You are on page 1of 140

ISSN 0169-6548

Communications on hydraulic
and geotechnical engineering
Modelling turbidity currents in reservoirs

May 1994 C.J. Slaff

-~'
T De If t Faculty of Civil Engineering

Delft University of Technology


Modelling turbidity currents in reservoirs

by

C.J. Sloff
May 1994

Communications on Hydraulic and

Geotechnical Engineering

Report No. 94-5

Faculty of Civil Engineering

Delft University of Technology


Abstract

A two-layer mathematical model is presented for sedimentation in reservoirs where


turbidity currents are to be expected. As the model is two-dimensional in plan, the
suspended-sediment concentration of the turbulent underflow is described by a
depth-integrated form of the convection-diffusion equation originally proposed by
Galappatti in 1983. The bed-level variations governing this model are described by
a depth-integrated sediment balance. For closure of the model various closure
relations are required based on the vertical distribution of flow and sediment of the
turbidity current. A semi-empirical model, presented to describe these distributions,
is used to quantify boundary shear stresses, suspended-sediment transport rate, and
the adaptation scales in Galappatti's equation. Additionally a discussion is given on
the applicability of various existing relations for interfacial mixing, as well as on
relations for the near-bed sediment concentration. The presented model, and the
proposed simplified models deduced from it, can be used for various types of
reservoirs and to more conventional computations such as for saline underflows.

Acknowledgment

This report has been written during my employment as a research assistant (AIO) in
the Hydraulic and Geotechnical Engineering Division of the faculty of Civil
Engineering of the Delft University of Technology. This study is carried out as part
of a research project to sedimentation in reservoirs, a joint cooperation between
Delft University of Technology and Delft Hydraulics, under supervision of Prof. Dr.
M. de Vries (Delft University of Technology).

Prof Dr. M. de Vries is gratefully acknowledged for his encouraging support and
critical reading during the writing of this report. Also Dr. Z.B. Wang is
acknowledged for his critical screening of this report and for his valuable
suggestions.

I would also like to express my appreciation to Prof. Dr. M.H. Garcia for sending
me his comprehensive works on turbidity currents, and to A. Sieben for his fruitful
discussions.

2
Contents

l. Introduction 5

2. Basic equations for a 2-DH two-layer model


2.1 Introduction 7
2.2 General basic equation for two-layer 2-DH models with mobile bed 7
2.3 Boussinesq equations for two-layer 2-DH models with mobile bed 12
2.3 Decoupling of external and internal flow 14

3. Galappatti's equation for a 2-DH two-layer model


3.1 Introduction 19
3.2 Convection-diffusion equation for suspended sediment 19
3.3 Derivation of Galappatti's equation 22
3.4 Suspended-sediment transport rates 25

4. Flow structure and sediment transport in turbidity currents: a semi-empirical


approach
4.1 Introduction 27
4.2 General flow structure and governing basic equations 28
4.3 Flow structure of the lower part of a turbidity current 34
4.4 Flow structure of the upper part of a turbidity current 36
4.5 Analysis of the velocity profiles for the full turbidity current 39
4.6 Suspended-sediment transport and equilibrium concentration 43
4.7 Galappatti's coefficients for the semi-empirical approach 52
4.8 Alluvial roughness 60
4.9 Summary and conclusions 61

5. Interfacial mixing
5.1 Introduction 65
5.2 Discussion of some empirical entrainment relations 69

6. Conclusions 77

Main symbols 81

References 85

Appendix A General derivation of basic equations


App.A. I Basic equations for 3-D sediment-laden flow in a reservoir 93
App.A.2 General basic equations for a 2-DH flow in a reservoir 96

3
App.A.3 Integration of pressure terms 98
App.A.4 Derivation of basic equations for a two-layer model 100
App.A.5 Components of gravity force in an inclined
coordinate system 108

Appendix B Derivation of a 2-DH Galappatti model for suspended-sediment


modelling in turbidity currents
App.B.l General derivation of the asymptotic solution 111
App.B.2 Galappatti's model for the 2-DH case 118

Appendix C Galappatti's coefficients for turbidity currents: tables and figures


App.C. l Regression coeficients 123
App.C.2 Figures 127

Appendix D Equation list


App.D.l Vertical distribution of velocity and concentration 133
App.D.2 Empirical relations for the reference concentration 136

4
Chapter 1

Introduction
The hydrodynamic and morphological behaviour of a reservoir, such as a
hydropower or water-supply reservoir, is characterised by a wide-range of physical
processes. Most of these processes are still poorly understood and rather
complicated, thus an exact mathematical simulation is still not possible (Sloff,
1991). However, if the processes are properly schematized it is possible to develop
mathematical models that can reasonably estimate the actual physical behaviour.

In this report a mathematical model is described for reservoirs where turbid density
underflows are to be expected. When a sediment-laden river inflow enters a
reservoir, it can plunge under the clear (almost stagnant) reservoir water to form a
turbidity current (see figure 1.1). This turbid underflow can propagate through the
reservoir transporting a part of its suspended sediment towards the dam. Field
measurements (Chikita, 1989 & 1990, Chikita and Okumura, 1987 & 1990, Fan,
1985a & 1986, Lambert and Giovanoli, 1988) show that these turbidity currents are
characterised by a distinct density interface, so that the flow can be considered as
two separate layers: one clear quiescent upper layer, and a turbid lower layer
carrying the greatest amount of sediment. Because of the time variation of the
incoming flood discharge the turbidity current is unsteady. Usually the turbidity
current moves along preexisting (river) channels to which its width is confined.

