/h, and then an effective viscosity u

,
is obtained as:

»e2 \ +
W- (2-40>

or, if u' is calculated from equ. (2.37) instead of equ. (2.38) :

V2 =
\ I [1 -

1.5(^) o
+
0.5(^)3] (2.41)
o

Now a conventional friction factor diagram can be used if the corres¬

ponding Reynolds number Re, is defined as:

Re, = ^ (2.42a)
we2

or Re, = ^ (2.42b)
Me2

Quian et al. (1980) performed flume experiments with a muddy slurry

and showed that their data in the laminar flow region can be represented
equally well by f-Re_-He and plot. the latter diagram the
an an
f-Re, In

points lie close the theoretical relation f

data to =
96/Re,. Zhang et

al. (1980) carried out similar open channel tests in flumes with dif¬

ferent cross sections. They presented their results in an f vs. Re, plot
and found the relation f = A /Ren for the laminar flow region, where A_
m 2 m

is 84 for the rectangular and the U-shaped and 74 for the trapezoidal
cross section. Wan (1982) measured the flow resistance of a bentonite

suspension in a closed rectangular channel (30 cm wide and 20 cm high),

in f the data points lie rather close the line
and an vs.
Re, diagram to

f =
96/Re2.
 -
43 -

2.2.2.3 Laminar-turbulent transition

Following the approach of Hanks (1963), who developed a criterion for

the laminar-turbulent transition of a Bingham fluid flowing in a pipe,

Naik (1983) established a similar criterion for open channel flow. His

proposed relations are given as:

ac He
=
(2.43)
(1-a')3 48'000
c

ReB,c "
I&- c
(1 "

U5a'c +
°'5ac3> (2.44)

where the subscript c denotes the critical conditions at transition, and

according to Naik the "constant" 48'000 may vary between 24'000 and

96 000, analogous to the critical Reynolds number in the Newtonian case

being in the range of 2'000 to 8'000. Naik found quite good agreement
between the equs. (2.43) and (2.44) and his experimental data obtained
from flume tests with a kaoline suspension. His data covered Hedstroem
numbers He up to 2-106 and stress ratios a' up to 0.75. Hanks (1963) in
his analysis, however, had found good agreement between theory and ex¬

periments only for He<10s or a'<0.55.

As an alternative the laminar-turbulent transition can be detected on

a conventional f vs. Re plot when a suitable Reynolds number is defined,

for example by using (2.42a).
as
Re, given by equ. In experimental
studies with a clay suspension or a fine slurry, values for the critical

Reynolds number Re, at transition were found to be between 3'000 ...

5'000 (Quian et al.', 1980), 4'000 ... 5'000 (Zhang et al., 1980), 4'000
... 6'000 (Wan, 1982, covered flume) and 6'000 ... 8'000 (Cao et al.,
1983, including field data). For his pipe flow data, Thomas (1963a) gave

critical Reynolds number of about 2'100, defined analogous

a to
Re,
using an effective viscosity. According to Straub et al. (1958), who

collected flume data on the laminar and turbulent flow of Newtonian

fluids, the lower critical Reynolds number seems to depend to some

extent on the channel shape and is generally larger than for pipe flow.
They also concluded that the cross-sectional shape is quite important in
rough laminar open channel flow.
 -
44 -

2.2.2.4 Smooth and rough turbulent flow

The velocity distribution in turbulent open channel flow of a New¬

tonian fluid is usually described by a loagrithmic relation (Yalin,

1977):

V 1
lnlf- B
=

k
+ (2.45)
X s
s

with B
s
= 2 5 ln(Re*) 5.5 if
Rek <= 5 (2.46)

B 8 5 if >= 70 (2.47)
s
=

Refc

where the roughness Reynolds number Re, is defined as Re. =

v^k p/u. The

value Re, = 5 is the upper limit of the hydraulically smooth turbulent

flow (or regime), and Re, = 70 is the lower limit of the fully developed
turbulent or rough turbulent flow (or regime); in between these values

the flow is in the transitional regime. Fig. 2.2 (taken from Yalin,
1977, his Fig. 2.5) illustrates how the value of B varies with Re, for

the experimental data obtained by Nikuradse.

Fig. 2.2 : Variation of B with roughness Reynolds number Re, (after

Yalin, 1977).

D.G. Thomas (1963b) performed pipe flow experiments with titania,

kaoline and thorium oxide suspensions. He correlated his data for smooth
 - 45 -

turbulent flow with a modified Blasius equation using the Bingham

Reynolds number Re„. If the Bingham yield stress x_ was less than 23.9

N/m2 the values of the friction factor f tended to approach those for
Newtonian flows with increasing Re_, but for x_ > 23.9 N/m2 f tended to

diverge from the Newtonian value with increasing Re_.

Hanks and Dadia (1971) developed a sophisticated theoretical analysis

for turbulent flow of a Bingham fluid in smooth pipes. They introduced a

paramter B ,
which is to be determined from experiments, representing
the effect of the boundary in damping the turbulence. Naik (1983) adap¬
ted this theory for smooth turbulent open channel flow of a Bingham
fluid. He found from his experiments with a kaoline slurry that for
values of the yield factor Y, smaller than 0.001 the flow is essentially
Newtonian (and B assumes its Newtonian value), and for Y,>0.001 the

flow behaviour shows increasing non-Newtonian character (and B in¬

creases linearly with Y,). His theoretical predictions were confirmed by

experimental data. In Fig. 2.3 a friction factor diagram computed by
Naik is shown. According to the theory the friction factor can be higher
or lower than in Newtonian flow, depending on the values of the Bingham
Reynolds number and Hedstroem number. For increasing values of Re. the

friction factor approaches the Newtonian value. This confirms one of

D.G. Thomas' (1963b) observations, and it seems that in Naiks tests x„

was always smaller than Thomas' critical value of 23.9 N/m2.

Gupta and Mishra (1974) showed that the conventional f vs. Re rela¬

tionship for Newtonian flow is also valid for pipe flow of a Bingham
fluid. Both Wan (1982) for his covered flume tests with a kaoline sus¬

pension and Zhang et al. (1980) for their open channel experiments with
fine slurries presented their data on a f vs.
Re, plot (using the effec¬

tive viscosity u ). There is a smooth transition on their diagrams

,

from the clear water data to the hyperconcentrated data points indica¬
ting that such a representation is possible not only in the laminar but
also in the smooth turbulent region. From pipe and open channel flow

experiments with a clay suspension, Yano and Daido (1965) concluded that
the conventional Newtonian relations can be used if a suitable Reynolds

number is defined, which is however given in their analysis by a more

complex expression than the one for Re,.

 - 46 -

01

10 Hedstrom number

Locus of transition

*y*f
0 01

-
Newtonian flow

/iVhc |
Bingham Reynolds number I—-
J

Fig. 2.3 : Friction factor diagram for a Bingham fluid as given by

Naik (1983); it was produced from the Bingham relations

for laminar flow and for smooth turbulent flow (with

Naik's approach), and also shows the limiting line se¬

parating the laminar and turbulent flow regime. (Note

that f is defined here as 2x /pV2 which is different from

the definition in equ. (2.30) used in this study.)

Torrance (1963) conducted experiments with pseudoplastic fluids in

smooth and rough pipes. His flow resistance equations can be given for
Bingham fluids (fluid index n=l) as:

j£
=
A1log(l-a') +
A2log(ReBi|f) -
A, for smooth pipes (2.48)

1 R
for rough pipes (2.49)
jl
=
A4log(j-) +
A5

where A., A,, A, and A. and R is the hydraulic radius.

A2> are constants

Yang and Zhao (1983) performed experiments with hyperconentrated

slurries in a rough flume and developed the following equations:
 -
47 -

^ =
21og(|-) +
Ag -

A7log(Re* B) for the transi- (2.50)

s
tional regime
1 R
4t =
21og(j—) + A„ for the rough turbulent regime (2.51)

where the constants A,, A-,, A_ depend on the roughness characteristics,

o/o
*
and Re. _
is the roughness Reynolds number defined with the Bingham
k,B ^
viscosity: Re. _
=
v^k p/f1„- It may be noted that these experiments
included flume slopes of 10 % and 20 % whereas all other open channel

studies mentioned here were limited to bed slopes of a few percent.

is defined, for example (equ.

If a suitable Reynolds number as
Re2
2.42), one could also imagine that the universal Colebrook equation,
valid both in turbulent smooth and rough flow, may be used. According to

Henderson (1966) it is expressed for open channel flow as:

ti -21°«<iis +
ii!i> <2-52>

where the Reynolds number is defined as Re = 4VRp/h . For flow in the

fully rough regime (large Re), this formula becomes identical to the

Nikuradse equation:

Jj . 21og(i^) (2.53)

As there is only little information available on turbulent flow of a

Bingham fluid in rough pipes or open channels, Naik (1983) developed a

theory for such flows in open channels. He assumed that the Bingham
Reynolds number is large so that the laminar shear stress due to the

Bingham viscosity »i_ can be neglected in comparison to the turbulent

D

shear stress. His equation can be given as:

^ =
0.88(l-a')lA +
ln(|-)] s
(2.54)

,~ (A +2)h2
with
Aq =
ln[(^) exp(-l- —^ )] (2.55)
cs

where A is a shape factor (depending on the depth to width ratio) and

A is the area of the flow cross section. Naik compared measured mean
cs

velocities for the flow of a kaoline suspension over a rough bed (made
 -
48 -

of a wire screen with a thickness of 3 mm) with theoretically predicted

values by equs. (2.54) and (2.55); he found good agreement between
measured and calculated velocities.

To distinguish between hyraulically smooth, transitional and rough

turbulent flow of a Bingham fluid, Naik (1983) suggested to use a

modified roughness Reynolds number which is also a function of the

yield/ wall stress ratio a':

Re*' =
vjtksp(l-a')/nB (2.56)

If, however, the effective viscosity u can be successfully used to

,

define a flow Reynolds number

Re, one may expect that also the particle

Reynolds number can be suitably expressed with u

,:

Re* =
v^d p/ye2 (2.57)

2.2.2.5 Summary

It is interesting that both equations proposed for hyperconcentrated

flow in the rough turbulent regime, equs. (2.49) and (2.51), are inde¬

pendent of any viscosity or yield stress parameter and are in fact

similar in form to the Newtonian relation (2.53).

To determine the flow resistance of a hydrodynamically smooth tur¬

bulent or a laminar flow two alternative methods may be valid: Either

using new relationships developed for Bingham fluids where f is a

function of Re_ and He, or defining an effective (representative) vis-

cosity Re, and applying conventional formulae for Newtonian fluids.

2.2.3 Flow and fluid characteristics

D.G. Thomas (1963b) pointed out that the use of Newtonian friction

factor relations (by defining a suitable Reynolds number) for non-New¬

tonian flow may be a too simplified approach. For it implies that the

scale and intensity of turbulence in a non-Newtonian system is essen¬

tially the same as in Newtonian fluids provided there is a Reynolds

 -
49 -

number similarity. There seems to be a general agreement that with in¬

creasing non-Newtonian behaviour the turbulence characteristics of the

flow must be damped in some way.

The velocity profile in turbulent flow is generally divided into

three zones: The laminar sublayer, the transition zone and the turbulent

core. Dodge and Metzner (1959) presented a generalised power law model

for the velocity distribution of pseudoplastic fluids and concluded that

the turbulent resistance law is very similar to the Newtonian friction

factor relation. According to Gust (1976), the existence of a Newtonian

flow structure is suggested by mean flow measurements in the logarithmic

part of the turbulent boundary layer of cohesive sediment suspensions

which behave like a Bingham fluid.

Richardson and Julien (1986) stated that large scale turbulence may

still be present in hyperconcentrated flow while small scale turbulence

may be rapidly damped. This is in agreement with van Rijn's (1983) con¬

clusion that turbulent fluid diffusivity is reduced in such flows. Both

van Riin and Richardson and Julien report that the effect of high sedi¬
ment concentrations on the von Karman constant K is not yet clear. The

change of a wake strength coefficient with increasing near bed con¬

centration of fines was used by Parker and Coleman (1986) to predict

changes in flow depth, velocity and resistance. Julien and Lau (1988)

theoretically concluded that the wake strength coefficient may still

remain the same as for clear water flow if the concentration of fine

sediment is uniform over the depth. They claim that Parker and Coleman
mistakenly assumed a decrease of the drag coefficient in dilute sus¬

pensions and questionned their predicted changes in flow parameters.

Bradley (1986) noted that in natural streams turbulent stresses

control the flow behaviour and the sediment transport for fine material

concentrations up to 20 % by volume. However, this limiting concentra¬

tion can be much smaller for highly viscous clay suspensions (as for
example with bentonite clay). Quian et al. (1980) gave a limiting con¬

centration of 15 to 20 % for hyperconcentrated flows. Fan and Dou (1980)

concluded from measurements that in hydrodynamically rough flow the

fluctuation energy remains strong within the boundary layer and that the

friction factor is close to the value for clear water flow.

 -
50 -

Bed forms may change with increasing fine sediment concentration

(Simons et al., 1963) and thus influence the flow resistance. The effect

of hyperconcentrated flows on bed forms seems however uncertain

(Bradley, 1986). In the upper flow regime grain roughness is mainly

important, and it seems unlikely that "small-scale" bed forms would

develop with high suspended sediment concentrations.

With increasing values of the Bingham parameters x. and fi_ the (ef¬

fective) viscosity of the fluid matrix also increases. This leads to an

increase in the thickness of the laminar sublayer (Gust, 1976; Woo,

1985; Ashida et al., 1987) as well as to a decrease in the fall velocity
(Thomas, 1963a; Wan, 1982; Woo 1985).

For a suspension with a fine sediment concentration of 20 % by vo¬

lume, Bradley (1986) calculated a reduced fall velocity according to

D.G. Thomas' (1963a) method to be about 30 % of its clear water value.

Wan (1982) used a more sophisticated approach to determine the settling

velocity in a Bingham fluid. He measured the fall velocity W of dif¬

ferent particles in a clay suspension of various concentrations and

determined the universal Reynolds number Re. as well as the drag coef¬
ficient C_ given by Ansley and Smith (1967):

pW2
R^=1 =
" : (2-58)
v, v,A
+ Tx„n7/24n7/0A
TigW/d

4 (o--p)gd .

4> "

3 pW2
u-3y>

Wan plotted his data in a C_ vs. Re. diagram, which shows his points to

lie close to a line for natural sand settling in a Newtonian fluid. Wan

balanced the integral of the drag force due to the yield stress over a

sphere surface, given as x_it2d27/8 by Ansley and Smith, with the

submerged weight of a spherical particle, itd3(a-p)g/6, to obtain the

critical Bingham yield stress x_ at non-settling conditions:

D, C

xn =
0.067(<j-p)gd (2.60)
B,C

Wan experimentally verified this non-settling criterion for gravel par¬

ticles, glass and lead balls up to 1.66 cm in bentonite suspensions of

different concentrations.
 -
51 -

2.3 Sediment transport

2.3.1 Sediment transport in hyperconcentrated flow

2.3.1.1 Effect of combined increase in fluid density and viscosity

The effects of high concentrations of fine material in suspension on

the fluid and flow properties are discussed in section 2.2. If the fine

material is uniformly distributed over the flow depth then the suspen¬

sion can be considered as a pseudo-homogeneous fluid having both an

increased density and viscosity as compared to clear water. While the

increase in fluid density obviously reduces the submerged weight of

transported (coarser) particles, the effect of a changed viscosity on

the mechanics of sediment transport is more complex.

Beverage and Culbertson (1964) observed from field data that the sand
concentration increased in hyperconcentrated flows with increasing
amounts of fine material in suspension. They concluded that water and

fine material act as lubricants and that the decreased fall velocity

helps to keep coarser material in suspension by less turbulence as

compared to clear water flows. Colby (1964) developed an empirical

method to calculate the increase in sediment transport rates as a func¬

tion of suspended sediment concentration and flow depth. His correction

factor, giving the increase in transport as compared to a corresponding
clear water flow, can be as high as 100, with a mean value of about 10,

and is based on field data (Rio Puerco, New Mexico) as well as flume

data (Simons et al., 1963) covering fine material concentrations C, up

to 8.7 X. To compute the correction curves he included the effect of

reduced fall velocities. Bradley (1986a) reported field data from the

Mount St. Helens mudflow. In this case no bed load discharges were

measured. For the hyperconcentrated (turbulent) flows suspended sediment

transport rates increased by about a factor of 10 as compared to clear

water flow, and for the laminar mud flows the transport rates were even

another order of magnitude higher.

Only a limited number of flume studies have been performed to con¬

sider the effect of a suspension of cohesive sediments on the sediment

transport capacity of the flow. Simons et al. (1963) conducted flume

 -
52 -

experiments with a bentonite clay suspension. The maximum clay concen¬

tration C, was 2.3% at which the apparent viscosity (which they did
however not define) was increased by a factor of about 4 and the fall

velocity reduced by a factor of about 2 as compared to clear water. The

bed material used in the sediment transport tests had a fall diameter of

0.47 mm and 0.54 mm. Increasing clay concentrations resulted in the

following effects: For the lower flow regime, with ripples and dunes as

bed forms, they noted a stabilization of the bed and a decrease in the

flow resistance. The transported bed material (bed load and suspended
load) decreased in flows over a dune bed, which they attributed to a

reduced fluid shear due to a smaller f-value. In the upper flow regime
(plane bed, standing waves, antidunes) they observed an increase in flow

resistance as well as in the sediment transport rate. They plotted their

data in terms of dimensionless transport and shear parameters similar to

those given by Bagnold (1956):

qt
**t =

b [(s-l)gcosgd3]"2
(2-61)

1/2 2/3
and eA ]
=

i(.e-ec)e (2.62)

with 6
Pg(sT-l)dm <2-63>

where q is the volumetric transport rate per unit width of bed load and

suspended load, d refers to the mean grain size of the transported

material, b is a constant for a given bed material and can be related to

d ,
6 denotes the dimensionless bed shear stress, and 6 is the critical
m c

value of 6 at initiation of motion. In the of Simons'

*A vs.
6^ plot et

al. there is much less scatter of the data points if the fine sediment

(clay) is not included in the calculation of

i^ .

Kikkawa and Fukuoka (1969) performed flume experiments using very

fine sand as washload and sand of 0.18 mm mean diameter as bed material.

With increasing washload concentrations they observed increasing bed

material transport rates both for lower and upper regime flows. With
regard to bed forms they noted a change from dunes to a flat bed and

from a flat bed to antidunes when the washload concentration in the flow

was increased.
 -
53 -

Wan (1982) recirculated a bentonite suspension with increasing con¬

centrations C, up to 0.97% in a covered, rectangular flume, and used

cylindric plastic particles of 3.45 mm volumetric diameter as bed ma¬

terial. The change in the grain-fluid density ratio s was negligibly

small; at the maximum clay concentration the Bingham viscosity riB was

about 4 cps. and the Bingham yield stress x_ about 0.56 N/m2. For an

assumed shear rate of approximately 20 l/s, this would result in an

effective viscosity u of about 30 cps. (or 30 times that of water).

,

Wan plotted his sediment transport data in terms of the dimensionless

parameters * and 6, with:

1/2
(2.64)
t =

qt/[(s-l)gd3r'z

but noted that most of the transported bed material moved as bed load.

As compared to the corresponding clear water flows the values for the

bentonite suspension runs were lower if 6 <= 0.4 and higher if 0 >= 0.4.

Wan calculated bed load transport rates for his experimental conditions
with the formula of Engelund and Fredsoe (1976), including some adjuste-
ments: He took into account the effect of a Bingham fluid (i.e. his

bentonite suspension) on the drag coefficient using equs. (2.58) and

(2.59); and he used a new relation to determine the dimensionless shear

stress 6 for incipient motion of a particle in a Bingham fluid (s.

section 2.2.3). After having corrected some of his measured transport

rates for particles that moved in suspension he found fair agreement

between calculated and measured values. The results indicate that, com¬

pared to clear water flow, the bed load transport rates were smaller in

the bentonite suspension, especially in the low flow intensity region

(due to a larger 0 -value); *_ was reduced by about 20% at the highest
clay concentration (C, - 97%), at a shear stress 6 = 1.0. The suspended
load, on the other hand, is supposed to be larger due to the reduced

fall velocity. As a result, the total sediment transport rate in a clay

suspension flow would be smaller in the low flow intensity region and
larger in the high flow intensity region, as compared to clear water

flow.

With respect to bed forms Wan observed that dunes are lower and

smoother in the bentonite flows and that the transition to a flat bed

occurred at lower flow intensities than in clear water flow. Wan only
presented flow resistance data for runs with a fixed bed and no sediment

the data points plotted in f and

transport; were a vs.
Re, diagram, they
 -
54 -

showed a similar behaviour as with a Newtonian fluid (s. also section

2.2.2). It may be noted that by using plastic particles with a density

of 1.29 g/cm3, the density factor (s-1) was only about 14% of the value
for quartz grains in water, and that the transport rates calculated with

the dimensionless sediment transport formula of Engelund and Fredsoe

agreed fairly well with the measured ones.

Wan and Song (1987) conducted flume experiments with a clay suspen¬

sion and using PVC particles with densities of 1.27 and 1.34 g/cm3 as

bed material. They measured flow and sediment parameters and found that

the total transport rate (bed load and suspended load) of the plastic

particles increased with increasing clay concentration. They plotted the

concentration of the transported PVC particles as a function of the flow

intensity parameter V3/(gR,W), where R, is the hydraulic radius cor¬

rected for side wall influence. If the fall velocity was calculated

according to Ansley and Smith (equ. 2.58 and 2.59), the flume data of

Wan and Song for the turbulent flow runs followed more or less a

straight line on a log-log plot. For the laminar flow runs they observed

higher bed material transport concentrations than in turbulent flow but

the dependence on the flow intensity parameter (shown in a plot using
the fall velocity in clear water) was much weaker. The difference in

sediment transport capacity between the laminar and turbulent runs was

very pronounced (several orders of magnitude) at low flow intensities

(V3/(gR,W) =
1), whereas at higher flow intensities (V3/(gR.W) =
10)
there was almost no difference.

Bradley (1986) conducted flume experiments with a bentonite clay

suspension of different concentrations up to C, =

3.1%, at which the

fluid viscosity was about 870 times larger than that of water. He

measured (total) sediment transport rates using sand of 0.18 mm mean

diameter as bed material. At a bentonite concentration C, = 2.2% the

fall velocity of this sand was reduced by about a factor of 2. As bed

forms he observed both dunes and plane beds. In general sediment trans¬

port rates over plane beds were an order of magnitude larger than those
over dunes, for the same fluid discharge. For the plane bed case the

turbulent bentonite suspension flows transported about 3 to 5 times as

much bed material as did the corresponding clear water flows. Two runs

at the highest clay concentrations were in the laminar flow regime, one

having dunes and the other a plane bed; the sediment transport capacity

was increased by about a factor of 10 as compared to the turbulent

 -
55 -

hyperconcentrated flows (with the same fluid discharge). No clear and

significant change in flow resistance was found for the turbulent flows

as compared to the clear water flows, but for the laminar flows f in¬

creased considerably. Bradley noted that a fluidized layer of bed ma¬

terial was in motion in the plane bed laminar flow run; this layer was

about 25 grain diameters thick.

Woo (1985) made a theoretical study on sediment transport in hyper¬

concentrated flows. He considered the effect of fine suspended material
on bed material transport rates by applying a modified form of the

Einstein (1950) sediment transport equation to the flume data of Simons

et al. (1963). Woo included three modifications to the Einstein formula:

Use of the measured average channel velocity instead of a calculated

one; a correction factor for laminar flow effects which is related to

the particle Reynolds number (Einstein and Chien, 1953); and a correc¬

tion to the von Karman "constant" (Einstein and Abdel-Aal, 1972), which

is used for the suspended load calculations. He compared calculated bed

load discharges obtained with the original and the modified Einstein

equation for the flow conditions of the experiments of Simons et al.

(they measured only total bed material concentrations). These calcula¬

tions showed almost no difference in the results. Theoretical con¬

siderations led Woo to the conclusion that a decrease in the density

factor (s-1) will increase the bed load discharge, and an increase in

fluid density generally will decrease the bed load discharge. He stated

that for the flow conditions of the Simons' et al. experiments these
effects seem to have counterbalanced each other. Woo further concluded

that the increase in total bed material transport in a clay suspension

must be mainly due to an increase in the suspended bed material dis¬

charge. He calculated the total bed material discharge and found that by
accounting for the changed density and viscosity in the Einstein

equation, the measured discharges of Simons et al. can be slightly

better predicted than without this adjustment. Woo also computed the

total bed material discharges with Yang's (1979) equation but he only
found a marginal improvement over the calculated discharges with the

Einstein equation.
 -
56 -

2.3.1.2 Effect of change only in grain-fluid density ratio

Woo's (1985) considerations are limited to the experimental range of

the Simons et al. (1963) flume data for which the highest clay concen¬

tration was 2.3%. Bradley (1986) theoretically examined the effect of

increased fluid density for a fine suspended sediment concentration of
20.1%. He found that the Meyer-Peter and Muller (1948) equation predicts
an increase in bed load discharge by a factor of 1.67, and the Einstein

(1950) equation an increase by a factor of 1.51. In the calculations

with the latter formula Bradley also accounted for the decreased fall

velocity of the transported particles.

Shields (1936) performed sediment transport experiments using water

and particles of different densities. The specific densities s (= cr/p)

varied from 1.06 up to 4.25. His bed load transport equation can be

written as:

10qJ(x -x )
(2-65)
%
-

o(s-l)H
m

where q_ is the volumetric bed load transport rate per unit width, q is
D

the volumetric water discharge per unit width, and x is the critical

shear stress at incipient motion. It may be noted that the density

factor (s-1) in equ. (2.65) appears with an exponent of 2 and not in a

linear form like in many bed load transport equations.

Mizuyama and Shimohigashi (1985) made steep flume tests to study the
effect of fine material in suspension on bed load transport. For a slope

range between about 10 and 20 %, with pearl-clay as fine sediment and

sand with d = 1.9 mm as bed material they found the following relation-
m

ship to be valid:

A' ' S2 (2-66)

«B
"

(i=l)T

where A'= 20 for clear water and A'= 25 for the experiments with the

fine sediment suspension. They also performed some tests with flyash

(being somewhat coarser than the pearl-clay) as fine sediment; in a plot

in terms of the dimensionless parameters *_ and 6 the flyash data points
lie close to comparative points from clear water tests. As compared to

the pearl-clay experiments (performed with a movable bed), the transport

 -
57 -

rates with the flyash suspension are about twice as high, because they
were obtained from tests with a fixed, rough bed (Mizuyama, 1988). The

viscosity of the fine material suspensions was not determined but Mi¬

zuyama (1988) estimated that it only slightly increased with increasing

concentration.

Chee (1988) reported on flume tests on the washout of a gravel dam

under a constant water discharge. In the experiments the density ratio s

was varied between 2.65 and 4.50 by using particles with different

specific weights; the initial downstream slope of the dam was fixed

between 33% and 45%. Chee determined a dimensionless equation in which

the ratio q„/q depends, among other parameters, on the factor l/(s-l)2.

Meyer-Peter and MUller (1948) performed bed load transport experi¬

ments and also included some runs using particles lighter (coal) and
heavier (baryt) than ordinary quartz grains. Their flume data is given
in Smart and Jaggi (1983), and the highest measured *_ value was 0.46.

In a simplified form, the Meyer-Peter/Muller bed load formula can be

expressed as (Yalin, 1977):

*B =
8(6-9c)1-5 (2.67)

where *_ is defined with q„:

*B =
qB/[(s-l)gd3]1/2 (2.68)

Another flume study on sediment transport using particles of dif¬

ferent densities was carried out by Luque and van Beek (1976). Using
water as transporting fluid, the density ratio s covered a range from

1.34 up to 4.58. The experiments were made in a closed rectangular

channel, and the slope of the movable bed was varied between 0° and 22°,

one objective being to study the iniation of motion at steep slopes.

Measured bed load transport rates were correlated in terms of the dimen¬

sionless parameters #_ and 0 to yield:

*B =
5.7(9-ec)1,5 (2.69)

The highest measured value of *_ was about 0.08. Equ. (2.69) can be

rearranged as:
 -
58 -

5.7 . o c-1.5
., 7n.
(2-70)
<b (iris (-p-}

As compared to equ. (2.65) from Shields there is only a linear depen¬

dence on the density factor (s-1) in equ. (2.70). Luque and van Beek

also measured the average particle velocity of single grains, U,, at

relatively low shear intensities and found the following empirical

relation:

U. = 11.5(v*-0.7v*) (2.71)
b c

where v* denotes the shear velocity at initiation of motion. A similar

c

equation for the mean speed of bed load grains was theoretically derived
by Bridge and Dominic (1984) (s. section 2.3.2, equ. 2.98). According to

equ. (2.71) the particle velocity may depend on the density ratio s only
via v*, and if the above relation is also valid at higher shear inten¬

sities then U. must ultimately become independent of s. Yalin (1977)

concluded that s is not an important variable for the uniform motion of

grains which is approximately the case for grains flowing "en masse")
but that s can not be neglected if the individual, irregular motion of a

saltating grain is considered.

It is interesting that in the dimensionless form of both equs. (2.69)

and (2.67) is only function of 9 and 6 and that effects
*R a , density
within the given experimental ranges can be accounted via the dependence
of the dimensionless parameters on the factor (s-1).

This fact is also illustrated in Fig. 2.4 (s. page 49), taken from

Yalin (1977), where bed load transport data from various sources are

shown on a log-log plot of *_ vs. 6 including particles heavier and

lighter than ordinary quartz grains. In this plot only the Wilson (1966)
data points for nylon particles in water (s =
1.138), covering a range

of 6 values between 1 and about 3.2, are located apart from the general
trend defined by the rest of the data. From the limited data it seems

that only at higher shear stresses the effect of the density ratio s

becomes noticable. Considering the weight of the moving layer and its
average velocity, Yalin (1977) theoretically developed a bed load trans¬

port equation in which $„ is given as funtion of 6, 9 ,

and s:

*„ = 0.635 90-5 m [1 -
—ln(l+zm)] (2.72)
D ZU1
 -
59 -

with 1 (2.73)
m =
(9/9c) -

and z = 2.4590,5/s0,A (2.74)

A check of Yalin's equation was made for one of Wilson's data points
9 2.3, 9 0.05, and 2.68 (for sand) 1.14 (for
using = = s = or s =

nylon), respectively. While Yalin's equ. (2.72) predicts a slightly

smaller transport rate for the sand case, the measured *_ value (at 9 =

2.3) for sand is about 2.4 times higher than that for nylon.

Wang and Zhang (1987) presented a new bed load transport equation

that is based on a probabilistic approach similar to the concept of

Einstein's formula. The bedload transport rate in their equation is

proportional to the probability of detachment, the average jump length

and height, and the average grain velocity. They showed that both the

jump characteristics and the grain velocity also depend on the density

ratio s. They plotted their equation in terms of the dimensionless

parameters #_ and 8, presenting three curves for different density

ratios s, together with selected flume data mainly taken from Yalin

(1977), s. Fig. 2.4, covering a range of *_ from 0.001 up to 100 and of

9 from 0.04 up to about 5. For a given shear rate 9 larger than about

0.1 their equation predicts decreasing #_ values for decresing density

ratios. Most of the flume data shown in the plot is for the transport of

quartz grains in water but some data from Wilson (1966) for nylon par¬

ticles (with s=1.138) in the higher shear intensity region (8 « 1 to

3.2) seems to support the equation of Wang and Zhang also for s values

smaller than 2.68. For 9 values below about 0.1 the influence of s on *_

appears to be negligible. Thus their theory does not contradict the data

for particles of different densities obtained by Luque and van Beek

(1976) and by Meyer-Peter and Muller (1948) (since only a few data

points of the Swiss investigators have 9 values above 0.1, with a

maximum of 0.25). However, the prediction of smaller dimensionless

transport rates with decreasing s values at high shear stresses is not

in agreement with Yalin's equ. (2.72), which predicts larger *_ values.

o

Low (1989) conducted a large number of experiments with lightweight

sediments to investigate the effect of the the density ratio s on bed

load transport rates. He used cylindrical plastic particles with various

amounts of lead inside them to obtain s values between 1.17 and 2.46.

Based on dimensional analysis and a regression of his experimental

 -
60 -

results he found that the bed load transport rate should be related to

the ratio of the shear velocity to the grain fall velocity; he proposed
the equation:

110
^li (2.75)
qB
=

where q_ is in [m3/s.m], v* in [m/s], d in [m] and W in [m/s]. He incor-

porated an expression by Rubey (1933), in which the fall velocity is

'

proportional to the factor [gd(s-l)] for his experimental conditions.

Then he obtained the relationship #D «
93, which is identical to the

Einstein-Brown (Henderson, 1966) formula.

Low also compared his measured transport rates graphically with those

predicted by different bed load transport equations; these included the

formulae of Shields (equ. 2.65), of Meyer-Peter and Muller (equ. 2.67),

of Bagnold (equ. 2.97) with the constant b = 4.25, of Yalin (equ. 2.72),
and of Smart and Jaggi (equ. 2.83). It is interesting that with all the
considered equations there was no systematic grouping of the data points
belonging to a given density ratio s, except for the equation of Smart

and Jaggi. Thus it appears that the first four formulae adequatly ac¬

count for a change in the grain-fluid density ratio. Low performed a

regression analysis in terms of the parameters used by Smart and Jaggi,

and he proposed that their equation (2.83) should be multiplied on the

(s-l)~
'

right hand side by a factor of ,

in order to better account for

a change in s.

2.3.1.3 Effect of increase in fluid viscosity, laminar flow

Willi (1965) studied the erosion of a gravel bed with time at steep

slopes. He considered the effect of a change in the eddy viscosity v'

which is often defined as m' =

l2|dv/dy| (where 1 is Prandtl's mixing
length), thus being analogous to the laminar kinematic viscosity v; the

total shear stress can then be expressed as:

t -

pg <v+ V) (2.76)

Using a relationship between the von Karman konstant K and the grain
concentration developed by Einstein and Chien (1955), Willi found that

the bed load transport rate q„ should be proportional to 1/-Jv'. He also

 -
61 -

calculated V from his measurements and concluded from a regression

analysis that in his measurements there was a much weaker dependence,

i.e. q_ was found to be proportional to 1/V

Hong et al. (1984) examined the effect of a water temperature change

from 30° C to near 0° C on the transport of sand in an open channel. For

the given temperature variation the viscosity of water was varied by a

factor of 2.5. For their flat bed runs they found only a slight increase
in the bed friction factor but an increase in the bed layer discharge
concentration by about a factor of 3 for the highest viscosity increase.

The results of the individual runs are, however, not directly comparable
because the experiments were conducted so as to maintain approximately a

constant depth while the energy slope was higher for the lower tempera¬
ture runs (Lau, 1987). Furthermore, Woo (1985) questionned the increase
in the bed layer discharge on account of the uncertain definition of the

reference concentration which Hong et al. used to distinguish between

bed load and suspended load.

With regard to bed load transport Yalin (1977) stated that with

decreasing Reynolds number the average distance travelled by individual

particles is reduced. Both Davies and Samad (1978) and Coleman (1967)
performed flume experiments concerning the fluid dynamic lift force on a

particle near the bed at low Reynolds numbers. The results of both

studies indicate negative lift force coefficients if the particle

Reynolds number is less than 5, or (as an alternative criterion) if the

laminar sublayer is larger than the particle size.

For bed load transport in turbulent flow Bagnold (1956) found that

the grain concentration at the bed, C ,

reached a maximum of about 0.53

if 9 >= 0.4. From theoretical considerations, on the other hand, he

calculated C to be about 0.3 for bed load transport in laminar flow at

9=2. These findings may suggest that the bed load transport rate could

be greater in turbulent flow than in laminar flow at the same dimension¬

less shear stress. In experiments with a very low particle-fluid density

difference (a-p =
0.004) in a closed rectangular channel, Bagnold (1955)
observed that with increasing grain concentration turbulence was more

and more suppressed; near the bed C decreased from its maximum value in

turbulent flow while the grain concentration at the top of the flow

increased. At a mean concentration of C =0.35 the flow was completely

laminar and the grains were dispersed uniformly over the flow depth.
 -
62 -

Later Bagnold (1973) stated that bed load transport by saltation can

also occur in laminar flow and must therefore be a different process

from that of the turbulent transport of suspended solids. Bridge and

Dominic (1984) concluded that turbulence modifies the saltation trajec¬
tories, especially at the higher flow stages, and that the fluid-trans¬

mitted shear stresses become more important at higher shear velocities;

this implies that bed load transport should be different in laminar and
turbulent flow. Yalin and Karahan (1979) described that grains of the

uppermost layer are dragged in laminar flows in the form of a "grain

carpet", which is similar to Bradley's (1986) description of a moving
"fluidized layer" near the bed.

2.3.1.4 Hydraulic transport of solids in pipes

In section 2.3.2 it is shown that the experimental and theoretical

derivation of a bed load transport formula for turbulent pipe flow by
Wilson (1966, 1984, 1986) resulted in similar equations as the ones

commonly used for open channel flows. Therefore the findings from pipe
flow experiments may also be relevant for the prediction of the sediment

transport behaviour in an open flume under similar flow conditions.

Brflhl (1976) reported on a study in which he examined the effect of a

slurry of fines on the hydraulic transport of sand in a pipe. According

to his literature survey, previous studies indicated partly contra¬

dictory effects. He found with preliminary experiments that the pressure

gradient, AH/L, for the transport of sand in a bentonite slurry was

reduced at lower velocities but increased at higher velocities, as com¬

pared to the case with clear water as carrier fluid. The reduction of

the pressure gradient was the more pronounced, the finer was the slurry
material and the larger was the sand concentration.

