INTRODUCTION TO SEDIMENT TRANSPORT AUTUMN 2016

1. OVERVIEW
1.1 Introduction
1.2 Particle properties
1.2.1 Diameter, d
1.2.2 Specific gravity, s
1.2.3 Settling velocity, ws
1.2.4 Porosity, P
1.2.5 Angle of repose, 
1.3 Flow properties
1.3.1 Friction velocity, uτ
1.3.2 Mean-velocity profile
1.3.3 Eddy-viscosity profile
1.3.4 Formulae for bed shear stress

2. THRESHOLD OF MOTION
2.1 Shields parameter
2.2 Inception of motion in normal flow
2.3 Effect of slopes

3. BED LOAD
3.1 Dimensionless groups
3.2 Bed-load transport models

4. SUSPENDED LOAD
4.1 Inception of suspended load
4.2 Turbulent diffusion
4.3 Concentration profile
4.4 Calculation of suspended load

EXAMPLES

Recommended Textbooks

Chanson, H., 2004, Hydraulics of Open Channel Flow: An Introduction, 2nd Edition,
Butterworth-Heinemann, ISBN 0-750-65978-5

Chadwick, A.J., Morfett, J.C., and Borthwick, M., 2013, Hydraulics in Civil and
Environmental Engineering, 5th Edition, CRC Press, 978-0415672450

Hydraulics 3 Sediment Transport - 1 Dr David Apsley

1. OVERVIEW OF SEDIMENT TRANSPORT

1.1 Introduction

In natural channels and bodies of water the bed is not fixed but
is composed of mobile particles; e.g. gravel, sand or silt. These
may be dislodged and moved by the flow – the process of
sediment transport.

Many large rivers are famed for their sediment-carrying capacity: the Yellow River in China
is coloured by its sediment content, and for centuries Egyptian farmers have relied on the rich
deposits carried by the Nile. In many cases erosion occurs but is not noticed because there is
a dynamic equilibrium established whereby, on average, as much sediment is supplied as is
removed from an area. However, short-term events (such as severe storms) and man-made
structures (such as dams) can severely disrupt
this equilibrium. Chanson’s book contains
salutary details of reservoirs rendered useless
by siltation and bridge failures because of scour
around their foundations. The Nile delta is
eroding because the sediment supply from
upstream is being held up by the Aswan Dam.
Bridge piers are highly susceptible to short-
term or long-term scour around their
foundations (see the CFD simulation right),
whilst road and rail transport are often
disrupted by surface damage in flash floods.

Models exist to address three basic questions.

 Does sediment transport occur? (“Threshold of motion”).
 If it does, then at what rate? (“Sediment load”).
 What effect does an imbalance have on bed morphology? (“Scour vs accretion”).

In general, two modes of transport are recognised:

 bed load: particles sliding, rolling or saltating (making short jumps), but remaining
essentially in contact with the bed;
 suspended load: finer particles carried along in suspension by the turbulent fluid flow.
The combination of these two is called the total load.

To predict sediment transport we need to consider:

 particle properties: diameter, specific gravity, settling velocity, porosity;
 flow properties: bed stress, velocity and turbulence profiles.

Finally, we note that, although our primary interest here is in sediment transport by water,
similar processes can be observed in air (“aeolian” or wind-borne transport). Examples are
the raising of dust clouds and sand-dune movement across deserts.

Hydraulics 3 Sediment Transport - 2 Dr David Apsley

Besides overall transport of material, sediment transport gives rise to some classic bedforms:
 ripples (fine particles; Fr<<1; wavelength depends on particle size, not flow depth);
 dunes (Fr < 1; migrate in the direction of flow);
 standing waves (Fr = 1; bed undulations in phase with free-surface standing waves);
 antidunes (Fr > 1; migrate in the opposite direction to flow).

FLOW
Dunes

erosion deposition

bedform migration

FLOW Standing waves

FLOW

erosion deposition
Antidunes

bedform migration

Hydraulics 3 Sediment Transport - 3 Dr David Apsley

1.2 Particle Properties

1.2.1 Diameter, d

Since natural particles have very irregular shapes the concept of diameter is somewhat
imprecise. Common definitions include:
 sieve diameter – the finest mesh that a particle can pass through;
 sedimentation diameter – diameter of a sphere with the same settling velocity;
 nominal diameter – diameter of a sphere with the same volume.