Due to the turbulent nature of the underflow it entrains clear water from the upper
layer through its interface. Meanwhile it entrains or deposits sediments at the
bottom. The velocity and sediment distributions over the depth are related to
interfacial and bottom shear-stresses and the rate of turbulent energy. When in a
depositing turbidity current the density difference to the ambient water diminishes,
it will loose its transport energy and will finally disappear. When the current meets
an obstacle to. the flow (such as a dam or a contraction) a backwater effect may be
produced and a wave of reflection may progress upstream resulting in an increase
of its thickness. Then sediments may settle out, for example forming a muddy lake
at the dam site when the bottom outlets are closed. Clearly it is possible to reduce
the deposition of these sediments by sluicing the current through the bottom outlets.

The water movement and sediment distribution in the reservoir is assumed to be


primarily dependent on river in- and outflow. Other secondary influences, such as
wave action, temperature gradients and rare events, are neglected. The motion of
water is described by Reynolds equations for turbulent flow, with a variable density
which is defined by the sediment concentration. The transport of suspended
sediment is described by the convection-diffusion equation (mass-conservation). For
the mobile bed situation the equations are closed with a sediment-balance equation

5
which relates bed morphology to sediment transport.

The two-layer model is derived by depth-integration and combination of the 3-D


equations of mass and momentum of four contiguous sub-layers: a bottom layer, a
bed-load layer, a suspended-load layer, and a (clear water) upper layer. To obtain a
relation for the depth-averaged suspended-sediment concentration from the
convection-diffusion equation we adopted the asymptotic approach proposed by
Galappatti (1983,1985) instead of direct integration (chapter 3). The advantage of
Galappatti's asymptotic solution is that an empirical entrainment function at the bed
is not required.

Since field-data and laboratory experiments indicate that density differences


between clear reservoir water and the turbid underflow are small, the general model
is further simplified in chapter 2, using the Boussinesq approximation and
decoupling of internal and external flow.

Closure of the two-layer model requires the quantification of boundary shear


stresses, sediment transport and interfacial mixing. Furthermore the coefficients in
Galappatti's equation must be computed from the vertical distribution of flow
velocity and sediment concentration in the turbidity current. In chapter 4 general
semi-empirical equations are derived for these distributions. The relevant parameters
governing these relations are used to determine regression equations for the
coefficients in Galappatti's equation, and are used to express alluvial roughness and
suspended-sediment transport.

Interfacial mixing is considered here as the net entrainment of clear water into the
underflow, and expressed as an entrainment velocity. In chapter 5 a discussion is
given on the applicability of some entrainment relations reported in literature.
Finally, in chapter 6, some conclusions are drawn with respect to the presented
theory and its use to modelling turbidity currents in reservoirs.
Plunge point
I

Deltaic \\
deposits ------
Turbidity current

Figure 1.1 Turbidity current in a reservoir

6
Chapter 2

Basic equations for a 2-DH two-layer model

2.1 Introduction

The dynamics of a turbidity current into a reservoir are determined by a complex 3-


D interplay of turbulent flow and sediment suspension. Simultaneously the mobility
of the bed blends into these processes by sediment entrainment and deposition
processes. In this chapter a two-layer depth-averaged model is presented for a turbid
underflow, spreading on the bottom of a reservoir. Turbidity currents in reservoirs
are non-conservative gravity currents in the sense that they loose their buoyancy due
to settlement of suspended sediment. They are closely related to conservative
gravity currents. Generally the conservative currents are salt-water density currents
such as salt water wedges and lock exchange flows. In literature we can further
subdivide the gravity currents in those on horizontal and those on sloping bottoms.
All these groups of currents are related and cannot be treated separately. For large
Reynolds numbers (> 1000) a gravity current can be considered fully turbulent. Then
viscous forces become small compared to buoyancy forces, and a turbulent mixing
pattern can be seen.

The 2-layer model is derived by means of depth-integration of the 3-D equations of


momentum and mass for fluid and sediment (appendix A). Firstly, in section 2.2,
general equations are presented which allow for high concentration (provided that
the flow remains turbulent). Secondly, in section 2.3, the model is further simplified
by adopting the Boussinesq approximation for turbidity currents with low
concentrations, which are common in nature. A third simplification, for low-
concentration internally-subcritical turbidity currents, is introduced in section 2.3 by
decoupling of external flow, internal flow and bed morphology. Finally the
necessity of using a two-layer model instead of a one-layer model (for deep water)
is clarified by means of the characteristic celerities of the models.