Bruhl performed main tests with very fine limestone and quarzit
particles (because they are chemically less reactive than bentonite

clay). The limestone and quarzit slurries showed a slightly plastic

(Bingham type) behaviour; a linear dependence of the viscosity on the

slurry concentration was observed up to C, = 13 % (where the maximum

viscosity was about 3 cps.). In the presentation of the results for

different slurry concentrations, the change in pressure losses were

plotted as a function of the mean fluid velocity. The conditions at the

critical deposit velocity (below which sand starts to settle out of the
 -
63 -

flow) are probably most similar to equilibrium transport conditions in

an open channel flow. It is noted that most sand particles moved in the

layers near the bottom of the pipe, and in some cases there was a

stationary deposit layer. For the transport of sand with a mean diameter

of about 0.35 mm, Briihl found that the reduction in the pressure

gradient at critical conditions increased with increasing fine material

concentration. Again a larger reduction resulted for higher sand

concentrations. This means that for a given bed slope in open channel

flow, the sediment transport capacity should increase under the above

conditions, if the pipe flow results are transferable.

A few other findings of Briihl's study may also be of interest. Ex¬

periments with a coarser sand of about 1.0 mm mean diameter indicated a

somewhat smaller reduction in the pressure gradient for the transport in

a slurry of fines. For the 0.35 mm sand, a reduction of AH/L was ob¬

served for the tests with a velocity V smaller than 4.5 m/s. For the

tests with V = 5.0 m/s, however, AH/L was larger than for the clear

water case.

Briihl concluded that main effect of the slurry of fines is to make

the flow more homogeneous, which was also confirmed by measured velocity

profiles. He attributed this homogenizing effect to the increased

densitiy and viscosity of the fluid, resulting in decreased settling

velocities W of the coarser particles. For the sand of 1.0 mm this

reduction in W was less pronounced than for the smaller sand. The higher
is the slurry concentration, the larger are the particles which make up

the new homogenoeous fluid and which should no longer be looked at as

"transported" sediment. At high velocities the transport mode is largely

homogeneous also without any fine particles.

Briihl used existing methods, i.e. the approach of Durand and the one

of Fiihrboter, to calculate the AH/L values for his experimental condi¬

tions. He modified the density dependent parameters in these calculation

procedures, and he also accounted for the reduced settling velocities in

the fine material slurry. Thus he obtained much better predictions of
the measured values than if he used the settling velocities in clear

water. This too confirmed his hypothesis that the change in W is at

least partly responsible for the change in the pressure gradient AH/L.
 - 64 -

A.D. Thomas (1979a,b) performed experiments to find the critical

deposit velocity V, and the gradient at depsosition J, for the hydraulic

transport of solids in pipes. He used both fluids and particles of dif¬
ferent densities such that the density ratio s varied between 2.04 and

9.71 while the fluid viscosity differed as much as by a factor of 70.

For large pipe sizes and for the case that the particle size d was

larger than the thickness of the laminar sublayer 8, he suggested to use

the following equations given by Wilson and Judge (1976):

Jd =
KlgP(s-l) (2.77)

Vd = [2.0 +
0.31og^-] l2gD(s-l)J1/2 (2.78)

where D is the pipe diameter and K. is a constant. It can be seen that

V,, which can be looked at as an index of the transport capacity of the

flow, is only affected by the fluid viscosity via the drag coefficient
C_. An increase in C_ will result in a slight decrease in V,, implying
an increase in sediment transport capacity. However, the influence is

only weak in equ. (2.78), as there is a logarithmic dependence on C_.

The opposite effect is predicted by equs. (2.77) and (2.78) as a result

of a change in the density factor (s-1): The smaller is the density dif¬

ference between the fluid and the transported solids, the higher will be

the transport capacity of the flow.

For the other case, if d <= 8 and for solids concentrations C up to

about 20%, A.D. Thomas (1979a,b) developed semi-empirical expressions

for the critical conditions at deposition, now denoted by V,, and for

J,, respectively, as:

V* = l.l[gti(s-l)/p]1/3 (2.79)

and p/D
Jd =
4.4[l+Cs(s-l)][gn(s-l)/p] (2.80)

From these two equations it can be concluded that the higher the fluid
viscosity the lower the transport capacity of the flow will be in the

pipe. Obviously the influence of the fluid viscosity is larger for the

case d <= 8 than for d >= 8. A.D. Thomas (1979a) further showed that by
combining the theories for the two cases he could improve the prediction
 -
65 -

results for the cases where the particle size was close to the thickness

of the laminar sublayer.

A.D. Thomas (1979b) also reported on experiments carried out in pipes

under laminar flow conditions, where coarse particles were transported

mainly in the sliding bed. He found that the pressure gradient required
to prevent deposition will always be greater than in turbulent flow,

implying a decrease in transport capacity in laminar flow.

Many experiments were carried out by Lanzendorf (1984) to study the

motion of solid particles in a pipe at conditions where the thickness of

the laminar sublayer is of the order of the grain diameter d. He studied

the movement of isolated cylindrical particles along the smooth bottom
of the pipe and found that the ratio of the particle to the local fluid

velocity decreased with an increase in the relative thickness 8/d of the

laminar sublayer. This again suggests that the transport capacity may be

expected to decrease with increasing fluid viscosity, for otherwise

constant flow conditions.

2.3.1.5 Summary

There seems to be general agreement that in turbulent hyperconcen¬

trated flows the suspended load transport rates increase with increasing
fine material concentration; this was concluded from field observations

(Beverage and Culbertson, 1964; Bradley, 1986), from laboratory experi¬

ments (upper regime tests of Simons et al., 1963; Kikkawa and Fukuoka,
1969; Wan, 1982; Bradley, 1986; Wan and Song, 1987), and from theore¬

tical considerations (Woo, 1985). What regards the effect on bed load

transport rates, an increase in q_ due to a decrease of the density

factor (s-1) is predicted from experimental results by Shields (1936),

by Meyer-Peter and Muller (1948) and by Luque and van Beek (1976), and
from theoretical considerations by Yalin (1977) and by Woo (1985).
However, Wan and Song's (1987) theory predicts a decrease in *„ for
B

decreasing s values at higher shear stresses (9 >= 0.1). Low (1989)

found that many bed load transport formulae adequatly account for a

change in the density ratio s, but that the Smart and Jaggi (1983)

equation should be modified. Information on the effect of an increasing

viscosity alone on q„ in open channel flow is less conclusive (Hong et

al., 1984; Woo, 1985; Lau, 1987). Wan (1982) concluded from his

experiments that *_ was smaller in the bentonite suspension flows than

 -
66 -

in the corresponding clear water flows, particularly in the low flow

intensity region.

Results from pipe flow experiments indicate the following trends for
the hydraulic transport of solids (at conditions that are probably
similar to equilibrium bed load transport in flumes) in a slurry of

fines: An increase in transport rates will result for the case where the

thickness of the laminar sublayer is (still) smaller than the repre¬

sentative grain size, but a decrease can be expected if the flow is in

the hydrodynamically smooth regime (Briihl, 1976; A.D. Thomas, 1979a,b).

An increase in the thickness of the laminar sublayer (without any change
in the density ratio s) will probably lead to a decrease in the trans¬

port rate of particles along the bottom of a pipe (Lanzendorf, 1984).

There is not yet much information available on sediment transport in

laminar flow. In their experiments, both Wan and Song (1987) and Bradley
(1986) observed an increase in the total bed material discharge; in

Bradley's study the transport rates in laminar flow were about an order

of magnitude higher than in the corresponding turbulent flow (with the

same fluid discharge). Pipe flow tests reported by A.D. Thomas (1979b)

may suggest that the bed load transport rates decrease in laminar flow,

assuming that bed load is the predominant mode of transport at a velo¬

city slightly above depositional conditions. Negative lift forces for

particles near the bed (Coleman, 1967; Davies and Samad, 1978) and theo¬
retical considerations by Bagnold (1956) seem also to indicate that bed

load transport should be smaller in laminar flow than in turbulent flow.

2.3.2 Bed load transport at steep slopes and high shear stresses

It is only recently that interest on sediment transport problems has

also focussed on steep streams. A limited number of laboratory studies

have been made to study the bed load transport in steep flumes. But

there is almost no field data available since it is very difficult to

measure transport rates of coarse material in steep streams.

Smart and Jaggi (1983) performed flume experiments including slopes

from 3 to 20 % and using gravel-sand mixtures as bed material. The mean
 -
67 -

grain size d was 4.2 and 10.5 mm for the two relatively uniform mate¬

rials, while d was 2.0 and 4.3 mm for the two materials with a wide
m

grain size distribution. Their measured dimensionless transport rates *_

ranged from 0.13 up to 83. In the analysis of their experiments they

also included the extensive data set of the Meyer-Peter/Miiller experi¬
ments (given in Smart and Jaggi, 1983) which cover the low intensity
transport region including flume slopes between 0.04 and 2%. A regres¬
sion analysis of these two data sets resulted in the following bed load
transport formula:

4 ,d90>0.2 S
.1.6
„
6c,
„
_..

«B
"

<i=l) ^
" (1
"

9"> <2-81)

where d„- and d,~ are characteristic grain sizes, than which 90 % and

30 %, respectively, of the material by weight is finer. At high trans¬

port rates the sediment in motion occupied a significant part of the

mixture flow depth H which is then larger than the fictitious fluid

(water) flow depth Hf (=q/V). For this increase in flow depth Smart and
Jaggi found an empirical relation as a function of slope and dimension¬

less transport rate:

H,/H = 1 -
1.41 S1-14 *?-18 (2.82)
i m B

A transport formula in terms of the well-known dimensionless parameters

can also be derived from equ. (2.81):

*B - * ^>0-2 *0-6«>:-5
30
<w £ m

where c is a flow resistance coefficient defined as c =

V/v*, v* and 9
m

are to be calculated with the (corrected) mixture depth h while 9,

,

should be determined with the (corrected) fictitious fluid depth h f.

Equ. (2.83) differs from their corresponding form given in Smart and
Jaggi (1983) and in Smart (1984), where the factor 9,/9 had not been
t m

considered. Thus equ. (2.83) overpredicts the transport rates at the

steeper slopes (high transport rates). A reinterpretation of their data

and a discussion of the applied sidewall procedure is given in chap¬

ter 4.
 -
68 -

Mizuyama (1977) carried out similar sediment transport tests with

flume slopes up to 20 X and presented the following bed load transport

equation for flow over a plane bed:

»B =

^ •(1,5-*) [1-^][1-M/)1/2] (2.84)

where k is a function of the slope.

Takahashi (1987) compared bed load transport formulae with several

data sets from steep flume experiments. He slightly modified an equation

developed by Ashida, Takahashi and Mizuyama in order to achieve better

agreement also with the flume data of Smart and Jaggi, and proposed the

following version of Mizuyama's equation:

*B "

^^iT (f)0-V-5|l-k'^][l-k'(J)1/2] (2.85)

where k' is also a function of the slope. It is not clear, however,

whether 6 is defined with h or h in equ. (2.84) and (2.85).

Daido (1983) also reported on bed load transport tests in steep chan¬

nels. Based on experimental data including flume slopes up to 10 X and

on theoretical considerations about the average transportation velocity

and the mean thickness of the transported sediment layer he gave the

following equation:

*B =
B261-5(l -

Q^)1,5 (2.86)

where B„ is a proportionality constant that may be a function of the

slope, the ratio 8 /6, the relative depth h/d, and the Froude number Fr
c
0 5 '

=
V/(gh) . For his experimental conditions Daido found B. = 3.7.

From data of Japanese steep flume tests that cover a slope range
betweentween isis much
6whereconditionsforandX,25andX5
6whereconditionsforandX,25andX5
larger than 6 ,

Mizuyama (1981) proposed a rather simple equation:

—
= 5.5 S (2.87)
q
 -
69 -

Ward (1986) collected available data on sediment transport in steep

flumes. In addition to the 77 measurements of Smart and Jaggi he found

46 with information the ratio slope,
another observations on
qR/q>
median grain size, and gradation. A regression analysis resulted in a

equation similar to (2.87) but including also grain size characteris¬

tics:

£ = 7.16 (d50)-°-25 (^)0-2 S2 (2.88)

q ju
o3Q

where in [mm] is the characteristic grain size, than which 50% of

d,-n
material by weight is finer. The disadvantage of this formula is that it

is not dimensionless on the right hand side, and thus application to a

field situation seems questionable.

Takahashi (1987) introduced the term "immature" debris flow (s. also

sec. 1.3) where the moving grain layer is in the domain of sediment

gravity flow and the concentration over the whole thickness of the

moving layer is approximately constant and equal to 0.4 C^. It is dif¬

ferent from a debris flow in that there is still a clear water layer
above the moving sediment carrying almost no grains; and it is different
from ordinary bed load by the fact that gravity rather than fluid shear
is the main force acting on the grains. If an appropriate bed load

transport and flow resistance formula is used, the limiting condition

between ordinary bed load movement and an immature debris flow can be

given in terms of the slope and a dimensionless water discharge q* =

'

q/(g d ). Takahashi developed the following equation for immature

debris flows; it is based on a part of the experimental data of

Mizuyama, Smart and Jaggi, and Takahashi, and on observed grain velocity

profiles:

qB 2 4.2-0.3^ 2 9c 2
6 (1 (2,89)
9_)
= "

v*~d 3 cos2B(tana-tan6)2

Equ. (2.89) can be rearranged to give:

, 4.2-0.3C* , '
, 9 -

*B 6 <X "

T> (2'89b)
3 cos2P(tana-tanP)2

From Takahashi's plot of the various debris flow types in a diagram of S

vs. q* it can be seen that immature debris flows generally occur at

 -
70 -

higher slopes and higher discharges q* than ordinary bed load transport
flows. It appears from equ. (2.89b) that under these conditions i„

should depend more strongly on 9 than under ordinary bed load transport
conditions.

Wilson (1966) intended to study the bed load transport of fine bed
material at high shear stresses. Previous flume data only covered *„
values up to about 10 and 9 values up to about 1. He performed bed load

transport tests with sand of 0.7 mm diameter in a pipe, because it was

easier to achieve high shear stress conditions in a pressurized conduit.

He measured *_ values up to 280 and 9 values up to 8.4. This still seems

to be the only comprehensive data set on bed load transport at such high
shear intensities. From a correlation of his experimental results Wilson

(1966, 1986) found the empirical relationship:

% grSi) <!V>1'5 (2-90>

where x is the shear stress at the surface of the deposit, corrected

for ripple drag. It is evident that equ. (2.90) is basically similar to

the Meyer-Peter/Muller formula, equ. (2.67), except that the constant is

12 instead of 8, which might be due to the fact that the equations are

valid in different shear intensity regions and/or for different flow

conditions. Fig. 2.4 (taken from Yalin, 1977) shows the experimental

range covered by various studies on bed load transport including those

of Wilson and Meyer-Peter/Miiller.

Wilson (1984,1986) also made a theoretical analysis of bed load

transport in turbulent pipe flow. From measured concentration profiles

of coarse particle slurries he assumed a linear concentration distribu¬

tion across the transported layer. Together with considerations on the

grain velocity gradient and the thickness of the sheared layer he ob¬
tained the following theoretical equation:

Lli (^)1,5 (2.91)

„
qB ^p *.< ••>>)
=
'
g(s-l)Ktana

At high shear stresses x is negligible in comparison to x . Wilson

(1986) assumed K = 0.4 and tana = 0.32 (Bagnold's "inertial" value)

which gives 1.51/(Ktana) = 11.8. It is interesting that equ. (2.91) is
 -
71 -

very similar in form to the empirical equ. (2.90), and that -

with the

above assumptions -
also the numerical constants are almost the same.

Fig. 2.4 : Comparison of bed load transport formulae with experi¬

mental data, after Yalin (1977).
 -
72 -

Before reviewing selected sediment transport formulas Yalin (1977)

presented theoretical considerations based on dimensional analysis. The

following set of characteristic parameters may be used to define the

two-phase motion of cohesionless granular material of specified geometry

in a fluid:

p, y, a, d, h, v*, g(o--p) (2.92)

Choosing the seven independent parameters indicated in (2.92) and selec¬

ting d, o", and g(o-p) as basic quantities, a general expression for

sediment transport is obtained:

*= [g(s-l)d3/v2, 8, h/d, s] (2.93)

where v =
y/p is the kinematic viscosity of the fluid and ^ is a func¬

tion describing the dependence of * on the dimensionless parameters in

the square brackets. The first term in the brackets is equal to Re /9

and reflects the influence of y; it has the advantage that it is inde¬

pendent of the flow stage.

From an inspection of the Meyer-Peter/Miiller bed load transport for¬

mula, equ. (2.67), it can be shown (Yalin, 1977) that for very high
shear intensities, i.e. 9 » 6 ,
the following proportionality relation
is obtained:

*B ~ 61-5 (2.94)

h -pv* V* (2-95)

where i„ is the bed load transport rate by immersed weight of solids per
o

unit width.

Based on experimental data of both transport of quartz grains in

water and in air, Bagnold (1956) developed a theory on bed load trans¬

port. His equation can be given as:

*^B = (6 -

ec)90-5 (2.96)

1/2
where *.„= q„/b[g(s-l)d3cos3] ,
6 =
9/cos0 and 6 = 9 /cosB. The factor
"D D HI c c

cos$ accounts for the reduction in shear resistance at higher bed

 -
73 -

slopes. Yalin (1977) rederived Bagnold's equation in a shorter manner

and obtained a slightly simplified version neglecting the influence of

cosB (which effect is smaller than 0.5% for slopes less than 10%):

•B =
b(9 -

9c)90,5 (2.97)

Yalin showed furthermore that for very high shear stresses equ. (2.97)

ultimately results in the same proportionality relations (2.94) and

(2.95) that he also had obtained from the Meyer-Peter/Miiller formula.

Bagnold (1956) found that extrapolated curves of the bed load portion of
various sediment transport data on a log-log plot are asymptotic to the
3/2
relation
**R= 9-6 .In the same plot he also showed data of his

experiments with a small density difference (o--p) =

0.004; the transport
of the light weight grains occurred in suspension but the data,

including 6 values up to about 4, plots parallel to the asymptotic

relation for bed load transport.

Yalin (1977) also examined the bed load formula of Einstein (1950).

For high shear stresses the original Einstein equation follows the

proportionality relation *_ ~
9. Yalin agreed on the principles of
Einstein's formula but proposed some modifications in the details of its

derivation. The modified version was then shown by Yalin to become

3/2
asymptotic to *_~ 9 at high shear stresses, the same behaviour as

demonstrated for the Meyer-Peter/Miiller and the Bagnold formula.

In their theoretical development of a bed load transport equation

Bridge and Dominic (1984) used an approach similar to the one of Bagnold

(1956, 1973). They derived expressions for the mean velocity of the
moving grains, U_, and for the immersed weight of grains moving over a

unit bed area, W', as:

UB =
a(v* -
v*) (2.98)

y
with a = I ln(-"0 (2.99)
K
V1

W' =
(t -

xc)tana (2.100)

where y is the distance of the effective fluid thrust from the boundary
 -
74 -

and y. is the "roughness height". Combining equs. (2.98) and (2.100)

yields:

*„ = t3— (<l8 - le )(9 -

9 ) (2.101)
B tana c c

For high shear stresses (i.e. 9 » 9 ) the proportionality relation *„ ~

c
3/2
9 is obtained again. Bridge and Dominic theoretically discussed the

influence of the parameters in equ. (2.99) on the value of 'a' and con¬

cluded that 'a' should increase for both fixed and mobile beds if the

bed shear stress increases. They analyzed available data both on grain
velocity and on sediment transport rates for lower and upper stage plane

beds. They found the parameter 'a' to increase with transport stage

(defined as v*-v*), which they attributed to increasing suspended sedi¬

ment concentrations, tana, on the other hand, was found to decrease with

increasing bed load concentration, grain size and grain shear stress.

The overall result is an increase in the value (a/tana) from lower stage
plane beds (mean value 9.5) to upper stage plane beds (mean value 17.1).
It may be interesting to note that in the Smart/Jaggi equation (2.83)
'

for bed load transport on steep slopes there is the factor cJ which

is close to cJ =
Fr; thus there seems to be a dependence of *„ upon

the Froude number which varied in their experiments between

approximately 1 and 3.

Hanes and Bowen (1985) theoretically developed a granular fluid model

for intense bed load transport. They considered the moving grain layer
to be composed of two zones: A region of collision dominated grain flow
which they analyzed based on Bagnold's concept of dispersive stresses,
and a saltation region where both grain and fluid stresses are impor¬
tant. They found that their theoretical equation for the total bed load

transport can be fairly well approximated by a simple expression; Hanes

(1986) later gave a corrected version as:

*„ = 6 92-5 (2.102)
o

The theory of Hanes and Bowen allows to calculate *_ separately for the
two bed load transport regions; they concluded that about 90 % of the

total transport is contributed by the saltation zone.

 -
75 -

From a limited number of flume experiments on bed load transport at

slopes S steeper than about 10 X, it has been found that the ratio q„/q
16
'

is proportional in first place to the factor S (Smart and Jaggi,

1983) or to S2 (Mizuyama, 1981; Ward, 1986).

Semi-theoretical considerations on bed load transport lead to the

conclusion that -
in terms of the conventional dimensionless parameters
3/2
-
the equation *_ = A 9 becomes a good approximation at high shear
D

stresses (Bagnold, 1956; Yalin, 1977; Daido, 1983; Bridge and Dominic,

1984). This dependency is also confirmed from the analysis of hydraulic

transport of solids in pipes (Wilson, 1984, 1986). For the case of an

"immature" debris flow, where a grain layer starts to move as a whole, a

2 5
relationship with the proportionality *_ ~
9 has been proposed, based
on experimental data (Takahashi, 1987). A theoretically developed granu¬

lar fluid model also suggests that at very high transport stages the
2 5
relationship *_ -
9 might be valid (Hanes and Bowen, 1985).

The studies of Smart and Jaggi (1983), Daido (1983) and Bridge and
Dominic (1984) indicate that the bed load transport on steep slopes may

further depend on flow conditions; experimental data suggests that the

Froude number is an important parameter.

2.3.3 Initiation of Notion

Many studies about the beginning of sediment transport have confirmed

in principle the well-known Shields (1936) curve (Yalin, 1977). In such

a diagram the dimensionless shear stress at initiation of motion, 9 ,

is

plotted against the particle Reynolds number, Re . The Shields curve

shows the following characteristics: For Re >= 400 measured 9 values

vary between about 0.03 and 0.06, and often an average value of 0.05 is

assumed; when the flow near the bed is the transitional regime between
hydrodynamically rough and smooth flow, 9 reaches a minimum at about

Re = 10; below Re of about 2 it has long been assumed that 9 =

*"
*
0.1/Re based on the few data points then available. Recently, some

investigations have been made to study especially the incipient motion

in the laminar boundary flow region (see below).
 - 76 -

Slope effect and large relative roughness

For flows at steep slopes the calculation of 9 has to be modified to

account for changed gravity effects. Ashida and Bayazit (1973) intro¬

duced a new definition for the critical, dimensionless bed shear stress

for incipient motion at steep slopes, 9 :

9sc =
(2.103)
cosBtana-(s/(s-l))sinB

Exactly the same definition was proposed by Mizuyama (1977) and by Daido
(1983), and a similar one, neglecting the lift force influence, by
Bathurst et al. (1982a). Ashida and Bayazit (1973) performed experiments
on the initiation of motion in a steep flume at high relative roughness
values d/h. They
' plotted their data in terms of 9 vs. d/h and showed
sc

that 9 increases from 0.04 for d/h <= 0.4 to about 0.12 at d/h * 1.7.
sc

However, in their experiments the values of d/h increased with increa¬

sing slope. Thus the increase in 9 seems to be mainly a slope effect

appearing in the definition of 9 ,
and the general trend in their plot
may be due to spurious correlation because d/h is also a function of the

slope.

An alternative way to express the critical, dimensionless bed shear

stress for initiation of motion at steep slopes, 9', is to introduce a

modification factor so that 9' can be compared directly to 9 for low

slope conditions. Based on a balance of moments acting on single grain

on an inclined bed surface, Luque and van Beek (1976) suggested the

equation:

sln(-g)
_
v '
c c sma

They showed that the above equation represents their experimental data
for slope angles 6 =
12°, 18°, and 22°, if the angle of internal

friction a was taken to be 47°. If the factor sin(a-B) in equ. (2.104)

is expanded, one obtains an equivalent formula, which was independently
given by Stevens et al. (1976):

9' = 9 cosB(l -
~^) (2.105)
v
c c tana'
 -
77 -

and 32° then is reduced

For example, if 6 = 11.3° (S = 20 %) a =
9c ac¬

cording to equ. (2.104) or (2.105) by a factor of 0.67.

Ikeda et al. (1982) started from a force balance to calculate the

reduction factor 9'/9 . They arrived at a rather complex equation which

can be simplified if the lift force is assumed to be zero:

9'/9 =
(1 -
tan2B/tan2a)°'5cosP (2.106)

For the same example as given above (B =

11.3°, a =
32°) the reduction

factor is calculated by equ. (2.106) as 0.93. From their experimental

data on the initiation of motion on side slopes and on plane level beds

they concluded that the influence of the lift force can be neglected

although inclusion of this effect results in a slightly lower prediction

of 9 or 9'.
c c

Low values of Re ; 8 in a Bingham fluid

Yalin and Karahan (1979) reported on an extension of the Shields'

diagram by Mantz (1977) for the region Re <= 1, for which the following
relation is obtained based on new experimental data:

9 = 0.1/Re*0-3 (2.107)
c

According to Yalin and Karahan it is not important for the initiation of

motion whether the whole flow is laminar or in the hydraulically smooth

region (with 8 > d) because the viscous flow near the bed is essentially

the same if x is also the same. But the conditions for the detachment
o

of a grain are expected to be different for hydraulically transitional

or rough turbulent flow as compared to laminar flow; for these two cases

two different curves should exist for incipient motion. They performed

experiments to determine 9 both in laminar flows as well as in

turbulent flows being inbetween the hydraulically smooth and rough

regime. The experimental data seems to prove their hypothesis. For

example
"
at Re = 6.5, 9 = 0.07 for laminar flow and 9 = 0.035 for
c c

hydraulically transitional turbulent flow, while the two curves tend to

approach each other for Re <= 1. Similar experiments were carried out

by Lin and Sun (1983) with a glycerol and water solution to determine 9

in a viscous flow; they basically confirmed the existence of different

curves for the incipient motion.
 -
78 -

Daido (1971) proposed to use the shear stress ratio a' = x_/x as an
o o

additional parmater in calculating 9 in a clay suspension (Bingham

fluid). He showed theoretically that for Re <= 7, 9 should become

larger with increasing non-Newtonian characteristic of the flow. For

example if a' =
0.5, 9 is increased by about a factor of 1.5 to 2.

In his study on the transport and the initiation of motion of plastic

particles in a clay suspension, Wan (1982) theoretically developed a re¬

lation to express 9 in a Bingham fluid flow:

8 = 0.047 +
K'^5 B
(2.108)
v '
c 2 pg(s-l)d

He determined the constant K' = 0.4 from his experiments, which con¬

firmed the linear dependence of 9 on

xR. The highest measured value for
8 was about 0.23. Wan did not indicate whether the flow in the cor-
c

responding experiments was laminar or turbulent, but the theoretical

reasoning to obtain equ. (2.108) is based on the assumption of laminar

flow. Similar to Daido's (1971) theory equ. (2.108) predicts an increase

in 9 with an increase of the Bingham yield stress x_.

Critical flow discharge

Combining the Shields type criterion for initiation of motion with

the Manning-Strickler flow resistance equation and a relation to charac¬

terise the grain roughness, Schoklitsch (1950) developed a formula for

the critical flow rate (per unit width) at beginning of bed load

transport, q :

qcr
« 0.26 (s-1)1'67 d1^5 S-1-17 (2.109)

where is be used in [m] and is in [m3/s.m]. (2.109) is

d,n to q Equ.
based on laboratory and field data. Bathurst et al. (1985b) proposed a

non-dimensional version of Schoklitsch's equation. They examined a great

number of experiments with regard to critical conditions at initiation

of bed load transport. For a slope range 0.25% < S < 20% and for essen¬

tially uniform bed material, they introduced the equation:

0.15 g0,5 d^5 S-1-12 (2.110)

qcr
=
 -
79 -

Equ. (2.110) may be modified to include a density factor (s-1), with

the exponent given by equ. (2.109):

c-1-12
n n^ tt. is1-67 „0-5 j1-5 S ,, 111N
g
qcr
=
0.065 (s-1) (2.111)
d5Q

Whittaker and Jaggi (1986) performed experiments on the stability of

block ramps, using rather uniform blocks at slopes between 5 X and 25 %.

Based on similar considerations as above, they determined the following

equation which defines the critical conditions at which the block ramp

is beginning to be destroyed:

qcr
- 0.257 (s-1)0"5 g0-5 d1-5 S"1-167 (2.112)

where d,_ is the characteristic grain size, than which 65 % of material

by weight is finer. Retaining the same exponent of the density factor as

in equ. (2.111), equ. (2.112) may be transformed into (using s=2.65):

„ ... .
,vl.67 0.5 ,1.5 --1.167 ,„ ..,.
(s-1) S
qcr
= 0.143 g dfi5 (2.113)

By comparing equ. (2.113) with equ. (2.111) it can be concluded that

the critical discharge predicted for the block ramp situation is roughly
twice as high as for the beginning of bed load transport data which is

the basis of the constant in equ. (2.110) and (2.111). The relative

depths in the block ramp tests were in the range 0.5 < h/d,,- < 5, and

thus they were possibly somewhat lower than in the flume situations

analysed by Bathurst et al., but there may be other reasons for the

different constant.

It is noted again that the formulae for the critical flow discharge

were derived from the concept of a critical shear stress, and thus there

is no basic difference in using one approach or the other.

 -
80 -

3 EXPERIMENTS

3.1 Experimental program

As mentioned in section 1.4 the objective of the study was to deter¬

mine the bed load transport capacity of a slurry flow in a channel at

steep slopes. The muddy slurry flowing in a torrent was simulated in the

laboratory by recirculating a clay suspension. Various amounts of clay

were mixed into the water circuit in order to examine the effect of an

increasing density and viscosity of the carrier fluid on the sediment

transport capacity.

The flow behaviour and sediment transport characteristics were

examined for five different clay concentration (or fluid density)

levels; the highest concentration of the clay suspension was about 22 X

by volume, for which the flume system could be reasonably well operated.

For each concentration level two kinds of tests were performed. First

the flow resistance was measured for flows of the clay suspension with¬
out sediment transport. This allowed to study the effect of the non-

Newtonian rheological properties of the clay suspension on the flow

behaviour, independent of any interference with transported sediment in

a two phase flow. In a second step, both the bed load transport capacity
of the clay suspension and the flow resistance of the fluid-gravel
mixture were determined for each concentration level.

One objective of the study was to compare the new experiments with
the results obtained by Smart and Jaggi (1983) who had used clear water

as transporting fluid. It was decided to use the same sediment as the

bed material No. IV of Smart and Jaggi. Thus their results could serve

as reference conditions, being equal to the case of a suspension with 0%

clay content (or a fluid with the density of water).

The flume slope, the fluid discharge and the fluid density (clay
concentration) were set for each experimental run; together with the

rheological properties they formed the independent parameters. The fluid

velocity, the (mixture) flow depth and the equilibrium bed load

transport rate were measured as dependent parameters.

 -
81 -

3.2 Flume apparatus

Sediment transport tests in a steep flume had already been performed

at the same hydraulic laboratory by Hanger (1979) and by Smart and Jaggi
(1983). For the new tests with the objective of recirculating a clay

sespension, the existing laboratory setup was slightly modified. A

schematic sketch of the flume system is shown in Fig. 3.1.

Slurry pump ,

&J=^-^^
>\;vsV\vv\vy^\\\\!

Fig. 3.1 : Schematic sketch of the flume system.

The water or clay suspension was recirculated in a independent fluid

circuit. First a conventional water pump was used, but after having
reached a certain concentration of the clay suspension (about 15 % by
volume), a special slurry pump had to be installed. Settling of clay
particles had to be expected during times when the pump was not

operated. Therefore the volume of the sump was made as small as

 -
82 -

possible, in order to limit the amount of added (dry) clay and to mini¬

mize concentration fluctuations after resuspension of settled clay par¬

ticles. The overall volume of the circuit was about 10 m3, including the
sump, the constant head tank and the pipelines.

The flume had a usable length of 5 m, with a width of 20 cm and 40 cm

high perspex side walls. The bed material consisted of relatively uni¬

form gravel. It was fed from two pairs of exchangable sediment hoppers
via a conveyor belt which discharged the gravel into the slightly
steeper upstream portion of the flume. There roughness elements were

installed in order to avoid scouring of the movable bed at the top of

the flume. The movable bed was maintained by an end sill of adjustable
height. The roughness elements consisted of three tooth-shaped metal
"lates, fixed about 10 cm apart from each other across the bottom of the

flume. They could be tilted at any angle against the flow, in order to

adjust the upstream roughness as required by flow conditions.

Before performing a sediment transport test the desired flow rate was

set by a regulating valve. Then a movable bed was built up by initially

reducing both the sediment input and the flow rate, using a gate valve.
This could be opened quickly to set the desired flow rate again; the

sediment feeding rate was increased almost simultanously, so that

initial scouring of the bed could be avoided.

Fig. 3.2 : Longitudinal section of the fixed rough bed, made of

gravel particles glued on a tooth-shaped PVC-plate.

 -
83 -

In order to determine the flow resistance for the clay suspension

alone, without any sediment transport but under similar roughness con¬

ditions, a fixed rough bed was mounted on the flume bottom. The fixed

bed was made of gravel particles glued on a tooth-shaped PVC-plate (s.

Fig. 3.2), so as to obtain a slightly irregular plane bed which is more

like a natural bed than a completely plane bed. The grain size distribu¬
tion of the gravel particles was the same as for the bed material used

in the sediment transport tests (with d = 10.0 mm, s. section 3.4).

3.3 Clay material and clay suspension

Since the fluid volume of the circuit was about 10 m3 and some part

of the added clay was "lost" (for the suspension) having settled out in

dead zones of the system, several tons of clay were necessary to achieve

a volume concentration of the flowing suspension of over 20 X. A clay

called Opalit (brand name) was used for the experiments. It is commer¬

cially available in Switzerland, and is ground from the so called

"Opalinuston". Opalit can serve as a sealing material in earth dams or

other works. This type of clay had been previously used for hydraulic

experiments by Abdel-Rahman (1963) and in soil mechanics research by H.

Einstein (1966).

The grain size distribution curves for Opalit as given by the above
two authors are shown in Fig. 3.3, and soil mechanical and mineralogical

properties are listed in Table 3.1. The values given by H. Einstein and

by Abdel-Rahman differ somewhat from each other. This could be due to

either variations in the composition of the used Opalit or to different

methods used in analyzing the samples. In this study, the particle den¬

sity of the Opalit was taken as 2.65 g/cm3.

Determination of the particle size distribution can be very diffi¬

cult, and for certain clay types very differing results may be obtained

(Grim, 1968). If the clay fraction of Opalit is determined with a more

modern method than wet sedimentation, it is about 40% (Kahr, 1989). This
value corresponds better with the overall mineral content for the clay

groups given in Table 3.1.

 -
84 -

100
s*

y
1^ s
A.'
80

//
/ /

Al>del-Ra hma n(1 963). V/

1 60

>
Ah Einstein (1966)

'/
c
40 <

&*
y&

20

0001 0.002 0 01 01

Particle size
[mm]

Fig. 3.3 : Grain size distribution curves for Opalit clay.

Property: H. Einstein (1966): Abdel-Rahman (1963):

Liquid limit [%] 47.3 41.0 -

43.0

Plastic limit [%] 16.6 18.0 -

19.5

Plasticity index [%] 30.7 23.0 -

23.5

Particle density [g/cm3] 2.71 2.60 -

2.65

Clay fraction < 0.002mm [%] 24.2 22.0 -

25.0

Mineral content [%]:

Illite (Muskovite) 20 -
25

Kaolinite 20 -
25

Chlorite 5-10

mixed-layer type clay 5-10

Calcite 10

Carbonate 6

Table 3.1 : Soil mechanical and mineralogical properties of Opalit clay.

 -
85 -

The clay was added into the operating circuit via the conveyer belt

of the sediment feeding machine. Thus the addition rate was regular and

could be kept small, so that good mixing of the dry clay with the flow
was obtained and no particle aggregation or clusters formed.

Once the pump had been started it took no longer than half an hour

until the flow had established a new equilibrium between the resuspended

particles and the ones still resting in the dead zones of the system;

then the fluid density remained practically constant until the pump was

shut off again. After a day or two of not operating the system, the

density of the flowing clay suspension had decreased by a few tenths of

a percent at the lower concentration levels, and up to a few percent at

the higher concentreation levels. This change affected the rheological

properties of the fluid only about as much as the uncertainty of the

viscometric measurements. However, if one series of experiments for a

given concentration level lasted more than about a week, or if the

system was not operated for a similar period, the fluid density

decreased by a greater amount, changing the rheological parameters x_

and tv by more than 10 %.

It should be noted, however, that there was no experimental require¬

ment to keep the Bingham parameters constant at a given clay concentra¬

tion level, since they were measured separately for each experiment.

For the examined steep slopes, flow rates and relative roughnesses,
and with the roughness conditions at the flume entrance, the turbulence

intensities were found to be sufficient to result in a uniform concen¬

tration distribution of the suspended clay particles over the flow

depth.

3.4 Bed material

One of the four bed materials used by Smart and Jaggi (1983) was also

taken for this study. For experimental reasons their material no. IV was

chosen. As it is the coarsest sediment mixture they had employed and

comperatively uniform, the separation of the transported gravel in the

downstream sediment hoppers from the clay suspension could be best
achieved. The clay slurry is drained and separated from the grains in

the hoppers through a plastic mesh on one side, with quadratic holes of
 -
86 -

2 mm width. If a finer bed material with a wider grain size distribution

had been used, drainage of the clay suspension would have become
difficult or impossible already at lower clay concentrations than with

the selected material. The grain size distribution of the material used
is shown together with material no. IV of Smart and Jaggi in Fig. 3.4.