A typical size classification is given below.

Type Diameter
Boulders > 256 mm
Cobbles 64 mm – 256 mm
Gravel 2 mm – 64 mm
Sand 0.06 mm – 2 mm
Silt 0.002 mm – 0.06 mm
Clay < 0.002 mm (cohesive)

Where there is a range of particle sizes the cumulative percentage is attached to the diameter;
e.g. the median diameter d50 is that sieve size which passes 50% (by weight) of particulate,
whilst a measure of spread is the geometric standard deviation σ g  (d 84.1 /d15.9 )1 / 2 . (The
percentiles assume a lognormal size distribution.)

This introductory course (and most models) will simply refer to a diameter d.

1.2.2 Specific Gravity, s

The specific gravity (or relative density) s is the ratio of the density of particles (ρs) to that of
the fluid (ρ):
ρ
s s (1)
ρ
Quartz-like minerals have density  2650 kg m–3, so the relative density (in water) is 2.65.

drag
1.2.3 Settling Velocity, ws

The settling velocity is the terminal velocity in still fluid and may
be found by balancing drag against submerged weight; i.e.
drag = weight – buoyancy
buoyancy, mw g
πd 2 πd 3 ws
2
c D ( 2 ρws )(
1
)  (ρ s  ρ) g
4 6
where cD is the drag coefficient. This rearranges to give
1/ 2
 4 ( s  1) gd 
ws    (2)
3 cD 
weight, mg

Hydraulics 3 Sediment Transport - 4 Dr David Apsley

Stokes’ law for the force on spherical particles at small Reynolds numbers gives
24 wd
cD  , Re  s  1
Re ν
whence (after some algebra):
1 ( s  1) gd 2
ws  (3)
18 ν
or, in non-dimensional form:
1/ 3
ws d 1 ( s  1) gd 3 1  ( s  1) g 
  d *3 , where d*  d  
ν 18 ν2 18  ν
2


However, this is valid only for very small, spherical particles (diameter < 0.1 mm in water).
For larger grains (and natural shapes) a useful empirical formula is that of Cheng (1997):
ws d
ν
 
 (25  1.2d *2 )1 / 2  5 3 / 2 (4)

One of the major uses of the settling velocity is in determining whether suspended load
occurs, and the concentration profile in the water column that results (see Section 4).

1.2.4 Porosity, P

The porosity P is the ratio of voids to total volume of material; i.e. in a volume V of space
there will actually be a volume (1–P)V of sediment and volume PV of fluid.

Porosity is important in modelling changes to bed morphology and the leaching of pollutants
through the bed. For natural uncompacted sediment P is typically about 0.4.

1.2.5 Angle of Repose, 

The angle of repose is the maximum angle (to the horizontal) which a pile of sediment may
adopt before it begins to avalanche. It is easily measured in the laboratory.

The resistance to incipient motion may be quantified by

R
an effective coefficient of friction μf. Consider the basic F
mechanics problem of a particle on a slope. Incipient
motion occurs when the downslope component of weight bed
equals the maximum friction force (μf × normal reaction).
Then
mg
mg sin   μ f (mg cos ) 
or
μ f  tan 

Although the mechanism for causing motion is not the same, μf can then be used to estimate
the effect of gravitational assistance on sloping beds, where both fluid drag and downslope
component of weight act on the particles of the bed (see Section 2).

Hydraulics 3 Sediment Transport - 5 Dr David Apsley

1.3 Flow Properties.

1.3.1 Friction Velocity, uτ

The bed shear stress τb is the drag (per unit area) of the flow on the granular bed. It is
responsible for setting the sediment in motion.

As stress has dimensions of [density]×[velocity]2 it is possible to define from τb an important

stress-related velocity scale called the friction velocity (or shear velocity) uτ such that:
τ b  ρu τ2 or uτ  τb / ρ (5)

1.3.2 Mean-Velocity Profile

It may be shown (elsewhere!) that a fully-developed turbulent boundary layer adopts a

logarithmic mean-velocity profile. For a rough boundary this is of the form
u z
U ( z )  τ ln(33 ) (6)
κ ks
where uτ is the friction velocity, κ is von Kármán’s constant (a famous number with a value
of about 0.41) and ks is the roughness height (typically 1 – 2.5 times particle diameter). z is
the distance from the bed.