2.2. General basic equations for two-layer 2-DH models with


mobile bed

In appendix A the general basic equations for two-layer 2DH flow on a mobile bed
are derived. The derivation follows the approach proposed by Sloff (1993) for 1-D
sediment-laden flows in unstratified channels. Firstly general equations of mass and

7
momentum for fluid and sediment are determined by means of depth integration of
Reynolds equations over the depth of a flow layer. Secondly we defined four layers
which are respectively: a bottom layer consisting of uniform sediment and pores
filled with stagnant water; a bed-load layer close to the bed in which sediment is
transported as bed-load; a suspended-load layer extending up to the density interface
in which sediment is transported as suspension; and the upper layer above the
turbidity current which is almost free of sediment. The two-layer model follows
from this schematization after application of the general basic equations to each
layer and after combining the equations of the three lower layers to one turbid-
underflow layer. To account for diffusion of suspended sediment in the lower layer
we extended the mass-balance equations in appendix B and used a depth-integrated
model for suspended sediment as presented by Galappatti (1983) and Wang (1989).
In this way we can account for the adaptation time and length of the suspended-
sediment concentration in case of erosion or sedimentation.

For all these derivations the following assumptions and methods apply:
a)- The density difference between the layers is small (<5%). On the other hand
it must be sufficiently large (with a minimum value related to the velocity
difference) so that the layer model remains stable.
b)- The turbidity current is assumed to be fully turbulent. Turbulence breakdown
at large sediment concentrations, e.g., due to sediment entrainment, can be
examined with an turbulent-energy balance (Parker et al.,1986).
c)- The vertical accelerations of the flow are small compared to the gravity
acceleration. The pressure is therefore hydrostatic. There are no abrupt
changes in the flow parameters (large obstacles, rapid slope changes, etc.).
This assumption excludes the precise computation of the structure of a
density front or an internal hydraulic jump. However, in Chapter 6 is shown
that the approximate computation of the motion of a front is still possible.
d)- The equations are derived for a coordinate system along the (average)
reservoir bottom slope. It is therefore necessary to use the components of the
gravity acceleration (for momentum and mass conservation) relative to the
inclined coordinate system. Hence the gravity-acceleration vector is rewritten
to allow for flow on steep slopes. In appendix A, App.A.5, is shown how the
gravity vector can be written in the tilted coordinate system.
e)- The sediment concentration in the upper layer is neglected, so that its density
equals that of clear (fresh) water. The density of the lower layer is defined
completely by the concentration of suspended-sediment particles (bed-load
transport only effects a small layer near the bottom without influencing the
net average momentum and mass transfer in the flow).
f)- Discharges (and flow velocities) in the upper layer are small. A turbidity
current in a relatively shallow reservoir may cause a return flow in this layer
by entrainment of fluid. The assumption of small discharges may need to be
verified in run-of-the-river reservoirs with a significant mean throughflow.
g)- The similarity assumption applies for all horizontal directions, i.e., the shape

8
of the velocity- and concentration profiles (depth distributions) are similar at
every location in all directions (Parker et al., 1986). This assumption
excludes the influence of secondary flow.
h)- Velocity and density profiles (depth distribution) can be described by the
depth-averaged value multiplied with dimensionless shape functions (e.g.,
u(t,x,y,z) = Uz(t,x,y) ·1.VJr1) with ri = zlai(t,x,y)). The integral of the shape
functions over the depth of the respective layer yields unity. When different
shape functions are multiplied (e.g. for u2 ·p2, or other convection terms) the
values of the integral of these terms may still be taken approximately equal
to unity (Ellison and Turner, 1959).
i)- At the density interface mixing of fluids can be expected due to turbulence
and related shear stresses. An interfacial mixing layer is formed in which
clear water is entrained from the upper layer, until the instability of the flow
in this layer is suppressed by the density gradient. We assume in accordance
to existing theory for turbidity currents that the boundary between upper and
lower layer is located at the upper edge of the mixing layer. Entrainment of
clear water through this edge is expressed by means of an entrainment
velocity. Details on the turbulent structure of a turbidity current is given in
Chapter 4, while interfacial mixing (entrainment) is treated in Chapter 5.
j)- For very large water bodies it is common to introduce the effect of the earth
rotation on the flow with the Coriolis coefficient (the coefficient of
geostrophic acceleration).
k)- The convection-diffusion equation is integrated over the depth with the
method proposed by Galappatti (Galappatti, 1983, Galappatti and
Vreugdenhil, 1985) which is extended for 2DH-models by Wang (1989). The
depth-integrated form of the equation is found by substituting an asymptotic
solution of the depth-integrated concentration into it, and by using a
concentration or gradient type bed-boundary condition. The resulting method
accounts for the spatial and temporal adaptation of the average suspended
sediment concentration in the time and space to the changing flow conditions
(bed-load transport is assumed to adapt instantaneously to changing flow
conditions and is described by a sediment-transport formula). The
redistribution process of the concentration profile (adaptation of Cs to Cs,) is
governed by adaptation time Ta and lengths L0 • The derivation of Galappatti's
equation is treated in Chapter 3.

The general 2-DH model for sediment-laden turbidity currents is derived in


appendix A, section App.A.4 (without using the Boussinesq approximation for small
concentrations, see section 2.2). The model can be rewritten in terms of discharges
and sediment-concentrations, instead of velocities and densities. Furthermore we
neglect interfacial sediment exchange and longitudinal velocities at the interface
level. A definition sketch of the model is presented in Figure 2.1.
If the sediment concentration in these equations is considered as a constant, the
model can be recasted into the more familiar two-layer models which are used to

9
Figure 2.1 Definition sketch

compute saline density currents. For example the model presented by Vreugdenhil
(1979) or in one dimension the model of Schijf and Schonfeld (1953).