They claimed that the slight variation in grain size distribution during
the tests did not significantly affect the results.

Table 3.2 gives the characteristic diameters of the original mixtures

of these two bed materials. The new mixture was made from the same,

naturally river rounded grains that had been previously used by Smart

and Jaggi and also by Hanger (1979). The specific density of the

material had been determined by Hanger as 2680 kg/m3, and the angle of

repose as 32.5°; these values were also adopted for this study.

100

Ml
m

A 'A
ii

// '/,'
i\ ni Mat m Smart/jaggi
// / i
/

20
-'
/
'
/
L'' /
<<

___/* -*-*
y
—"

18 20

Grain size [mm]

Fig. 3.4 : Grain size distribution curves of the bed material used

in this study, in comparison with material no. IV of

Smart and Jaggi (1983); the thick line refers to the

original distribution, the thin lines represent samples

taken during or after the tests.

 -
87 -

Material: d
m d90 d30 d90/d30
[mm] [mm] [mm]

of this study 10.0 12.0 8.7 1.38

no. IV of

Smart and Jaggi 10.5 12.1 9.0 1.34

Table 3.2 : Grain size characteristics of bed material.

3.5 Measuring methods

3.5.1 Slope

The flume slope, which was preset before each test, was one of the

independent parameters. It was measured by a spirit-level. The maximum

error AS„ in the slope setting of the flume bottom was determined by

levelling to be about ±0.3 X (absolute slope value).

At equilibrium transport conditions both the slope of the movable bed

and the fluid surface slope should be approximately equal to the flume

slope. The maximum deviation of the fluid surface slope from the flume

slope was less than AS„ =

±0,4 X. Thus the maximum, absolute deviation

of the energy slope from the flume slope S amounts to (AS- + AS_) = ±0.7
d r

X. This results in a maximum relative error of (AS,, + AS„)/S = 0.10 for

D r

7 % of 0.035 for slope S 20 %.

a slope S = or
(ASg +
ASp)/S = a =

3.5.2 Fluid discharge

A magnetic flow meter was used to monitor the fluid discharge. Accor¬

ding to the manufacturer the relative accuracy of the instrument should

be within +1 % of the actual flow rate. Fluctuations of the fluid dis-

 -
88 -

charge during a test were generally also less than ±1 X, as indicated by

relatively slow changes in the reading.

However, at the two highest clay concentration levels the reading was

observed to vary by as much as about 10 %. This might have been due to

local and periodic clay accumulations in front of the plate of the

regulating valve, influencing also the flow in the magnetic flow meter

about 2 m upstream of the valve. Therefore the instrument was placed

further upstream in a vertical part of the pipeline where no adverse

effects due to deposited and partly consolidated clay material had to be

expected. This measure, however, did reduce the fluctuations in the

reading only partly, indicating that they probably originated in the

head tank.

3.5.3 Sediment discharge

The sediment feeding rate could be controlled by regulating the speed

of the conveyer belt, which is driven by rollers, and by adjusting a

slot through which the gravel was discharched into the flume.

The sediment feeding machine was originally calibrated by Hanger. He

determined the roller speed in function of the position of the regula¬

ting knob (s. Fig. 3.5). Having installed an additional regulating knob

at a more convenient place, Hanger's function was checked by new mea¬

surements, which are also shown in Fig. 3.5. At the highest speeds the
new points indicate a slight deviation from the original, straight line.

The result of the calibration for the bed material no. IV used both
'
by Smart and Jaggi and by Hanger is given in Fig. 3.6. Because the

gravel mixture of this study was slightly different from the original
material no. IV, a few check measurements were made with the new mix¬

ture. The corresponding points are also shown in Fig. 3.6.

1) In the original Fig. 8 of Smart and Jaeggi the ordinate is incorrectly

labelled "GB(s-l)" instead of "G„(s-l)/s" which represents the weight
measured under water.
 -
89 -

30
*7

*/
01
25

20
/
15
A
10

0 20 40 60 80 100

Position of regulating knob ['/•/

Fig. 3.5 : Relation between the position of the regulating knob and

the roller speed (expressed in revolutions per minute) of

the sediment feeding machine; the stars indicate control

measurements.

Both Fig. 3.5 and 3.6 indicated that the original calibration curves

could also be adopted for this study, if the position of the regulating
knob would not exceed a value of 70 %. For the experiments of this

study, the position of the regulating knob was in one case at 71 %,

otherwise it was lower. The deviation of the control measurements from

the originally calibrated feeding rates is estimated to be less than a

few percent.
 -
90 -

90

80 / /
r
if
*y

70
£
/ 7 y

1
Q
O
60

/
Je

Ol
c-

IS 50

\ 1\
a
Ol
tu
V.

o 40

//
c
o
^

ui
o 30
U.

20
///
10
10 20 30

GB lllU [kg/.]

Fig. 3.6 : Resulting calibration lines of the sediment feeding

for bed material IV of Smart and Jaggi and of

system no.

Hanger; the stars indicate control measurements of this

study.

3.5.4 Fluid velocity

In a sediment transporting flow at steep slopes the solids occupy a

considerable of the flow cross-section; the heavier particles do

part

not move with the same velocity as the carrier fluid except for debris

flows. Therefore the fluid velocity cannot be determined from a know¬

ledge of the mixture flow depth and the flow rate alone.
 -
91 -

A method to measure the fluid velocity in such a two phase flow is

the salt-velocity technique (see for example Davies and Jaggi, 1981). It

was already employed by Smart and Jaggi (1983). Also in this study three

pairs of electrodes were fixed on the flume walls (s.Fig. 3.7). By in¬

jecting a slug of salt solution the conductivity increase due to passage

of the salt cloud can be recorded at each measuring cross-section. The

conductivity readings were transmitted via an analog/digital converter

to a personal computer where they were stored and analysed.

detection of bed level

Fig. 3.7 : Measurement installations along experimental flume to

monitor fluid velocity V„), fluid surface level and

(V1,
bed level (h. and h, are the mixture flow depths); Al,
A2, A3 and Bl, B2 refer to locations of measuring cross-

sections.

Fig. 3.8 shows three different types of small tubes which were used

for an optimal injection of the salt solution into the flow at the top

of the flume. At the lower clay concentrations of the suspension the

tracer was injected from a glass bottle, in which a higher pressure was

produced manually. At higher clay concentrations above about C, = 10 %

this method did no longer provide sufficient mixing of the salt solution

with the fluid. Therefore the glass bottle was replaced by a steel tank
 -
92 -

which allowed to use higher pressures. With pressures between 1.5 and

2.5 bar the tracer could then be injected much faster too, and mixing
was sufficient again.

V=^

Fig. 3.8 : Three types of tubes used to inject the salt solution

into the flowing clay suspension.

If the salt solution is well mixed with the flow, the recorded con¬

ductivity increase is representative for the whole measuring cross-

section. Then the average fluid velocity can be calculated as the

travelling time of the salt cloud between two cross-sections divided by

the corresponding interdistance of the electrode pairs. Having installed

three electrode pairs, the velocity in the upper and lower reach of the

flume could be determined.

The sampling frequency of the conductivity readings was 33 Hz. This

is clearly above the frequency of the fluid surface undulations, which

were estimated to be between 1 and 10 Hz (with wave amplitudes of up to

about 1 cm). An example of the conductivity readings is shown in Fig.

3.9.

The calculation of the travelling time of the salt cloud is illus¬

trated in Fig. 3.10. The starting point (t.) of the conductivity in¬

crease is generally clearly defined; the determination of an end point

(t_) may be more difficult because the tail of the curve can be quite
long. To separate the base conductivity signal from the peak due to the
 -
93 -

salt tracer, a base line had to be determined. Similar to the procedure

of Smart and Jaggi, the mean value of all conductivity readings was

calculated and multiplied by a certain factor which was found from ex¬

perience. A moving average value was then used to compute the starting

and end point of the passing salt cloud.

EC
Al

jH/S/tAjvJAT

A2

Base conductivity level

!'^^vV^Ji^^

V,= 177 m/s V, =180 m/s V, *1 79 m/s

Method
V2 = 175 m/s V2 = 17S m/s V2-1 76 m/s

Fig. 3.9 : Example of salt velocity measurement. Conductivity rea¬

dings (EC = electric conductivity) in function of time

(t) are shown for the three measuring cross-sections (Al,

A2 and A3). The average fluid velocity in the upper (V^
and lower flume reach calculated by the three
(V,) was

methods a, b, and c. (The experimental conditions for

this run were: S = 20 %, Q = 10 L/s, p = 1.08 g/cm3; no

sediment transport.)

Three different methods were employed to find a reference time which

is representative for the passage of the tracer (s. Fig. 3.10). The

first method (a) is the same as the one used by Smart and Jaggi,
 -
94 -

determining the centroids of the whole area under the peak; Fig. 3.10a

shows the line through the centroid defining the time t . With the

second method (b) the reference time (t, ) is determined by the line

dividing the area under the conductivity curve into two equal halves, s.

Fig. 3.10b. The third method (c) is one described by Cao (1985): The

reference time (t ) is computed for the centroid of the shaded area in

Fig. 3.10c; the shaded area is calculated via an intermediate

conductivity value (EC„,) which is a function of the difference between

the base conductivity level (EC,) and the peak conductivity level (ECp).

Line through centroid

of shaded area

Base conductivity level

"WyyKT

la)

Line dividing shaded area

into two equal halves

.
Line through centroid
1 ^^ of shaded area

Fig. 3.10 : Three methods to determine the reference time (t , t, ,

t ) for calculating the travelling time of the salt

cloud, using the salt velocity technique (EC = electric

conductivity).
 -
95 -

From a knowledge of the flow rate and the measured flow depth for a

flow over the fixed bed without any sediment transport, the average

also be calculated; it is denoted by This

fluid velocity could Vh-
allowed an independent check of the above three methods to be made; the

corresponding velocities refered by Fig- 3.11 shows the

are to
^mes-
ratio V /V, in function of the relative depth for all three methods,
mes h

'mes

Vh

I OS

o 0
0 o fl
.<•/. 0
0 0
o •
A
b
s
-iV. 0

o 8
*

8
<s

09 Method applied a b C

Symbol D o

Fig. 3.11 : Ratio of velocity measured with salt tracer technique to

the one obtained from the flow rate and flow depth mea¬

surement (V /V,), in function of the relative depth

mes h

(h/d ); the three different methods (a, b, c) are com-

m

pared that were used to determine V from the con-

mes

ductivity readings.

It can be seen from Fig. 3.11 that the salt velocity technique tends
to give too lower velocities than the depth-discharge method, for rela¬

tive depths values less than about 4. (In the sediment transport tests
 -
96 -

the values h /d were always above 4, so that this peculiarity did not

affect the corresponding velocity measurements.) A comparison of the

three analyzing methods shows that method c seems to give slightly too

high velocities while methods a and b yield comparable values. A very

similar picture regarding the average velocities determined at lover

relative depths was also obtained by Bathurst et al. (1984), s. Fig.
3.12. It is difficult to decide whether the salt-velocity technique or

the depth-discharge method is more reliable in this region; the deter¬

mination of the flow depth gets more uncertain and three-dimensional

effects become more important with decreasing h/d values.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 K 15 16 17 18 19 20

Ratio of depth to sediment size d/Ds0

Fig. 3.12 : Figure taken from a study by Bathurst et al. (1984)

showing the ratio of velocity using the salt tracer

method to velocity measured using the point gauge method,

in function of the ratio of depth to sediment size,

h/drf,; data are for flows without sediment transport

(sediment I, II and III) or with low rates of sediment

transport (sediment III only); calibration curves (solid

lines) were fitted by eye.

Because of the relatively strong fluctuations in the conductivity

signal (s. Fig. 3.9), the velocities for the upper flume reach (V,) and
 -
97 -

the lower reach differed sometimes by than 5 X. During the

for (V,) more

analysis it was decided to use only those measurements with (V.-V,)/V, <

5 % for the determination of a mean value. The mean value for each test

was generally determined from a minimum of 7 individual velocity mea¬

surements.

Having found that method (a) gave the largest number of useful

measurements under all flow conditions, it was selected for the final

analysis of the experiments.

Because of corrosion problems in the pipe system it was desirable not

to use sodium chloride as tracer salt. Also, the applied salt should
influence the rheological properties of the clay suspension as little as

possible. One possible problem is that cations in the clay structure may

be replaced by ions of the added salt. According to Grim (1968) the

replacing power of Na is much weaker than that for other common

cations. It was therefore decided to use sodium acetate as tracer salt.

A few viscometric measurements were carried out to check the effect of

the added salt on the Bingham parameters of the clay suspension. The
results indicated that the rheological parameters did not change noti-

cably if the amount of sodium acetate was kept below about 20 % by

weight of clay in the suspension.

3.5.5 Flow depth

Smart and Jaggi (1983) determined the mixture flow depth (H ,

s. Fig.

3.14) by visual observation of the flow through the perspex flume wall.

By using a muddy clay suspension in this study, the motion of the grains
could no longer be visually detected. Furthermore it was desirable to

have an automatic recording and display of the flow depth on a personal

computer, so as to have a faster and more accurate method to find the

equilibrium transport conditions.

Two kinds of measurements were required: to record the fluid surface

level and, in addition, to determine the bed level in the case of the

sediment transport tests.

Two ultrasonic distance measuring devices were installed, one at an

upstream and one at a downstream cross-section 3.5 m apart (s. Fig.

 - 98 -

3.7). They were operated at a measuring frequency of about 30 Hz, which

was found to be sufficient for fluctuations of fluid surface undulations

with a frequency of about 1 to 10 Hz; the sampling frequency was about

40 Hz. The instruments gave a digital distance reading with a resolution

of 1 mm. The air temperature was recorded before each series of measure¬

ments, and a numerical correction was applied for the change of sound

velocity with temperature. The fluid surface level was measured during
five to seven seconds, before an average value was displayed and stored

on the computer.

In the case of tests without any sediment transport, the flow depth
(H) could then be determined. The bed level of the fixed rough bed was

taken as the mean elevation of the roughness elements (grains) above the

flume bottom (s. Fig. 3) at the measuring cross-sections (Bl and B2, s.

Fig. 3.7). For a fluid flow over the fixed rough bed, the standard

deviation of repetitive fluid surface measurements (determined as an

average value over 5 second periods) was generally less than 0.5 mm

which gives an idea of how reproducible these measurements were. The

maximum amplitude of the surface undulations was about ±1 cm.

Yi + 1
f
Fig. 3.13 : Scematic sketch of flume cross-section illustrating the

method to determine the level of the fixed rough bed.

 -
99 -

In the sediment transport tests the thickness of the erodible but not

moving grain layer in the flume depends mainly on the height of the end

sill and to some extent also on the flow conditions. The bed level was

defined as the height above the flume bottom where the largest gradient
in grain velocity over depth could be detected. In transport tests with

clear water as a carrier fluid, it was observed that a rather marked

boundary exists between fast moving grains of the bed load transport
zone and the almost stationary gravel particles in the bottom layer
underneath.

The following technique was employed to detect the bed level: At the

same two measuring cross-sections (Bl and B2) where the ultrasonic
devices were installed, a number of "point" electrodes (screws) were

fixed at different levels above the flume bottom on both side walls,

each successive level (screw) being 5 mm higher (s. Fig. 3.14). During a

Steel bottom / Fixed, rough bed

Quasi stationary
-

grams
of erodable bed

Fig. 3.14 : Scematic sketch of flume cross-section illustrating the

method to determine the bed level, i.e. the boundary
between the fast moving transport layer and the quasi-
stationary zone of the erodable bed.
 -
100 -

sediment transport test, the electrical conductivity across the flume

(width) was recorded simultanously at six consecutive levels. Examples

of such measurements are presented in Fig. 3.15. The sampling frequency
was again about 40 Hz, and each measurement lasted about 7 seconds. At a

level where grains of the bed load layer were moving fast through the

cross-section, the conductivity signal showed marked fluctuations with a

frequency of a few Hz. On the other hand there were less frequent and
lower amplitude fluctuations (or almost none at all) in the conductivity
readings from a level where the bed was quasi-stationary.

Cross-section B1 Cross-section B2

Fig. 3.15 : Example of bed level measurement: Time series of electric

conductivity (=EC) readings are shown for the upper (Bl)

and lower (B2) cross-section of the flume; the numbers on

the left indicate the height (in cm) above the flume

bottom. (The experimental conditions for this run were: S

= 15 %, Q = 30 L/s, p = 1.23 g/cm3.) In this case, the

bed level Y' was taken as 7.0 cm.

First an attempt was made to determine the bed level Y£ by a numeri-

cal analysis of the conductivity readings for each sampled level (e.g.

counting number of fluctuations about the mean value, calculating the

sura of the absolute amplitude values with respect to the mean value, or

by using information from a Fourier transformation). The independent

determination of Y' was made by visual observation of the bed for tests
 -
101 -

with clear water as transporting fluid. The numerical analysis proved to

be successful for (only) about 90 X of the cases. So it was finally

decided to definitely determine the bed level by visual judgement of the
conductivity curves.

3.5.6 Fluid density

Prior to each experiment the density of the clay suspension was

determined by taking a sample of one liter from the flow at the end of

the flume. For the lower clay concentrations the density could be mea¬

sured with a hydrometer. Samples taken in periodic intervalls showed

almost no variation in the density over two or three hours. Therefore

the density was usually determined once before each test.

At clay concentrations above about 4 % by volume the shear strength

of the suspension was so large that the hydrometer no longer moved
freely in the fluid. Therefore the density had to be calculated by
measuring the volume and weight of a sample.

3.5.7 Rheological parameters

For each set of experimental runs (for which the fluid density re¬

mained practically constant), at least one sample (of 1 liter) of the

clay suspension was taken. The samples were then analysed with a vis¬

cometer, which allows to determine the shear stress in a fluid as a

function of the shear rate. An example of the measurements is shown in

Fig. 3.16.

The principle of a rotational viscometer is shown schematically in

Fig. 3.17. The fluid is sheared between two cylinders at a given shear

rate; from the applied torque, which is measured, the corresponding

shear stress can be calculated. In the present analysis, a "Rheomat 30"

(manufactured by Contraves AG) was used as a viscometer. With this

instrument the clay suspension was sheared in a 1.8 mm wide gap between

two concentric cylinders. The shear rate was increased continously and

automatically in one minute from 0 to 650 l/s and decreased at the same

rate. This comparatively short measuring time (the shortest one avai¬

lable on the instrument) avoided that the rheologic properties changed

 -
102 -

P 1 27 g/Cm3

24 00

22 00

4 000' —

2 000

ooooi"-l—I—I—I—I I 1 1 1 1 1 1—
0 20 40 60 BO 100 150 200 300 400 500 600 700

Shear rate
j± [|]

Fig. 3.16 : Example of rheologic behaviour of clay suspension ob¬

tained from measurement with a viscometer; the Bingham
shear stress x„ and the Bingham viscosity t|_ can be

determined from the above diagram.

due to the settling of clay particles during the analysis. If a sample

was left undisturbed for three or four minutes, a redetermination of the

rheological curve indicated already a decrease in the Bingham shear

stress x_ and in the Bingham viscosity <i_. With increasing shear rates,
measured shear stresses were generally higher than with decreasing shear
rates (s. also the two slightly shifted curves in Fig. 3.16). This be¬

haviour is commonly observed with clay suspensions, and it is associated

with the fact that the arrangement of the plate-like particles also
depends on the stress history.
 103

'Torque

Rotating inner

cylinder

Sheared fluid

Fig. 3.17 : Schematic cross section through rotational viscometer.

From the applied torque the shear stress can be calcu¬

lated which is necessary to maintain a given shear rate.

The straight line that approximates the true rheologic behaviour (s.

Fig. 3.16) was positioned so, that it matched the curve at shear rates

above about 150 l/s. For in this region the measured lines are already
rather straight; furthermore, it was not known a priori which range of

shear rates would be most representative for the experiments to be

performed.

3.6 Equilibrium transport conditions

The flume slope and the flow rate were set before starting each test.

The third independent parameter, the fluid density, was given by the
actual concentration of the clay suspension (in the flume) after

remixing the suspension in the circuit.

 -
104 -

In the case of the flow resistance experiments without any sediment

transport, about 12 to 20 sets of measurements were performed for each

test. A measurement consisted of a flow depth determination at two

cross-sections (Bl and B2) and of a velocity measurement for the upper

and lower reach of the flume and Fig. 3.7).

(V1 V,; s.

In the case of the sediment transport experiments, the bed level was

also determined, apart from recording the fluid surface level and

measuring the mean velocity.

Equilibrium transport conditions require that the flow of the sedi¬

ment-fluid mixture is steady and uniform. The following criteria were

used to check whether this requirement was satisfied: The fluid surface

slope should be approximately equal to the flume slope; the (mixture)

flow depth should be about the same at both measuring cross-sections

(implying that also the movable bed slope is equal to the flume slope);

and the velocities in the upper and lower flume reach should not differ

much from each other. The equilibrium transport conditions for each test

had to be found by trial and error. The sediment feeding rate was first

set arbitrarily or by trend extrapolation of previous results. If the

gravel input was too small, the movable bed was quickly scoured, begin¬
ning downstream of the adjustable roughness elements, and the erodable
bed was washed down the flume. On the other hand, if there was too much

sediment input, the bed aggraded and the bed slope increased. Once the

equilibrium conditions were established, the movable bed had reached a

quasi steady state, and a set of measurements could be made.

To determine average values for each test, only those measurements

were considered for which both the fluid surface and the bed level at

the upstream cross-section did not differ by more than about 10 mm from

their corresponding values at the downstream cross-section. In general,

the velocity measurements which were used in the final analysis did not

differ by more than 5 % between the upper and lower flume reach, except
for extreme flow conditions; with the highest clay concentrations,
successful application of the salt velocity technique became more and

more difficult.
 -
105 -

3.7 Reproduction of some tests of Smart and Jaeggi

In order to ensure that the sediment transport experiments performed

by Smart and Jaggi (1983) could serve as reference conditions, some of

their tests with bed material No. IV were reproduced with the slightly
modified experimental setup. The reproduction of some earlier experi¬

ments also allowed to check the new measuring techniques described

above.

The six tests which were reproduced are shown in Table 3.3 together
with the corresponding tests of Smart and Jaggi. The tests which were

chosen to cover mainly the higher transport domain of their test series.

Table 3.3 also includes the experimental values obtained and gives the

relative deviation between both sets of measurements.

meas. Smart/Jaggi meas. by author rel. deviation

h V Ah AV
S Q
s h V
GB *S
[%] [L/s] [kg/s] [cm] [m/s] [kg/s] [cm] [m/s] IX] m m

10 10 0.84 5.0 1.07 0.98 4.40 1.07 17 -12 0

15 10 2.30 4.8 1.18 2.78 4.85 1.17 21 1 9

15 15 4.08 6.0 1.57 4.34 6.22 1.54 6 4 -2

15 20 5.83 7.6 1.96 5.97 7.00 1.81 2 -8 -8

20 30 14.91 8.5 13.02 8.15 2.30 -13 -4

Table 3.3 : Comparison between selected tests of Smart/Jaggi (mat.

IV) and experiments repeated by the author.

Another possibility to compare the quality of the experimental re¬

sults, is to plot the measured values against the calculated ones (s.

Fig. 3.18). The bed load transport rate q„ was computed with a new

equation presented in section 4.6.1 (equ. 4.19), since the Smart/Jaggi

equation was found to underpredict the higher transport rates. Equation

(2.82) was used to calculate the mixture flow depth H (the fluid flow
 -
106 -

PBcate[m3/sm]

• a

•a

Smart/Jaggi »
i •
mat W
4 4

Tests repeated •

by author

0 02 0 03

cWa- [m3/sm]
Fig. 3.18 (a)

010

A
A * A

• A

:
i

* A
0 06
* A

Smart/Jaggi A

mat IV

1 A Tests repeated •

by author

0 04 0 06 0 08 010

[m]
Fig. 3.18 (b)
 -
107 -

*
1
1

• 4 4

" 4
4
A

A
A
•A

A
Smart/Jaggi 4
A
mat IV

Tests repeated •

by author
*

10 14 18 22 26

(c) Vmeas [m/s]

Fig. 3.18 : Comparison of tests repeated by the author with the

Smart/Jaggi tests with material IV, shown as calculated

against measured values: (a) bed load transport rates,

(b) mixture flow depth, (c) mean fluid (water) velocity.

depths as inferred from the flow rate and velocity measurements were

considered to be more reliable), and the mean water velocity V was

computed using the iterative procedure described in Smart and Jaggi

(1983).

It can be seen that the values measured by the author are generally
within about +/- 10 X of the calculated ones, and they indicate about
the same amount of scatter as the values obtained by Smart and Jaggi.
The measured velocities by the author are somewhat smaller than the ones

of Smart and Jaggi; this could be due to the fact that they let the salt

solution fall into the flow surface from a flat plate while the author

injected the solution in the middle of the flow cross section from one
 -
108 -

tube or several smaller tubes. Thus the mixing in the first case might
not have been sufficient (at higher velocities) to prevent that the

major part of the solution travelled with the faster moving upper fluid

layers. While the bed load transport rates obtained by the author are

close to the calculated ones, the (new) measured depth values appear to

be somewhat smaller than the calculated ones.

 -
109 -

4 RESULTS AND ANALYSIS

4.1 Methods used in data analysis

4.1.1 Correction for sidewall friction

The flow depth in the steep flume of 20 cm width varied between 2 cm

and 10 cm. It was therefore necessary to apply a sidewall correction to

account for that part of the energy loss which is due to friction of the

fluid along the flume walls.

A procedure originally proposed by H.A. Einstein (1936) was adapted

by Smart and Jaggi (1983) to determine the effect of the sidewall in¬

fluence. The flow cross section is divided into a part where the flow

acts on the bed, and into two parts where the flow acts on the side-

walls; the three areas are separated by straight lines (Fig. 4.1a).
Assuming equal mean velocities in all three subareas, the following
relationship can be used:

where R is the hydraulic radius belonging to one sidewall flow subarea

F .,
and k is the Strickler value characterising the flume wall mate-
w w

rial. The total area can be expressed by (Fig. 4.1b):

H-B (2
=
(hr B) +
Rw H) (4.2 )

where H denotes the measured flow depth, h is the reduced flow depth
corrected for sidewall influence, and B is the flume width. Combining
equs. (4.1) and (4.2), h is determined as:

hr -H -2
(^53>1'5 I
w
<4-3 >

The reduced flow rate,

qr> acting on the bed is calculated accordingly
 -
110 -

qr
=
q -
V (H -

hr) (4.4 )

(a) lb)
2F
F
'
F * '
w w w

/ \ \
t\
\
1 1
1

1
1
fr \ a:
Fr
1 \
1
1

,1
0
*
,
*_
1

Fig. 4.1 : Sidewall correction procedure applied to the steep flume

experiments. The trapezoidal subarea (a), which represents

the flow section acting on the bed, is replaced by a rec¬

tangular section (b).

Smart and Jaggi examined the influence of the choice of the flume

wall Strickler coefficient the difference

kw on between calculated and

measured bed load transport rates. Apart from their main experiments in

a flume of 20 cm width they also performed some comparative tests in a

narrower flume of 10 cm width. They concluded that the choice of the k

w

value was important for the 10 cm flume, but not important in the case

of the 20 cm flume. For their main experiments they chose a k value of

w
1/3
110 m /s. Since a similar flume (with 20 cm width) was used in the

present study, the same Strickler coefficient was adopted for the side-

wall correction procedure.

4.1.2 Regression analysis

A computer program was applied to perform a regression analysis with

the measured parameters or combinations thereof. The logarithmic values
of the considered parameters were used in the computational procedure of

the implemented standard multiple linear regression analysis. Informa¬

tion on the statistical procedure can be found in Draper and Smith

(1966).
 -
Ill -

To judge the quality of a regression equation, the same method was

applied as described in Smart and Jaggi (1983). The correlation coeffi¬

cient r is obtained as:

L(X -
I) a - X )
c c m m
(4 5 }

[r(xc -

Xc)2(XB Xn)2]0-5 -

where X denotes the calculated (predicted) value and X the measured

value; the bars indicate mean values over the total number of observa¬

tions, N. The standard error S_ is defined as the standard deviation

divided by the mean value:

2 0 5

^-0-5 (4.6 )
SEE
(N-i)0,5 t
m

It should be noted that r and S_ are calculated with linear, and not

with logarithmic values. Using these statistical parameters in the

analysis of the present experiments, the results were directly compa¬

rable to those obtained in the (reference) study of Smart and Jaggi.

4.2 Rheology of clay suspension

Before performing the main experiments a few preliminary test were

carried out in a smaller flume apparatus. One objective of these tests

was to check how well the Opalit clay suspension could be recirculated
in a flume system.

Samples of the clay suspension of various concentrations were taken

during these tests and analysed with a viscometer (as described in sec.

3.5.7). From the viscometric measurements the Bingham parameters x„ and

rw, were determined. The same procedure was applied to the samples ob¬

tained during the experiments in the main flume. The relation between

the Bingham parameters and the clay concentration of the suspension is

presented in Fig. 4.2 and 4.3.

 -
112

TB[N/m>]
40 -*

Author's Opalit clay suspension

* in mam flume
i
+ in smaller flume

20
-—kaoline suspension (Wan, 1982) *

kaolinite suspension (Higgins et al, 198O)

+

T
Debris flow slurries
10
0 Mount St Helens (Higgins etal, 1983) *
8
Jangj a Ravine (Chen, 1986b) *

t
1
1°
1
Y 1
1 T
*
/

/
/
10
/
/
0.8
r
/
06
+

*
*
OX

02
+

01 —n

0.01 002 004 006 008 010 0.20 030

Cf

Fig. 4.2 : Relation between Bingham yield stress x_ and volume con-
o

centration C, of fine material suspensions; measurements

of this study are shown together with data from other

sources.
 113

Hb [°ps1

60
T

40

20

T
+

to
V •
B
/'*
6
y •

y
s Author's Opalit clay suspension
a in main ffume
4
%+ + in smaller flume

*
kaoline suspension (Wan, 1982)

kaolinite suspension (Higgins et al, 1983)

+

+
Debris flow slurries
+
0 Mount St Helens (Higgins at al, 1983)

o Jangjia Ravine (Chen, 1986b)

i

01 02 0.3

Fig. 4.3 : Relation between Bingham viscosity h_ and volume concen¬

tration Cf of fine material suspensions; measurements of

this study are shown together with data from other

sources.

It can be seen from Fig. 4.2 that the samples from the main flume

apparatus show a larger x_ value than those from the smaller flume for a

given clay concentration. This fact can be explained by the relatively

much larger fluid volume present in the main flume system; a bigger
portion of the added clay material settled out in the dead zones of the

system and was not resuspended during the experiments. Thus the samples
from the main flume contained a larger fraction of the very fine par¬

ticles of the Opalit clay.

 -
114 -

Fig. 4.2 further shows that in this study, x_ is approximately pro-

D

portional to
C|; this proportionality is also indicated by the dashed
line, which represents measurements made by Wan (1982).

The data points from the main flume apparatus in Fig. 4.3 seem to

indicate a discontinuity at a clay concentration of about 15%. A break

of a few months occurred at this concentration level because a special

slurry pump had to be installed in the system. At the end of this period
a substantial part of the coarser clay particles had settled into the
dead zones, resulting in a new suspension containing more very fine

particles than before. For a clay concentration up to 15%, the data from

the two flume systems show that the Bingham viscosity of the Opalit
suspension seems to be less sensitive to particle composition than the

Bingham yield stress.

In both figures, the data points representing samples of mudflow

material from the Mount St. Helens lie relatively close to values ob¬

tained in experimental studies. Regarding the samples from debris flow

slurries in the Jangjla Ravine (China), the Bingham yield stress is

clearly lower than for the other data points shown.

It may be noted that for the Bingham type approximation, the straight
line in the rheological diagrams (e.g. Fig. 3.16) was fitted at shear

rates above about 100 s~ for most of the data shown in the above

figures; no information in this respect was available for the Jangjla

Ravine data.

4.3 Flow resistance of clay suspension on a fixed rough bed, without

sediment transport

In order to examine the flow behaviour of the Opalit suspension

without any interference by transported bed load grains, velocity- and
depth- measurements were made for the flow of the suspension over a

fixed, rough flume bed. The rough bed was made up of the same gravel as

that used in the sediment transport tests (s. section 3.4 for grain size

characteristics). For each of the examined clay concentration levels,

 -
115 -

Ci, twelve combinations of flow rate and flume slope were considered. A

list with all performed measurements is given in Appendix I.

Summarised, the fluid flow rate was varied between 10 l/s and 40 l/s,
and the slope was set between 5% and 20%. The range of the examined clay
concentration levels is shown in Table 4.1, together with the correspon¬

ding Bingham parameters; the figures represent mean values for a given
clay concentration level. Also shown is the effective viscosity (which
depends on the flow conditions).

Symbol/Ci P [g/cm3] Cf [X] Xg [N/m*] \ [cps] Pe2 [cps]

0.998 0.0 0.0 1.02 1

A
H20
* CI 1.078 4.7 0.44 1.93 5 -
10

C2 1.165 10.0 2.82 3.60 25 -

55

+ C3 1.238 14.4 7.29 5.92 60 -

140

* C4 1.324 19.6 23.6 20.0 180 -

400

Y C5 1.365 22.1 40.8 34.3 250 -

1800

Table 4.1 : Experimental range of clay concentration levels (for

experiments without sediment transport), shown with the
corresponding Bingham parameters x_ and »i_. The effective

viscosity u
,
was calculated according to equ. (2.40); it

depends on the flow conditions.

It may be mentioned that the velocity measurement by the salt tracer

technique became unreliable for flows which were at transition between

the turbulent and laminar regimes (some tests at the levels C4 and C3);
this was indicated by relatively large differences in the values V and

q/H. At the highest clay concentration level, application of the salt

velocity technique was no longer possible. For all these experiments the
velocity determined as q/H was used to calculate f, Re, and Re_ (s. also
Appendix I).
 -
116 -

Jaggi (1983) proposed a new flow resistance equation for a sediment

transporting flow over a movable bed at smaller relative depths h/d-0,

because many conventional formulae tend to overpredict the velocities in
the range 5 < < 20. equation is given
h/d„0 His as:

v
o2.5s n „»„/•
"l*1 mO-5 t ,12.27 h, ,,
,
,
—0),
= -

^ [1 exp(- (4.7)

ln(^^)

where the coefficients p\ and cu depend on the grain size distribution,

the packing and the shape of the bed material. For their steep channel

bed load transport tests, Smart and Jaggi (1983) determined a. = 0.05

and 1.5.
Bx =

The results of the flow resistance measurements with the clay suspen¬

sion are presented in Fig 4.4 in terms of the resistance coefficient c =

V/v* as a function of the relative flow depth h/d„n; note that the flow

depth corrected for sidewall influence, h ,

was used to determine v* and

h/dgfl for the experimental data. Equ. (4E4) and the Nikuradse equation
for open channel flow, equ. (2.53), are included for comparison.

It can be seen from Fig. 4.4 that most of the data points show no

significant change in flow resistance with increasing clay concentration

level Ci. The data points with 5% slope and the largest flow rate (or
relative depth) indicate some variation of c. At the highest clay con¬

centration level (C5), the experimental results show a trend for c, at a

given slope, to become independent of the parameter h/d-.; this is an

indication that the roughness elements no longer dominate the flow be¬
haviour.

The comparison with equ. (4.7) shows that the flow resistance (at

steep slopes) is smaller in a flow over a fixed bed than in a flow over

a mobile bed. This conclusion is in agreement with a study by Zagni and

Smith (1976) who found that the friction factor for turbulent flow over

a permeable bed is higher (by about a factor 1.5 to 2) than for a flow

over an impermeable bed with the same roughness elements. Further, the

macrostructure of the particles on the fixed bed might have been still
somewhat more regular than the natural arrangement on a mobile bed.
 117 -

Symbol Ci »

a H,0

* CI

o C2

+ C3

A
V
ft
o C4
s"'
s

r C5

\pj*y s

/
/
ft

V_
^^y&^yx
'

v*
All

«£''"
/ \\ /z
/^
1
h yy& I
y^y^
yy^

_h_
d90

Fig. 4.4 : Flow resistance measurements of clay suspension flows over

a fixed rough bed without sediment transport, in terms of

the resistance coefficient c = V/v* as a function of the relative

flow depth h/dqo; data points with equal slope and for a given clay
concentration level are connected by straight lines. The Nikuradse

k d„») and the equation used by Smart and Jaggi

equation (with =

(with ou« 0.05 and P.= 1.5) are shown for comparison.
(a)

Fig. 4.5 : Flow resistance measurements of clay suspension flows on a

fixed rough bed without sediment transport, in terms of

the friction factor f vs. the Reynolds number

Re?; data points with
equal slope are connected by straight lines. The smooth lines

represent f values calculated with the Colebrook equation (with k =

Also shown is the relationship for laminar Newtonian flow, f

dg0). =

96/Re, in a rectangular channel, (a) Data points for H„0, C2 and C4,
(b) data points for CI, C3 and C5.
 -
119 -

I
/(
// /
^/
/ '//
/// '/
/// / s
>

/A '/
M'
«y v/// l'>
*' / /' i
/ /f/ A
/ //^ / /
/ u

/// /
^
j / / / y
y

<•>' *//f//A y

>/J///.1
V
t yj<rf /
'
/j£f7

*\i y///yf
X • ''yy\
*

'
^ <* »*

io B * n w — Soooooo o o
OOOO O OOOOOOOO O O

Fig. 4.5 (b)

 -
120 -

To illustrate the effect of the increasing viscosity on the flow

resistance with increasing clay concentration, the experimental results

are presented in a Hoody type diagram. In Fig. 4.5 (a) and (b) the

friction factor f is plotted as a function of the Reynolds number Re.