1.3.3 Eddy-Viscosity Profile

A classical model for the effective shear stress τ in a turbulent flow is to assume, by analogy
with laminar flow, that it is proportional to the mean-velocity gradient:
dU dU
τ  μt or τ  ρν t (7)
dz dz
μt is the eddy viscosity. (νt = μt /ρ is the corresponding kinematic eddy viscosity). μt is not a
true viscosity, but a means of modelling the effect of turbulent motion on momentum
transport. Such models are called eddy-viscosity models and they are widely used in fluid
mechanics. In a fully-turbulent flow μt is many times larger than the molecular viscosity μ.

At the bed, the shear stress is τ  τ b  ρu τ2 . At the free surface (z = h), in the absence of
significant wind stress, τ = 0. In fully-developed flow it may be shown that the shear stress
varies linearly across the channel; hence, to meet these boundary conditions it is given by
τ  ρu τ2 (1  z / h) (8)
Differentiating the mean-velocity profile (6) we find:
dU u τ
 (9)
dz κz
Substituting these expressions for stress and mean-velocity gradient into (7) gives
u
ρu τ2 (1  z / h)  ρν t τ
κz
whence:
ν t  κu τ z (1  z / h) (10)

Hydraulics 3 Sediment Transport - 6 Dr David Apsley

The kinematic eddy viscosity νt therefore has a parabolic profile (in channel flow).

z z z

U tb nt

1.3.4 Formulae For Bed Shear Stress

Many sediment-transport formulae rely on knowledge of the bed shear stress τb, which is
what sets the particles in motion and determines the bed load.

Recall that in normal flow there is a balance between the downslope component of gravity
and bed friction, leading to:
τ b  ρgRh S (Rh is the hydraulic radius; S is the slope) (11)
Alternatively, by definition of the (skin-)friction coefficient cf:
τ b  c f ( 12 ρV 2 ) (V is the channel-average velocity) (12)
If you are lucky, cf may be given; (probably the only possibility in air). Otherwise one could
adopt one of the following approaches. Both assume fully-developed flow (although this
assumption is regularly stretched.)

In normal flow, if the discharge is known then Manning’s equation:

1
V  Rh2 / 3 S 1 / 2 (13)
n
may be used to determine the hydraulic radius and thence, from (11), the bed shear stress. A
useful correlation when the bed consists of granular material is (the dimensionally-
inconsistent) Strickler’s equation:
d 1/ 6
n (14)
21.1
where d is the particle diameter in m. This gives n = 0.015 m–1/3 s for a grain size of 1 mm.

Alternatively, one can find an analytical expression for depth-averaged velocity V by

integrating the velocity profile (6) to get (exercise: do it!):
h
1 u 12h
V   U ( z ) dz  τ ln( ) (15)
h 0 κ ks
The skin friction coefficient cf is then, from its definition:
2
τ ρu 2 u 
cf  1 b 2  1 τ2  2 τ 
2 ρV 2 ρV V 
whence
0.34
cf  (16)
ln(12h / k s )2
Typical values of cf are in the range 0.003 – 0.01.

Hydraulics 3 Sediment Transport - 7 Dr David Apsley

2. THRESHOLD OF MOTION

2.1 Shields Parameter

In general, a granular bed will remain still until the flow is sufficient to move it. This point is
called the threshold of motion and the bed stress that initiates it the critical stress, τcrit.

According to a simple frictional model, on a flat bed:

critical stress × representative area = friction coefficient × normal reaction
πd 2 πd 3
τ crit  c  μ frict  (ρ s  ρ) g
4 6
where c is a constant of order unity. Hence,
τ crit 2μ frict

(ρ s  ρ) gd 3c
i.e.
τ crit
 dimensionless function of particle size and shape (17)
(ρ s  ρ) gd
In practice, the dependence on shape is not very significant.