2DH Basic equations for the upper lqyer ofa reservoir

Continuity equation (from equation A45, appendix A)

- w.,e = 0 (2.3)
2DH Basic equations for the turbidity current in a reservoir
If we define:

DCS
-- =acs
- q2x acs
+-- +--
q2y acs
Dt at a2 ax a 2 oy

and p2 = p1{l+cr'C,) then we may write the equations as follows:


Momentum equations (from equations A37 and A38, appendix A):

aq2x
at
+ j_(q;)
ax a2
+ j_(qhq2y)
oy a 2
+ ( Pj)gza2 aal
p2 ax
+ gza2 a(a2 +zb)
ax + (2.4)

+ 1 2CJ '(Pj)acs
-gzGi - --
(•xb-,xi) -
+ -'-----'- f q2y - gxa2 + qhCJ '(Pi)DCS
- -- = 0
2 p2 ax p2 p2 Dt

+ (2.5)

= 0

Mass balance equation of mixture (from equation A41, appendix A):

oa2 + oq2x + 0%y + ozb + W. = Q (2.6)


at ax ay at "'
Sediment balance (from equation A42, appendix A)

(2.7)

Depth-averaged suspended-sediment is governed by the convection-diffusion


equation, which is replaced by Galappatti's equation as defined in Chapter 3.

The following variables are defined:


a 1,a2 depth of upper layer and lower layer respectively
c. depth-averaged suspended-sediment concentration
f Coriolis coefficient for earth rotation
gx g-sina/cosay' (See appendix A, App.A.5)
gy g-cosa/sinay' (See appendix A, App.A.5)
gz g-cosa/cosay' (See appendix A, App.A.5)
qlx>qly discharge per unit width in x,y-direction in upper layer
q2x>q2y discharge per unit width in x,y-direction in lower layer

11
Sbx,Sby bed-load transport in x,y-direction
ssx,ssy suspended-load transport in x,y-direction
W;, entrainment velocity at the interface
ws fall velocity of sediment particles (relative to z-axis)
zb bed level
a/,ay' average bed slope in x,y-direction
f,p porosity of the bed material
Pt density of water (upper layer)
P2 density of lower layer= pJcr'C5+l)
Ps density of sediment particles
cr' relative sediment density= (Ps·P1)/p1
•x;,'ty; shear stress at density interface in x,y-direction
'txb> 'tyb bed shear-stress in x,y-direction

Although the simplicity compared to a fully 3-D model is obvious, the price to be
paid for simplicity is the requirement of closure relations. These have to describe
physical processes which are still rather obscure. However, without depth-averaging,
the results presented by Eidsvik and Brnrs (1989) indicate that the problems with
turbulence closure and quantification of suspension do not necessarily prove the
superiority of 2-DV or 3-D models.

The primary closure for the presented model consists of relations for velocity and
sediment profiles, boundary shear stresses (notably bed-shear), entrainment
velocities (interfacial mixing), bed-load transport and sediment fall-velocity. These
closure relations are treated in the following Chapters.

The system of hyperbolic partial differential equations presented above has time and
space varying coefficients, so that the characteristic celerity at each point of the
domain (x,y,t) depends on the solution itself, and the characteristics of the same
family may intersect one another (Cunge et al., 1980). Furthermore this system is
not in divergence form (i.e. it is not a system of conservation laws according to
Lax, 1957). Friction and entrainment act as source and sink terms in the equations,
so that the momentum or mass contained in any domain G of x,y-space does not
change at a rate equal to the flux of the vector field into G (this vector field is not
divergence free). More details on these mathematical properties is given in Chapter
6 in with respect to the computation of frontal motion.

2.2 Boussinesq equations for two-layer 2-DH models with mobile


bed

In the previous section a general two-layer model is presented for turbidity currents
in a reservoir. However, for small sediment concentrations (say Cs<0.05) the
Boussinesq approximation is commonly used in turbidity-current modelling. The

12
approximation states that density differences only affect body forces. Hence it
affects the basic equations by taking the density p2 equal to p1 everywhere except
when multiplied with g.

After adopting the momentum and mass-balances according to the Boussinesq


approximation, again the equations can be integrated over the respective layers in
the same manner as presented in appendix A for the general model. Clearly the
basic equations for the upper layer remain unchanged compared to those for the
general model. The resulting equations for the turbid underflow are presented
below.

2DH Boussinesq equations for the turbidity current

Momentum equations (neglect Coriolis force):

(2.8)

2
1
-g CJ
/ acs'½
-- +
2 z oy (2.9)

where h, = water-surface level = zh + a 2 + a 1

Mass balance equation of mixture:

+ w.,e = 0 (2.10)

In most practical situations it suffices to use the Boussinesq equations, which are a
more simple version of the previously presented general model, and hence which
are much easier to work with. In the following section these equations are used to
achieve an even more simple approach based on an distinction between external and
internal phenomena.