(as defined by equ. 2.42a). Again the reduced flow depth h (corrected
for sidewall influence) was used to determine f and Re, for the

experimental data points.

The flow resistance values predicted by the Colebrook equation (2.52)

are also shown in Fig. 4.5. For a given clay concentration level, the

effective viscosity u » was determined with the Bingham parameters of

Table 4.1, which allowed one to calculate the Reynolds number Re.. Not

much information is available to decide which representative grain size

should be taken for k at small relative flow depths. Kamphuis (1974)

performed flow resistance measurements with a fixed rough bed, and
determined k for different values of h/d<._. He suggested to use k =

2d_. for larger relative depths (h/d.. >= 15); a few experimental points
in the range 2 < h/d.. < 10 indicate that k is approximately equal to

value of k used calculate f with

0.5d9Q — 1.5d._. A mean =
d„0 was to

the Colebrook equation in Fig. 4.5.

According to Fig. 4.5 the transition from turbulent to laminar flow

occurs at a critical Reynolds number Re, between about 500 and 1500. The
change in Re. with increasing clay concentration seems to be reasonably
well reproduced by using the Colebrook equation in combination with the
effective viscosity u
,.

To determine the limit below which viscous effects begin to take

place, the experimental results are presented in another form in Fig.

4.6, plotting the values [l/4f -

2-log(h_/d0_)] = B' on the ordinate and

* *
Re. on the abscissa (analogous to Fig. 2.2). It is seen that for Re. >=

10, B' is approximately constant implying that no viscosity influence

can be detected; in this region the following relation can be given:

hr
1
4 -
2 log (-rM = 2.4 (4.8 )
it
a90
 -
121 -

The above equation may be transformed into:

1 12hr
2 log (4.9 )
Tfa (5^ra^>

Comparing equ. (4.9) with the Nikuradse equation (2.53), it follows that
a mean value of k = 0.76-d.. would best describe the conditions of the

clay suspension experiments.

It is further noted from Fig. 4.6 that B' does not change below Re, »

54 (corresponding to v*k /u
,
= 70). This value marks the upper limit of

the hydraulically transitional region in Newtonian flows; in this

region, an increase of B' occurs for flows over uniform sand roughness,
while B' should gradually decrease in the case of a nonuniform sand

roughness (Rouse, 1960). There are only two data points for the clay
concentration level C4 which may indicate an increase in B'.
s

• «
* *
* * m
*+ +

>
*
•• + ++
*
*
4
4*A
*
>

r
Y
*
1 i
Y

Symbol Ci

H20
* C1
C2
+ C3
o ** 1 * C4
II ii Y C5

0) of
tr cn ]
3 4
log (Re,*)

Fig. 4.6 : Flow resistance data shown in terms of the parameters B'

=
[1/-Jf -
2-log(h /d_.)J and Re,, for the clay suspension

flows over a fixed rough bed.

 -
122 -

The transition from turbulent to laminar flow can also be judged

using the criterion for Bingham fluids developed by Hanks (1963) for
pipe flow and adopted by Naik (1983) for open channel flow. The critical
conditions at transition were calculated using equs. (2.43) and (2.44);
the corresponding curve is shown in Fig. 4,7, together with the experi¬
mental results for the three higher clay concentration levels C3, C4 and
C5 (using the reduced flow depth h ). In Fig. 4.7, all those experiments
which indicated turbulent flow are grouped together, while those which

were laminar or at transition are marked with a different symbol. How¬

ever, it was difficult to judge from visual observation whether a flow

was already laminar or only in a transitional stage. At the more viscous

flows, the formation of a plug flow could be observed in the middle of

the flume, the plug becoming broader with decreasing turbulence inten¬
sity. In these cases, the flow behaviour was classified as transitional

in Fig. 4.7, although this may not be evident from Fig. 4.5. Data from

preliminary experiments in a smaller flume (with 5 cm width) is also

included in Fig. 4.7.

The experimental data shown in Fig. 4.7 seems to support the theore¬

tical criterion for the transition between laminar and turbulent flow of

a Bingham fluid in an open channel. It may be noted that the data points
marked "laminar or transition" in Fig. 4.7 all have a Reynolds number

Re, smaller than approximately 1500 which is about the upper limit for

the critical value of Re, as predicted from Fig. 4F4. Thus it appears

justified to use the effective viscosity u - together with conventional

formulae developed for Newtonian fluids.
 123

j;os

V
9
oB»^
T
^
V

Re„
0
y*
v » v-f

+
.* 7
+
+
y? 7 1 r/l eoretical curve)

*
/ "I
^^-^
•
J it \
< i *

laminar or turbulent
Flow behaviour
transition

f C5 0

Data from J ci m 0

main flume 1 ._
V

Data from « +

smaller flume
1 1 1 I I II

(0
=
)0o ;o
He

Fig. 4.7 : Experimental results obtained at the higher concentration

levels, shown in terms of the Bingham Reynolds number Re_

B
and the Hedstroem number He; the data points are grouped
in tests with turbulent flow, and in those with

transitional or laminar flow. The theoretical line of

transition was calculated with equs. (2.43) and (2.44).

4.4 Bed load transport experiments

Similar to the tests without bed load transport on a fixed rough bed,
experiments were performed at five different clay concentration levels
(Ci). At each level, the fluid flow rate was varied between 10 and 30

L/s, and the slope was set between 7% and 20%. A list with all performed
 -
124 -

measurements is given in Appendix II. The range of the examined clay

concentration levels is shown in Table 4.2, together with the correspon¬
ding Bingham parameters x„ and ru and the effective viscosity u
,.

Symbol/Ci [g/cm3] [%]

p
Cf xB [N/m*] •Tg [cps] Ue2 tcps]

A 0.998 0.0
H20 0.0 1.02 1

* CI 1.072- 4.4- 0.43- 1.87- 10 -

15
1.096 5.8 0.67 2.17

C2 1.141- 8.6- 2.20- 3.00- 35 -

65
1.160 9.7 2.77 3.33

+ C3 1.201- 12.2- 4.31- 4.35- 100 -

200
1.246 14.9 8.55 6.74

* C4 1.257- 15.6- 12.8- 11.2- 200 -

450
1.293 17.8 20.1 16.2

Y C5 1.356- 21.6- 33.6- 28.8- 800 -

1000
1.363 22.0 40.8 34.3

Table 4.2 : Experimental range of clay concentration levels Ci for

the sediment transport tests, shown with corresponding

Bingham parameters x„ and Iru. The effective viscosity u
,
was calculated according to equ. (2.40). The symbols of
this table are also used in the following figures.

The experimental results clearly indicate a change in the bed load

transport rate with increasing clay concentration of the suspension.

Fig. 4.8 shows the change of the ratio
qB/qB H20
with increasing fluid

density p. For a given slope and flow rate, q_ increases with p up to

the concentration level C3 or C4, and then decreases again at higher

fluid densities. Vhile an increase in q may be expected with a decrease

in the solid-fluid density ratio s, the decrease in q„ with increasing

clay concentration is not likely to be associated with the changing
fluid density; this decrease is more likely to be a viscosity effect.
 -
125 -

*
6
+
ft

3K
*
!>
*

yt
** * +

y
*

Y
08H
1 0 1 1 1 2 13 1 4
r , «,
P [g/cm3]

Fig. 4.8 : Ratio of the bed load transport rate measured in the clay

suspension to the value obtained in clear water,

q./q„ as a function of the density p of the clay

„,_,

suspension.

In Fig. 4.9 the ratio q„/q is plotted against the particle

B MB,H20
Reynolds number Re,, which accounts for the change in fluid viscosity

with increasing clay concentration; is defined v*d.„p/u It

Re, as
,.

can be observed that there is an increase in q_ with decreasing Re, (or

increasing clay concentration) down critical value of below

to a
Re,,
which the bed load transport capacity of the flow clearly decreases
again. At critical conditions the value of Re, is about 10 to 15.

For open channel flow, the thickness of the laminar (viscous) sub¬

layer, i, is given as (Yalin, 1977): 8 = 11.6v/v* =

11.6(r)/p)/v*. This

implies that in a flow with Re =12, the laminar sublayer is of the

same magnitude as the equivalent sand roughness k . It is assumed that

the relationship for S is also valid in the case of the clay suspension,
using the effective viscosity u
,.
With regard to the critical value of

Re, shown in Fig. 4.9, this means that the thickness of the laminar
 -
126 -

sublayer is of the order of the grain size of the transported bed ma¬

terial. Thus it appears that once the flow around the grains becomes
laminar, the bed load transport capacity starts to decrease, for other¬

wise equal flow conditions.

qB 4

^B,H20
>

ff

i
*
«
2 0-1
+
6

4
*
* <&

+ *

*
Y
1
+ * g

31
Y
X
Y

0.8- . _

1.0 10.0 15
100.0 1000.0
Re*

Fig. 4.9 : Ratio of the bed load transport rate measured in the clay
suspension to the value obtained in clear water,
*
function of the grain Reynolds number Re,.
VqB,H20 ,
as a

Also other experimental studies show an increase in q_ with a de¬

crease in the density ratio s (s. sec. 2.3.1.2); in these studies the

viscosity of the fluid did not change. It is therefore reasonable to

assume that at least a part, if not all of the increase in the ratio

lo^ln
dm
ls due to the increasing density of the clay suspension. Some
B B,HZ0
additional tests were carried out in a smaller flume apparatus (with a

width of 5 cm), using a cellulose solution so as to change only the

viscosity but not the density of the fluid. The results of these quali¬
tative experiments indicated decrease of the ratio with
a
1B/<JB s2n
increasing fluid viscosity.
 -
127 -

The change in flow resistance with increasing clay concentration is

depicted in Fig. 4.10, where the friction factor f is plotted against

Reynolds number Both f and determined with the

the
Re,. Re, were

measured fluid velocity , V, and the mixture flow depth corrected for

sidewall influence, h . A slight increase in f is observed with

r, m

increasing clay concentration (i.e. decreasing Re,), followed by a

slight decrease. This behaviour is qualitatively similar to the change

in q_ with increasing clay concentration levels. A decrease in f,

however, already at the level C3, while still increases at the

occurs
qR
same clay concentration level.

0 80 I
\
0 70 \ 96
,-
0 60

0 50
\ Re2
1
\
^
>s

K T
•-* \ S-20% »>
0 40 \
V^ \ N
*

\ \\
15%

I
V \ \
030 \
\ \
0 20
\
^^ k
K> \\ \^ \
^
\
\ \
^ \

\
7%»v\
*c
010 \
0 09 \
-

0 08
o2 103 1 3' 1 0B 106

Fig. 4.10: Flow resistance measurements of clay suspension flows with

bed load transport. The friction factor f is shown as a

function of the Reynolds number data points with

Re,;
equal slope are connected by straight lines. Also given is

the relationship for laminar Newtonian flow, f =

96/Re, in

a rectangular channel.

4.5 Case I: Thickness of laminar sublayer is smaller than grain size

In the following analysis only those experiments were considered for

which an increase in q was observed with respect to the q_ value mea¬

ls B

sured at the next lower clay concentration level (with a lower fluid

density). Thus 12 individual experiments were available for each of the

 -
128 -

levels H20 (data taken from Smart and Jaggi), CI, C2; 10 tests at the

level C3; and 4 experiments at the level C4 (s. also Appendix II). This

resulted in a total number of 50 tests which were treated as a homo¬

geneous data group. This data set is labelled RI in the following

sections.

It can be inferred from Fig. 4.10 that there was no significant

change in flow resistance with increasing clay concentration, despite
the very pronounced change in the bed load transport rate. This is also

confirmed by Fig. 4.11a and b, where the change of the corrected mixture
flow depth, h ,
and of the fluid velocity, V, respectively, are shown
r, m

as a function of the density factor (s-1). There is a tendency for only

a slight increase in both h and V with increasing suspension concen¬

tration (i.e. decreasing values of s-1).

* +

D +

+
•
+ «.

X
*
X *

+
»
X

*
+

+
*
+ X
*
** X
+ X X

*
X

07-
11 1 2 1 3 1 4 15 16
(a) 1 0
(S-1)

Fig. 4.11: Variation of flow parameters measured for the clay

suspension with bed load transport, as a function of the

density factor (s-1): (a) ratio of the corrected mixture

flow depth to the corresponding clear water value,
h,- /h mn>
and (b) ratio of the the fluid velocity to
r,m r,m,nzu
the corresponding clear water value, V/V.
H20*
 129

1 J-

•
+
•
+
X
+

k
+
*1 **"
X

"
*
+
«
*
»

07-
1 0 11 12 13 14 15 16

(S-1)

Fig. 4.11 (b)

« +

E>

»
* <>

+ *
«
*
X *

X X +
X

*.
X

0 8-
08 09 1 0 1 1 1 2 13

Fig. 4.12: Ratio of the bed load transport rate measured in the clay
suspension to the value obtained in clear water,

function of the ratio of the corresponding

qB/qB J,,.)
as a

(corrected) mixture flow depths,

'
h /h „.„.

r,m r,m,H20
 -
130 -

Considering Fig. 4.11a, it may be expected that the change in q_ is

not associated with a change in the flow depth. This is in fact

confirmed by Fig. 4.12 which shows the ratio q,,/q,, „,„

as a function of
o o,HZU
the ratio h /h „„.; no correlation can be recognised between the
r,m r,m,HzU

two parameters.

In previous studies on bed load transport at steep slopes it was

found that is proportional to the flow rate q (s. sec. 2.3). In Fig.
qR
4.13 the ratio q_/q is plotted against the density factor (s-1) for the
experimental results of this study; q denotes the reduced flow rate

corrected for sidewall influence. It is seen that there is generally a

stronger than a linear dependence of q„/q on the factor (s-1). It is

d r

further observed that the slope is an additional parameter determining

the value of the ratio q„/q .

a r

A regression analysis was performed to obtain an equation of the

form:

qB
=
B2 q*1 Se2 (s-l)e3 (4.10)

where is The resulting exponent el close to 1.0,

B, a constant. was very

confirming the findings of other studies. It was decided to fix the

value el = 1.0, with the advantage that equ.(4.10) is then dimensionally

correct. Another regression analysis yielded the equation:

25.2 .2.3 rz = 0.964 ,.

....

S
qB .
.,2.0 qr Sp = 14.9%
(4-U)

(s-1) E

A comparison between the measured q„ values and those calculated by equ.

(4.11) is shown in Fig. 4.14.

To check the influence of neglecting the conditions at initiation of

motion, a critical slope S was determined similarly as described by

Smart and Jaggi (1983):

Scr "

9c <S-X) dm /
hm <4-12>

They calculated the Shields parameter 9 according to a procedure

developed by Iwagaki and Tsuchiya (1959), which basically accounts for

the change in 8 at particle Reynolds numbers below about 400; at higher

 -
131 -

S=20%

S=15%

}S=10%
} S=7%

Fig. 4.13: Ratio of bed load transport rate to corrected fluid dis¬

charge, in function of the density factor (s-1),

qB/Qri
showing the slope S as additional parameter; data points
with equal flow rate in the flume (Q) and equal slope are

connected by a straight lines.

 -
132 -

0 04 » /

20
qrS23(s
\
25 2 1
qB
-
=

*
/
qBca,Jm3/sm]

0 03 *
_.

+ / »

*/
/ *
> /

0 02

A
/
x/ A

/*

Clay
+
concentration
0 01
A i Data
level

Smart/Jaggi & (H20)

Rickenmann x C1
C2
*v
*+ C3
C4

0 00
0 00 0 01 0 02 0 03 0 04

qBmeas[m3/Sm]

Fig. 4.14: Comparison between measured bed load transport rates,

and those calculated, using equ.

qBqB,meameas.' qB c lc>»
(4.11).

values of Re the Shields parameter has a constant value of about 0.05,

it
while 9 assumes a minimum of about 0.03 at an Re value of about 10.

Daido (1971) extended the above theory to incorporate the effects of a

Bingham fluid (clay suspension) on the critical dimensionless shear

stress at initiation of motion, 9 . In the present analysis, Daido's

(1971) procedure was used to determine appropriate 9 values for the

experimental conditions. These 9 values were then reduced for the slope
effect with the equation of Stevens et al. (1976) (equ. 2.105) to obtain

the 6' values which were used in equ. (4.12) to compute S .

 -
133 -

Using now the term (S -

S ) instead of S alone, another regression

analysis resulted in the equation:

16.7 ,. .
.2.1 r2 = 0.959 ,, ,,.

% qr Scr> (A-13)
"
"
(S
15.9%
1.6
SE =

The quality of the correlation is practically the same for equ. (4.13)
and equ. (4.11). However, the introduction of the critical slope S

changes the exponent of the density factor (s-1) markedly. It is noted

that for most of the experiments S > 7-S ,

i.e. the flow conditions

were clearly above critical conditions for beginning of transport.

According to dimensional analysis (Yalin, 1977), bed load transport

relationships may be expressed in terms of the dimensionless parameters

*_ and 6, and possibly other parameters. In the present study, the

D

Froude number Fr and the density factor (s-1) were chosen as additional

dimensionless parameters. In the bed load transport equation of Smart

'

and Jaggi, equ. (2.83), the factor cS appears, which is very close to

(1989) showed that the density factor (s-1) should also

*

Fr = cS . Low

appear in equ. (2.83) as additional parameter; this may also be expected

when comparing equ. (4.13) with the Smart/Jaggi equ. (2.82) or (2.82b).
Therefore the following alternative parameter set was used:

*B =
*(e, 9c, Fr, s-1) (4.14)

A regression analysis performed with these parameters yielded the

following relationships:

2.9 1.5 1.1

*b Terrors em Fr <---15>
(s-1)

T^ <V ecr>1-5 Fr°-8 <4*16>

*B "

(s-1)

6m-5 Fr°'9 <*<-17>

*B =
-^33 <V V
(s-1)

In all three equations, both 9 and Fr were determined with the

m

corrected mixture flow depth, h ,

and 9 was calculated according to

the procedure of Daido (1971), and then corrected for the slope effect
with the relation of Stevens et al. (1976), as described above. It
 -
134 -

should be noted, however, that the assumption of a constant 9 value,

for example 0.05, does not greatly affect the correlation equations.
(Note that 9 >= 7-9^.)

One may argue that a regression analysis between #_ and 9 leads to a

d m

spurious correlation because the density factor (s-1), which was varied

during the experiments, is contained in both * and 9 . Therefore, the

quality of the correlation was also checked between the measured trans¬

port
r
rates (q_ ) and those calculated (q_
. ) with a transformed
VMB,meas. ^B,calc.
version of equs. (4.15) to (4.17). This resulted in the following
correlation parameters:

qB,calc.
"

*B,calc.[g(s-1)dml0'5 SE

with equ. (4.15) 0.967 15.0%

with equ. (4.16) 0.975 13.0%

with equ. (4.17) 0.972 13.9%

It can be concluded that there is a similar quality of correlation

with both parameter
r
sets (q„, q
, S, S , s-1) and (*„,
v
9,9 ,
Fr,
'
X^B Mr cr B m cr

s-1). From the point of view that the experimental determination of the

flow rate Q (and thus q ) is more reliable than the measurement of the

flow depth H (and thus h ) and the fluid velocity V, a better correla¬

tion might be expected when using the first parameter set. But this

first set does not include any parameter describing the flow behaviour

(except through the term S ,

the influence of which is however small),
whereas the second set contains the flow parameters h and V. It is
r,m
also interesting that the exponent of the density factor (s-1) is less

sensitive to the inclusion of 9 in equs. (4.15) to (4.17) than of S

cr cr

in equs. (4.11) and (4.13).

 -
135 -

4.6 Comparison and analysis with further data

4.6.1 Experiments of Smart and Jaeggi

In the analysis of their steep flume data, Smart and Jaggi (1983)

compared their findings with the experiments of Meyer-Peter and Muller

and concluded that the bed load transport rate can be predicted by a

common expression for both data sets (s. sec. 2.3.2).

During the analysis for the present study it was found that a dif¬

ference exists in comparison to the procedure that lead to the bed load

transport euqation of Smart and Jaggi (1983). To adjust their steep

flume data for sidewall friction they used the Einstein procedure as

described in section 4.1.1 . For the bed load transport regression

analysis, however, they used a fictitious mean velocity (=q/H) derived

from the measured flow rate Q and mixture flow depth H. If a significant

part of the flow cross section is occupied by transported bed material,

this velocity is lower than the effective fluid velocity (which in fact

was measured and cannot be derived from the mixture flow depth). Thus

their reduced flow rates q are higher (for the steeper slope condi¬
tions) than if these values are determined with the measured fluid

velocities. But it should be pointed out that this peculiarity did only
affect the bed load transport equation (2.81). In the other parts of

their analysis they used measured values (not corrected for sidewall
influence), and in the verification calculations that followed the

development of the necessary equations, they used the measured fluid

velocity in the sidewall correction procedure.

Their data set includes 77 bed load transport experiments and is

labelled SJ henceforth. Four types of bed material were used in their

experiments: Two with a relatively uniform grain size distribution (mat.

II and IV), and two consisting of a mixture of grain sizes (mat. I and

III). These two classes are marked with different symbols in the fol¬

lowing figures.

In this study their bed load transport data was reanalysed using the
measured fluid velocities in the sidewall correction procedure. It is

seen in Fig. 4.15 that a systematic deviation of the bed load transport

rates calculated with the Smart/Jaggi equation (2.81) exists for the
 -
136 -

higher transport intensity region. From Fig. 4.15 it may be estimated

that the higher measured transport rates are systematically underpre-

dicted by about 20% to 30%. In fact, these differences can also be

detected from the results of the verification calculations presented in

Smart and Jaggi (1983). Equ. (2.81) is rewritten here, using the term

S instead of 8 :
cr c

4 "90.0.2 .0.6 ,.
q* n
S <S S ) (2.81b)
<3^>
"

•B (ill) qr cr

qBc*^*11]
qB =
4( |2)02 qrSoe(S-Scr) (s-1)1 /X

0 03

/ X

x-5

X
0 02

x A
y
X

/ *
/ X 4

&
/ X

X/ XA
0 01
4x a

Uniform gram
Data set Mixture
A * size

Smart/Jaggi A X

0 0
00 0 01 0 02 0.03

Fig. 4.15 : Bed load transport rates of Smart and Jaggi, comparison
between measured values, q„ ,
and those calculated

with equ. (2.81), q . (using the measured fluid

velocities in the sidewall correction procedure).
 -
137 -

Another difference results from the simplified transformation of equ.

(2.81) into an equation in terms of *_ and 9, because the the (fluid)

flow rate q should be substituted by V and h,, and not by V and h . The
I m

correct transformed form of equ. (2.81) is equ. (2.83); the simplified

form given in Smart and Jaggi (1983), omitting the factor h,/h or

8,/9 ,
is:
f m

e°-5 <W
*b -

(s^i) (of)0"2 s°-6 «

Bed load transport rates q_ were calculated with equ. (4.18) for the

Smart/Jaggi data to estimate the resulting deviation (the corrected

mixture flow depth,
r '
h ,
was used to determine 6 ). It can be seen from
r,m' m

Fig. 4.16 that the calculated values systematically overpredict the

measured transport rates by about 50% at the highest transport inten¬

sities.

It may be noted that the first difference (resulting in an under-

prediction of transport rates) partly compensates for the second dif¬

ference which would otherwise lead to an even stronger overprediction of

the transport rates. However, it should be pointed out that the bed load

transport rates predicted by the equations of Smart and Jaggi only devi¬
ate substantially from the measured ones for slopes steeper than 10% to

15%, within the given range of experimental conditions. The validity of

both equations is not affected by the differently applied sidewall
correction in the lower slope ranges where the bed load grain concentra¬

tions are relatively small.

A new regression analysis was performed with the Smart/Jaggi data,

which yielded the equations:

9.4 d90,0.2 „ _
r* = 0.958 1Q,
(rf—>
n
S ,„ , ,.
%
" -

(iTI) 1r (S
Scr)
gE = 205%
(4.19)

8.1 d90.0.2 .. _
.1.9 rJ = 0.958
,. ,.,

(d^> Scr>
n

qB (S
"

(S=I)
"

(4*20)
qr 20.4%
SE =

U.O ,d90.0.2 .2.1

S
r* = 0.955
,.
...

qB
n
(4-21>
(iriy (a^> qr
sE = 21.2%
 -
138 -

It is again noted that for the majority of the data points S >
5-Scr,
i.e. conditions clearly above beginning of transport. As found for the

clay suspension data, the equs. (4.19) to (4.21) show that the inclusion

of the term S does not greatly affect the quality of the correlations.

0.06 X

qB.caic[m3/sm] •b =
<@0!S"cC! (em-ecr)
>
0.05

0.04
X

a
K
/
X
X
A
0 03

*
X

0 02 A 4 /
A

< /
A

X
X
Uniform grain
0.01 *&/ Data set Mixture
X
t size

Smart/Jaggi A X

0.0
0.01 0.02 0.03 0.04

qB,meaJm3/Sm]

Fig. 4.16 : Bed load transport rates of Smart and Jaggi, comparison
and those calculated
measuredbetween
measuredbetween
values
values,
q_ ,
B,meas.

with equ. (4.18), qfi cal

 -
139 -

It is observed that the exponent of the slope factor is now higher

than in equ. (2.81b) of Smart/Jaggi, for which the statistical para¬

meters are also given for comparison:

4 ,d90.0.2 „
.0.6
S ,„
„ .
r2 = 0.952
.,
(2*81b) ...,
qB Scr>
"

(iTT)
"

qr (S

<a*^> SE = 29.3%

The difference in the exponent might be expected to be due to the dif¬

ferently applied sidewall correction procedure. To check this effect, a

comparative regression analysis was made, using the measured flow rates

q and ignoring any sidewall effect; the resulting equation is:

7.2 ,d90.0.2 „
.0.90
S (S,«,,..
r* = 0.956
,. 17.
(4*17b) ,
qB
"

(ill)
'

q
Scr>
<d^> SE = 27.3%

It is noted that the exact exponent of S in equ. (4.17) is 0.95 but it

was rounded to 1.0 (adjusting the constant correspondingly). The com¬

parison shows that the effect of including a sidewall correction is not

very pronounced.

The different exponent in equ. (4.17) and equ. (2.81b) must therefore

have another reason. It is presumably due to the fact that Smart and

Jaggi wanted to establish a single relationship for both their steep

flume data and the experiments of Meyer-Peter and Muller (s. also sec¬

tion 4.6.2).

Comparing the new equations for the Smart/Jaggi data set (SJ) with
the relations obtained for the data set of the author (RI), it is re¬

cognised that the exponent of the slope factor is not much different,

but that there is stronger dependence upon the density factor in the
formulae for the clay suspension data. In Fig. 4.17 the measured bed

load transport rates of the RI data set are compared with those pre¬

dicted by equ. (4.20); the q„ values for the higher concentrations

levels are seen to be systematically underpredicted, which is due to the

exponent of the density factor (s-1).

 140

0 04

qBMlc[m3/sm]

0 03

0 02

0 01

0 00
0 00 0 01 0 02 0 03 0 04

Fig. 4.17 : Comparison between measured bed load transport rates,

q„ ,
with those predicted by equ. (4.20), q_ ,
,
B,meas. B,caic.
for the new experiments.

Since the density factor (s-1) did not vary in the Smart/Jaggi tests

it was decided to fix the exponent at -1.6, the value obtained for the

clay suspension data. Another regression analysis was then performed

with both data sets SJ and RI, with the resulting formula:

13.0 "90.0.2 2.0 = 0.956

(S S ) (4.22)
(s-1)1'6 V
''
-

cr = 19.5%

The correlation coefficient r2 and the standard error S„ are similar to

the values for equ. (4.20). In Fig. 4.18 the measured bed load transport
 -
141 -

rates are compared with those predicted by equ. (4.22); it is seen that

there is no systematic deviation for both data sets.

If the conditions at initiation of motion are neglected, the fol¬

lowing regression equation is obtained for the SJ and RI data sets:

17.3 ,"90,0.2 .2.1 0.949

2~q (g—)
Qr
r
S
20.9%
(4.23)
(s-l)z-u Q30

qB = 13 0 ($*•)<>•• qr(S Scr)*(s1)i6

3 m] x A

A
*
A
+

A >
+

> /
*
/

&
/

+ y%

Data
Clay concentration
level
Smart/Jaggi a (H20)
/+ Rickenmann * C1

A
« C2
+ C3
• C4

oon-
ZL
0 00 0 01 0 02 0 03 0 04

%meas[m3/Sm]

(a)

Fig. 4.18 : Comparison between measured bed load transport rates,

and those Preoicted by (*-22).

qB meas.' e1u- <JB caic.'
for (a) the clay suspension data and (b) the Smart/Jaggi
data.
 142

U Uf X

qB = 13 0(g)" q,(S -

Scr) *(s-1)
<<-

W^3' s m]
X

(J Uo"

/x *

U \Jd~ /
X

/A
A
/

X/* A
Vx
.
&

*
/ X
A
/
* ££

yA X

Uniform gram
Data set Mixture
size

Smart/Jaggi A X

OOOi
0 00 0 01 0 02 0 03 0 04

qBmeas[m3/sm]
Fig. 4.18 (b)

Equation (4.23) has the practical advantage that it only requires the

knowledge of the slope, the flow rate, the grain size characteristics,

and the fluid-solid density ratio s. But no information is needed on the

flow depth, in contrast to the formulae where the critical slope term

S is included,
cr

In section 2.3.3 it was shown that the conditions at beginning of bed

load transport may as well be expressed by a critical discharge, the

determination of which does also not require a knowledge of the flow

depth. Since the constant in equ. (2.111) is partly based on the

Smart/Jaggi tests, this relation shall be used here:

. ... ,
,,1.67 0.5 ,1.5 .-1.12
0.065 (s-1) g d5Q S (2.111)
qcr
=
 -
143 -

Another regression analysis was performed with the data sets SJ and RI,

using the parameter set (q_, q -q , S, s-1), where q was calculated

with equ. (2.111); the resulting equation is:

12.6
,d90s0.2 „2.0 r2 = 0.951
(j^)"" , x
S-" I (4-24)
(qrr qcr)
-

j-j J 20zv.iz1%
"

(s-1)1*6 d30 cr
!>E -

Since critical conditions at transport begin have only a weak effect on

calculated q_ values for the steep flume data under consideration (q >
JJ r

6-q for most of the data), there is practically no difference in the

performance of equs. (4.24) and (4.22); both formulae show a similar

quality of correlation.

In steep torrents with a very rough bed, the critical discharge at

beginning of transport might be underestimated by equ. (2.111). It may

be more appropriate in this case to use equ. (2.113) which was developed
to judge the stability of steep block ramps.

4.6.2 Experiments of Meyer-Peter and Mueller

The comprehensive data set of the bed load transport experiments

performed by Meyer-Peter and Muller (labelled MPM data set below) is

given in Smart and Jaggi (1983). Most of the 137 experiments were

carried out with ordinary quartz grains as bed material. Coal (a = 1.25

g/cm3) was used as lightweight sediment in twelve tests, and baryt

(<j = 4.22 g/cm3) as heavy sediment in eight tests. Meyer-Peter and

Muller (1948) had shown that their formula adequately accounts for a

change in the grain-fluid density ratio s = tj/p. This is also true for

the simplified form of their equation, given by Yalin (1977), equ.

(2.67) (s. section 2.3.1.2, analysis of Low, 1989).

Low (1989) demonstrated that the Smart/Jaggi formula (equ. 2.83 with

1, i.e. small bed load concentrations, 4.18) does

9f/8 = =
equ. not

correctly predict bed load transport rates for varyng s values (s. sec.

1) In Smart and Jaeggi (1983), the transport rate G_ [g/s] should be

divided by a factor of 10 for the 17 experiments with d.. = 4.00 mm,
which have a circle as symbol.
 -
144 -

2.3.1.2). This can also be seen from Fig. 4.19(a) where q_ values were

calculated by equ. (2.82) for all MPM experiments with a mean grain
diameter of 5.21 mm (N =
40), having a uniform grain size distribution,
including the coal and baryt runs. The data points with s « 1.25 and

with s = 4.22 plot systematically away from the line of perfect

agreement. Fig. 4.19(b) shows the same data but with q_ values computed
by a modified formula, equ. (4.25), which is presented below.

qBcalJm3/Sm]

|% =
4°(cli7)02qrS06(s-sa)(s-iv
(D

0.0002
CD A

CD A

K
o

0.0001 ©/
X

Meyer-Peter / Muller Uniform grain

%
data size
X

:a
s=4 22

S=2 68 o

X
S=125

0.0
00 0.0001 0.0002

^B.meas [m3/s m]

Fig. 4.19 (a)

 -
145 -

<kcalc[m3/sm]

qB =
2.7@^qrS0 5(S-Scr)(s-1)^ X
/
X
/

ID

A *

Meyer-Peter / Muller Uniform grain

data size

s=4 22

^0 s-2 68 o

y
s-1 25

0.0-
0.0 0.0001 0.0002
(b) qB,meaJm3/S N

Fig. 4.19 Bed load transport rates of selected experiments by

Meyer-Peter and Mflller (N =

40), comparison between
measured values, qB and those calculated,
meas-»
(a) with equ. (2.83 or 4.18) and (b) with
qB calc..
equ. (4.25).

An analysis was performed with a subdata set of the coal, baryt and
some of the quartz grain experiments of the MPM data. A better agreement

between predicted and measured bed load transport rates was found if the

exponent of the density factor (s-1) in equ. (2.83 or 4.18) was put to

-1.5 instead of -1.0. This modification is also confirmed by the study

of Low (1989).
 -
146 -

A new regression analysis was performed for the MPM data set only,
and the exponent of the density factor (s-1) was fixed at -1.5. The

following equation best fitted the data set:

2.7 ,d90.0.2 0.5 .„ _ .

r2 = 0.960 ...
,,
3-5 (—)
S S ) (4.25)
qB
=
qr (S -

5 9%
5E
_
"

(s-1) 30

S was calculated by equ. (4.12), and a constant 9 = 0.047 was used, a

mean value for the MPM experiments. Smart and Jaggi gave the correlation

parameters for equ. (2.82) applied to the MPM data set as: r2 =
0.97, S„
= 66%, and for the original formula developed by Meyer-Peter and Muller

(1948) as: r2 = 0.92, S_ = 72%. It appears that the standard error S„ is

c. cj

further reduced by using equ. (4.25).

Equ. (4.25) can be transformed into a relationship in terms of the

dimensionless parameters (*_, 9, Fr, s-1):

2 7 90 0 2 0 5
*B ^ V (6r ecr> Fr (4-26>
"

-^-O
-

B r r cr
(s-ir--5 Q30

where again a constant value 9 = 0.047 was used. (Note that at the

lower slopes no distinction between 6. and 9 is necessary). Equ. (4.26)

is confirmed by a separate regression computation using the dimension¬
less parameters. The performance of equ. (4.26) applied to the MPM data

set is illustrated in Fig. 4.20. It may be seen that the relative error

between measured and calculated bed load transport rates decreases with

increasing transport intensity. As for the SJ data, the MPM experiments

performed with a uniform bed material and with a mixture of grain sizes,

respectively, are marked by different symbols.

 -
147 -

*B 2.7©02er°5 (8r. ecr)Fr{s.1)05

/
A
=

'09 (<W)
a f4 «

»x
0 ©
yx
X < i
/
1
X z
(
'A ?
z£ 0°/ 3 [
%
x

z X
X 0

*/*
*
X
: *
X

3
X
o
/ X

Uniform grain
Data set

/
z Mixture
size

Meyer-Peter 0 x

/
X 0
0

-7 0 -6 0 -5 0 -4 0 -3 0

l0g(qBmeas)

Fig. 4.20 : Bed load transport rates of the Meyer-Peter and Muller
data set (N =
137), comparison between measured values,
q. ,and those calculated with equ. (4.26), q„ .

o,meas. o,caic.

If the exponent of the density factor is adjusted in the original

equation (2.81) of Smart and Jaggi (1983) from -1.0 to -1.5, the formula

can be transformed to (using s =

1.68):

5.2 "90.0.2 .0.6

,.
. .

qB =
(A~) n
S Scr) (4.27)
T~5
"

(S
D qrr
(s-l)1-0 °30 C

Comparing equ. (4.27) with equ. (4.25) it is seen that there is only a

difference in the constant and in the slope factor; this difference is

tabulated below for some slope values within the range of the MPM

experiments:
 -
148 -

in equ. (4.25): in equ. (4.27):

2.7-S0*5 5.2-S0-6 A[%]

S = 0.02 0.382 0.497 30

S = 0.01 0.270 0.328 21

S = 0.002 0.121 0.125 3

S = 0.001 0.0854 0.0824 -4

It is observed that the maximum difference at a slope of 2% is about 30%

(which is still within the calculated standard error for equ. 2.25). It

is noted, that equ. (4.27) shows a stronger slope dependence, because it

was developed by Smart and Jaggi to also predict the steep flume data.
However, an equation based on the Smart/Jaggi data only, equ. (4.19), is

seen to be even more dependent on the slope (factor S instead of S ).

Thus the exponent of S in equ. (4.27 or 2.81) appears to be a compromise
for both the low and steep slope data.

The difference in the slope factor S suggests that in terms of the

parameter set (q_, q , S, S , s-1) two separate equations should be

used to predict transport rates for low and steep slopes. Thus a better

agreement between measured and predicted values is obtained than in the

case of one equation covering both slope ranges. It should be noted that

in view of the scatter between predicted and measured values, the Smart/

Jaggi formula, equ. (2.81 or 4.27), is equally valid as equ. (4.25) over

the whole range of the MPM data set but their equation tends to under-

predict the higher transport rates of the SJ and RI data sets (i.e. at

slopes steeper than about 10%).