The non-dimensional stress

τb
τ*  (18)
(ρ s  ρ) gd
is called the Shields parameter (in the sediment-transport literature often denoted θ) after
American engineer A.F. Shields, who, in 1936, plotted his results on the initiation of
sediment motion in the form of a graph of τ*crit against the particle Reynolds number Rep:
τ *crit  f (Re p ) (19)
in what became known as a Shields diagram. Here,
ud
Re p  τ (20)
ν

In practice, the choice of dimensionless groups in the original Shields diagram is not
convenient for predicting the threshold of motion because the bed stress τb effectively appears
on both sides of the equation. (Remember that the friction velocity in Rep is given by
u τ  τ b / ρ .) You will recall from dimensional analysis in Hydraulics 2 that one is at liberty
to replace either of the dimensionless Π groups by another independent combination. In this
case we can eliminate the dependence on bed stress (or uτ) by forming
Re 2p ( s  1) gd 3

τ* ν2
where s = ρs/ρ is the relative density. Taking the cube root in order to give a dimensionless
parameter proportional to diameter:
1/ 3
 ( s  1) g 
d*  d  2  (21)
 ν 
Shields’ threshold line may then be replotted as a function of τ*crit against d*:
τ *crit  f (d *) (22)

Hydraulics 3 Sediment Transport - 8 Dr David Apsley

 0.15

0.10
t*crit

0.05

0.00
1 10 100 1000
d*

A convenient curve fit to experimental data is provided by Soulsby (1997)1:

0.30
τ *crit   0.055 [1  exp( 0.020d * )] (23)
1  1.2d *

For large diameters the critical Shields parameter tends to a constant value; this is 0.055 for
Soulsby’s formula, although 0.056 is actually a more popular figure in the literature.

Summary of Calculation Formulae For Threshold of Motion

0.30
τ *crit  f (d *) e.g. graphically, or τ *crit   0.055 [1  exp( 0.020d * )]
1  1.2d *
where
τb
τ*  (Shields parameter)
(ρ s  ρ) gd
1/ 3
 ( s  1) g 
d*  d  2  where s = ρs/ρ
 ν 

Example. (Exam 2007 – part)

An undershot sluice is placed in a channel with a horizontal bed covered by gravel with a
median diameter of 5 cm and density 2650 kg m–3. The flow rate is 4 m3 s–1 per metre width
and initially the depth below the sluice is 0.5 m. Assuming a critical Shields parameter τ*crit
of 0.06 and friction coefficient cf of 0.01:

(a) find the depth just upstream of the sluice and show that the bed there is stationary;

(b) show that the bed below the sluice will erode and determine the depth of scour.

1
Soulsby, R., 1997, “Dynamics of Marine Sands”, Thomas Telford.

Hydraulics 3 Sediment Transport - 9 Dr David Apsley

2.2 Inception of Motion in Normal Flow

In the large-diameter limit, particles will move if

τb
 0.056
(ρ s  ρ) gd
But for normal flow, with hydraulic radius Rh and slope S:
τ b  ρgRh S
Putting these together, and noting that ρs/ρ = 2.65 for sand in water shows that, for large
particles in normal flow, the bed will be mobile if
d  10.8Rh S (24)
–4
For example, a uniform flow of depth 1 m on a slope of 10 will move sediment of diameter
about 1 mm or less.

Note that this is an estimate, applying for large particles of a particular density in water and
assuming normal flow. If possible it is better to compare τ with τcrit (or τ* with τ*crit) to
establish whether the bed is mobile.

2.3 Effect of Slopes

The main effect of slopes is to add (vectorially) a downslope component of gravity to the
fluid stress acting on the particles. For a downward slope the gravitational component will
assist in the initiation of motion, whereas for an adverse slope it will oppose it.

flow
flow

 mg mg

Let τcrit be the critical stress on a slope and τcrit,0 be the critical stress on a flat bed. Let the
slope angle be β and the angle of repose .

For slopes aligned with the flow:

sin(  β)
τ crit  τ crit ,0 (where β is positive for upslope flow)
sin 

For slopes at right angles to the flow:

tan 2 β
τ crit  cos β 1  τ crit ,0
tan 2 

For arbitary alignment see, e.g., Soulsby (1997)1 or Apsley and Stansby (2008)2.

2
Apsley, D.D. and Stansby, P.K., 2008, Bed-load sediment transport with large slopes, model formulation and
implementation within a RANS flow solver, Journal of Hydraulic Engineering, 134, 1440-1451.