13
2.3 Decoupling of external and internal flow

For two-layer models the characteristic analysis showed that internal and external
flow may be decoupled if density differences are small (Sloff, 1992). The external
flow phenomena correspond to those in unstratified open-channel flow (long waves,
backwater, etc.) and we may assume they are hardly affected by internal
phenomena, such as internal waves and internal bores. By decoupling internal and
external flow the model is significantly simplified. Since the propagation speed of
external waves is of the order of magnitude of 10 times (or more) the internal wave
speed, we even considered to adopt a quasi-steady computation for the external flow
with respect to internal flow and morphology. These simplifications are essential
from computational point of view since numerical solution requires a time step
adjusted to the largest characteristic celerity to prevent possible instability. By
decoupling (and possibly computing external flow as quasi-steady) we eliminate the
large external celerities and we may use time steps which are of the order of ten
times those for computing with the coupled model. Unsteady external flow
phenomena are not relevant for us since they are of practically no importance to the
morphology (acting on a different time scale). For derivation of the decoupled
model we started from the Boussinesq equations, recalling that density differences
have to be small and consequently that concentrations have to be small.

We can divide the system into equations for the internal and for the external
phenomena. The equations for the external phenomena influence the internal ones
by the relations:
(2.11)

(2.12)

(2.13)

The momentum equations for external flow are found by adding the basic
(Boussinesq) equations (2.1)+(2.8), and by adding (2.2)+(2.9). The continuity
equation is found by adding (2.3)+(2.10).

aqx
at
+ _§__[
ax
q;la + _§__( qxqy)
oy a
(2.14)

where h, is again the water surface level.

14
aqy g a ohs - g a = 0 (2.15)
+ _§_(qxqy) + _§_[q:) + "yb -
at ax a oy a z ;:,.,
V.)' Pt
y

(2.16)

The following relations were used to rewrite convection terms:


2 2 2 2
qlx q2x qx a1a2 2 qx (2.17)
+ = -
a
+ -(u -u)
a 1 2 "'
al az a

2 2 2 2
qly qzy qy a1a2 z qy (2.18)
+ + - ( v 1 - vz) "'
al az a a a

qlxqly + q2xq2y (2.19)


al az

The simplifications of these terms as shown above, follow from the stability
requirements for our two-layer model and the assumption of small density
differences, which states (Sloff, 1992):

(2.20)

where we define a relative density ei

(2.21)

Thus terms with (u 1-u2) and (v 1-vz) in the convection relations as above may be
neglected.

External flow is only weakly influenced by density differences because all terms
above are small compared to the others if ed <ii; 1. The equations for external flow
represent long-wave shallow-water equations for open-channel flow. Clearly any
interaction with internal waves or other internal phenomena is neglected.
Note that bed friction -rb oc u2 Iu2 I where u2 can often be approximated by q!a

15
similarly to the simplifications of the convective terms.

External flow influences internal flow by the relations (2.11), (2.12) and (2.13). The
terms which include the gradient of surface elevation, which is determined by
external flow, must be eliminated from the original momentum equations to find
those for internal flow. This can be achieved by combination of the momentum
equations (2.8) -(a/ a 1)-(2.1) and (2.9)-(a/ a 1)-(2.2).

The continuity equation for internal flow is equal to (2.3).


aaz aq2x aq2
+ - - + _ _Y + W . = 0 (2.24)
at ax ay "
The sediment continuity equation equals (2.7).

(2.25)

The possibility of decoupling is verified with computations of gravity current fronts


(to be published later). It is shown that the expected error in computing with the
decoupled model is of the order of ed which is reasonable if compared to the errors
introduced for example by the closure relations for sediment transport.

Another simplification which is usually proposed for modelling turbidity currents in


submarine canyons (e.g., Parker, et al. 1986), is the adoption of an infinite
'reservoir' depth. Through this assumption the water-surface gradient in the
Boussinesq equations can be discarded from the beginning, hence it is a rigid-lid
approach. Also upper-layer balance equations are not longer necessary. However, it
can be shown from theoretical considerations that only in a really infinite deep

16
reservoir this rigid-lid assumption is justified.

1 i :I 'I; i
I r I
<l>+[m/s] !
@I ~ ...
0.8 v """"
i~
0.6 V
I/

0.4 I

0.2
I Ii I
0
100 1000
a [m]
Figure 2.2 Comparison of 1-D positive internal-wave celerity for (1): decoupled
model or general model, (2): rigid-lid model.

In figure (2.2) the positive characteristic 1-D celerities for an internal wave derived
from the model with rigid-lid are compared to that from the decoupled model (the
latter gives an approximately equal celerity as the general model). For comparison
we used data from Sanmenxia reservoir in China reported by Fan (1991):

The real reservoir depth a was of the order of 15 m, but in figure (2.2) is plotted
how the celerities approach each other asymptotically for increasing reservoir depth.
Clearly for this type of shallow reservoirs the rigid-lid model is not appropriate.