If equ. (4.26) is compared with the corresponding formula obtained

for the clay suspension data, equ. (4.16), a remarkable similarity is
recognised. A similar equation also results from an analysis of the

Smart/Jaggi data set. In Fig. 4.21 all three data sets obtained at the

same hydraulic laboratory (MPM, SJ and RI), including a total number of

252 experiments, are plotted in terms of *_ vs. 9 (determined with h

o r r, m

for the SJ and RI data). It may also be inferred from this figure that a

common analysis of all experimental results is promising.

 149

CD

X X

x
2
x

X
X

*xx grain
size
Uniform
X 1
o »g 0 * H

t
<

*. /

*•%,
Jag i
Rickenma Meyr-Petr Muler
X

^ CO

co
/

i'

Smart
cd
x<V Q

Xx1*V
*>

<

Si* 1
•

'Y i'

H\ f
\ \
1 -

\XV .

<
X

< 1
0
1 <

«
"
N
X
g
0 *C x

.- o
o
o o
o
o
o

Fig. 4.21 : Experimental results of the three bed load transport data

sets MPM, SJ and RI (N= 252), in terms of the parameters

*_ and 9. The term k, /k represents the ratio of grain

friction to total friction (including form drag); the
average value for the MPM data is about 0.85, five tests

with an exceptionally low ratio are marked in the figure.

 -
150 -

The following bed load transport formula was obtained from a re¬

gression analysis with the data shown in Fig. 4.21:

3.1 90 0.2 0.5 1.1

(0r 9cr> Fr (4.28)
*B" 0 5
(d^> 6r "

(s-l)0*-' a30 r r cr

where h was used to determine 0 for the steep

r
flume data (SJ
x
and
r,m r

RI). Again, the correlation parameters are given in terms of q„

/\ c "
i meas.

0.961 and
'
and qn *_ r2 34.8%.
Sg
"

,
=
[g(s-l)d3] = =

MB,calc. B.calc.
,
m'

01

,0a(CtBca,c[m3/Sm]
J
/
*B =
3lO028r°5 (6r "

9cr) Fr"(S-1)05
t
/
•

0*
am

L
Jjr

c %A a
s

w%
O ©
~A*

X
X

x? o* *'
t
s
x

X
X

F » Uniform grain
X
* w Data set Mixture
X
x size
J»

Smart /Jaggi » x
I X
/ X
Rickenmann *

Meyer Peter /
o X
Muller

o
a

u.._
X

07 05 03 02 01

log(qBa.[m3/sm])

Fig. 4.22 : Comparison between measured bed load transport rates,

and thoSe Predlcted bV eclu- (4-28),

qB,meas.' 1B(Calc..
for the three data sets MPM, SJ and RI. Note that loga¬
rithmic values are used to cover the whole experimental

range in the same figure.

 -
151 -

A comparison between measured transport rates and those calculated with

equ. (4.28) is shown in Fig. 4.22. For ease of comparison, the

performance of equ. (4.28) is also shown in linear plots separatly for

the steep flume data in Fig. 4.23.

0.05
*>

qB.ca,c[m3/Sm] •J^s.Mig)02 V5 (em-ecr)Fr'1(s-D05

0.04

+ /

0.03 + /

+
y

0.02

A*
Clay concentration
A*
+

Data
0.01 level
y*
Smart/Jaggi a (H20)
Rickenmann * C1
o C2
+ C3
* C4

0.0 | |
0.0 0.01 0.02 0.03 0.04

qB,meas[m3/Sm]
(a)

Fig. 4.23 : Comparison between measured bed load transport rates,

and those Predicted bv e1u- (4-28)>

qB meas.' lB,calc.'
shown in a linear scale for (a) the clay suspension data
and (b) the Smart/Jaggi data.
 -
152

U.UO"
1 1 !
<J>b =
3M!o)02 <C5 (em- ecr)Frii(s-i)0
5
X

qB.ca,Jm3/ss

m]
m]
iB.calc

X <

X /

x
*
/
X /
*/a4
a/*

<y x

V* : :

Uniform grain
A^r^ Data set
size
Mixture

Smart/Jaggi A X

on- _| 1 1
0.0 0.01 0.02 0.03 0.04

PRmfiaJm3/sm]
Fig. 4.23 (b)

4.7 Summary of bed load transport formulae

the bed load equations that

The following tables summarise transport

Note that the correlation parameters r2 and S

were determined above. E

in of q„ also if the equa-

are always given terms , versus q_ ,

tion is presented in a *„ form.

 -
153 -

RI data set (N =
50): r2
SE[%]

25.2 „2.3
qr s .(4.11) 0.964 14.9
1bB =
TH r
(s_1}2.0

16.7 ..
.2.1
.(4.13) 0.959 15.9
.

%B =

"Tfi
n

qrr
(S "
Spr'
cr
(s-1)1*6

t._3i9 J^Fr0*8
(9-e .(4.16) 0.975 13.0
B cr'
^^0.4 m

» 3^9°*5(8-9 )Fr0*9 •(4.17) 0.972 13.9

v
B .
,.0.5 m m cr'
(s-1)

SJ data set (N = 77): r»

sEm

9.4 ,90.0.2 .

lu 1 S (S Srr) .(4.19) 0.958 20.5

T^TiT vd(h—>
= -
_

*B
(s-1)
30

8.1 "90.0.2 .1.9

.(4.20) 0.958 20.4
4B (s-1) vd
"

30

11.0 .90.0.2 .2.1

.(4.21) 0.955 21.2
4B (s-1) vd
"

30

MPM data (N 137): r2

set =

SE[%]

2.7 .90.0.2 „0.5 /r - . ,, ,c.

<B 7-^3 <ar;> "rS Scr)

•
(S- ..(4.25)

(s-1) 30
0.960 45.9

2l-'7 90 0 2 0.5
* =

n
. (-r^r"** e"*D (6 -
0 ) Fr ..(4.26) =
(4.25)
B r r cr
(s-l)u*° a30

Table 4.3 : Summary of bed load transport equations developed inde¬

pendently for the clay suspension data (RI), the Smart/

Jaggi data (SJ) and the Meyer-Peter/Muller data (MPM).

 -
154 -

Table 4.3 shows the bed load transport equations developed indepen¬
dently for the data sets RI, SJ and MPM.

In Table 4.4 the new formulae are listed which were developed for a

combination of the data sets RI, SJ, and MPM. They are compared to the

original formula of Smart and Jaggi, for which the influence of the

density factor was adjusted, equ. (4.27). It is recognised that those

equations valid only for the steep flume data (equ. 4.22, 4.24, 4.23) or

the one valid for the low slope data only (equ. 4.26 =
4.25) show a

better performance in their corresponding range of application than any

of the formula that is based on all data sets (equ. 4.28 and 4.27). It

is further observed that equ. (4.28) has generally smaller standard

errors S„ than equ. (4.27).

It is recommended that the first three equations are only applied

within a slope range 5% <= S <= 25 %, while equ. (4.26 or 4.25) should
be used only in the range 0.1 % <= S <= 2%.

It has also been checked how well the proposed equations perform for
a subdata set containing only experiments with an essentially uniform
bed material and those with a mixture of grain sizes. The correlation

parameters calculated for these subdata sets are given in Table 4.5. It

is noted that both subdata sets of the Smart/Jaggi experiments are best

predicted with the first three equations.

 -
155 -

Table 4.4: Comparison of bed load transport formulae for various combi¬
nations of the data sets of the author (RI), of Smart and

Jaggi (SJ), and of Meyer-Peter and Muller (MPM). The correla-

2
tion coefficient squared, r ,
and the standard error, S_,
K
were determined between measured and calculated values of q .
 156

(A
Q) w
vo to en
!3
3 Z "ST o
£ CM CM CM CM
cn

O o o

MPM
UI
u"l
en w
r— CO CO r-
en
1
•31
„
0
Z CO VD CM
H CM in
C en en
D
o o o

m CO CO •3"
<U r- w
ci to d o T
CM CM CM

a Z r- *T fM CO co
E CM m
CTi
en
°} en en

d o O o o

a
10
*T O
D- o w
t CO CXi
CM CN ro
E
*~
-

0
Z r^ VD a\ ^
H r^ r~ VD
C Oi en Cn en cn
""
a o O o o

eg
5 5 « £ P

TT «r <* -r
^ <r

n &
4-> Cn
Q) o ° U

~H to

1
U CD
+J M CD
T) U to
O CO
J3 1.
M
1 O1 O CD D
~.
CO Q
CO ui to
In ^
cr o u
cr O1 rj*
CD ©
CM CN CM

o O o ^ Q o

ol o Si 0 ol o
en f-i en -n en ci en -o
T> It) TJ I'D -O TO T) IT) tj m tj It)

VO vD o Ln
m u-l

o
"V
VD
^ ". CM
o^ O
V
r-
til
^

J] I/) r*l UI lT> in

CM in
" ~
*-*

CQ CQ CQ
cr cr a1
e

formulae for the subdata

Table 4.5: Comparison of bed load transport

sets with nearly uniform bed material and with mixtures of

experiments of Smart and Jaggi (SJ), and of

grain sizes;
The correlation coefficient
Meyer-Peter and Muller (MPM).
2
and the standard S_, determined
squared, r , error, were

between measured and calculated values of q_.

 -
157 -

4.8 Flov resistance and other aspects of the grain-fluid mixture

It is seen from Fig. (4.10) that the friction factor did not signifi¬

cantly change with increasing clay concentration. In the previous sec¬

tions it is shown that the bed load transport behaviour is essentially

the same as for the clear water case, if the flow around the grains is
not laminar (Re„ <= 15). For these conditions, the fluid velocity

slightly increased with increasing clay concentration, because the

higher bed load transport rates brought about a slight increase in the
mixture flow depth (Fig. 4.12).

In Fig. 4.24 the flow resistance coefficient c = V/v* is shown as a

function of the relative depth h/d,... The data points of the clay sus¬

pension tests were determined with the corrected mixture flow depth
h ; they are compared in Fig. 4.24 with equ. (4.7) used by Smart and

Jaggi (1983). There is a similar agreement with the calculated values as

found by Smart/Jaggi for their experimental results. In general, a

slight tendency can be observed that the experiments with the higher
transport rates (especially C2 and C3 tests with S >= 15 %) show a

somewhat higher flow resistance (smaller c value) than those experiments

with smaller transport rates.

To compare the fluid velocities measured in the clay suspension ex¬

periments (data set RI) with those of the Smart/Jaggi tests, the flow

resistance equation used in their study, equ. (4.7), was applied to both

data sets. Calculated and measured values are depicted in Fig. 4.25. It

appears that the two data sets may be analysed together. In general, the

predicted velocities are somewhat larger than the measured ones; this

was already noted by Smart and Jaggi.

 158 -

K\\ X
v^ -5^K
^NX \
\ \^ >
^*^
%
Ski A
\
*

o I* o
^
%
Symbol
|
< *

%x\
>l*
(a)

Fig. 4.24 : Comparison of data from clay suspension experiments with

the flow resistance equation (4.7) used by Smart and

Jaggi, shown in terms of V/v* vs. h/d--; data points

with equal slope are connected by straight lines. The

concentration levels (Ci) are: (a) H.O (mat. IV of SJ)

and CI, (b) C2 and C3, (c) C4 and C5.

 -
159 -

Fig. 4.24 (b)

-4 in
6 (J O

Symbol 0 >-

»|i
Fig. 4.24 (c)
 -
160 -

3.5

Vcalc [m/s] V=25(ghrmS)()5ln(8.2hrm/d9o) 4

=
•[ 1-exp (-0 05 hrm/(d90S° ))]"
=

3.0
*
X /
*

X
*

* 5 *
2.5
4

* X ,/
A
4
* 4/ '*
*
y x

2.0 4
*x*

*
A
r x

X
4*
Vx*
*
44 t

X
1.5 * X

A
x1 x

X
*
X

4
X

<
/*
A x
Uniform gram
1.0 A Data set Mixture
size
Ax.
X Smart/Jaggi & X

A *
Rickenmann *

0.5
0.5 1.0 1.5 2.0 2.5 3.0
V [m/s]

Fie.
6
4.25 : Comparison
r
between measured fluid velocites, V ,
and
meas.

those calculated with equ.

^ (4.7), V ,
,
for the data
calc.
sets SJ and RI.

The correlation parameters for the data shown in Fig. 4.25 were

determined between measured and calculated (with equ. 4.7) velocities

as: r2 = 0.68 and S„ = 20%. (It may be noted that a slightly better
correlation results for the SJ data set alone: r2 = 0.74 and S„ =
17%.)
Ei

A regression analysis was performed to obtain equations with dimension¬

less parameters in the form of:

 -
161 -

.8.0.5 .
Kn
1 . rjin.0.5 ,. ,Q.
(?) =

-q~29 (j-*-)
1.50 (4.29)
r
su.^ a90

V .0.30 qr ,0.5 ,. ,n.

anda , ., r
" 2*U S I
7T73 .3 ,0.5' (4*30)

(gd90) (gd90)

In terms of V versus V the corresponding correlation para-

n r ° r
meas. calc.

meters are: r2 =
0.68, S„ = 16.5% with equ. (4.29), and r2 =
0.88, Sg =

10.8% with equ. (4.30). Since both equations are based on a limited

experimental conditions, they might not be applicable at relative depths

h/d»0 larger than about 20.

The performance of equ. (4.29) is illustrated in Fig. 4.26; a similar

scatter of the data points as on Fig. (4.25) can be observed. According¬

ly, equ. (4.29) and equ. (4.7) show similar agreement between measured
and predicted velocities. However, the first relationship has a much

simpler form and is thus easier to apply.

Equ. (4.30) was developed as an alternative form because it allows to

determine the fluid velocity as a function of the design discharge,

without requiring the knowledge of the flow depth, which is difficult to

estimate in a steep torrrent.

The performance of equ. (4.30) with the data on which it is based is

shown in Fig. 4.27 (a). It may be noted that generally slightly higher
velocities were measured for the case with a bed material with a wide

grain size distribution. This might be due to a smoothening effect of

the finer particles lying between the coarser grains on the stationary
bed. It also be noted that in general is taken of
may d„n as a measure

an average, characteristic grain size of the armour layer; if there are

many fine particles present near the bed surface, this approximation

might become inadequate.

 -
162 -

/
Vca,>/S] V =
150S0^hrm(7
3 \0 5 A

x/

A
/ X

*
A* X

• * X*
4
• / *

/A A

^
A
A
«
*x
X/b. « x x

A A, / > X

.* •
* / x
* A/ "
A A
4 / X

** X

t
y
4
X
x
x
%y*y e
*

4
4
y X

^4 x

Uniform grain
Data set Mixture
size
X

x/x Smart/Jaggi a x

A x
*
Rickenmann

0.5-
0.5 1.0 1.5 2.0 2.5 3.0

Vmeas[m/S]

Fig. 4.26 : Comparison between measured fluid velocites, V ,

and
meas.'

those calculated with equ. (4.29), V .

,
for the data

sets SJ and RI.

The flow resistance measurements for the clay suspension without

sediment transport were used to check how well equ. (4.30) would apply
to flows over a rigid bed without sediment transport. It is seen from

Fig. 4.27 (b) that the measured velocities systematically deviate from

the predicted ones by about 16% on the average (the corresponding

correlation between measured and predicted velocities was calculated as:

r2 = 0.96). While the form of equ. (4.30) is confirmed by the indepen¬

dent data set, the constant should be a bit higher than 2.14. But it is

not unexpected that there is no unique constant for the two cases, i.e.

for a sediment transporting flow over a mobile, permeable bed and for a

fluid flow over a fixed, impermeable bed (s. also section 4.3).
 -
163 -

Theoretically, the knowledge of the increase in the flow depth due to

transported bed load grains should allow to compute the mean grain

velocity U_, if the fictitious fluid depth (q/V or q /V) is known:

a +
5m = H (4.31a)
v
uB

or =- (4.31b)
r,m

Vcal>/s]

[m/s]

(a)

Fig. 4.27 : Comparison between measured fluid velocites, V ,

and
meas.

those calculated with equ. (4.30), V .

,
for (a) the

sediment transport tests (SJ and RI) and (b) the clay
suspension experiments with a fixed rough bed.
 -
164 -

3.5

Vcalc.[m/s] V = 2.14 S°3° q,05^)025

3.0

-160 yU

y
y

/^'Ky
y
y
2.5
y
y
i-
y
/

2.0
y

/ y

1.5

y&x *

/ y

1.0
y

/ y
/ y
/ y
/ y
/ y
0.5
0.5 1.0 1.5 2.0 2.5 3.0 3.5
VmeaS>/S]
Fig. 4.27 (b)

Because of limited accuracy of the measurements (particularly of the

mixture flow depth H ,

and to a lesser extent of the fluid velocity V),
application of equ. (4.31a) or (4.31b) to the steep flume data results

in negative grain velocities Ug for some experiments. A regression equa¬

tion was therefore proposed by Smart and Jaggi (1983) to determine

smoothed values for H :

q/V 1.14 .0.18

1.41 S (4.32)
H *B

A combined analysis of the data sets SJ and RI showed that the constant

1.41 in equ. (4.32) should be replaced by the term 2.37/(s-l) to account

for the change in with increasing clay suspension concentration:

 -
165 -

2.37 1.14 .0.18

b
*B (4.33)
(s-1)

An alternative equation was devloped from both data sets using the
depths
r
values corrected for sidewall influence (h , and h ) and the
L j L L fill
instead of the terms
ratio
QR/qr #fi and (s-1):

%* - 1 -
1.64 S0*42 A0*63 (4.34)
hr,m qr

with r2 = 0.70 and S_ = 13%. It is noted that the correlation parameters

for equ. (4.33) and (4.34) are practically the same. A comparison
between measured and calculated depths ratios is shown in Fig. 4.28a for

equ. (4.33), and in Fig. 4.28b for equ. (4.34).

(a)

Fig. 4.28 : Ratio of fictitious fluid flow depth to mixture flow

depth: Comparison between measured and predicted values,

calculated (a) with equ. (4.33) and (b) with equ. (4.34).
(It may be noted that measured values > 1.0 were not used

in the regression analysis, because theoretically the

ratio should be < 1.0.)

 -
166 -

1 0

(hrAJca,c
hrl/hrm = 1
164S°42(^)063

*y<

0 8

06

Uniform grain
Data set Mixture
size

Smart/Jaggi
Rickenmann

0 4*-
0 06 08 1 0 12 14

(hr./h,m)meas

Fig. 4.28 (b)

It was decided to use equ. (4.31b) in combintion with equ. (4.34) to

determine a mean grain velocity and a mean bed load concentration (by
volume) of the moving grains for each experiment. First, the smoothed

mixture flow depth was calculated as:

q /V
JT
O0.42. (4.35)
r,m,calc. , ,.

1.64 S
.
,0.63
(qB/<lr)

and then the mean grain velocity, U_, was obtained as:

UB =

h
t
-

(q /V)
(4.36)
r,m,calc. r

Since the fluid and the grains do not move with the same (mean) velo¬

city, the bed load volume concentration C „

is not simply given by the
ratio q_/q; it is defined as that part of the cross section which is
B

occupied by the transported grains, and it was calculated as:

q /V
^r
*
Cv,B "
(4.37)
-

h
r,m,calc.
 -
167 -

Comparative calculations were also made using equ. (4.33) instead of

equ. (4.34) but this was not found to greatly affect the results presen¬

ted below. The effect of an increasing clay concentration (of the sus¬

pension) on the calculated UD and C _

values was examined using the
D V , D

procedure described above. The change in the mean grain velocity is

shown in Fig. 4.29 as a function of the density factor (s-1). It is

observed that U_ increases slightly with a decrease in (s-1), similarly

to the change of h and V with clay concentration shown in Fig. 4.11a

and Fig. 4.11b, respectively. The change in the mean bed load

concentration is illustrated in Fig. 4.30 as a function of (s-1). The

increase of the bed load concentration with decreasing clay

concentration is much more pronounced than the change of U_ in the

previous figure.

1.5 *.

\
S=7%
N

S=10%
1.4
S=15%
"•^ S=20%
\
1.3

^\
^-TK
1 /^ s

1.2
\ *^
f>2v\x
\
\
/
\A^ N

\ -^O-
1.1
/
/

i
s
1.0

y
0.9

X
0.8 X„

0.7
1.0 1.1 1.2 1.3 1.4 1.5 1.6

(S-1)

Fig. 4.29 : Ratio of mean bed load grain velocity in clay suspension
to corresponding value in clear water, UD/UD „„„, as a
a DftiZO
function of the density factor (s-1); experimental points
with equal slope and flow rate are connected by straight
lines.
 -
168 -

22
9*.
c„ S=7%
*s S=10%
2.0
«. \N sN S=15%
S-20%
\
\ \
1.8 ^

k.^ \\ V
-Ik
X

1.6
\ ^ XX Xx
^X\
1 4
\
\ .x>X s'.

\ ^
1.2
_„——-*""" <^ ;X
V
>**."
n*
\^
---^
1 0

>«

08
1.0 1.1 1.2 1.3 1.4 1.5 1.6

(S-1)

Fig. 4.30 : Ratio of mean bed load concentration (by volume) in clay
suspension to corresponding value in clear water,

Cv B/Cv B H20'
as a function or tne density factor (s-1);

experimental points with equal slope and flow rate are

connected by straight lines.

Calculated mean grain velocities for the clay suspension experiments

are shown in Fig. 4.31 as a function of the shear velocity (determined
with h ). The data points with 7% slope were omitted from the figure
for clarity, and because the calculated velocities are somewhat more

vague for these experiments with relatively large errors between mixture

flow depth and fictitious fluid flow depth. In Fig. 4.31 there is an

almost linear dependence of U_ on the shear velocity; it is further

D

evident that the U. values depend also on the flume slope.

 169

1.5

UB[m/s]

1.0

0.5

°'0.1 0.2 0.3 0.4

vMm/s]

Fig. 4.31 Mean velocity of bed load grains, function of

:
Ug, as a

the shear velocity, v*, for the clay suspension data;

experimental points with equal slope and flow rate are

connected by straight lines.

4.9 Case II: Laminar flow around bed load grains

In this section only those experiments of the clay suspension data

are considered for which the fluid showed a higher viscosity. It can be

seen in Fig. 4.9 that the bed load transport rate decreased with in¬

creasing clay concentration below a particle Reynolds number Re. of 10

*•
it
15. The physical meaning of this critical value of is that the
to
Re,
flow around the bed load grains laminar in the experiments with
was
Re2
<= 10 (s. sec. 4.4).
 -
170 -

The criterion used to separate experiments belonging to case I and

case II is also described in sec. 4.5. A decrease in the bed load trans¬

port rate q„ With respect to the corresponding value at the next lower

concentration level was observed for the following tests (s. also Appen¬
dix II): For three experiments at the level C4 and for all 5 tests per¬

formed at the highest clay concentration level C5. All these experiments
belong to the case II category.

It may be mentioned that the velocity measurements were not reliable

for the experiments at the level C5, because the mixing of the salt
solution within the viscous clay suspension was not sufficient. As an

approximation, the corresponding velocities measured at level C4 were

used to calculate the friction factor f and the Reynolds number Re. for

the C5 experiments (used in Fig. 4.10).

For the case II experiments, Fig 4.9 shows that the ratio q./q. „„.
B U,nZU
^
is related to the logarithm of the corresponding Re. values. The

decrease in q may also be examined with respect to the maximum

transport rate (for the same flow rate and flume slope) observed at the

level C3, This ratio, is plotted against the

q„ _,.
new
qB/QB c3i

logarithm of Re. in Fig. 4.32, and against the viscosity change,

U ,/u , „,,
in Fig. 4.33. (The correlation coefficient squared is for
ez sz joj
the line in Fig. 4.32: r2 = 0.81, and for the line in Fig. 4.33:

r2 0.89.) It be noted that also be related linear

=
may qR/qR ro
can to

values of Re. but then a worse correlation results.

Comparing the shear rates for the case II experiments with the cor¬

responding values for the C3 tests, the difference is seen to be less

than 15%. It can therefore be concluded that the decrease in the bed

load transport rate is mainly due to the increase in the fluid vis¬

cosity.

It can be hypothesised that the decrease in qD is, at least partly,

D

due to a change in the critical shear stress for initiation of motion

which increases for Re, values below about 10. The relation for 6 given
2 c
*
in Yalin and Karahan (1979), equ. (2.107), was used (with Re, instead of
^
it
Re ) to determine S with equ. (4.12); then q was calculated with equ.

(4.22). The q values calculated for the C4 tests (case II) were smaller
B
than the q„ values for the corresponding C3 experiments. But for the C5
o

tests, all calculated q. values were larger than the corresponding

 -
171 -

values for the C3 experiments. But it is questionable, whether an

equation such as equ. (4.22) may be applied to the case II experiments

it
(with Re. values below 10) where the viscous influence should not be

neglected; it cannot be decided therefore whether the 6 values are

correctly determined by equ. (2.107).

The increase in 0 was also calculated with the relationship proposed

c

by Wan (1982) for Bingham fluids, equ. (2.108). However, the computed 9c
values were larger than the measured 9 values, and thus no bed load

transport rates could be determined.

's

08 » /

Y A 1

A y

04 Y /

0.2

Oi 06 08 toglRe}) 10

Fig. 4.32 : Ratio of bed load transport rate for case II experiments
to the corresponding value at level C3, as a function of

the grain Reynolds number Re. (logarithmic values).

 -
172

1
la
\
\
\
\
\

\ •

0.8

0.6

X. Y

0.4

Y X

0.2
Ve2

He2 ,C3

Fig. 4.33 : Ratio of bed load transport rate for case II experiments
to the corresponding value at level C3, as a function of

the respective ratio of the effective viscosities,

Me2/,Je2,C3*
 -
173 -

5 DISCUSSION

5.1 Bed Load Transport

Note that in this section only those experiments (case I) of the

author are considered, for which viscous effects were found not to be

important.

Form of proposed equations

Based on the steep flume experiments with both clear water (Smart and

Jaggi, 1983) and the clay suspension (tests of this study), a new bed

load transport formula, equ. (4.22), has been developed in terms of the

following parameters:

qB
-
(qr, S, Scr, s-1, d90/d3Q) (5.1 )

Using the same parameters, a similar relationship, equ. (4.25), has been
established for the low slope experiments of Meyer-Peter and MOller
(1948; data given in Smart and Jaggi, 1983). Comparing equ. (4.25) with

equ. (4.19), it is evident that the exponent of the slope factor is

different for the two data sets. This suggests that two different for¬

mula for the two slope ranges will give better predictions than one

common equation. Therefore, the density adjusted version of the original

Smart/Jaggi equation (4.27) shows a larger scatter between predicted and
measured transport rates than any of the formulae developed only for a

limited data range (s. Table 4.4, section 4.7).

Many bed load transport equations based on the tractive force concept

be expressed with the dimensionless and 8 (Graf, 1971;

can parameters *R
Yalin, 1977). Choosing the additional parameters given below:

*B =
(6' 9cr' Fr' s_1' d90/d30) <5*2 >

a more universal formula, equ. (4.28), has been proposed which predicts
bed load transport rates reasonably well for both the low slope and the
 -
174 -

steep slope data analysed in this study. It should be pointed out that

in a relation of the form of (5.2), two parameters are included which

describe the flow "behaviour" (9 and Fr), whereas no such parameter is

contained in (5.1). It is therefore not surprising that equ. (4.28) can

be applied to a wider range of conditions than any equation of the form

of (5.1). When comparing equ. (4.28) and (4.27) in Table 4.4 (section

4.7), it should be remembered that that the experimental determination

of 9 in the steep flume situation is clearly less accurate than that of

q; despite that fact equ. (4.28) shows a better overall performance.

Based on dimensional analysis, Yalin (1977) proposed equ. (2.93) as a

general form of a bed load transport relation. If the term representing

viscosity influences is neglected, and if the parameter h/d is replaced
by Fr, using the "bridge relation" Fr = c[(s-l)-9-d/h]0*5 (Yalin, 1977),
then the following relationship is obtained from equ. (2.93):

* =
+[9, Fr, s-1] (5.3 )

The use of the Froude number implies that the bed load transport rates

depend also on the flow resistance. It is noted that relation (5.2) is

of a similar form, except for the additional inclusion of the grain size

distribution parameter d.0/d,..

Comparison with other formulae

The entire data set used in this study is replotted in Fig. 5.1. Also
shown are the three bed load transport equations of Meyer-Peter and
Muller (in the simplified version given by Yalin, 1977), equ. (2.67), of

Bagnold (1956; with an average value of b = 4 for d >= 0.5 mm), equ.

(2.97), and of Luque and van Beek (1976), equ. (2.69). It is seen from

Fig. 5.1 that equ. (2.67) predicts an upper limit of the Meyer-
Peter/Muller data; this is not surprising since any effects of (bed)
form drag and sidewall friction were neglected in the derivation of equ.

(2.67) from the original version of Meyer-Peter and Muller (1948). The

general tendency of Bagnold's equation is to overpredict the low slope

data, and to underpredict the steep slope data. The formula of Luque and
van Beek is based on lower shear intensity data (9 < 0.06) but there is

fair agreement with the presented data up to 9 = 0.4. It may be noted

that this value of 9 is an approximate limit above which bed load con
a. re
(1> .a » re
2" c re ts
•"I ft)

- a -
c

<-f
a-
•
S'
» B> •o a
i
re o

h; w

s. «
0.
o

« =r
o o
B S =r
o

/-*
ro *»
re
i-i
rt
o
„

o re

re cr

» «

0 0001
w to n

0 04 0 06 0 08 0 1
 -
176 -

centrations (in the steep flume tests) were no longer negligible, i.e.

the transport layer occupied more than about 10% of the flow depth.

Inclusion of the Froude number

In many bed load transport equations based on the tractive force

concept, *_ is given as a function of 9 and 9 . In equ. 2.83, Smart and

^
0 '
fi
Jaggi included the factor c-S ,
which is very close to the Froude
'

number Fr = c-S . Bridge and Dominic (1984) compared several data sets

concerning bed load transport rates for flows over plane beds. They

developed an equation in terms of the parameters (*_, 9, 9 ), equ.

2.101, and concluded that the value of the factor (a/tana) in their

relationship should increase from lower stage to upper stage plane beds

by about a factor of 2, on the average. There are a number of bed load

transport equations similar to equ. (2.101), but they have a constant

instead of (a/tano), probably because they are based on limited data

sets. In his analysis of steep flume tests, Daido

(1983) proposed that the "constant" B, In his equation (2.86) may be a

function of the Froude number in general. A flow resistance coefficient

and a slope factor is also included in a formula (equ. 2.85) presented

by Takahashi (1987) to predict transport rates in flows at steep slopes.

Thus it appears that the Froude number may be an important parameter

to predict bed load transport rates in flows over plane beds for both
the subcritical and supercritical flow regime.

The role of the Froude number in the experiments of Meyer-Peter and

Muller was discussed by Smart and Jaggi (1983). For the majority of the
tests, Fr was below or close to 1.0; in a few experiments Fr was as high
as 1.7. The mainly observed bed forms included plane beds and dunes of

various forms, while ripples were present only in very few tests. To

account for additional friction due to bed forms, Meyer-Peter and Muller
'

(1948) had introduced the factor (k,/k ) in their original bed load
transport equation; k denotes the Strickler k value for the total bed

resistance, and k the corresponding value for grain friction only. Thus
the factor k,/k represents a correction factor to reduce the shear

stress in account of form drag losses. Smart and Jaggi demonstrated that
'

their factor c-S has a similar effect on the calculated 4_ values as

15
the factor (k,/k ) in the original Meyer-Peter/Muller equation.
 -
177 -

However, it should be noted that there were only a few experiments with
a high form roughness (the tests with ripples, where k,/k =
0.5); the

mean value of the ratio k,/k was about 0.85.

b r

It was also shown by Mantz and Emmett (1985) that a flow resistance

coefficient may be required to predict sediment transport rates (at

lower slopes), if no other correction for form drag losses is included.
They presented numerous field measurements on bed load transport, using
Bagnold's (1966) stream power parameters i_ and (&) -
w ); where i_ is
D C o

the bed load transport rate by immersed of solids per unit width,
» = x -V is the stream power
r
and « the critical value of w at initia-
o c

tion of motion. For a given value of (<a -

w ), the bed load transport
rates i_ were shown to be about an order of magnitude larger, if the

friction factor f decreased by about 50%.

Effect of a change in the grain -

fluid density ratio

The influence of a change in the density ratio s on q_ was studied by

Low (1989). From an analysis of his experiments with lightweight sedi-
2 5
ment he found that q_ should be proportional to d-v* /W . He argued that
the shear velocity v* (or the mean flow velocity V) may be considered as

a measure of the flow strength, while the fall velocity V can be looked

at as a measure of the resistance of the grains to motion. For his ex¬

perimental conditions, he showed W to be a function of the factor

0 5 " 3
[gd(s-l)] ,
and found that the relation *_ 8 does also correctly
account for the density effect. He further demonstrated that several bed

load transport formulae in terms of (#_, 9 and 8 ) adequately account

for a change in the density factor (s-1). This is independently

confirmed by the experiments of Luque and van Beek (1976); their data,

including varying values of s, is well described by a transport equation

using the parameters (_, 9 and 9 ), equ. (2.69). According to the

analysis of Low, a correct reproduction of the density effects requires

that the ratio of the shear velocity to the fall velocity is correctly
represented.

Low (1989) proposed to include a modified density factor (s-1) in the

Smart/Jaggi equation (2.83). In the present study, this modification was

confirmed in the analysis of a subset of the Meyer-Peter/Mflller experi-

 -
178 -

ments, where the density ratio s was varied. It appears that the inclu¬
'

sion of the Froude number Fr =

V/(gh) requires the additional factor
'

(s-1) ,
which accounts for the change in fall velocity.

The regression analysis of the clay suspension data resulted in a

similar dependence of „ on (s-1), (s. equ. 4.15 to 4.17). For this data
D

set (case I experiments), the calculated fall velocity at the concentra¬

tion level C4 is about 25% of the corresponding clear water value. In a

Newtonian fluid, viscosity begins to affect the fall velocity in the

hyraulically transitional regime. From the analysis of the clay suspen¬

sion flow over a fixed rough bed, viscous effects were found to become

important only at Re, <= 10 (s. sec. 4.3). Interestingly, a distinct

change in the bed load transport behaviour was found in the clay suspen-
* *
sion tests at Re. =
10; in the region Re. >= 10, transport rates can be

well predicted by an equation which includes the density factor (s-1)

but no viscosity term.

Briihl (1976) studied the hydraulic transport of sand in pipes. He

found that the pressure gradient is reduced (i.e. the transport capacity
increased) if he used a slurry with fines as transporting fluid instead
of clear water. In analysing his results, he calculated the pressure

gradient with existing methods developed for the clear water case. In

these methods, the fall velocity of the transported solids appears ex¬

plicitly as a parameter. By adequately accounting for the change in the

density factor (s-1) and in the fall velocity W, he obtained fair

predictions for the slurry data. This again suggests that a correct

representation of the effects of W is a necessary requirement to

describe the (bed load) transport in the case of a varying solid-fluid

density ratio s.

Equations for steep flume data

For the Meyer-Peter/Miiller data, the density factor (s-1) in equ.

(s-1)"
*

(4.25) is equivalent to the density factor in equ. (4.26),

which can be shown by transforming one form of the equation into the

other. But for the steep flume situation with high bed load

concentrations, a formula in terms of the parameter set (5.2) cannot be

easily transformed into the parameter set (5.1), because the flow rate

can no longer be expressed as the product of the (mixture) flow depth

 -
179 -

and the fluid velocity. The regression analysis of the clay suspension
data (RI) resulted in similar exponents for the density factor (s-1) as

for the low slope data. However, if the term S is neglected in

c-2.0
parameter set (5.1), there is a dependence on (s-1) in equ. (4.11)
and in equ. (4.23):

17.3 ,d90.0.2 .2.1 ..

,,.
^rB =
5~0 (T~> n

1rr
S (4-23)

(s-1)2*0 d30

It is interesting that both Mizuyama and Shimohigashi (1985), and

Chee (1988) proposed from steep slope experiments with varying s values
-2 0
that q_ should depend on (s-1) . In fact, the bed load transport
o

formula of Mizuyama and Shimohigashi is very similar to the above

equation:

q = A' q S2 -^-j (2.66)

(s-1)

Mizuyama (1981) earlier presented a similar equation obtained from steep

flume tests with clear water as transporting fluid:

qB
= 5.5 q S2 (2.87)

It is not known whether a correction for sidewall effects was applied to

the experimental data that lead to either equ. (2.66) or (2.87).

In order to compare the steep flume data used in this study with
these two formulae, a regression calculation was made with the data sets

RI and SJ, fixing the factor S2; the result is:

,.
.

U'4
,d90.0.2 .2.0
S
1 r2 = 0.94 ...
.
,

Qb "' (
=

(Xl?7*
'
V SE - 21*5*

0 2
If the grain size distribution factor (d-./d,.) is to be neglected,
it should be replaced by the value 1.05 according to Smart and Jaggi
(1983). Comparing equ. (5.4) with equ. (2.66), this results in A' = 15.1

for the data of this study. Mizuyama and Shimohigashi (1985) gave A' =

20 for the case of clear water, and A' = 25 for the tests performed with
a fine material suspension. If equ. (2.87) is transformed into a rela¬

tionship of the form of equ. (2.66), the corresponding constant A' =

 -
180 -

15.5 is obtained (using s =

1.68). Thus it can be concluded that the

steep flume data used in this study (data sets RI and SJ) and the one

used by Mizuyama (1981) result in very similar bed load transport for¬

mulae. No explanation could be given by Mizuyama (1988) as to why the

experiments with a fine material slurry (in the study with Shimohigashi)
lead to a different value for the constant A'.