Hydraulics 3 Sediment Transport - 10 Dr David Apsley

3. BED LOAD

3.1 Dimensionless Groups

Bed-load transport of sediment is quantified by the bed-load flux qb, the volume of non-
suspended sediment crossing unit width of bed per unit time. The tendency of particles to
move is determined by the drag imposed by the fluid flow and countered by the submerged
particle weight, so that qb may be taken as a function of bed shear stress τb, reduced gravity
(s – 1)g (where s = ρs/ρ), particle diameter d, and the fluid density ρ and kinematic viscosity
ν. A formal dimensional analysis (6 variables, 3 independent dimensions) dictates that there
is a relationship between three non-dimensional groups, conveniently taken as
qb
q*  (dimensionless bed-load flux) (25)
( s  1) gd 3

τb
τ*  (dimensionless bed shear stress, or Shields parameter) (26)
ρ( s  1) gd
1/ 3
 ( s  1) g 
d*  d  2  (dimensionless particle diameter) (27)
 ν

3.2 Bed-Load Transport Models

Some of the more common bed-load transport formulae are listed in the table below. (For
those containing τ*crit, bed-load sediment transport is zero until τ* exceeds this.)

Reference Formula Comments

Meyer-Peter and
q*  8(τ*  τ*crit ) 3 / 2
Müller (1948)
Nielsen (1992) q *  12(τ*  τ*crit ) τ*
0.053 τ *
Van Rijn (1984) q*  0.3
(  1) 2.1
d* τ *crit
 K exp( 0.391 / τ * )
Einstein3-Brown  τ *  0.182 2 36 36
q*   0.465 K  3 
(Brown, 1950)  40 Kτ *3 3 d* d*
3

 τ *  0.182
1 τ* τ *crit
Yalin (1963) q *  0.635r τ* [1  ln(1  σr )] r  1 , σ  2.45
σr τ *crit s 0.4

Brown, C.B. (1950). “Sediment Transport”, in Engineering Hydraulics, Ch. 12, Rouse, H. (ed.), Wiley.
Meyer-Peter, E. and Müller, R. (1948). “Formulas for bed-load transport.” Rept 2nd Meeting Int. Assoc. Hydraul.
Struct. Res., Stockholm, 39-64.
Nielsen, P., (1992). Coastal Bottom Boundary Layers and Sediment Transport, World Scientific.
Van Rijn, L.C., (1984). “Sediment transport. Part I: bed load transport.” ASCE J. Hydraulic Eng., 110, 1431-
1456.
Yalin, M.S., (1963). “An expression for bed-load transportation”, Proc. ASCE, 89, 221-250.

3
Prof. H.A. Einstein was a noted hydraulic engineer – he was the son of Albert Einstein!

Hydraulics 3 Sediment Transport - 11 Dr David Apsley

Note.
(i) Models have been written here in a common notation, which may differ a lot from
that in the original papers.

(ii) The actual derivations of the original models vary considerably. Some are mechanical
models of particle movement; others are probabilistic.

(iii) Several models contain a factor ( τ*  τ*crit ) indicating, as expected, that sediment does
not move until a certain bed stress has been exceed. Formulae for the threshold of
motion:
τ*crit  f (d *)
have already been covered in Section 2.

Hydraulics 3 Sediment Transport - 12 Dr David Apsley

4. SUSPENDED LOAD

4.1 Inception of Suspended Load

Particles can not be swept up and maintained in the flow until the typical updraft of turbulent
motions exceeds their natural settling velocity ws. A typical turbulent velocity fluctuation is
of the order of the friction velocity uτ. Thus, suspended load will occur if
uτ
1 (28)
ws

In practice, what is deemed to be suspended load rather than bed load is not precise, and
different authors give slightly different numbers on the RHS.

For much gravelly sediment, suspended load simply does not occur and bed load is the only
mode of sediment transport.

4.2 Turbulent Diffusion

The sediment concentration C is the volume of sediment per total volume of material
(fluid + sediment).