17
81
Chapter 3

Galappatti's equation for a 2-DH two-layer model

3.1 Introduction

The transport mechanism for suspended-sediment particles can be represented by


means of the convection-diffusion equation which follows from conservation of
mass for a unit volume in a turbulent flow. It describes the process of convection
and turbulent diffusion in terms of the local sediment concentration. The concentra-
tion field in 2-DV and 3-DV flows can then be computed if appropriate boundary
conditions are specified. From that the sediment transport rate can be determined by
integration.

If the convection-diffusion equation is depth-integrated (in line with our approach)


the process of vertical adjustment of concentration profiles is replaced by an
entrainment function, different than the original bed-boundary condition. The
required calibration for this empirical or semi-empirical function decreases the
predictive power of these models. To overcome this problem Galappatti
(Galappatti,1983, Galappatti and Vreugdenhil, 1985) proposed an asymptotic
approach to the solution of the convection-diffusion equation, which can be used if
the deviation of the concentration profile from the equilibrium profile is small. The
depth-averaged concentration is then theoretically determined instead of empirically,
since the convection-diffusion equation is actually solved (although approximately).
Consequently the resulting equation for the depth-averaged suspended-sediment
concentration, which we call Galappatti's equation, does not involve any empirical
entrainment function, and is therefore much more robust. Wang (1989) extended the
originally 1-D· Galappati's equation to two-dimensions, and in this chapter a further
extension of this theory is presented for 2-D turbidity currents. The applicability
and validity of the original approach was studied by Wang (1992)

3.2 Convection-diffusion equation for suspended-sediment

In appendix A (App.Al) the (Reynolds) mass-balance equations have been


presented for fluid and sediment neglecting the horizontal diffusion of the sediment.
Applying Reynolds procedure to the full 3-D mass-balance equations and applying
the eddy-viscosity concept (e.g., as presented by van Rijn, 1987) the following full
3-D convection-diffusion equation is found (mass-balance equation for sediment):

19
(3.1)

Where: cs local sediment volume concentration


u,v local fluid velocity in x and y-direction
w local fluid velocity in z-direction
particle fall-velocity
sediment-mixing coefficients in x and y-direction
sediment-mixing coefficient in z-direction

The mass-balance for fluid can be written as

(3.2)

Where:
efeelY fluid-mixing coefficients in x and y-direction
eft fluid-mixing coefficient in z-direction

If we add the 3-D mass-balance for sediment and for fluid (assuming that fluid-
mixing and sediment-mixing coefficients are approximately equal for fine sediment)
we derive the following continuity equation for the fluid-sediment mixture:

au +
av +
aw (3.3)
ax d)l az

Using this equation we can rewrite the convection-diffusion equation as

(3.4)

We assume that (1-cs)ws = ws,eff = effective particle fall velocity. Richardson and

20
Zaki (1954) have shown that this fall velocity is not only affected by the return
flow due to the displaced fluid but also by additional effects such as particle
collisions, particle induced turbulence and modified drag forces. Therefore we may
express the effective fall velocity as
(3.5)

Where w, = particle fall velocity in clear, stagnant fluid


a = coefficient (::.4 to 5 for particles ranging from 50-500 µm)

For small concentrations we may assume that w,.,ff= w, (using a ➔ 0).

The mass-balance equations presented above can be integrated over the depth of a
turbidity current similarly as shown in appendix A. After combination of equations
for bottom, bed-load and suspended-load layer the following equations are found for
the lower layer of the two-layer model with a= 1:

(3.6)

in which
bed-load transport per unit of width in x-direction
= bed-load transport per unit of width in y-direction
Zi
'

oc
ssx =
fucsdz - f
Za Za
£
sx
5
-dz
ax (suspended-load)

reference level (lower boundary of suspended-load layer)


level of density interface between the layers (=zb+a2).

Note that these equations only differ in ssx and ssy from the version presented in
appendix A in which horizontal diffusion terms were neglected. This diffusion only
appears in the gradients of the suspended-load transport.

If we assume that there is no exchange of suspended-sediment at the density

21
interface than the right hand side of the sediment mass equation (3. 7) becomes
equal to zero. This part of the equation represents the (zero) exchange of sediment-
mass at the interface and will be used as the interfacial boundary condition for our
model.

Note that equation (3.7) is closed by using a sediment transport formula for the
bed-load transport and appropriate boundary conditions at level z;, Furthermore to
compute the suspended-load transport the concentrations cs (and the depth-averaged
concentration Cs) must be computed from the convection-diffusion equation.
Practically the crucial problem in the latter computation is the bed-boundary
condition, for which still no well defined solution is available. Furthermore the
computational effort required in solving the convection-diffusion equation for long
term morphological computations is a major drawback for using this approach.

Instead of solving the full convection-diffusion equation, Galappatti's asymptotic


solution can be used to determine the depth-averaged concentration and sediment-
transport rates. In section 3 .3 the suspended-sediment vector is written in terms of
the depth-averaged concentration Cs following from Galappatti's approach. The
derivation of Galappatti's equation and its application to turbidity currents is
discussed in the following sections.

3.3 Derivation of Galappati's equation

In 1983 Galappatti (see also Galappatti and Vreugdenhil, 1985) proposed that the
solution of the convection-diffusion equation can be expressed in terms of an
asymptotic expansion, provided that the deviation of the concentration profile from
the equilibrium profile is small. A great advantage of Galappatti's equation is the
absence of an empirical relation for deposition/pick-up rate (entrainment rate) near
the bed. The elaborate derivation of the model is given in appendix B, and is based
on Wang's (1989) 2-D extension of the original 1-D approach.