Effect of increasing clay concentration on grain transport

In the clay suspension experiments (RI data set), the same slope and
flow rate combinations were used for all clay concentration levels (Ci).
According to equ. (4.11), the bed load transport rate can be described

by the three parameters q, S and (s-1). Therefore, the effect of (s-1)

on q„ or related parameters is easily seen on appropriate diagrams.
Considering Fig. 4.29 and Fig. 4.30, it is observed that the increase in

q. with a decrease in (s-1) is mainly due to an increase in the mean bed

load concentration (C „), while there is a much weaker increase in the

v,B
mean transport velocity of the grains (U_).

It is seen from Fig. 4.31 that the calculated mean grain velocity U_

depends on the shear velocity in an approximately linear way. This is in

agreement with the empirical relationship for the velocity of single

grains U, found by Luque and van Beek (1976), equ. (2.71), and also with
the theoretical equation (2.98) for the mean velocity of bed load grains

proposed by Bridge and Dominic (1984). A linear dependence of U, on v*

was also observed by Abbott and Francis (1977); they further showed that
U, was slightly higher for smaller grain densities o (or smaller values
of s, since p was constant). In Fig. 4.31 there seems to be a similar

tendency for the data from the tests with 10% and 15% slope, whereas the

data points with 20% slope show no clear grouping.

It is also observed from Fig. 4.31 that the mean grain velocity UD is
D

not only a function of v* but also of the slope. A higher shear velocity
is required for a steeper slope to result in the same U_ value. Since

the bed load concentration are clearly higher at steeper slopes, grain
to grain interactions may be expected to become more important, possibly
resulting in an increased loss of grain momentum.
 -
181 -

General flow characteristics

In the steep flume tests of Smart and Jaggi (1983), the bed forms

present were flat beds and antidunes; for the tests with mat. IV (used
as reference conditions) there were only flat beds. In the clay suspen¬

sion experiments, flat beds also existed at the lowest clay concentra¬

tion level; no direct observation was possible for the tests with a

denser suspension. However, some indication was obtained from the opera¬

tion of the flume system: With increasing suspension concentration it

was found easier to adjust the upstream roughness elements to establish

equilibrium transport conditions; the bed stability seemed to be less

influenced by slightly varying conditions at the flume entrance.

White (1987) presented a criterion to delineate the occurrence of

different bed forms. The dimensionless unit stream power U_ is re¬

arranged here as:

v s ^/3 )
U -
"
(5l3':>5 '
UE 2/3 . .
,.1/3
g d (s-1)

White proposed that for flat beds and antidunes U„ should be greater
than 0.02. This criterion is satisfied for the clay suspension experi¬
ments (using v =
u_./p). It is interesting that the value of U_ in¬

creases with a decrease in (s-1) and with an increase in v. For the clay
suspension experiments, a stabilising effect on the bed forms under

consideration may therefore be expected.

Kresser (1964) established the following criterion for a minimum flow

velocity above which there should be suspended load: V >= 360-g.d. For

the clay suspension experiments this would require a velocity of about

6 m/s; this value was clearly not reached. But a number of grains cer¬

tainly moved in saltation as was observed by Smart and Jaggi in their

clear water experiments.

To distinguish between bed load and suspended load, another criterion

is given by Bagnold (1966): If W/v* >= 0.8, then only a negligible part
of the sediment is supposed to be carried as suspended load. According
to this criterion some tests at the concentration level C3 and all

experiments at the levels C4 and C5 should have carried a non-negligible

part of grains in suspension (if u .
is used to calculate W). However,
 -
182 -

Murphy and Aguirre (1985) suggested that suspended load and bed load can

only be distinguished, if effects of the fluctuating fluid forces are

considered, apart from the trajectory length and height of the grains
which can be obtained from mean flow parameters.

The maximum packing concentration (by volume) CA for a static gravel

layer is between 0.63 and 0.74 (O'Brien and Julien, 1984). For the clay

suspension experiments with 20% slope, the calculated mean bed load

grain concentrations (C .) range between 30% and 55% by volume. If the

v,u
"packing" concentration in the transport layer is assumed to be about

0.60 (in a sheared layer it must be less than CA), it results that this

layer occupied more than half of the total (mixture) flow depth. In

other words, the transport layer had a thickness of several grain dia¬
meters while the "clear" fluid layer above was only a few grain
diameters thick. Thus it seems questionable whether a suspended load
should be defined for the given flow and transport conditions.

At the higher transport rates it was observed by Smart and Jaggi

(1983) in their tests that the grains close to the grain surface moved

in a suspended like manner because inter-particle contact prevented

saltating particles from returning to the bed. They also noted then a

tendency towards a more uniform mixing of water and grains. These

conditions may be termed debris-flow-like sediment transport. In the

case of the clay suspension experiments this corresponds to type 1

debris flow as defined by Davies (1988) (s. Table 1.2, section 1.3).

The role of the Froude number was discussed by Smart and Jaggi (1983)
with respect to the Meyer-Peter/Miiller experiments; it may be looked at

as an indicator for friction losses other than grain friction. In the

steep flume tests the bed forms were flat beds and antidunes, and bed

load transport rates were generally large. Therefore, bed form drag may

be considered negligible in comparison to the part of the flow energy

required to transport the grains. The dimensionless shear stress 9 is a

measure of the flow strength at the bottom of the flowing grain-fluid

mixture. However, in the steep flume tests the transport layer extended
over more than half the (mixture) flow depth. Therefore, the use of 9

alone may not be sufficient to predict bed load transport rates. The

Froude number (or the fluid velocity) may be looked at as a measure of

the forces acting on the grains within the flow. Since both Fr and q„

increased with slope and flow rate, it seems reasonable to use Fr as an

 -
183 -

additional parameter to predict bed load transport rates at steep

slopes.

Comparison with stream power approach

'

Putting Fr = Fr and considering conditions with 8 » 9 , equ. 4.28

simplifies to:

*B "
7V5 e1"5 Fr <5*6 >
(s-1)

This equation can be rearranged in terms of the stream power approach:

iB S0*5 (5.7 )
=

B3 co

Bagnold (1973) postulated that the work rate performed by the trans¬

ported solids is equal to the available power (stream power) times an

efficiency factor. If the effects of an appreciable gravity slope cannot

be neglected, the following relationship can be given according to

Bagnold (1973), to express the maximum bed load transport efficiency:

(5.8 )
(tana -
tang)

where the first factor on the right hand side is the efficiency factor.
Many attempts were made to determine the angle of internal friction for
the dynamic case of bed load transport. Bridge and Dominic (1984) sum¬

marised values proposed in the literature: A range of tana = 0.4 ... 0.7

was found by various authors for inertial grain shearing conditions

(G2 > 1500), while tana = 0.75 was proposed by Bagnold (1954) for macro¬

viscous conditions (G2 < 100). It was concluded by Bridge and Dominic

that tana should also increase with increasing bed shear stress (in the
inertial region).

The steep flume data (sets SJ and RI) are shown in Fig. 5.2 in terms

of the stream power parameters. Also shown are two lines defining the
100% efficiency according to Bagnold (1973), equ. (5.8); both lines were

determined for the steepest slope of 20%, line (A) represents the case

of inertial shearing (with tana =

0.4) and line (B) the case of

macroviscous shearing (with tana =

0.75). It is noted that the flows of
 -
184 -

the clay suspension experiments performed at the clay concentration C3

approached the limit of viscous grain shearing (Re, =

10; s sec. 2.1.3).
The largest bed load transport rates were measured at the level C3, and
it is interesting that the corresponding points in Fig. 5.2 lie close to

the 100% efficiency line (B).

y^
^y^
*

xC**4
i

iJN/s m]
^<S\
^^ ^l\
*
^
A

^y # t\
x
^y r • -»
'
r*
'

A'
\A^ X *

A
x
.* »**

el
Uniform grain
Data set Mixture
size

a Smart/Jaggi » X

Rickenmann •

1-1 1 III
100 1000
<o[N/s m]

Fig. 5.2 : Experimental data from steep flume tests (data sets SJ

and RI), shown in terms of the bed load transport rate by

immersed weight per unit width,
iR, and the stream power,

co = x -V. The straight lines represent the 100% effi¬

ciency relationship, equ. (5.8), for a slope of 20% and

(A) for inertial grain shearing (tana =
0.4) and (B) for
viscous grain shearing (tana =
0.75), respectively.

9 relationships for high shear stresses

Yalin (1977) demonstrated that many bed load transport formulae

3/2
predict the proportionality relation #_ ~
9 at high shear stresses.

This same result is also obtained from equ. (4.28) and (4.26), for con¬

stant Fr and s. If the thickness of the transport layer is small with

 -
185 -

respect to the total flow depth and if a flow over a plane bed is con¬

sidered, then Fr may have no influence on #_. In Fig. 5.1, the simpli¬
fied Meyer-Peter/Miiller equation (2.67) is seen to predict *_ values
within the range defined by the steep flume data. A correlation calcu¬

lation with the data sets SJ and RI gives:

°*87
» 8 (e e X5 r2
*B =
8

(9r
-

8cr) _ 37Z
a

However, as is also seen from Fig. 5.1, the steep flume experiments
alone define a relationship in terms of (9-8 ) with an exponent somewhat

higher than 1.5; a regression computation results in:

B - 8.9
(9r -

9cr)1-9 l':%lA (5.9)

Takahashi (1987) defined an "immature" debris flow as a flow with a

clear water layer and a transport layer underneath, in which dispersive

forces between the grains dominate. According to Takahashi, some of the

Smart/Jaggi tests fall within the category of immature debris flows.

Based on experiments with quartz sand in water, he gave a critical con¬

centration within the transport layer as

0.4C^ below which the grains
cannot disperse throughout the flow depth. Assuming a value of C^ = 0.75

for the steep flume experiments, the critical concentration would be

about 0.30. However, larger concentrations (defined with respect to the

total depth, and not to the thickness of the transport layer) were

determined for some of the steep flume tests. For the clay suspension
experiments there was no indication that the grains were transported in

significant concentrations within the upper fluid layer; this is con¬

cluded from the conductivity readings of the bed level measurements, and
from several checks by holding one hand into the muddy grain-fluid
mixture.

Considering the mode of transport, it may be concluded that some of

the steep flume tests (SJ and RI data) were in the regime of Takahashi's

immature debris flows, although calculated bed load concentrations are

not quite in agreement with the values proposed by Takahashi. The bed
load transport equation proposed by Takahashi (1987), equ. (2.89b), is

shown in Fig. 5.1 for comparison with the steep flume data; the values

tang = 0.20 (S =
20%), C^ =
0.65, 9 =0.05 were used to determine the
 -
186 -

lines (a) with tana =

0.6, and (b) with tana = 0.8. It was noted by
Takahashi (1987) that his equ. (2.85), which is similar to the

Smart/Jaggi equation, is also valid for immature debris flows.

According to Takahashi's (1987) equ. (2.89b), *„ should depend on

2 5
8 at high shear stresses (9 » 9 ). This is a somewhat stronger de¬

pendence than suggested by equ. (5.9), which was fitted to the steep

flume data. Hanes and Bowen (1985) theoretically developed an equation

for intense bed load transport, defining a transport layer zone domi¬

nated by grain collisions and a saltation zone where both grain and
fluid stresses are important. Their results also lead to the conclusion
2.5
that *_ may become proportional to 8 at high shear stresses.

5.2 Flow resistance

Analysis of a Bingham fluid with Newtonian formulae

It is shown in section 4.2 that the clay suspension used in the ex¬

periments can be characterised as a Bingham fluid. The tests without

sediment transport are analysed in section 4.3. The analysis suggests

that the flow characteristics can be reasonably well described by using
an effective viscosity y . (defined by equ. (2.40)) in combination with

flow resistance formulae developed for Newtonian fluids.

The flow resistance of a laminar flow of a Newtonian fluid is given

by:

f (5.10)
-

Klan/Re

where for pipe flow, K. = 96 is a constant. It is seen in Fig. 4.5

that the data points in the laminar flow region lie below the theore¬
tical relationship for Newtonian pipe flow. Straub et al. (1958) noted
that the channel geometry is an important factor determining the value

of K, in the above equation. They showed theoretically how K. varies

for a rectangular channel with the aspect ratio (W/h). For the experi¬

mental conditions of the data shown in Fig. 4.5, K. would be expected

to assume a value between 60 and 77. This is in good agreement with the
 -
187 -

lines defined by the data in Fig. 4.5. Zhang et al. (1980) also analysed
the flow of a Bingham fluid using y .
and Re.; for experiments made in a

rectangular channel, they found that a value of K, = 84 fitted their

lam

laminar flow data the best. Kozicki and Tiu (1967) presented a theore¬

tical analysis of the laminar flow of Non-Newtonian fluids, and they

demonstrated that the aspect ratio has an influence on the flow of a

Bingham fluid through a rectangular channel.

With regard to the transition from the turbulent flow of the clay

suspension to laminar flow, the hydraulically smooth turbulent flow

region seems to be hardly detectable for the given flow conditions.

According to Fig. 4.6 no viscous effects on the flow resistance could be

detected in the region Re. >= 10 (which corresponds to hydraulically

transitional or rough turbulent flow for a Newtonian fluid). A plug flow
zone already started to form, however, for a few flows in this region,

extending from the middle of the flow surface out- and downwards; this
*
is a characteristic feature of a Bingham fluid. In the region Re. <= 10,

the flows should be first in the hydraulically smooth turbulent regime,

before changing to laminar flow (according to Newtonian fluid mecha¬

nics). According to the experimental results, some flows (at steep

slopes, with small depths) appeared to have become laminar with in¬

creasing visosity, without a clear notion of hydraulically smooth tur¬

bulent flow (Fig. 4.5 and Fig. 4.6). This particularity could be ex¬

plained by the fact, that there is not only a growing viscous sublayer
at the bottom, but also a (laminar) plug flow zone near the surface,

increasing with increasing Bingham yield stress; if the (relative) depth

is small, the two laminar zones may soon merge with growing viscous

effects in a Bingham fluid, and may thus confine the hydraulically

smooth turbulent region somewhat.

The fact that any viscous effects may be neglected in the region Re.

>= 10, is also supported by the analysis of the bed load transport ex¬

periments: A bed load transport formula was developed which describes

equally well both clear water data and clay suspension tests with Re.

values down to about 10.

 -
188 -

Steep flume data with bed load transport; comparison with other data and

flow resistance equations

In the combined analysis (sec. 4.8) with the experiments of Smart and

Jaggi (1983), only those bed load transport tests with the clay suspen¬

sion are considered for which viscous effects may be neglected

(Re* >= 10).

In the analysis of their steep flume experiments, Smart and Jaggi

(1983) proposed to apply a logarithmic flow resistance formula, equ.

(4.7). In another study, Jaggi (1983) had found that velocities tend to

be overpredicted by Nikuradse type equations for lower relative flow

depths and steeper slopes in flows over mobile gravel beds. He therefore

introduced a correction factor to account for form drag losses. This

correction is a function of the slope and of grain characteristics; it

is given by the term in square brackets in equ. 4.7.

Recently, Griffiths (1989) presented an approach to predict the flow

resistance in gravel channels with mobile beds. Similarly to the cor¬

rection of Jaggi (1983), he suggested to divide the total energy slope

or friction factor, f, into a part representing pure grain friction, f,

and a part accounting for form drag, f" ; the part f" can be due to

either bed form drag or viscous drag on transported grains. He calcu¬

lated the flow resistance due to grain friction, f', with a Nikuradse

type equation:

^ = 2.12 + 2.03
log(j-)
m
(5.11)

Using the data sets of Meyer-Peter/Muller and of Smart/Jaggi, he deter¬

mined f" = f -
£', where f is given by the experimental values. He then
showed how the ratio t"II' varied with total dimensionless bed shear

stress, 6, and relative depth, h/d. He found that different patterns

should exist for the subcritical and the supercritical flow regime. For

the supercritical case Griffith (1989) gave the relation:

, ..
,-0.066,11 ,1.30, XX
,,
f = f (1 + exp[ ] (-5—)
,_
(5.12)
,,,

' 455(g-)
9 m c

The procedure of Griffiths (1989) was applied to the steep flume data

(SJ and RI), and a comparison between calculated and measured velocities
 -
189 -

is shown in Fig. 5.3. It may be noted that there is a systematic

deviation between V and V .

for the clay suspension data with
meas. calc.

increasing velocity. This may be due to the fact that the relatively

high transport rates in these tests result in a too high a factor (9-

9 )/9 in equ. (5.12).

o.u-

Vcalc^S]

d.O'
' =

Data set
ni+exp[-|^(^)^o](Mo)}
Uniform grain
size
i

Mixture
i
/
Smart/Jaggi & X

*
Rrckenmann

X
X

d.\)~
X X
A c

X 4
/

Xx
X
4
X

X
/ 4,
Xx X

a
4
x

X /
4 1
/ 4 X
X /

4 X* ,
i » X
a

* x X4 • * &
/ .

4.
" 4 X4
*
X
/ 4
.
X *.
"a
/ x •
* *
* * *
*
» *
» • * X «
Xx < 4 *
*
* • •

0.5h
0.5 1.0 1.5 2.0 2.5 3.0

5[m/s]

Fig. 5.3 : Comparison between measured velocities, V and those

calculated according to Griffiths (1989), V .

,
for the

steep flume data.

It may be noted that according to Yalin (1977), the validity of the

logarithmic velocity distribution in open channel flow is questionable

at relative flow depths below about 6 to 7. For this case, measured

velocity profiles indicate a deviation from the logarithmic distribution

 -
190 -

at the bottom layer. Kamphuis (1974) stated that it may be doubtful to

use the equivalent sand roughness k as a measure of flow resistance for

relative depths below about 10.

Bathurst et al. (1982b) reported on a study about bedforms and flow

resistance in steep gravel bed channels. For flume slopes between 3% and
9%, bed load transport was generally initiated at Froude numbers above

1. For a flow with constant discharge, and only minor variations in

depth and velocity, they found the flow resistance coefficient c =

'

(8/f) to decrease with increasing slope and bed load transport. Since
the bed forms changed from plane bed to antidunes and then weak anti¬

dunes, they concluded that the increase in resistance is due to grains

in transport. On the other hand, for flows with no or little bed load

transport, c was shown to be mainly related to the relative depth.

A similar conclusion was drawn by Cao (1985). He obtained flow resis¬

tance measurements from experiments with a mobile gravel bed, with vary¬

ing bed load transport rates. He presented his data in a plot of c

versus the parameter V/(gd.„) ,

called grain Froude number or sediment

mobility number. In such a plot, c is seen to decrease with increasing

sediment concentration in the flow, at a given value of the sediment

mobility number. The same behaviour was shown to apply to data taken

from Cooper and Peterson (1969), and to the data of Smart and Jaggi

(1983).

A tendency towards an increase in flow resistance (decrease in c)

with increasing transport rates (at a given relative depth h/d..) can be

inferred from the clay suspension experiments at slopes steeper than

about 10 X (s. Fig. 4.24). To judge from the steep flume data, a de¬

crease in c with increasing q already at moderate slopes (with weaker

transport rates) must appear somewhat doubtful.

Bathurst (1985) reported on a study on the flow resistance of gravel

and boulder-bed rivers with slopes between 0.4% and 4%. The relative
depths of the examined flows varied between 0.6 and 11.5, and all flows

were subcritical. No sediment transport occurred in these flows. He

presented his data in diagrams in terms of c versus h/d.,, together with

field and laboratory data from other sources with h/d., < 50, including
also flows with sediment transport. In the region of the sediment trans¬

porting flows (approximately defined by h/d., >= 5), the envelope of the
 -
191 -

available data shows that c may vary by about as much as a factor of 2

for a given value of h/d... If the steep flume data of this study is

plotted in a similar diagram it is clearly seen that the relative depth

alone cannot explain the variation of c.

In a study on the hydraulics of torrents, Meunier (1988) performed a

regression analysis with the field data given by Bathurst (1985), and

presented the following equation:

3*9
"
S°*289h (5.13)
.0.56
,.
(d84)

for which he determined the correlation coefficient squared as r2 =

0.87. Putting (d84)°*56 =

(d90)°'5 and substituting 3.9 = 1.25-g0*5, a

dimensionally correct version of equ. (5.13) can be obtained which can

then be transformed into:

,8,0.5 =
1.25 ,h ,0.5 ... ,.,

^OT <d^>
(5*14)
<f>

Interestingly, this formula is very similar to equ. (4.29). Both equa¬

tions are of the form:

,8,0.5 -
, ,h,0.5 ,. 1K,
»'
(5.15)
(3) (g)

where m' is a function of the slope. Equ. (5.14) and equ. (4.29) are

compared in Table 5.1, where the corresponding values for m' are given
for the different slope values.

For the slope range of this study, equ. (4.29) is seen to predict
velocities about 40% to 60% higher than those calculated with equ.

(5.14). However, it is not this difference that is surprising but rather

the fact that both equations are of very similar form although one was

developed from subcritical flows over a rigid bed while the other

applies to a supercritical, bed load transporting flow over a mobile

bed.
 -
192 -

Slope S Steep flume data Bathurst (1985) data,

of this study, analysis by Meunier
equ. (4.29) (1988), equ. (5.14)
rai
mi m'
mi

0.01 5.7 3.3 1.73

0.05 3.6 2.3 1.57

0.10 2.9 2.0 1.45

0.15 2.6 1.86 1.40

0.20 2.4 1.75 1.37

Table 5.1 : Comparison of flow resistance formulae, derived from

steep flume data and field observations by Bathurst,

respectively.

From steep flume experiments classified as "immature" debris flows,

Takahashi (1987) derived the empirical expression:

(8)0.5 _ 1>5 (h)0.56 (5.16)

where d probably refers to d (no exact definition is given). Putting

^
0 '
56
(h/d) b
(h/d ), equ. (5.16) is again of the same form as equ.

(5.15). The following formula is obtained from a regression of the steep

flume data (SJ and RI), using d instead of d..:

0.45
(8)0.5 0.27 M
; (5.17)

for which the correlation parameters between V and V n are: r2 =

meas«, calc.

0.66 and S„ - 18%. The factor m' = 0.45/S is compared with

Takahashi's constant 1.5 in Table 5.2, for slope values at which

"immature" debris flows may occur.

 -
193 -

Slope S Steep flume data

of this study,
equ. (5.17)
ra3
m3 1.5

0.10 2.3 1.53

0.15 0.82 1.40

0.20 0.87 1.30

0.25 0.92 1.22

Table 5.2 : Comparison of a flow resitance equation from this study

with a formula developed by Takahashi (1987) for intense

bed load transport conditions, equ. (5.16), for which

m' = 1.5.

It is observed from Table 5.2 that the equation developed from the

steep flume data of this study (equ. 5.17) predicts velocities that are

20% to 50% higher than those calculated with Takahashi's (1987) formula
(equ. 5.16) for intense bed load transport conditions. It cannot be

decided whether this difference is a result of different measuring tech¬

niques, or whether it is caused by different flow conditions.

In conclusion, it may be noted that there is still a considerable

scatter between predicted and measured velocities for the steep flume
experiments of this study, irrespective of which approach is used to

calculate the flow resistance. It is therefore recommended to use the

flow resistance equation given by Smart and Jaggi (1983), equ. (4.7), if
the fluid velocity is to be estimated as a function of the mixture flow

depth (and the slope). On the other hand, if the flow rate is given or

assumed (as in the design case), it is more reliable to use equ. (4.30),
in a slope range between 5% and 20% (s. also sec. 6.2), or equ. (5.20,

presented below) in a slope range between 5% and 50%.

 -
194 -

Comparison with Takhashi's debris flow velocity equation

Based on the concept of dispersive stresses between the moving

grains, Takahashi (1978) developed an equation for the mean velocity of

a quasi-steadily moving debris flow front (integrated form of equ. 2.12,
section 2.1.3). Replacing concentration and density terms together with

grain shearing coefficients by one single parameter A*, the following

expression is obtained (Takahashi, 1978):

[q /(gd )]

where U is the mean velocity of the moving grains and q is the fluid

discharge per unit with. Takahashi plotted his experimental results in

terms of the nondimensional velocity (given by the left-hand side of

equ. 5.18) vs. the bed slope tang. As predicted by equ. (5.18), there is

only a slight dependence on tan6 in his figure. From his diagram (Fig.
9, Takahashi, 1978), it can be determined that A* = 1.3. Using this
experimental value for the parameter A*, equ. (5.18) can be transformed

into:

q /
• a\0.2 0.6 0.2 .
,0.4
,. ..,
ii
U =
i
1.3 (smB) q g / d (5.19)

Comparing this relation with the empirical equ. (4.30) derived in this

study, a remarkable similarity is evident. In fact, Fig. 5.4 shows very

good agreement for the steep flume data between measured and calculated

fluid velocities; putting sing = tang =

S, d = d.. and U =
V, equ.

(5.19) becomes:

.0.2 0.6 0.2 0.4 r2 = 0.79 ...

„ ,1.3, S /
V
=
q g
d9Q (5.20)
s = 13_g %

where the reduced discharges q were used to compute the statistical

parameters r2 and S„. Interestingly the fluid velocity of a bed load

transporting flow at slopes 0.05 <= S <= 0.20 may be predicted by the
same formula as the velocity of the front of a debris flow. The cor¬

responding debris flow experiments of Takahashi (1978) cover a slope

range 0.17 <= S <= 0.47.

 195

o.u-

4
X
X /
calJm/S] | V =
1.3S02qr06g02d9004| X

4 /4

X
** *
A1
4
/ x
x
•
•» Xx 4
X* .
X
<
44 4
y
.

4 / *

X X 4

4 X 4 *

4' **X^
*

4 +*/£ *

Data set
Uniform grain Mixture
4
size
O 4/1"
4 X
<X
Rickenmann *

/•*.

0.5-
0.5 1.0 1.5 2.0 2.5 3.0

Vmeas[m/S]

Fig. 5.4 : Comparison between measured velocities, V ,

and those
meas.

calculated with equ. (5.20), V for the steep flume

, ,

data.

Analysis of clay suspension experiments

Increasing clay concentrations in the suspension used by the author

produced an interesting result: For a given slope and flow rate, the bed

load transport rates increased considerably with increasing fluid

density, particularly at steeper slopes (Fig. 4.13). This change was

associated with only a minor increase in flow depth and in flow velocity
(Fig. 4.11), implying that the flow resistance did not significantly
change (Fig. 4.10). It can be concluded that the fluid velocity is

largely determined by the slope and the flow rate. Since these two
 -
196 -

parameters also determine the bed load transport rates (s. for example

equ. 4.23), any effects of a bed load on the fluid velocity could be

implicitly accounted for by q and S. This may explain the rather good

performance of equ. (4.30), in which the fluid velocity is related to

the slope, the flow rate and a characteristic grain size.

Equ. (4.30) is based on flume tests with relatively high bed load

transport rates. Somewhat larger velocities were measured for the flows

over a fixed rough bed without sediment transport (for the same values

of q and S); however, the velocities were found to be only about 16%

higher on the average for the case of a rigid bed (s. Fig. 4.27). A

closer examination of Fig. 4.27a and b reveals the following trend: For

lower values of q and S (imlying weaker transport rates), the measured

velocities are smaller in the movable bed case while for larger values
of q and S they seem to approach those of the rigid bed case.

As a rough approximation, it is proposed that the turbulent flow in a

steep rough channel is mainly governed by the slope and the flow rate

and the fluid density. For a given q and S, larger depths were observed

for the flows over a movable bed than for the flows over a fixed bed.

These larger flow depths (or shear velocities) are associated with the

space required for the transported grains. The only minor change in

fluid velocity suggests that the adjustement of the (mixture) flow depth
could be the primary mechanism by which the flow supplies the energy

required for grain movement. It may be hypothesised therefore that the

fluid -
grain mixture is a self -

regulating system, adjusting flow

depth and transport rates in such a way that the fluid velocity is not

greatly affected, but instead is largely determined by the slope and

flow rate.

5.3 Effects of high fine material concentrations; implications for

field situation

Effect of increasing viscosity

At clay concentrations C, above about 16%, a decrease in the bed load

transport rates was observed with further increasing Cf values (s. also
 -
197 -

Appendix II). This decrease was found to be associated with a grain

Reynolds number Re, below about 10 (Fig. 4.9) which implies that the
thickness of the laminar sublayer, S, is greater than the grain size, d.

A similar conclusion can be drawn based on a study about the hy¬

draulic transport of solids in pipes by A.D. Thomas (1979a,b). The

transport capacity of solids can be expected to decrease if 8 >= d;

then, according to equ. (2.80), the critical gradient at deposition
increases linearly with increasing fluid viscosity. Interestingly, a

linear decrease in bed load transport rates with increasing viscosity

y - of the clay suspension is suggested by the case II experiments of
this study (Fig. 4.33).

Negative lift forces may be an explanation for this decrease. Based

on experiments, Davies and Samad (1979) found the lift forces to de¬

crease with decreasing particle Reynolds numbers; they measured negative

values for Re <<* 5. Similar experimental results were reported by
Coleman (1967).

It is noted that the case II experiments of this study were in the

hydraulically smooth turbulent regime. For a given pressure gradient in

pipe flow, A.D. Thomas (1979b) concluded that the transport capacity
should be smaller in laminar than in turbulent flow. Since the flow

resistance increases strongly in laminar flow with increasing viscosity,

larger flow depths will result in open channel flows (for a given flow
rate), implying an increase in shear stress. With increasing viscosity,
the bed load transport rate is expected to decrease; however, a part of
the grains may then be transported in "suspension" (not supported by
turbulence but due to a very small settling velocity), and the total

load was found to be increased in laminar flow (sec. 2.3.1). It is

difficult to predict what the combined effect will be on the bed load

transport capacity in laminar open channel flow. Experimental evidence

(Bradley, 1986a; Wan and Song, 1987) suggests that the total transport

capacity may be considerably higher in laminar flow.

At steeper slopes, a mechanism called "autosuspension" may become

important (Bagnold, 1962). When the the bed slope is higher than the
ratio W/U_, then the gravity component of the moving grains contributes

enough energy to maintain the grains in "suspension". Using calculated

settling velocities (W = 0.75 m/s in clear water, W = 0.14 m/s at
 -
198 -

concentration level C4) and mean grain velocities (UR)j this mechanism

seems possible for the steepest slopes at the the clay concentration
levels C3 and C4.

Scaling considerations

In the case II experiments, the bed load transport rates decreased by

about a factor of 3; this decrease is assumed to be a result of a cor¬

responding viscosity increase by about the same factor (s. also Fig.
4.33). The effective viscosity y .
is given as:

V
y
e2 \ +
fv (2-40)

For the given "mean" shear rates (2-V/h) and fluid properties of the

case II tests, it can be concluded that y .

is mainly determined by the
Bingham yield stress T_ because the Bingham viscosity t|_ accounts for

less than 10% of the right hand side of equ. (2.40). Therefore, the

effective viscosity may be approximated in this range as:

TBh
»e2
-
o (5-21>

The experimental flows in the turbulent regime can be scaled to the

field situation according to Froude scaling. If the length scale is

'

defined as h„ = h„ /hu then the velocity scale is V_ =

(h„)
, ., ' v .

R „
Nature Model R R
Given the same Tn value in the model and in nature and using equ.
0.5
(5.21), the viscosity scale becomes: (y ,). = (hD) . This allows the
. ez k k

scale of the grain Reynolds number Re, to be determined:

(hR>0'5
* * hR
<Re2>R= 0.5 -\ <5-22>
(hR)

Considering a field situation with the same fluid and flow properties,
it can thus be concluded that Re. will be greater in nature, despite a

larger effective viscosity y „. At the same fine material concentration

in the flow, somewhat smaller values of the Bingham parameters can be

expected in the field situation because of a smaller proportion of very

fine particles. Therefore, higher fine material concentrations will be

 -
199 -

required in nature to produce a critical grain Reynolds number Re. below

10. This conclusion is in general agreement with a critical value of

Cf = 30% suggested by Davies (1988) above which a flow in nature might

*
it
be in the macroviscous regime (i.e. have Re. <= 10); this value of
Cf
was estimated by Davies based on field data concerning the Bingham
parameters of fine material suspensions.

Wide grain size distribution

It should be pointed out, however, that the field situation is com¬

plicated by the very wide grain size distribution of the material

available for transport, in the case of flood conditions in a torrent

catchment. Viscous effects will influence the transport of smaller

particles already in a less concentrated slurry of fine material. With

increasing fine material concentrations in the flow, the small particles

may soon become transported in suspension, while intermediate sized

grains might be deposited when the flow around them becomes laminar. It

may also be noted in this context that the grain size distribution

parameter d../d,. assumed a maximum value of 8.5 in the experiments of

Smart and Jaggi (1983). It is recommended to use only values of roughly

d-./d,. <= 10 in the proposed bed load transport equations.

Pulsing behaviour of debris flows

Three mechanisms were proposed by Takahashi (1981) for the initiation

of a debris flow: (a) a landslide from a hillslope may transform into a

debris flow, (b) at sufficiently steep slopes (> =

27%) and with large
bed shear stresses, material deposited in a channel bed may be mobilized

and eventually form a debris flow, (c) accumulation of debris material

from the sideslopes can cause a temporary dam to be built up in the

channel, and a debris flow may be initiated when this dam breaks.

It was noted by Davies (1988) that a temporary blockage of large

grains can occur in a uniform flow of a grain -

fluid mixture in a

steep, narrow channel. He further pointed out that the development of

roll waves may cause the pulsing of a debris flow; the critical Froude

number for the occurrence of roll waves is much smaller in laminar than

in turbulent flow. Based on field data, Davies showed that the limiting
 -
200 -

conditions developed for clear water flows, also apply generally for
debris flow conditions. Recently, Savage (1989) demonstrated theore¬

tically that the critical Froude number for the occurrence of roll waves

decreases with increasing cohesion, viscosity and particle interaction.

In the light of the experimental results of this study, the pro¬

nounced decrease of the bed load transport rate in macroviscous flow may

also be seen as an element enhancing the instability of a uniform flow.

In the range of fine material concentrations C, between 20% and 50%, the

Bingham yield stress T_ increases strongly with C. (sec. 2.2.1). The

transition to macroviscous flow may be expected to occur in this con¬

centration range (Davies , 1988). For a given flow rate in a torrent, a

local input of fine material might thus cause a sudden decrease in bed

load transport capacity, causing a part of the grains to be deposited.

If such events are frequent and/or large enough, the deposited material
could form a temporary dam. In contrast to direct deposition of material
from the sideslopes, a partial or full blockage of the flow might there¬

fore also be caused by a change in flow properties.

 -
201 -

6 SUMMARY AND CONCLUSIONS

6.1 Summary

In a debris flow, big boulders and stones are transported in a slurry

of finer material, which typically consists of a mixture of sand, silt

and clay particles. It has been shown that these slurries often show a

Bingham type rheological behaviour, which is different from that of a

Newtonian fluid such as clear water.

The objective of this study was to investigate the transport of

coarse bed material in a slurry of fine material. This mode of transport

may be expected in the transition region between a "normal" flood with

clear water and approximately uniform flow on the one side, and the

unsteady, pulsing debris flow on the other side. In the experiments, a

clay suspension of various concentrations was recirculated in a 20 cm

wide and 5 m long flume. For slopes ranging from 7% to 20% and for flow

rates between 10 l/s and 30 l/s, the equilibrium bed load transport
rates were determined. A rather uniform gravel mixture with a mean

diameter of 1 cm was used as bed material. The clay suspension showed

increasing non-Newtonian characteristics with increasing concentration.
Its rheological behaviour was approximated as a Bingham fluid. The

maximum density of the suspension was about 1.36 g/cm3 (corresponding to

a volume concentration of 22%), and the maximum effective viscosity

(defined analogous to the Newtonian viscosity) reached 1800 cps.

In the hydraulically rough turbulent or transitional regime (Re. >=

10), no viscous effects on the bed load transport rates could be de¬

tected. However, the decrease in the grain -

fluid density ratio brought
about a marked increase in bed load transport rates with increasing clay
concentration, by as much as a factor of 3 as compared to the clear

water case. According to a study by Low (1989) it appears that this

change can be correctly predicted if the ratio of the shear velocity to

the fall velocity is adequatly represented in a bed load transport for¬

mula; if viscosity effects can be neglected, the density factor (s-1)

may be substituted for the fall velocity. A regression analysis of the
present experiments showed that the observed bed load transport rates
 -
202 -

can be fairly well predicted by an adequate inclusion of the density

factor. For a given slope and flow rate, the calculated mean bed load

concentrations increased considerably with increasing clay concentration

while there was only a minor increase in the corresponding mean grain
velocities.

A regression analysis of the bed load transport measurements (with

the clay suspension) resulted in a relationship in terms of q. =
<|>(q ,

S, S , s-1). A reanalysis was made of the steep flume experiments of

Smart and Jaggi (1983), using a slightly different procedure in the data
preparation. The resulting equations were of a very similar form, and

therefore a formula could be derived which applies to both data sets; a

grain size distribution parameter, d../d,., was also included in the

form as proposed by Smart and Jaggi. A separate regression analysis was

performed with the Meyer-Peter and Muller (1948) bed load transport

experiments which were carried out at lower slopes (data given in Smart
and Jaggi, 1983). The exponent of the slope factor was found to be

different, and it was therefore concluded that the steep and low slope
data should be described by two separate equations in the q„(q) -
form.

In a second step, a more general relationship was developed in terms

of the parameters *_ =
<K9, 9 , Fr, s-1, d../d,.). The resulting bed
load transport formula can be applied to all three data sets considered

(Meyer-Peter/Miiller, Smart/Jaggi and author). This relationship appears

to be valid over a wider range of conditions than an equation in terms

of the first parameter set, because it also contains parameters des¬

cribing the flow behaviour (i.e. 9 and Fr). It should however be pointed
out that the role of the Froude number is different for the low and

steep slope case. For the Meyer-Peter/Miiller experiments, in which bed

forms were noted, Fr can be considered as a measure of that part of the

(total) flow resistance which is due to bed forms and is thus not avai¬

lable for bed load transport. In the case of the steep flume experiments
(of Smart/Jaggi and of the author), antidunes and plane beds were pre¬

sent; they caused probably not much form drag. In the tests with high
Froude numbers, bed load grain concentrations were so large that several

grain layers were in motion, occupying a considerable part of the total

flow depth. Since 9 is only a measure for the shear stress at the

interface to the quasi-stationary bed, the Froude number may in this

case be a measure for the velocity of the upper, faster moving grain
layers.
 -
203 -

In the hydraulically smooth turbulent regime (Re, <=* 10), the bed

load transport rates decreased with increasing clay concentration. This

decrease was found to be linearly related to the increase of the fluid

viscosity. No bed load tests could be carried out in the laminar flow

regime.