Once sediment has been swept up by the flow it is distributed throughout the fluid depth.
However, since it is always tending to settle out there is a greater concentration near the bed.
Because the concentration is greater near the bed, any upward turbulent velocity fluctuation
will tend to be carrying a larger amount of sediment than the corresponding downward
fluctuation. Thus, where a concentration gradient exists turbulent diffusion will tend to lead
to a net upward flux of material, whereas settling leads to a net downward flux. An
equilibrium distribution is attained when these opposing effects are equal.

dC
C+l
dz
C ws
dC
C-l
dz

Suppose the average concentration at some level z is C. In a simplistic model an upward

turbulent velocity u′ for half the time carries material of concentration (C – l dC/dz), where l
is a mixing length – an order of size for turbulent eddies. The corresponding downward
velocity for the other half of the time carries material at concentration (C + l dC/dz). The
average upward flux of sediment (volume flux × concentration) through a horizontal area A is
1  dC 1 dC dC
2 u A(C  l )  2 u A(C  l )  u l A
dz dz dz
The quantity u′l is written as a diffusivity K, so that the net upward flux is
dC
K A (29)
dz

Hydraulics 3 Sediment Transport - 13 Dr David Apsley

This is referred to a gradient diffusion (because it is proportional to a gradient!) or Fick’s law
of diffusion. The minus sign indicates, as expected, that there is a net flux from high
concentration to low.

At the same time there is a net downward flux of material wsAC due to settling. When the
concentration profile has reached equilibrium the upward diffusive flux and downward
settling flux are equal in magnitude; i.e.
dC
K A  ws AC
dz
or, dividing by area:
dC
K  ws C (30)
dz

4.3 Concentration Profile

Because it is the same turbulent eddies that are responsible for transporting both mean
momentum and suspended particulate it is commonly assumed that the diffusivity K is the
same as the eddy viscosity νt; i.e. for a wide channel and fully-developed flow:
K  ν t  κu τ z(1  z / h)
where κ is von Karman’s constant (0.41), uτ is the friction velocity (  τ b / ρ ), z is the
vertical height above the bed and h is the channel depth. Hence, substituting in (30):
dC
 κu τ z (1  z / h)  ws C
dz
Separating variables and using partial fractions,
dC w 1 1 
 s   dz
C κu τ  z h-z 
Integrating between a reference height zref and general z gives, after some algebra:
C w z ref (h  z )
ln  s ln
C ref κu τ z (h  z ref )
ws
 h / z  1  κuτ
 ln  
 h / z 1
 ref 
Hence,
ws

C  h / z  1  κuτ
  (31)
C ref  h / z ref  1 

This is called the Rouse profile, and the exponent

ws
(32)
κu τ
is called the Rouse number after H. Rouse (1937). Typical concentration profiles are shown
below.

Hydraulics 3 Sediment Transport - 14 Dr David Apsley

To be of much predictive use it is necessary to specify Cref at some depth zref, typically at a
height representative of the bed load. There are many such formulae but one of the simplest is
that of Van Rijn (see, e.g., Chanson’s book):

 0.117  τ *  
C ref  min   1, 0.65

 d *  τ*  
  crit  
1/ 2
z ref  τ* 
 0.3d *0.7  *  1 (33)
d τ 
 crit 
where d* and τ* are the dimensionless diameter and stress defined in Section 2.

0.8

0.6
z/h
Rouse number
0.4 0.1
0.2
0.2
0.5
1
0
0 0.2 0.4 0.6 0.8
C/Cref

4.4 Calculation of Suspended Load

Equation (31) describes the concentration profile, but usually the most important quantity is
the total sediment flux.

The volume flux (per unit width) of fluid through a depth dz in a wide channel is
u dz
Since concentration C is the volume of sediment per volume of fluid, the volume flux of
sediment through the same depth is
Cu dz
Hence, the total suspended flux through the entire depth of the channel is (per unit width):
h

q s   Cu dz (34)
zref
This must be obtained from C and u profiles by numerical integration – see the Examples.

Hydraulics 3 Sediment Transport - 15 Dr David Apsley

Summary of Main Suspended-Load Formulae

Inception of suspended load:

uτ
1 (typical)
ws

Concentration profile:
ws

C  h / z  1  κuτ
 
C ref  h / z ref  1 
where:

 0.117  τ *  
C ref  min   1, 0.65

 * 
 d *  τ crit
  

1/ 2
z ref  τ* 
 0.3d * 0.7   1
d  τ* 
 crit 

Velocity profile:
u z
U ( z )  τ ln(33 )
κ ks

Sediment flux (per unit width):

h

q s   CU dz
zref

Hydraulics 3 Sediment Transport - 16 Dr David Apsley