In appendix Bis shown how the convection-diffusion equation (in normalized form)
can be written in terms of differential operators:
(3.8)

with transformed coordinates

and with normalized velocity profiles

22
\Jllr1) = u(TJ)lu = nonnalized main flow-velocity profile in x-direction
\Jl.CrJ) = v(ri)lv = nonnalized main flow-velocity profile in y-direction
At the density interface of a turbidity current the interfacial boundary condition, as
stated in section 3 .1, becomes

[• ~ : + w,c, - w0 c,t, • [•~ ~ + c, - c,w; L• 0 (3.9)

Where = w;/ws = nonnalized fluid entrainment coefficient


= e,)(w,.as) = nonnalized diffusion coefficient

At reference level za (i.e., ri=O) two types of bed-boundary conditions are proposed:
* concentration type / Dirichlet type:
(3.10)

* gradient type/ Neumann type:

(3.11)

where c.(ri) = Cs, tlo(rJ) is the equilibrium concentration, a'0(rJ) is the equilibrium
concentration-profile function, and ca = c,(O) is the equilibrium bed concentration.
The gradient type condition assumes that at level za the upward diffusive flux is
only determined by local conditions.

Now the theory of Galappatti states that the (approximate) solution of the
convection-diffusion equation can be presented as an asymptotic expansion

(3.12)

where cj is one order of magnitude smaller than cH After substitution of this


asymptotic solution into (3.8), tenns of the same order of magnitude can be
collected:

(3.13)

Formally also the boundary conditions should be treated in this manner (see Wang,

23
1989, 1992), and a bed-boundary condition should be used for each term ci'
However, Galappatti made the important assumption that only the zeroth-order term
contributes to the depth-averaged concentration so that the bed-boundary condition
only has to be used once (for c0). Alternatively Wang (1989) proposed a more
general approach by introducing a set of test functions, which is discussed in detail
in appendix B.

The solution of the problem involves the application of the inverse operator 0· 1 on
the different shape factors (a0 , \Jiu ti'0, etc.) related to concentration (50 ('11), .. ) and
velocity profiles (\Jflr1), ..). The inverse operator is derived in appendix B and
becomes equal to that given by Galappatti if w,' equals zero (no interfacial mixing).
For complicated flow and sediment profiles, as can be found in turbidity currents,
the solution to this inverse operator requires numerical integration. In chapter 4
regression coefficients for turbid underflows have been determined from such
numerical computations for a large number of possible shape factors, under various
conditions. Galappatti has carried out a similar exercise for open channel flows.

For unsteady 2-DH problems no more than the first-order solution can be applied in
practice (Wang, 1989): c, = c0 + c 1•
Furthermore we neglect horizontal diffusion and assume similarity of concentration
and velocity in all horizontal directions (no helical flow). Then the resulting
Galappatti equation in terms of an dimensionless adaptation time and length can be
expressed as (equation B40, appendix B):

I as acs I Uas aCs I Vas aCs


Cse - Cs = Ta-- + L -- + L -- (3.14)
ws at a ws ax a ws ay

where

La' =
Y2 "2 = dimensionless adaptation length
Yo «o

Y1 a, dimensionless adaptation length if a concentra-


Ta 1 =
Yo «o tion-type bed-boundary is used

Ta, = (Y 1 + 1) a1 dimensionless adaptation length if a gradient-


Yo «o type bed-boundary is used

The coefficients in these adaptation scales are defined as

24
I

aj = f Wu(ri)a/r1)dri
0

Since a0(ri) represents the equilibrium concentration profile, and \!fu(Tt) the velocity
profile, the use of Galappatti's model is only possible if mathematical relations can
be defined for these profiles. This closure problem of Galappatti's equation is
tackled in the following chapter. Galappatti (1983) already showed for open-channel
flow, by using simple profile functions, that the adaptation time depends on the
ratio wju., the boundary level Tia, and the roughness scale Tio (or Chezy value).
These results also indicate that the difference between the model derived by using a
gradient bed-boundary, versus that derived by using a concentration bed boundary,
becomes more pronounced for low values of wju•. In general the adaptation time
computed with the gradient-type bed-boundary is larger for all values of wju•.

For 1-D models the same Galappatti equation can be used just by discarding the
derivatives in y-direction (or taking v = 0).

3.4 Suspended-sediment transport rates

The suspended-sediment transport rate can be found by substituting the asymptotic


solution for c, from equation (B42) or (B43) in appendix B into the suspended-
sediment transport vector of equation (B7) (see Wang, 1989, or Ribberink, 1986).
The horizontal diffusion terms can be included, but, as stated before, in practice
their magnitude is very small. Although they have a stabilizing effect on the
computations they are neglected in the following.