To study the effect of the increasing non-Newtonian properties with

increasing clay concentration (i.e. increasing Bingham yield stress and

Bingham viscosity) on the flow, additional experiments were performed

with the suspension flowing over a fixed rough bed without any sediment

transport. The flow resistance analysis indicated that Newtonian for¬

mulae can still be used if an adequate viscosity is defined, such as the

effective visosity y ».
No viscous effects could be detected in the

hydraulically transition region, which is in agreement with the analysis

of the bed load transport tests. The hydraulically smooth turbulent

regime appears to "obscured" by the formation of a plug flow. This is a

zone with no shearing between adjacent fluid layers, extending from the

flow surface downwards; with increasing Bingham yield stress of the

fluid, this zone occupies increasing proportions of the flow cross-

section. When it merges with the viscous sublayer, the flow becomes

laminar.

As noted by Smart and Jaggi (1983), the flow resistance of the sedi¬

ment transporting flows cannot be determined by the relative depth

alone. Including a slope dependent correction term, their formula re¬

duces the available (energy) slope in order to account for additional

resistance due to transported grains. However, there is still a large

scatter between predicted and calculated velocities, which can be only
partly attributed to inaccurate depth measurements. An alternative

equation for the fluid velocity was developed as a function of the flow

rate, the slope and a characteristic grain size. A theoretical equation

was proposed by Takahashi (1978) to describe the velocity of a debris

flow front; introducing an experimentally determined parameter, this

formula becomes very similar to the one of the author. If applied to the

steep flume data of this study, it shows even a better performance than
the author's equation. It is therefore recommended to use this semi-

theoretical formula if the velocity is to be calculated as a function of

the flow rate.

 -
204 -

6.2 Recommended calculation procedure

The equations presented in this study allow the unknown parameters of

a bed load transporting flow to be estimated. In engineering applica¬
tions, usually either (A) the flow rate or (B) the mixture depth is
given to calculate the other required parameters; in both cases, the

slope and the grain size characteristics can be obtained in the field,
while an appropriate value for the fluid density has probably to be

assumed.

Different equations were developed to determine the unknown parame¬

ters of the grain-fluid flow. If one particular set of formulae is used,

the quality of the predicted parameters will not be the same whether the

calculations are made for case (A) or (B). It is therefore proposed to

use two different sets of equations, which should give best estimates of

the unknown parameters for each of the two cases. The two recommended

calculation procedures are presented in the following figures.

If two out of the three parameters (q, V, h,) are known, then the

fluid flow is determined. In the case (A), q is given, and the calcu¬

lation according to Fig. 6.1 results in a similar agreement between

predicted and measured mixture flow depths, as when applying the

iterative procedure given in Smart and Jaggi (1983). In the case (B),
however, the computations are based on a parameter which characterises

the flow of the grain -

fluid mixture. Then, two fluid flow parameters

have to be calculated, s. Fig. 6.2, and the agreement between predicted

and measured q values is not as good as in case (A) for h . Since it is
m

the fluid flow rate that largely determines the flow of the mixture, a

different quality of the results can be expected, depending on the ap¬

plied procedure.

It is pointed out that the procedure (A) should only be applied to a

slope range between 5% and 20%, because the bed load transport equation

(4.24) was derived from the steep flume data. It has been shown that an

equation of the form of (5.1) is less general than one in terms of

(5.2). Furthermore, the velocity formula (5.20) is based on steep flume

experiments that cover a limited range of relative flow depths.

 -
205 -

Procedure (B), on the other hand, may be used both for steep and low

slope situations (slope range 0.2 X <= S <= 20 %). The bed load trans¬

port equation (4.28) was developed from the Meyer-Peter/Muller and the
steep flume data. The logarithmic flow resistance formula (4.7) can be

applied generally; the correction factor in square brackets becomes

unity for low slope values and for large relative flow depths. Also, no

distinction between the mixture depth and the fluid depth is necessary

at low slopes and low bed load concentrations, i.e. equ. (4.23) and

(4.34) can be omitted for slopes less than about 5 %.

In a torrent situation, procedure (A) can be used if the the flow

section of check dams or if an artificial channel on the fan are to be

designed according to a maximum flood discharge. The procedure may be

further applied to estimate transported sediment volumes during a flood

event if a hydrograph is known or determined from rainfall intensity

data.

A typical application of procedure (B) is given, when the (mixture)

flow depth at a defined cross-section can be determined from flood marks

but no information is available on the flow rate. Of course, one has to

be sure that these flood marks are due to a more or less steady uniform

flow and were not caused by a debris flow.

In both procedures, it is assumed that there is a full, unlimited

supply or availability of bed material. This requirement might not be

satisfied particularly in situations where an armour layer of coarse

material is not yet broken up. In this case, equ. (2.113) may be used
instead of equ. (2.111). But it should be remembered that the bed load

transport rates calculated by either of the two procedures give values

corresponding to the (maximum) transport capacity.
 -
206 -

(A) "Design case": given S, (s-1),

: q,
d9Q, d5Q, d3Q
unknown : qn, h ,
h,, V
d m t

Calculation scheme:

n r^ , 1x1-67 0.5 ,1.5 .-1.12

S (2.111)
qcr = 0.065
(s-1) g
d5Q

12.6 ,90,0.2 .2.0

, ,
q° S (4.24)
TO
"

(Tr) (q
qcr)
(s-1)' 30

.
.0.2 0.6 0.2 .0.4
„ , .

S q g /
(5.20)
d9Q
V =
1.3

hf =
q/V

SL

h - hf / [
l
1 -
1.64 S°-42 (V*63 (4.34)
m f yq
'

Fig. 6.1 : Proposed calculation procedure for the "design case" (A).
Recommended range of application: 5 % <= S <= 20 %,
q >= 5-qr h
m
/d.-
90
<= 20. (A sidewall correction procedure
r

may be applied additionally.)

 -
207 -

(B) "Field case": given :

hm, S, (s-1), i9Q, dm, d3Q
unknown : q, h,, V
qR,

Calculation scheme:

•lV.0.5, ,12*3hm
V = 2.5
(ghmS)0*5 [1 -
exp( 0.5
)] ln(
ei d90
)] (4.7)
d90S

*B = -1^3 (?)°'2 •30'5 (e -

ec) Fr1*1 (4.28)
(s-1)"*-5 a30
c

17.3 ,90,0.2 .2.1

(4.23)
(S-ir*u a3o

h.
f
= h
m
[1
'
1.64 S0*42 (^)°*63 ] (4.34)

H> q = V-h <H

Fig. 6.2 : Proposed calculation procedure for the "field case" (B).
Recommended range of application: 0.2 % <= S <= 20 %; ax
=

0.05 and 1.5 for S >= 5 %;

^ =
equ. (4.23) and equ. (4.34)
can be omitted for S < 5 % (then h, = h ); d >= 1 mm. (A
r m m
sidewall correction procedure may be applied
additionally.)
 -
208 -

6.3 Example calculation

Consider a channel reach in a torrent, where the following

information on slope and grain size characteristics is given:

S = 0.15

0.50
d9Q = m

d..= d = 0.15 m
50 m

0.04
d3_ = m ;

assume p = 1.15 t/m3, with cr = 2.65 t/m3 —> (s-1) = 1.30 ;

since
d.0/d,. =
12.5, take maximum recommended value of 10.

a) In the first case, the fluid discharge shall be given as q =

3.5 m3/s.m, and procedure (A) can be applied:

(2.111) q = 0.065-1.301*67.9.81°*5.0.151*5 / 0.151*12

= 0.153 mVs.m

(4.24) q. = 12.6-100*2-(3.5 -
0.153)-0.152*0 / 1.301*6
= 0.985 m3/s

- 2610 kg/s

(5.20) V = 1.3-0.15°*2.3.5°-6-9.810-2 / 0.450'4

= 4.10 m/s

f
= 0.854 m

(4.34) h = 0.854 / [1 -
1.64-0.150-42-(0.985/3.5)0'63]
= 1.28 m

b) In the second case, it is assumed that the mixture flow depth has
been measured as h = 1.4 m, and procedure (B) can be applied:

(4.7) V = 2.5-(9.81-1.4.0.15)°-5.[l -

exp( -O-05'1-^ 5)1°-5

0.45.(0.15)^
,12.3-1.4,
•
inr
*ln<1.5-0.i5)
3.84 m/s
 -
209 -

9 = h -S / [(s-l)-d ] = 1.4-0.15 /[l.30-0.15] = 1.08

m m

Fr = V / (g h )0-5 = 3.84 / (9.81-1.4)0*5 = 1.04

assume 9 =0.05
c

(4.28) *R = 3.1-10°'2-1.08°*5.(1.08 -
0.05)-1.041-1 / 1.300*5
= 4.78

qB
=
*B-[g(s-l)d3l0*5 4.78-[9.81-1.30-0.153]0*5=

= 0.996 m3/s

= 2640 kg/s

1K2.1
,A2.0
(4.23) qB/q = 17.3-10"*lyj0.2.n-CIS"--1 ,
/
,
1.30

= 0.30

(4.34) h{ =
1.4-[1 -
1.64-0.15°'42-0.300*63]
= 0.916 m

3.84-0.916
q =
V-hf =

= 3.52 m3/s.m

In comparison, it is observed that procedure (A) predicts a slightly

higher fluid velocity and a lower mixture depth than procedure (B), at

about the same flow rate and bed load transport rate. This is not sur¬

prising, because different sets of equations are used. It may be noted

that a sidewall correction could be included in both procedures.

6.4 Suggestions for further research

In this study, the effect of an increasing fluid density and vis¬

cosity on the bed load transport capacity in flows at steep slopes has
been examined. Little is still known about how flows carrying a heavy
sediment load can become unsteady and eventually turn into a pulsing
debris flow. The following aspects of sediment transporting flows in

steep channels should be studied further:

 -
210 -

-
How does the bed load transport capacity change from the hydrauli¬

cally smooth turbulent regime to the laminar regime ?

-
A wide grain size distribution can considerably affect the rheolo¬

gical properties of the slurry. What will the effect be on bed load

and suspended load transport rates, particularly in the transition

region between turbulent and laminar flow ?

-
According to the theoretical analysis of Savage (1989), roll waves

are more likely to occur in laminar flows and fluids with a high
cohesion (or Bingham yield stress) than in turbulent water flows. The

stability of such flows should also be studied experimentally.

-
In order to gain a better understanding of the mechanics of these

flows, further attempts should be made to measure velocity profiles

both of the fluid and the grains, or to track the paths of individual
particles.

-
If heavily sedimented uniform flows at steep slopes appear to become

easily unstable because of the inherent mechanical characteristics,

it would be interesting to study the effect of disturbances such as

slug inputs of debris material at different positions along a

channel. For in the field situation, failures of sideslopes may be an

important factor to trigger a flow instability. The effect of a

varying channel geometry on the flow stability could also be

examined.

-
More experiments should be performed with flows over a mobile bed at

slopes steeper than about 25%. Under these conditions, small shear

stresses are already sufficient to destabilise a loose bed down to

considerable depths; the theoretical analysis of Takahashi (1981)

suggests that the whole layer of loose material may start to move en

masse. Similar conditions are indicated by observations made by Smart

and Jaggi (1983) at flume slopes of 25%. Such experiments will have
to be carefully designed because the "erosion" depth will be very

sensitive to the shear strength of the bed (underground) material.

Recently, Tanigushi et al. (1988), reported on an experimental study

concerning the erosion capacity of a debris flow front at a slope of

40%.
 -
211 -

REFERENCES

Abbott, J.E. and Francis, J.R.D. (1977):

"Saltation and suspension trajectories of solid grains in a

water stream", Phil. Trans. Royal Soc. London, Vol. 284 A.

Abdel-Rahman, N.M. (1963):

"The effect of flowing water on cohesive beds", Ph.D. thesis
No. 3263, ETH Zurich.

Ackerman, N.L. and Shen, H. (1979):

"Rheological characteristics of solid-liquid mixtures", AIChE
Journal, Vol. 25, No. 2.

Ansley, R.W. and Smith, T.N. (1967):

"Motion of spherical particles in a Bingham plastic", AIChE
Journal, Vol. 13, No. 6.

Ashida, K. and Bayazit, M. (1973):

"Initiation of motion and roughness of flows in steep
channels", Proc. 15th IAHR Congress, Istanbul, Turkey, Vol. 1.

Ashida, K. , Miyamoto, K. and Kanda, M. (1987):

"Characteristics of hyperconcentrated flows", Proc. 22nd IAHR
Congress, Lausanne, Switzerland, Techn. Session A.

Bagnold, R.A. (1954):

"Experiments on a gravity-free dispersion of large solid
spheres in a Newtonian fluid under shear", Proc. Royal Soc.
London, Vol. 225A.

Bagnold, R.A. (1955):

"Some flume experiments on large grains but little denser than
the transporting fluid, and their implications", Proc. Inst.
Civil Engrs., London, 4(3).

Bagnold, R.A. (1956):

"The flow of cohesionless grains in fluids", Phil. Trans. Royal
Soc. London, 249A.

Bagnold, R.A. (1962):

"Auto-suspension of transported sediment; turbidity currents",
Proc. Royal Soc. London, 265A.

Bagnold, R.A. (1966):

"An approach to the sediment transport problem from general
physics", U.S. Geol. Survey Prof. Paper 422-1.

Bagnold, R.A. (1973):

"The nature of saltation and of 'bed-load' transport in water",
Proc. Royal Soc. London, Vol. 332 A.

Bathurst, J.C. (1985):

"Flow resistance estimation in mountain rivers", ASCE,
J.Hydr.Eng., Vol. Ill, No. 4.

Bathurst, J.C, Graf, V.H. and Cao, H.H. (1982a):

"Initiation of sediment transport in steep channels with coarse
bed material", Proc. Euromech 156: Mechanics of sediment trans¬
port, Istanbul.
 -
212 -

Bathurst, J.C, Graf, W.H. and Cao, H.H. (1982b):

"Bedforms and flow resistance in steep gravel-bed rivers",
Proc. Euromech 156: Mechanics of sediment transport, Istanbul.

Bathurst, J.C, Cao, H.H. and Graf, W.H. (1984):

"The data from the EPFL study on hydraulics and sediment trans¬

port in a steep flume", EPFL, Lausanne.

Bathurst, J.C, Graf, W.H. and Cao, H.H. (1987):

"Bed discharge equations for steep mountain rivers", in:
load
Sediment transport in gravel bed rivers, eds. Thorne, Bathurst,
Hey, John Wiley and Sons.

Beverage, J.P. and Culbertson, J.K. (1964):

"Hyperconcentrations of suspended sediment", ASCE, J.Hydr.Div.,
Vol. 90, No. HY6.

Bingham, B.C. (1922):

"Fluidity and plasticity", McGraw-Hill, New York.

Bradley, J.B. (1986):

"Hydraulics and bed material transport at high fine suspended
sediment concentrations", Ph. D. Dissertation, Colorado State
Univ., Fort Collins, Colorado.

Bradley, J.B. and McCutcheon, S.C (1987):

"Influence of large suspended-sediment concentrations in
rivers", in: Sediment transport in gravel bed rivers, eds.
Thorne, Bathurst, Hey, John Wiley and Sons.

Brauer, H.B. (1971):

"Grundlagen der Einphasen- und Mehrphasenstromungen", Verlag
Sauerlander, Aarau und Frankfurt.

Bridge, J.S. and Domonic, D.F. (1984):

"Bed load grain velocities and sediment transport rates", Water
Res. Res., Vol. 20, No. 4.

Briihl, H. (1976):
"Einfluss von Korngemischen auf den hydrau-
Feinststoffen in
lischen Rohrleitungen", Mitt, des
Feststofftransport in
Franzius-Instituts fur Wasserbau und Kusteningenleurwesen der
Tech. Univ. Hannover, Heft 43.

Caldwell, D.H. and Babbitt, H.E. (1941):

"Flow of muds, sludges, and suspensions in circular pipe",
Industrial and Engineering Chemistry, 33, 2.

Cao, H.H. (1985):

"Resistance hydraulique d'un lit de gravier mobile a pente
raide; etude experimentale", these No. 589, EPFL, Lausanne.

Cao, R., Chen, S., Lu, W. and Ren, X. (1983):

"The law of hydraulic resistance in density current with hyper-
concentration", Proc. 2nd Int. Symp. on River Sedimentation,
Nanjing, China (in Chinese with English abstract).
 -
213 -

Chee, S.P. (1988):

"Models and scale effects related to erosion of granular dams",
Proc. int. symp. on Modelling soil-water structure inter¬
actions, Kolkman et al. (eds.), Balkema, Rotterdam.

Chen, C. (1983):
"On frontier between
rheology and mudflow mechanics", Proc. of
the Hydraulics Division Speciality Conference on "Frontiers in
Hydraulic engineering", ASCE, Massachusetts Institute of
Technology, Cambridge, Mass.

Chen, C. (1986a):
"Bingham plastic of Bagnold's dilatant fluid as a rheological
model of debris flow?", 3rd Int. Symp. on River Sedimentation,
Univ. of Mississippi.

Chen, C (1986b):
"Chinese concepts of modeling hyperconcentrated streamflow and
debris flow", 3rd Int. Symposium on River Sedimentation, Univ.
of Mississippi.

Chen, C. (1988a):
"Generalized viscoplastic modeling of debris flow", ASCE,
J.Hydr.Eng., Vol. 114, No. 3.

Chen, C. (1988b):
"Generalized solutions for viscoplastic debris flow", ASCE,
J.Hydr.Eng., Vol. 114, No. 3.

Chu, J. (1983):
"Basic characteristics of sediment-water mixture with hyper-
concentration", Proc. 2nd int. symp. on river sedimentation,
Beijing, China, Vol. 1.

Colby, B.R. (1964):

"Practical computations of bed material discharge", ASCE,
J.Hydr.Div., Vol. 90, No. HY2.

Coleman,N.L. (1967):
"A theoretical and experimental study of drag and lift forces
acting on a sphere resting on a hypothetical stream bed", Proc.
12th IAHR Congress, Fort Collins, USA, Vol. 3.

Cooper, R.H. and Peterson, A.V. (1969):

"A review of data sediment transport experiments", Rep.
from
No. HY-1969-ST2, Dept. of Civil Eng., Univ. of Alberta.

Costa, J.E. (1984):

"Physical geomorphology of debris flows", Developments and
applications of geomorphology (eds. Costa and Fleisher),
Springer-Verlag, Berlin and Heidelberg.

Daido, A. (1971):
"On the occurrence of mud-debris flow", Bull. Disaster Pre¬
vention Research Institute, Kyoto Univ., Vol. 21, Part 2.

Daido, A. (1983):
"Incipient motion and bed load of sediment in steep channel",
10th IAHR Congress, Moscow, Soviet Union, Vol. VII.
 -
214 -

Davies, T.R.H. (1985):

"Mechanics of large debris flows", Proc. int. symp. on erosion,
debris flow and disaster prevention, Tsukuba, Japan.

Davies, T.R.H. (1986):

"Large debris flows: a macroviscous phenomenon", Acta Mecha-
nica, Vol. 63.

Davies, T.R.H. (1988):

"Debris Flow Surges -
A Laboratory Investigation", Mitt. Nr. 96
der Versuchsanstalt fur Wasserbau, Hydrologie und Glaziologie,
ETH Zurich.

Davies, T.R.H. and Samad, M.F.A. (1978):

"Fluid Dynamic Lift on a Bed Particle", ASCE, J.Hydr.Div.,
Vol.104, HY8.

Davies, T.R.H. and Jaggi, M.N.R. (1981):

"Precise Laboratory Measurement of Flow Resistance", Proc. 19th
IAHR Congress, New Dehli, India.

Dodge, D.W. and Metzner, A.B. (1959):

"Turbulent flow of non-Newtonian systems", A.I.Ch.E. Jour.,
Vol. 5, No. 2.

Drake, T.G. and Shreve, R.L. (1986):

"High-speed motion pictures of nearly steady, uniform, two-
dimensional, inertial flows of granular materials", Jour.
Rheology, Vol. 30.

Draper, N.R. and Smith, H. (1966):

"Applied regression analysis", John Wiley and Sons, New York.

Du, R., Kang Z., and Zhang S. (1986):

"On the Classification of Debris Flow in China", Third int.
Symp. on River Sed., Mississippi, USA.

Einstein, H. (1966):
"Die Scherfestigkeit kohasiver Boden in Abhangigkeit vom La-
gerungszustand und von der Materialart", Diss. ETH Nr. 3811,
(= Mitt, der Versuchsanstalt fur Wasserbau und Erdbau Nr. 71),
Zurich.

Einstein, H.A. (1934):

"Der hydraulische oder Profilradius", Schweiz. Bauzeitung,
No. 8.

Einstein, H.A. (1950):

"The bed-load function for sediment transportation in open
channel flows", U.S. Dept. of Agriculture, Techn. Bull. No.
1026.

Einstein, H.A. and Abdel-Aal, F.M. (1972):

"Einstein bed load function at high sediment rates", ASCE,
J.Hydr.Div., Vol. 98, No. HY1.

Einstein, H.A. and Chien, N. (1953):

"Transport of sediment mixture with large ranges of grain
sizes", U.S. Army Corps of Eng., M.R.D. Sediment Series, No. 2.
 -
215 -

Einstein, H.A. and Chien, N. (1955):

"Effects of heavy sediment concentration near the bed on velo¬
city and sediment distribution", Univ. of Berkley, California.

Engelund, F. and Fredsoe, J. (1976):

"A sediment transport model for straight alluvial channels",
Nordic Hydrology 7.

Fan, J. and Dou, G. (1980):

"Sediment transport mechanics
(including hyperconcentration
transportation)", Proc. Int. Symp. Riv. Sed., Beijing, China.

Fei, X. (1981):
"Bingham yield stress of sediment-water mixture with hypercon-
cetration", Jour. Sediment Research, Beijing, China, No. 3.

Fei, X. (1983):
"Grain composition and flow properties of heavily concentrated
suspensions", Proc. 2nd int. symp. on river sedimentation,
Nanjing, China, Vol. 1.

Graf, W.H. (1971):

"Hydraulics of sediment transport", McGraw-Hill, New York.

Grim, R.E. (1968):

"Clay mineralogy", 2nd ed., McGraw-Hill, New York.

Griffiths, G.A. (1989):

"Form resistance in gravel channels with mobile beds", ASCE,
J.Hydr.Eng., Vol. 115, No. 3.

Gupta, S.N. and Mishra, P. (1974):

"Turbulent flow of inelastic non-Newtonian fluids in pipes",
Indian Jour, of Technology, Vol. 12.

Gust, G. (1976):
"Observations on turbulent-drag reduction in a dilute sus¬
pension of clay in sea-water", J. Fluid Mech., Vol. 75, Part 1.

Hampton, H.A. (1975):

"Competence of fine-grained debris flow", Jour. Sedimentary
Petrology, Vol. 45.

Hanes, D.M. (1986):

"On the use of grain-flow dynamics to model intense bedload
sediment transport", ASCE Spec. Conf. on advances in aerody¬
namics, fluid mechanics, and hydraulics, Minneapolis.

Hanes, D.M. and and Bowen, A.J. (1985):

"A granular-fluid model for
steady intense bed-load transport",
J. Geophys. Res., Vol. 90, No. C5.

Hanger, M. (1979):
"Geschiebetransport in Steilgerinnen. Pilotstudie fiir feste und
glatte Sohle und Gefalle von 3 bis 30%", Mitt. Nr. 38 der
Versuchsanstalt fiir Wasserbau, Hydrologie und Glaziologie, ETH
Zurich.
 -
216 -

Hanks, R.V. (1963):

"The laminar-turbulent transition for fluids with a yield
stress", A.I.Ch.E. Journal, Vol. 9, No. 3.

Hanks, R.V. and Dadia, B.H. (1971):

"Theoretical analysis of the turbulent flow of non-Newtonian
slurries in pipes", Am. Soc. Chem. Eng. Jour., Vol. 17, No. 3.

Henderson, P.M. (1966):

"Open channel flow", MacMillan Publ. Co., New York.

Hong, R.-J., Karim, M.R. and Kennedy, J.F. (1984):

"Low-temperature effects on flow in sand-bed streams", ASCE,

J.Hydr.Eng., Vol. 110, No. 2.

Howard, CD.D. (1963):

"Flow of clay-water suspensions", ASCE, J.Hydr.Div., Vol. 89,
No. HY5.

Ikeda, S. (1982):
"Incipient motion of sand particles on side slopes", ASCE,
J.Hydr.Div., Vol. 108, No. HY1.

Innes, J.L. (1984):

"Debris flows", Progress in Physical Geography.

Iverson, R.M. and Denlinger, R.P. (1987):

"The physics of debris flows a conceptual assessment", Proc.
-

Symp. on Erosion and Sedimentation in the Pacific Rim, Cor-

vallis, IAHS Publ. No. 165.

Iwagaki, Y. and Tsuchiya, Y. (1959):

"An analysis of the stable cross section of a streamchannel",
Bull. Disaster Prevention Research Institute, Kyoto Univ.,
No. 29.

Jaggi, M.N.R. (1983):

"Alternierende Kiesbanke", Diss. ETH Nr. 7208, ZUrich (= Mitt.
Nr. 62 der Versuchsanstalt fur Wasserbau, Hydrologie und Gla¬
ziologie der ETH).

Johnson, A.M. (1965):

"A model for debris flow", Pennsylvania State Univ., unpub¬
lished Ph.D. Dissertation.

Johnson, A.M. (1970):

"Physical processes in geology", Freeman Cooper and Co.

Johnson, A.M. and Rodine,

J.R. (1984):
"Debris Flow", Chapter 8 in "Slope Instability", ed. D.
Brunsden and D.B.Prior, John Wiley and Sons.

Julien, P.Y. and Lau, Y. (1988):

Discussion of "Simple model of sediment-laden flows" by Parker
and Coleman (1986); ASCE, J.Hydr.Eng., Vol. 114, No. 2.

Kahr, G. (1989):
personal communication
 -
217 -

Kamphuis, J.W. (1974):

"Determination of sand roughness for fixed beds", J.Hydr.Res.,
Vol. 12, No. 2.

Kang, Z. and Zhang, S. (1980):

"A preliminary analysis of the characteristics of debris flow",
Proc. Int. Symp. on River Sedimentation, Beijing, China (in
Chinese with English abstract).

Kikkawa, V. and Fukuoka, S. (1969):

"The characteristics of flow with wash load", Proc. 13th IAHR
Congress, Tokyo, Japan, Vol. 2.

Kozicki, W. and Tiu, C (1967):

"Non-Newtonian flow through open channels", Canadian J. Chem.
Eng., Vol. 45.

Kresser, V. (1964):
"Gedanken zur Geschiebe- und Schwebstoffiihrung der Gewasser",
Oesterreichische Wasserwirtschaft, 16. Jhg., Heft 1 und 2.

Krieger, I.M. and Dougherty, T.J. (1959):

"A mechanism for non-Newtonian flow in suspensions of rigid
spheres", Trans. Soc. of Rheology, Vol. 3.

Kurdin, R.D. (1973):

"Classification of mudflows", Soviet Hydrology: Selected
Papers, Issue No. 4.

Lau, Y.L. (1987):

Discussion of "Low temperature effects on flow in sand-bed
streams" by Hong et al. (1984), ASCE, J.Hydr.Eng., Vol. 113,
No. 1.

Lanzendorf, V. (1984):
"Zum Bewegungsverhalten Feststoffpartikeln im Bereich der
von

laminaren Unterschicht einer

RohrstrShmung", Mitt, des Leicht-
weiss-Instituts fflr Wasserbau, Techn. Univ. Braunschweig,
Heft 83.

Lin, C. and Sun, M. (1983):

"A remark on the shields diagram", 2nd Int. Symp. on River
Sedimentation, China (in Chinese with English abstract).

Low, H.S. (1989):

"Effect of sediment density on bed-load transport", ASCE,
J.Hydr.Eng., Vol. 115, No. 1.

Luque, R.F. and van Beek, R. (1976):

"Erosion and transport of bed-load sediment", IAHR,
J.Hydr.Res., Vol. 14, No. 2.

Mantz, P.A. (1977):

"Incipient transport of fine grains and flakes by fluids -

extended Shields diagram", ASCE, J.Hydr.Div., Vol. 103,

No. HY6.
 -
218 -

Mantz, P.A. and Emmett, W.W. (1985):

"Analysis of United States Geological Survey sediment transport
data for some California streams", Proc. Euromech 192: Trans¬

port of suspended solids in open channels, Neubiberg.

Meyer-Peter, E. and Mueller, R. (1948):

"Formulas for bed-load transport", 2nd meeting IAHSR, Stock¬
holm, Sweden.

McTigue, D.F. (1982):

"A nonlinear constitutive model for granular materials:
application to gravity flow", J. of Applied Mechanics, ASME,
Vol. 49(2).

Mills, S.V. (1983):

"An experimental study of the rheology of mudflows", M.S.
thesis, College of Eng., Washington State Univ.

Mizuyama, T. (1977):
"Bedload transport in steep channels", Ph.D. Dissertation,
Kyoto Univ.

Mizuyama, T. (1981):
"An intermediate phenomenon between debris flow and bed load
transport", Symp. on Erosion and Sed. Transp., Pacific Rim,
Christchurch, New Zealand.

Mizuyama, T. (1988):
written communication

Mizuyama, T. and Shimohigashi, H. (1985):

"Influence of fine sediment concentration on sediment transport
rates", Jap. Civil Eng. Jour. 27-1, 1985 (in Japanese).

Moshev, V.V. (1979):

"Viscosity relationship for heavily filled suspensions", Fluid
mechanics Soviet research, Vol. 8, No. 2.
-

Murphy, P.J. and Aguirre, E.J. (1985):

"Bed load or suspended load", ASCE, J.Hydr.Eng., Vol. Ill,
No. 1.

Naik, B. (1983):
"Mechanics of Mudflow treated as the Flow of a Bingham Fluid",
Ph.D. Thesis, Washington State University, USA.

O'Brien, J.S. and Julien, P.Y. (1984):

"Physical properties and mechanics of hyperconcentrated sedi¬
ment flows", Proc. Specialty Conference on Delineation of
Landslides, Flashfloods and Debris Flow Hazards in Utah, Logan,
Utah.

O'Brien, J.S. and Julien, P.Y. (1986):

"Rheology of non-Newtonian fine sediment mixtures", ASCE,
Speciality Conference on advances in aerodynamics, fluid
mechanics, and hydraulics, Minnesota.

Parker, G. and Coleman, N.L. (1986):

"Simple model of sediment laden flows", ASCE, J.Hydr.Eng., Vol.
112, No. 5.
 -
219 -

Pierson, T.C. and Costa, J.E. (1984):

"A rheological classification of subaerial sediment-water
flows", Abstracts with Programs, 97th Annual Meeting GSA, Vol.
16, No. 6.

Quian, Y., Yang V., Zhao V., Cheng X., Zhang L., (1980)Xu V.
"Basic Characteristics of Flow with
Hyperconcentration of
Sediment", Int. Symp. on River Sed., Beijing, China.

Richardson, E.V. and Julien, P.Y. (1986):

"Bedforms, fine sediments, washload and sediment transport",
3rd Int. Symp. on River Sedimentation, Univ. of Mississippi.

Rouse, H. (1960):
"Elementary mechanics of fluids", John Wiley and Sons, London.

Rubey, W.W. (1933):

"Settling velocities of gravel, sand and silt particles",
American J. of Science, 5th Series, Vol. 25(148).

Savage, S.B. (1989):

"Flow of granular materials", in: Theoretical and Applied
Mechanics, eds. Germain, Piau, Caillerie, Elsevier, Amsterdam.

Schoklitsch, A. (1950):
"Handbuch des Wasserbaus", 2. Auflage, Springer Verlag, Wien.

Shen, S. and Xie, S. (1985):

"Structure and rheologic property of mud debris flow",
mode
Int. Symp. on Erosion, Debris Flow and Disaster Prevention,
Tsukuba, Japan.

Shields, A. (1936):
"Anwendung der Aehnlichkeitsmechanik und der Turbulenzforschung
auf die Geschiebebewegung", Mitt. d. Preuss. Vers.anst. fflr
Wasserbau und Schiffbau, Berlin, Nr. 26.

Simons, D.B., Richardson, E.V. and Haushild, W.L. (1963):

"Some effects of fine
sediment on flow phenomena", U.S.G.S.
Water-Supply Paper 1498-G.

Smart, G.M. (1984):

"Sediment transport formula for steep channels", ASCE,
J.Hydr.Eng., Vol. 110, No. 3.

Smart, G.M., and Jaeggi, M. (1983):

"Sediment Transport on Steep Slopes", Mitt. Nr. 64 der Ver¬
suchsanstalt fiir Wasserbau, Hydrologie und Glaziologie, ETH
Zurich.

Stevens, M.A., Simons, D.B. and Lewis, G.L. (1976):

"Safety factors for riprap protection", ASCE, J.Hydr.Div., Vol.
104, No. HY8.

Stiny, J. (1910):
"Die Muren", Verlag der Wagner'schen Universitatsbuchhandlung,
Innsbruck.
 -
220 -

Straub, L.G., Silberman, E. and Nelson, H.C (1958):

"Open-channel flow at small Reynolds numbers", ASCE, Trans¬
actions, Vol. 123.

Syanozhetsky, T.G.V., Beruchashvili, G.M. and Kereselidze, N.B. (1973):

"Hydraulics of rapid turbulent and quasilaminar (structural)
mud-streams in deformed beds with abrupt slopes", Proc. 15th
IAHR Congress, Istanbul, Turkey, Vol. 1.

Tang, C. (1981):
"Formula for the Bingham limit shear stress in sediment-water
mixture", Jour, of Sediment Research, Beijing, China, No. 2.

Takahashi, T. (1978):
"Mechanical characteristics of debris flow", ASCE, J.Hydr.
Div., Vol. 104, No. HY8.

Takahashi, T. (1980):
"Debris flow on prismatic open channel", ASCE, J.Hydr.Div.,
Vol. 106, No. HY3.

Takahashi, T. (1981):
"Debris flow", Ann. Rev. Fluid Mech., Vol. 13. (1987):
"High velocity flow in steep erodible channels", 22nd IAHR

Congress, Lausanne, Switzerland, Techn. Session A.

Taniguchi, Y., Takahashi, T. and Sassa, K. (1988):

"Determination of the sheared thickness of torrent sediment
deposit by debris flow", Interpraevent, Graz, Austria, Vol. 2.

Thomas, A.D. (1979a):

"Predicting the deposit velocity for horizontal turbulent pipe
flow of slurries", Int. J. Multiphase Flow, Vol. 5.

Thomas, A.D. (1979b):

"The role of transition in determining the
laminar/turbulent
critical deposit velocity and the operating pressure gradient
for long distance slurry pipelines", Proc. 6th Int. Conference
on the Hydraulic Transport of Solids in Pipes, BHRA Fluid
Engineering, Cranfield, Bedford, England.

Thomas, D.G. (1961):

"Laminar flow properties of flocculated suspensions", A.I.Ch.E.
Jour., Vol. 7, No. 3.

Thomas, D.G. (1963a):

"Non-Newtonian suspensions, Part I: Physical properties and
laminar transport characteristics", Ind. and Eng. Chemistry,
Vol. 55, No. 11.

Thomas, D.G. (1963b):

"Non-Newtonian suspensions, Part II: Turbulent transport cha¬
racteristics", Ind. and Eng. Chemistry, Vol. 55, No. 12.

Torrance, B.Mck. (1963):

South African Mechanical Engineer, Vol. 13, p. 89.
 -
221 -

Tsubaki, T., Hashimoto, Suetsugi, T. (1982):

H. and
"Grain stresses and flow properties of debris flow", Proc. Jap.
Soc. of Civ. Eng., Vol. 317 (in Japanese; a condensed version
appeared in Trans, of Jap. Soc. of Civ. Eng., Vol. 14).

VanDine, D.F. (1984):

Debris flows and debris torrents in the Southern Canadian
Cordillera", Can. Geotech. Jour., Vol. 22.

van Rijn, L.C (1983):

"Sediment transportation in heavy sediment-laden flows", Proc.
2nd Int. Symp. on River Sedimentation, Nanjing, China.

Wan, Z. (1982):
"Bed Material Movement in Hyperconcentrated Flow", Ser. Paper
31, Inst. Hydrodyn., T.U. Denmark, Lyngby.

Van, Z. and Song, T. (1987):

"The effect of fine particles on vertical concentration dis¬
tribution and transport rate of coarse particles", Proc. 22nd
IAHR Congress, Lausanne, Switzerland, Techn. Session A.

Vang, S. and Zhang, R. (1987):

"A new equation of bed load transport", Proc. 22nd IAHR Con¬

gress, Lausanne, Switzerland, Techn. Session A.

Ward, T.J. (1986):

Discussion of "Sediment Transport Formula for Steep Channels"
by G.M. Smart, ASCE, J.Hydr.Eng., Vol.112, No.10.

Ward, T. and O'Brien, J.S. (1981):

"Hydraulic and rheologic modelling of mud and grain flows",
Proc. Erosion and sediment transport in Pacific Rim Steeplands,
IAHS Publ. No. 132, Christchurch, New Zealand.

White, V.R. (1987):

"Engineering aspects of fluvial morphology", Proc. 22nd IAHR
Congress, Lausanne, Switzerland, Techn. Session A.

Whittaker, J, and Jaggi, M.N.R. (1986):

"Blockschwellen", Mitt. Nr. 91 der Versuchsanstalt fiir
Wasserbau, Hydrologie und Glaziologie, ETH Zurich.

Villi, V. (1965):
"Zur Frage der Sohlenerosion bei grossen Gefalien", Mitt. Nr.
68 der Versuchsanstalt fiir Wasserbau und Erdbau, ETH Zurich.

Wilson,K.C. (1966):
"Bed-load transport at high shear stress", ASCE, J.Hydr.Div.,
Vol. 92, No. HY6.

Wilson, K.C. (1984):

"Analysis of contact-load distribution and application to
deposition limit in horizontal pipes", Journal of Pipelines,
Vol. 4.

Vilson, K.C. (1986):

"Analysis of bed-load motion at high shear stress", ASCE,
J.Hydr.Eng., Vol. 113, No. 1.
 -
222 -

Vilson, K.C. and Judge, D.G. (1976):

"New techniques for the scale-up of pilot plant results to coal
slurry pipelines", Int. Symp. on Freight Pipelines, Washington.

Voo, H. (1985):
"Sediment transport in hyperconcentrated flows", Ph.D. Disser¬
tation, Colorado State Univ.

Yalin, M.S. (1977):

"Mechanics of sediment transport", 2nd ed., Pergamon Press,
Oxford.

Yalin, M.S. and Karahan, E. (1979):

"Inception of sediment transport", ASCE, J.Hydr.Div., Vol. 105,
No. HY11.

Yang, C.T. (1979):

"Unit stream power equations for total load", Jour, of Hydro¬
logy, Vol. 40,No. 1/2.

Yang, V. and Zhao, V. (1983):

"An experimental study of the resistance to flow with hyper-
concentration in rough flumes", Proc. 2nd Int. Symp. on River
Sedimentation, Nanjing, China (in Chinese with English
abstract).

Yano, K. and Daido, A. (1965):

"Fundamental study on mud-flow", Bull. Disaster Prevention
Research Institute, Kyoto Univ., Vol. 14, Part 2.

Zhang, H., Ren, Z., Jiang, S., Sun, D. and Lu, N. (1980):
"Settling of sediment and the resistance to flow at hypercon-
centrations", Proc. Int. Symp. on River Sedimentation, Beijing,
China (in Chinese with English abstract).

Zagni, F.E. and Smith, K.V.H. (1976):

"Channel Flow over permeable beds of graded spheres", ASCE,
J.Hydr.Div., Vol. 102, No. HY2.
 -
223 -

LIST OF SYMBOLS

A dimensional constant

A' constant in Mizuyama and Shimohigashi's (1985) bed load

transport equation

A* parameter in velocity equation for debris flow front

A area of the flow cross section

cs

A parameter in the flow resistance equation proposed by Naik

0
(1983)

A shape factor in the flow resistance equation proposed by

s
Naik

A., A., A,, A. in the flow resistance equations

A3, constants

proposed by Torrance (1963) (pipe flow)

A,, A., A„ constants in the flow resistance equations proposed by

b ' H
Yang and Zhao (1983)

B [m] flume width

B' "intrinsic" viscosity (constant)

B ,
B' nondimensional parameter in the logarithmic velocity law

B. constant in Daido's (1983) bed load transport equation

C_ drag coefficient =( 4(u-p)gd/3pW2 )

C grain concentration at the bed

C, volume concentration of fine material

C volume concentration of solids (grains)

C
f weight concentration of fine particles smaller than
SIW
0.025 mm

C _ (calculated) volume concentration of bed load grains

v,B

C. parameter that depends on depth, concentration, sediment

size and boundary roughness

CA maximum possible packing concentration of particles

Ci clay concentration level

Fr Froude number =( V/(gh)"2 )

G dimensionless grain flow parameter =( <rd2T/Xrl2 )

G„ [kg, ._/sl sediment feeding rate

H [m] measured flow depth

 -
224 -

fictitious fluid flow depth =( q/V )

measured mixture flow depth

change in the pressure gradient (pipe flow)

Hedstroem number =( pT_16h2/t|i )

critical gradient at deposition, for case &<d (pipe flow)

critical gradient at deposition, for case S>=d (pipe

flow)
factor accounting for the interaction of the colliding
particles (usually taken as
K=l/C^)
constant in Van's (1982) equ. for 8

K,, Kc, K,, K., K.... dimensional and

K,_, constants,

K. constants, in empirical equations for Bingham parameters

number of observations

2] dipersive pressure

] flow rate in the flume

hydraulic radius

hydraulic radius corrected for side wall influence

hydraulic radius belonging to sidewall subarea

Reynolds number =( 4VRp/ri ) or ( 4Vhp/n )

Bingham Reynolds number =( 4Vhp/ri_ )

Reynolds number defined with the effective viscosity y

*",
=( 4Vhp/ue2 )

particle Reynolds number =( v*dp/t| )

particle Reynolds number defined with the effective

viscosity y =( v*dp/y
. ) „

roughness Reynolds number =( v*k p/tl )

roughness Reynolds number defined with the Bingham

viscosity =( v*k p/t|_ )

roughness Reynolds number as defined by Naik (1983)

=( v*ksp(l-a')/nB )

universal Reynolds number as defined by Ansley and

Smith (1967) =( pW2/[nRW/d +
TBn7/24] )

-] slope (=tanB)

-] critical slope value at initiation of motion

 -
225 -

S_ [%1 standard error

T [N/m2] shear stress acting between different grain layers

T [N/m2] total shear stress

yy

T [N/m2] normal stress

l '
yx

U [m/s] mean velocity of debris flow front

U. [m/s] average velocity of a single particle

U_ [m/s] average velocity of bed load grains

a

V [m/s] average fluid velocity

V, [m/s] critical deposit velocity, for the case 5<d (pipe flow)
it
V, [m/s] critical deposit velocity, for the case 6>=d (pipe flow)

W [m/s] particle settling velocity

W' [N/m2] immersed weight of grains moving over a unit bed area

Yf Yield factor =( 2tB/pV2 )

a. constant
1

a. constant

a constant

a' ratio between Bingham yield and bed shear stress

-< Vo >

a' critical value of a' at laminar -

turbulent transition
c

b constant in Bagnold's (1956) bed load transport equation

c flow resistance coefficient defined as c=V/v*

c' [N/m2] cohesion

d [m] grain size

d [m] mean grain size

m

d30 [m] characteristic grain size, than which 30% of the material
by weight is finer

[m] characteristic grain size,

d50 than which 50% of the material
by weight is finer

[m] characteristic grain size,

d90 than which 90% of the material
by weight is finer
du
[l/s] grain velocity gradient perpendicular to the flow
dy direction
 -
226 -

dv
[l/s fluid velocity gradient (shear rate) perpendicular to the
dy flow direction

f Darcy-Weisbach friction factor =( 8t /pV2 )

f Darcy-Weisbach friction factor, part due to grain

roughness

f" Darcy-Weisbach friction factor, part due to form drag

g acceleration due to gravity

h Im] flow depth

h [m] (reduced) flow depth corrected for sidewall influence

r

h fictitious fluid flow depth in flow of grain-fluid mixture

r, f[m]
=( Q /V ), corrected for sidewall influence

h
-'' mixture flow depth (including space occupied by moving
r,
grains), corrected for sidewall influence

h [N/s bed load transport rate by immersed weight of solids per

unit width =( qBg(o-p) )

k slope parameter in Mizuyama's (1977) bed load transport

equation

k' slope parameter in Takahashi's (1977) bed load transport

equation

k, [m1,3/s] Strickler coefficient for bed (grain and form drag)

k [mI,3/s] Strickler coefficient due to grain roughness only

k [m"3/s] Strickler coefficient for flume walls

w

k [m] equivalent sand roughness

k
r- relative roughness

1 [m] Prandtl's mixing length

m parameter in Yalin's (1977) bed load transport equation

n fluid or flow behaviour index (empirical constant)

p [N/m2] pressure

p [N/m2] normal stress

q [m3/s.m] volumetric water or fluid discharge per unit width

q [mVs.m] critical flow discharge at beginning of bed load transport

q* dimensionless water or fluid discharge

=( q/[g"2d3'2] )

q„ [mVs.m] volumetric bed load transport rate per unit width

 -
227 -

[m3/s.m] volumetric transport rate per unit width of bed load

and suspended load

correlation coefficient (between measured and calculated

values)

s ratio between grain and fluid density =( alo )

t [s] time

u [m/s] local grain velocity

V [m/s] local fluid velocity

V* [m/s] shear velocity =( [ghS]"2 )

V* [m/s] shear velocity at initiation of motion

c

X [m] coordinate in flow direction

y [m] coordinate perpendicular to the flow direction, measured

upwards from the bed

y' [m] coordinate measured downwards from the flow surface

y [m] distance from the boundary of the effective fluid thrust

'n
on the bed load grains

'1
[m] rougness height

z parameter in Yalin's (1977) bed load transport equation

a dynamic angle of internal friction

a' angle of internal friction

static angle of internal friction

s

coefficient in flow resistance formula of Smart and

"1
Jaggi (1983)

slope angle

6, coefficient in flow resistance formula of Smart and

Jaggi (1983)

5 [m] thickness of laminar (viscous) sublayer

t| [kg/m.s] dynamic fluid viscosity

tv [kg/m.s] Bingham viscosity

H relative viscosity of a suspension = ti I1\ or n /h

11 [kg/m.s] dynamic viscosity of a suspension

ri [kg/m.s] dynamic viscosity of water

 -
228 -

dimensionless bed shear stress parameter

=
( [(9-9c)9"2]2'3 )

dimensionless bed shear stress

=( To/Pg(s-l)dm ) =( hS/(s-l)dm )

critical value of 9 at initiation of motion

critical value of 9 at initiation of motion, used in

data analysis

dimensionless bed shear stress defined with the fictitious

fluid depth (used in data analysis) =( h ,S/(s-l)d )

dimensionless bed shear stress in grain-fluid mixture

(used in data analysis) =( h S/(s-l)d )

dimensionless bed shear stress, corrected for sidewall

influence =( h S/(s-l)d ) or =(
'
h S/(s-l)d
v )
r m r,m m

critical dimensionless bed shear stress for initiation of

motion at steep slopes (modified definition)

critical dimensionless bed shear stress for initiation of

motion at steep slopes (corrected Shields parameter)

dimensionless bed shear stress used by Bagnold (1956)

=( 6/cosP )

dimensionless critical bed shear stress used by Bagnold

(1956) =( 6 /cosp )

von Karman constant

von Karman constant as modified by the presence of grains

linear grain concentration =( 1/[(C./C )"3-l] )

* s

consistency index (dimensional parameter)

cross-consistency index (dimensional parameter)

m] effective viscosity =( fL + t_h/2V )

m] effective viscosity =( rW[l-1.5(-r /t )+0.5(t„/t: )3] )

D DO DO

kinematic viscosity

eddy viscosity (also called 'turbulent' viscosity)

] fluid density

] grain density

shear stress in the fluid

critical shear stress at incipient motion

fluid shear stress at the bed =( gpSh )

 -
229 -

total shear stress (caused by fluid and grains)

Bingham (yield) stress

critical Bingham stress at non-settling conditions

dimensionless bed load transport rate

=( qB/[g(s-l)d3]"2] )

dimensionless sediment transport rate

=( qt/[g(s-l)d3]"2] )

dimensionless bed load transport rate

=( qB/b[g(s-l)dmcose]"2] )

dimensionless sediment transport rate

=( qt/b[g(s-l)d3cos6]"2] )

general function in bed load transport relationships

stream power =( Vx ) =
( Vpgh S )
Leer -
Vide -

Empty
 -
231 -

APPENDIX

I : Experimental results of clay suspension flows without

sediment transport

Note: The fixed rough bed consisted of gravel particles glued to the

flume bottom, with the same grain size characteristics as for the

bed material used in the sediment transport tests.

Columns (1) to (7) refer to measured parameters, columns (8) to

(14) give derived parameters.

a: In these tests, a velocity measurement by the salt tracer

technique was not reliable (C3, C4) or not possible at all

(C5); the velocity determined as V =

Q/(B-H) was used here.
 - 232 -

CD
DC

<D O CI rH

DC V

rH O rH

.o
a>
T3

> E

a
rH i-H CO

o
04
o X O
 233 -

A *

a CM
01 O
i DC

~
=1

<n >

H rH CN
w
 - 234

3 t?

i cc

? I

rH O

CJ rH

£ O

o
 235 -

oont

CN 10 V
10 io en Ol m *J" rH Ol O tn
« r- OJ Ol rH OJ r-
-. CD Ol m P-
rH Ol m v m 40 r- 00
m 10 CO Ol m
rH rH
rH rH Ol OJ

to rH lO CN m <N Ol
CN CO Ol <N H H o
o 40 o en 10 r- 10
cn
- <D rH tn Ol to

40 00 in r- o Ol 10
10 00 en en in rH en m tn
rH rH Ol
H rH CN rH CN

r- tn m tN m O o P* tn
m P~ O w m CO m

10 CM m r- 10 f- <H
** o. r» rH m rH o o H Ol rH
Ol rH w o r» m m 00 00
Ol rH en CO o Oi r- CO m
a H OJ Ol rH CN rH t- o 00 r- 10 m m N (N Ol
rH

Ol o 00 Ol m CO Ol tn Ol 00 CO
en to en Ol 10 Ol
o p* CO O 40
en CO Oi rH o p- 00 m 00 00 m r-
o Ok rH tn M Ol
o Ol r- Ol o 40 r- o m
"*-*
rH rH ri rH fH rH rH O
rH rH o o rH rH o rH rH Ol

o o O O o o o O O
o o o O o o o o o

Ol tn P- m 40 Ok Ol
o rH m OJ CN CO Ol rH
p* 40 in 40 OJ m o
CO O o H Ol CO OJ

o O 40 r- CO p- CO Ol
CO o V u> CO m r- CO
rH CO Ol o CO
rH rH

00 r- en rH p- tn O rH
Ol U> OJ 00 00 o o
m <N tn Ol
o r- rH o o CM Ol 00 40 Ol Ol

Sjz" a
tn en OJ en tn
OJ en Ol en m en en tn OJ
OJ

o r-t oi m Ol 00 10 o
m m in o p- CO m r-
e-t r- OJ 00 tN 10 Ol r-
m r- co en 00 GO
10 en p-
c'"
s Ol en Ol en 40 m m *r en en tN en

m m m m m m in m O o o
OJ Ol CN Ol OJ <N v> 10 10 10 40 40 10 10 40
°*
£ en en en en en en en en tn tn tn tn en tn tn en tn tn

o o o o o o o o o o o o o o o o o o
o o o o o o en en tn en en en tn tn en o o o

£ t? o o o o o o lo lo is

Ol OJ OJ OJ <N OJ tn tn tn tn tn tn en en tn tN OJ Ol

o o o o o o o o o o o o o o o O o o
10 10 10 10 10 00 00 CO 00 CO 00 CO CO CO O o o

5 t*° en en en en en en o o o o o o o o o o o o

OJ Ol CM CN CN Ol tn tn en

r- en OJ Ol Ol en tn en m r- o o r-

co r* 10 r- rH Ol 10 m CO CO en m Ol 00 00 tn

£ x
Ol en m Ol in to 10 in tn Ol m

CS a a a at a a eg 9 ctt m 9 «
a a
o 10 Ol o i0 m io en 10 OJ r- p- P- lo

r- o CO en r- rH m 00 r* o r- (M p-

5 >
Ol Ol rH Ol en o rH H rH H rH rH (M H Ol Ol

o o o O o o o O O o o O O O o o O o

st O o m o O o o o m o o m
rH
o
Ol
o
rH
m
H
o
CM
o O
Ol
o
tn
rH H en rH CN rH rH OJ

in m m O O o m m in o o o m m m o O o
rH rH H Ol Ol Ol o o o rH tH rH rH rH Ol OI Ol

5 w
O o o o O o o o o o O o o O o o O o

to
O
 - 236 -

II : Experimental results of clay suspension flows with

bed load transport

II.1 : Case I experiments

Note: Columns (1) to (8) refer to measured parameters, columns (9) to

(26) give derived parameters.

The experimental results labelled H.O were taken from Smart and

Jaggi (1983).

b: No velocity measured in this experiment; the

was
velocity
calculated by Smart and Jaggi was used here.
 -
237 -

(1) (2) (3) (4) (5) (6) (7) (8)

Ci S Q V H GB *B In P

P/s]_ [m/s] [cm] [kg/s] [N/m2] [cps] [g/cm3]

H20| 0.07 15.0 1.28 5.90 0.68 0.00 1.02 0.998

0.07 25.0 1.55 8.20 1.42 0.00 1.02 0.998

0.10 10.0 1.07 5.00 0.84 0.00 1.02 0.998

0.10 15.0 1.38 6.30 1.60 0.00 1.02 0.998

0.10 30.0 2.06 6.00 3.53 0.00 1.02 0.998

0.15 10.0 1.18 4.80 2.30 0.00 1.02 0.998

0.15 15.0 1.57 6.00 4.08 0.00 1.02 0.998

0.15 30.0 2.52 8.00 9.49 0.00 1.02 0.998

0.20 10.0 1.26 6.00 4.80 0.00 1.02 0.998

0.20 15.0 1.74 6.80 8.84 0.00 1.02 0.998

0.20 20.0 2.44 7.60 10.30 0.00 1.02 0.998

0.20 30.0 2.59b 8.50 14.91 0.00 1.02 0.998

C1 | 0.07 15.0 1.16 5.19 0.88 0.67 2.17 1.092

0.07 25.0 1.57 7.31 1.49 0.61 2.09 1.084

0.10 10.0 1.11 4.61 1.49 0.43 1.93 1.075

0.10 15.0 1.35 5.25 2.31 0.60 2.13 1.084

0.10 30.0 2.01 7.69 4.34 0.63 2.13 1.087

0.15 10.0 1.18 4.82 3.52 0.59 2.05 1.091

0.15 15.0 1.52 6.01 5.29 0.52 1.93 1.072

0.15 30.0 2.03 8.12 9.08 0.63 2.13 1.088

0.20 10.0 1.25 5.62 7.05 0.59 2.17 1.086

0.20 15.0 1.60 6.89 10.58 0.58 2.03 1.096

0.20 20.0 1.90 7.77 12.88 0.43 1.93 1.077

0.20 30.0 2.47 9.07 17.63 0.63 2.13 1.088

 - 238 -

cont

(i) (2) (3) (4) (5) (6) <7> (8)

S Q V H Gb *B -1b P

[l/s] [m/s] [cm] [kg/s] [N/m2] [cps] [g/cm3]

C2 0.07 15.0 1.27 5.83 1.15 2.70 3.33 1.152

0.07 25.0 1.46 7.80 1.97 2.56 3.22 1.149

0.10 10.0 1.12 4.80 2.03 2.50 3.32 1.160

0.10 15.0 1.40 5.96 3.12 2.77 3.33 1.153

0.10 30.0 2.18 8.54 5.08 2.26 3.06 1.145

0.15 10.0 1.18 4.95 4.75 2.20 3.16 1.153

0.15 15.0 1.58 6.30 7.05 2.20 3.16 1.152

0.15 30.0 2.22 8.52 10.85 2.41 3.19 1.147

0.20 10.0 1.29 6.58 9.63 2.50 3.32 1.158

0.20 15.0 1.83 7.70 14.37 2.38 3.22 1.159

0.20 20.0 2.19 8.25 16.27 2.26 3.00 1.141

0.20 30.0 2.75 9.79 20.74 2.26 3.06 1.145

C3 0.07 15.0 1.21 5.62 1.49 8.06 6.27 1.245

0.07 25.0 1.60 8.11 2.24 4.31 4.35 1.201

0.10 10.0 1.15 4.50 2.58 8.07 6.43 1.243

0.10 15.0 1.41 5.91 3.91 8.07 6.43 1.243

0.10 30.0 1.93 8.72 4.95 7.35 6.00 1.223

0.15 10.0 1.23 5.44 5.97 6.95 5.75 1.240

0.15 15.0 1.61 6.70 9.01 8.14 6.74 1.241

0.15 30.0 2.34 9.04 12.20 7.75 6.37 1.228

0.20 10.0 1.58 7.46 13.56 8.55 6.10 1.246

0.20 15.0 1.73 8.69 17.55 8.14 6.74 1.236

C4 0.07 25.0 1.69 7.06 2.39 18.60 16.20 1.286

0.10 30.0 2.08 8.15 7.18 17.60 15.70 1.275

0.15 10.0 1.32 6.21 6.38 20.10 17.50 1.266

0.15 30.0 2.49 8.53 13.32 16.20 14.80 1.257

 - 239 -

(l) (2) (9) (10) (11) (12) (13)

Ci| s; Q •hb a •Ha s„ (S-1)

[l/s] [.10-3m3/s] [m3/s] [m3/s]

H20 0.07 15.0 1.27 0.0676 0.0101 0.0148 1.680

0.07 25.0 2.65 0.1093 0.0101 0.0110 1.680

0.10 10.0 1.56 0.0471 0.0068 0.0157 1.680

0.10 15.0 2.99 0.0681 0.0068 0.0128 1.680

0.10 30.0 6.59 0.1249 0.0068 0.0102 1.680

0.15 10.0 4.29 0.0474 0.0043 0.0146 1.680

0.15 15.0 7.61 0.0684 0.0043 0.0120 1.680

0.15 30.0 17.71 0.1210 0.0043 0.0098 1.680

0.20 10.0 8.96 0.0469 0.0031 0.0104 1.680

0.20 15.0 16.49 0.0672 0.0031 0.0094 1.680

0.20 20.0 19.22 0.0795 0.0031 0.0089 1.680

0.20 30.0 27.80 0.1233 0.0031 0.0080 1.680

0.07 15.0 1.63 0.0699 0.0079 0.0119 1.454

C1

0.07 25.0 2.77 0.1101 0.0081 0.0090 1.472

0.10 10.0 2.77 0.0469 0.0056 0.0127 1.493

0.10 15.0 4.29 0.0692 0.0054 0.0115 1.472

0.10 30.0 8.06 0.1278 0.0054 0.0086 1.465

0.15 10.0 6.55 0.0472 0.0034 0.0108 1.457

0.15 15.0 9.82 0.0685 0.0036 0.0094 1.500

0.15 30.0 16.86 0.1321 0.0034 0.0070 1.463

0.20 10.0 13.09 0.0469 0.0025 0.0082 1.468

0.20 15.0 19.62 0.0682 0.0024 0.0069 1.445

0.20 20.0 23.90 0.0883 0.0025 0.0064 1.488

0.20 30.0 32.70 0.1240 0.0025 0.0058 1.463

 - 240 -

cont

(i) (2) (9) (10) (ID (12) (13)

S Q (s-1)
% Qr Q« s„

[l/s] [•10-3m: /s] [m3/s] [m3/s]

C2 0.07 15.0 2.13 0.0679 0.0068 0.0080 1.326

0.07 25.0 3.66 0.1119 0.0069 0.0062 1.332

0.10 10.0 3.77 0.0468 0.0045 0.0090 1.310

0.10 15.0 5.79 0.0681 0.0046 0.0075 1.324

0.10 30.0 9.43 0.1208 0.0047 0.0057 1.341

0.15 10.0 8.80 0.0471 0.0029 0.0078 1.324

0.15 15.0 13.09 0.0677 0.0029 0.0063 1.326

0.15 30.0 20.14 0.1269 0.0029 0.0046 1.337

0.20 10.0 17.88 0.0462 0.0021 0.0049 1.314

0.20 15.0 26.70 0.0647 0.0021 0.0043 1.312

0.20 20.0 30.20 0.0828 0.0022 0.0042 1.349

0.20 30.0 38.50 0.1144 0.0021 0.0037 1.341

C3 0.07 15.0 2.77 0.0689 0.0054 0.0067 1.153

0.07 25.0 4.16 0.1078 0.0060 0.0057 1.232

0.10 10.0 4.78 0.0467 0.0036 0.0079 1.156

0.10 15.0 7.26 0.0679 0.0036 0.00 62 1.156

0.10 30.0 9.19 0.1275 0.0038 0.0046 1.191

0.15 10.0 11.08 0.0465 0.0023 0.0058 1.161

0.15 15.0 16.70 0.0667 0.0023 0.0049 1.160

0.15 30.0 22.60 0.1222 0.0024 0.0039 1.182

0.20 10.0 25.20 0.0430 0.0017 0.0038 1.151

0.20 15.0 32.60 0.0648 0.0017 0.0033 1.168

C4 0.07 25.0 4.44 0.1077 0.0049 0.0053 1.084

0.10 30.0 13.31 0.1246 0.0034 0.0046 1.102

0.15 10.0 11.84 0.0453 0.0022 0.0049 1.117

0.15 30.0 24.70 0.1195 0.0022 0.0040 1.132

 - 241 -

(1) (2) (14) (15) (16) (17) (18) (19)

S Q
Ci •b 9 C,,b UB

[l/s]_ [m/s]

H20 0.07 15.0 0.290 0.213 0.045 1.77 0.044 0.524

0.07 25.0 0.606 0.285 0.045 1.85 0.052 0.692

0.10 10.0 0.358 0.268 0.042 1.57 0.073 0.451

0.10 15.0 0.682 0.329 0.042 1.83 0.087 0.636

0.10 30.0 1.506 0.413 0.042 2.44 0.098 1.004

0.15 10.0 0.981 0.389 0.038 1.76 0.163 0.549

0.15 15.0 1.740 0.475 0.038 2.12 0.185 0.767

0.15 30.0 4.053 0.582 0.038 3.07 0.220 1.306

0.20 10.0 2.047 0.652 0.034 1.68 0.294 0.578

0.20 15.0 3.770 0.720 0.034 2.20 0.344 0.814

0.20 20.0 4.393 0.766 0.034 3.00 0.341 1.139

0.20 30.0 6.359 0.847 0.034 3.03 0.326 1.205

0.07 15.0 0.443 0.230 0.039 0.050

QJ 1.69 0.510

0.07 25.0 0.728 0.304 0.039 1.98 0.053 0.708

0.10 10.0 0.723 0.291 0.037 1.70 0.105 0.559

0.10 15.0 1.126 0.329 0.038 1.96 0.108 0.691

0.10 30.0 2.120 0.452 0.039 2.50 0.109 1.035

0.15 10.0 1.731 0.474 0.034 1.75 0.213 0.603

0.15 15.0 2.560 0.561 0.035 2.04 0.218 0.782

0.15 30.0 4.450 0.746 0.035 2.41 0.202 1.024

0.20 10.0 3.450 0.734 0.030 1.72 0.373 0.586

0.20 15.0 5.210 0.898 0.031 2.01 0.380 0.752

0.20 20.0 6.250 0.965 0.031 2.27 0.366 0.891

0.20 30.0 8.630 1.100 0.032 2.79 0.360 1.159

 - 242 -

cont

(i) (2) (14) (15) (16) (17) (18) (19)

S Q % 9 9
cr
Fr
C,B Ub

[m/s]
[I/S]_
C2 °-07 i5-0 0.593 0.280 0.032 1.77 0.061 0.620

0.07 25.0 1.014 0.359 0.032 1.78 0.062 0.717

0.10 10.0 1.047 0.332 0.030 1.71 0.128 0.614

0.10 15.0 1.605 0.399 0.030 1.95 0.132 0.784

0.10 30.0 2.616 0.524 0.030 2.63 0.125 1.193

0.15 10.0 2.442 0.520 0.027 1.76 0.257 0.638

0.15 15.0 3.628 0.648 0.027 2.11 0.262 0.857

0.15 30.0 5.582 0.841 0.026 2.59 0.232 1.167

0.20 10.0 4.954 0.945 0.023 1.65 0.459 0.589

0.20 15.0 7.395 1.071 0.023 2.21 0.478 0.827

0.20 20.0 8.373 1.088 0.023 2.58 0.442 1.009

0.20 30.0 10.674 1.246 0.023 3.04 0.420 1.278

C3 0.07 15.0 0.821 0.313 0.030 1.70 0.071 0.638

0.07 25.0 1.196 0.402 0.033 1.92 0.069 0.831

0.10 10.0 1.422 0.366 0.029 1.79 0.148 0.676

0.10 15.0 2.155 0.470 0.029 1.93 0.153 0.837

0.10 30.0 2.690 0.637 0.029 2.23 0.119 1.028

0.15 10.0 3.290 0.668 0.026 1.73 0.299 0.688

0.15 15.0 4.970 0.803 0.026 2.07 0.309 0.902

0.15 30.0 6.630 1.000 0.026 2.66 0.255 1.263

0.20 10.0 7.480 1.222 0.023 1.90 0.596 0.626

0.20 15.0 9.670 1.390 0.023 1.94 0.541 0.739

C4 0.07 25.0 1.362 0.393 0.030 2.19 0.072 0.898

0.10 30.0 4.053 0.632 0.029 2.52 0.152 1.236

0.15 10.0 3.578 0.788 0.026 1.74 0.317 0.742

0.15 30.0 7.420 0.971 0.026 2.93 0.274 1.365

 - 243 -

(i) (2) (20) (21) (22) (23) (24) (25) (26)

Ci S Q »V Jr_ c f »* Re2 Re*2

d90
[l/s] [cm] [cps] [.10»1
H20 0.07 15.0 5.36 4.43 6.67 0.1797 1.0 2690.51 2276.5

0.07 25.0 7.19 5.94 6.98 0.1644 1.0 4370.39 2635.9

0.10 10.0 4.73 3.91 4.97 0.3242 1.0 1984.75 2555.2

0.10 15.0 5.80 4.79 5.79 0.2390 1.0 3138.82 2829.3

0.10 30.0 7.28 6.02 7.71 0.1346 1.0 5881.10 3169.7

0.15 10.0 4.58 3.79 4.55 0.3872 1.0 2119.37 3079.6

0.15 15.0 5.58 4.61 5.48 0.2665 1.0 3435.53 3398.7

0.15 30.0 6.85 5.66 7.94 0.1270 1.0 6769.41 3766.4

0.20 10.0 5.75 4.75 3.75 0.5685 1.0 2841.18 3984.7

0.20 15.0 6.35 5.25 4.93 0.3292 1.0 4332.94 4187.5

0.20 20.0 6.76 5.59 6.70 0.1782 1.0 6468.39 4320.4

0.20 30.0 7.47 6.17 6.77 0.1748 1.0 7587.18 4541.1

C1 0.07 15.0 4.78 3.98 6.39 0.1958 16.0 151.13 148.4

0.07 25.0 6.40 5.33 7.48 0.1430 14.5 299.28 187.5

0.10 10.0 4.35 3.63 5.36 0.2781 10.4 199.84 257.0

0.10 15.0 4.85 4.04 6.19 0.2085 12.9 220.24 219.9

0.10 30.0 6.62 5.52 7.90 0.1281 12.5 464.36 266.2

0.15 10.0 4.60 3.83 4.52 0.3909 13.6 174.00 250.9

0.15 15.0 5.61 4.68 5.28 0.2874 11.6 315.71 319.9

0.15 30.0 7.28 6.07 6.21 0.2074 13.4 480.32 318.7

0.20 10.0 5.39 4.49 3.85 0.5397 14.9 197.14 285.0

0.20 15.0 6.49 5.41 4.49 0.3964 13.8 331.19 340.8

0.20 20.0 7.18 5.98 5.07 0.3109 10.0 586.72 483.2

0.20 30.0 8.05 6.71 6.23 0.2063 12.4 700.64 419.3

 - 244 -

cont

(i) (2) (20) (21) (22) (23) (24) (25) (26)

S Q hr J\_ C f
r* Re2 Re*2
d90
[l/s] [cm] [cps] [-102]
C2 0.07 15.0 5.30 4.42 6.67 0.1797 59.5 52.22 44.3

0.07 25.0 6.83 5.69 6.72 0.1772 63.3 72.15 47.2

0.10 10.0 4.35 3.63 5.40 0.2746 52.1 43.21 55.2

0.10 15.0 5.29 4.41 6.15 0.2112 55.6 61.53 56.7

0.10 30.0 7.02 5.85 8.32 0.1156 39.4 178.15 91.5

0.15 10.0 4.59 3.83 4.54 0.3874 45.9 54.45 78.3

0.15 15.0 5.73 4.78 5.44 0.2705 43.1 96.78 93.2

0.15 30.0 7.49 6.24 6.69 0.1789 43.8 173.99 104.2

0.20 10.0 6.21 5.18 3.70 0.5857 63.5 58.44 76.4

0.20 15.0 7.03 5.86 4.94 0.3284 48.9 122.27 105.7

0.20 20.0 7.34 6.12 5.77 0.2402 40.9 179.49 127.1

0.20 30.0 8.35 6.96 6.80 0.1732 37.4 281.61 148.9

C3 0.07 15.0 5.15 4.29 6.43 0.1932 177.8 17.45 15.8

0.07 25.0 7.07 5.89 7.25 0.1523 99.8 54.38 31.8

0.10 10.0 4.23 3.53 5.65 0.2510 154.8 15.62 19.6

0.10 15.0 5.43 4.53 6.11 0.2143 161.8 23.52 21.3

0.10 30.0 7.59 6.32 7.06 0.1606 150.8 47.41 26.6

0.15 10.0 5.17 4.31 4.47 0.3997 151.3 20.91 27.1

0.15 15.0 6.21 5.18 5.34 0.2810 163.4 30.42 27.5

0.15 30.0 7.88 6.57 6.87 0.1694 136.9 66.18 36.7

0.20 10.0 7.03 5.86 4.25 0.4437 196.7 28.09 28.2

0.20 15.0 8.12 6.77 4.34 0.4239 197.3 35.28 30.0

0.07 25.0 6.08 5.07 8.28 0.1167 350.4 15.10

C4 9.0

0.10 30.0 6.96 5.80 7.96 0.1261 310.0 23.83 12.9

0.15 10.0 5.87 4.89 4.49 0.3966 464.4 8.45 9.6

0.15 30.0 7.33 6.11 7.58 0.1394 253.4 36.18 19.5

 -
245 -

II : Experimental results of clay suspension flows with

bed load transport

II.2 : Case II experiments

Note: Columns (1) to (8) refer to measured parameters, the other columns

give derived parameters.

c: In these tests, a velocity measurement by the salt tracer

technique was not reliable (C5); as an approximation, the

velocity of the corresponding flow at the next lower concentra¬

tion level (C4 or C3) was used here.

 - 246 -

(1) (2) (3) (4) (5) (6) (7) (8)

S Q V H GB TB nB p

m_ [m/s] [cm] [kg/s] [N/m2] [cps] [g/cm3]

^"* 0.07 15.0 1.36 5.83 1.20 18.60 16.20 1.287

0.10 10.0 1.16 4.90 1.76 20.10 17.50 1.293

0.10 15.0 1.43 6.00 3.35 16.20 14.80 1.275

C5 0.07 15.0 1.36c 6.85 0.59 40.80 34.30 1.363

0.10 10.0 1.16c 5.75 0.81 40.80 34.30 1.363

0.10 15.0 1.43c 6.50 2.17 40.80 34.30 1.363

0.15 10.0 1.32° 6.60 3.06 33.60 28.80 1.356

0.20 10.0 1.58c 7.80 7.65 33.60 28.80 1.356

 - 247 -

(i) (2) (9) (10) (13) (14) (15)

Ci S Q Qb (S-1) e
a •b

[l/s] [•10"3m3/s] [m3/s]

C4
0.07 15.0 2.22 0.0666 1.080 0.689 0.340

0.10 10.0 3.26 0.0463 1.070 1.017 0.430

0.10 15.0 6.22 0.0675 1.100 1.913 0.500

0.07
C5 15.0 1.10 0.0690 0.966 0.361 0.459

0.10 10.0 1.51 0.0460 0.966 0.496 0.561

0.10 15.0 4.03 0.0695 0.966 1.323 0.629

0.15 10.0 5.68 0.0482 0.976 1.855 0.987

0.20 10.0 14.20 0.0484 0.976 4.637 1.564

 - 248 -

(i) (2) (20) (21) (22) (23) (24) (25) (26)

Ci S Q h, hr f
C
Hc.2 Re2 Re*2
d90

[l/s] [cm] [cps]

C4
0.07 15.0 5.24 4.37 7.18 0.1552 374.0 9.82 7.8

0.10 10.0 4.60 3.83 5.47 0.2678 415.7 6.64 7.9

0.10 15.0 5.50 4.58 6.15 0.2117 326.8 12.26 10.9

C5 0.07 15.0 6.34 5.28 6.52 0.1883 985.3 4.77 3.5

0.10 10.0 5.42 4.52 5.03 0.3161 987.5 3.47 3.8

0.10 15.0 6.08 5.07 5.86 0.2333 901.7 5.26 4.4

0.15 10.0 6.42 5.35 4.29 0.4337 845.9 5.43 5.9

0.20 10.0 7.63 6.36 4.08 0.4797 840.1 7.78 7.5

 -
249 -

Curriculum vitae

Born in Zflrich on October 11, 1958, I grew up in a village

nearby where I visited the primary school. From the age of 12 On

I went to a high school in Ziirich (Literargymnasium Ramibiihl),

graduating with a Hatura Typus B. During this time I was playing

football with a local sport club.

In the following years I studied civil engineering at the

Swiss Federal Institute of Technology (ETH) in Zurich, before

receiving the Diploma in 1983. During the term holidays I did

some practical training with consulting engineers. Outdoor

sports and hiking in the mountains had meanwhile become my

favourite activities in the leisure time.

Thanks to a scholarship from the ETH I spent one year in

Canada at the University of New Brunswick. There I was involved

in graduate studies in the field of hydrology, hydraulics and

wastewater treatment.

Since October 1984 I am employed as a research engineer at

the Laboratory of hydraulics, hydrology and glaciology of the

ETH. My main subjects of interest are debris flows and hydrau¬

lics of mountain torrents.