The convective part of the suspension transport (in x-direction) is defined as


Z; I

J ucs dz = asiiJljrJr1)csCri)dri = a/iCs (3.20)


z. 0

and similar in y-direction:


Z; I

J VCS dz = as VJ ivvcsdTJ = as vCS (3.21)


z. 0

25
26
Chapter 4

Flow structure and sediment transport in turbidity


currents: a semi-empirical approach

4.1 Introduction

For derivation of the two-layer model in chapter 2 some assumptions have been made
based on the structure of the flow. Especially the computation of suspended-sediment
transport is based on this structure. In this chapter an approximative semi-empirical
approach is presented to quantify the flow structure and related sediment transport
capacity for turbidity currents in a reservoir.

In turbidity currents the turbulence related flow structure and the sediment-
concentration distribution are coupled in a complex way. Vertical exchange of
momentum through agency of turbulent mixing gives rise to local stresses in the flow
which affects the shape of the velocity profile and sediment-concentration distribution.
In turn the mixing processes are influenced by sediment concentration and
concentration gradients. Therefore water and sediment may be considered here to fully
interact. The spatial and temporal development of such currents require models such
as the k-e turbulence model proposed by Eidsvik and Bmrs (1989) or the model
proposed by Stacey and Bowen (1988a). A formulation of the flow by these models
is still in an early stage of development and does not fit into the choices we made
concerning the modelling. For practical reasons we only considered the steady uniform
flow formulation in combination with Prandtl's mixing length concept and empirical
parameters proposed in literature.

Hinze (1960) .formulated velocity and density profiles for steady uniform turbidity
currents using the mixing length concept. However he neglected the sub-layer below
the velocity maximum (near the bed) and assumed similarities to free turbulent flows
of homogeneous fluids (neglecting stabilizing effects of density gradients). At the same
time Lofquist (1960) presented equations for velocity and density profiles for saline
density currents supplemented with experimental data. More recent literature on this
subject is usually referring to saline or thermal density currents which have, as Lofquist
already emphasised, a different behaviour as turbidity currents.

In our approach we assumed the turbid flow to be almost unidirectional, inviscid,


steady and uniform corresponding to conditions in laboratory flumes reported in
literature. Furthermore we adopted the Boussinesq approximation for density currents
which states that density differences only affect bode forces (see section 2.2). This

27
latter approximation affects the basic equations by raking the density equal to p1
everywhere except when multiplied with g. Sediment is assumed to be uniform.

In section 4.2 a general description is given of the flow structure in a turbidity current,
emphasizing on turbulence and the interaction between fluid and density. Using the
Boussinesq approximation, basic equations are presented and rewritten for a steady
uniform turbidity current. In sections 4.3 and 4.4 the velocity profile and diffusion
coefficients have been derived from these equations applying the mixing length theory
and empirical relations. The similarity of velocity profiles of turbidity currents is used
to define relevant parameters. A distinction between flow in the sub-layer below and
the sub-layer above the velocity maximum has been made. In section 4.5 the velocities
in both layers have been combined to a depth-averaged velocity. It's sensitivity to the
relevant parameters, notably the bed-roughness scale, has been studied here.

From the formulation of velocity profiles and turbulence parameters we derived


relations for sediment concentration distribution in section 4.6 in a similar way as
proposed by van Rijn (1984b ). These relations are integrated over the depth to express
the depth-averaged equilibrium concentration in terms of flow and sediment parameters
, and the concentration at the bed. Additionally, in section 4.7, the derived flow and
sediment distributions are used to compute the coefficients for Galappatti's equation
in terms of regression equations for the governing parameters.

In section 4.8 some reflections on alluvial roughness with respect to the presented
results are given. Finally in section 4.9 a discussion and conclusions are presented.

4.2 General flow structure and governing basic equations

Turbidity currents are generally of a turbulent boundary-layer type, although due to the
presence of very large sediment concentrations they may be laminar. Laminar turbidity
currents such as fluid mud layers will not be considered here.

l
\
\

,\
..... .J
~

Figure 4. l Schematic diagram of two-layer flow

28
After frontal passage of the intruding current the reservoir may be considered as a two-
layer system with a dense lower layer (with depth a2) and a clear upper layer (with
depth a1). The layers are defined by the flow and density profile as schematized in
figure 4.1. The lower layer is defined as the suspended-sediment transport layer and
actually consists of two sub-layers: a dense turbulent sub-layer covered by a turbulent
mixing layer (interfacial sub-layer) with a strong density gradient. The upper layer is
defined as the clear quiescent reservoir water. These definitions are commonly used in
literature to analyze turbidity currents in laboratory. When the depth of the interfacial
mixing layer is small compared to the other layers it is possible to neglect its flow
structure and to assume a two layer model with almost uniform density distributions
(Kranenburg, 1983). However, in practice, especially with swift turbidity currents on
steep slopes, the underflow depth is greatly taken by the interfacial mixing layer.

For a unidirectional-flow situation (1-D flow) the depth a; of the interfacial layer is
found empirically, e.g., by Thorpe (1971), Gartrell (1980) and Chu & Baddour (1984):

ii -ii)1 2
a. "' 0.32 ( 2
l
(4.1)
ed g

where erl-(pifp2) and subscripts 1 and 2 denote averaged quantities of the layers
above and below the interfacial layer respectively. By assuming that this depth at least
must be smaller than the total depth the linear stability criterion for internal waves in
our two-layer model (equation 2.20) is found again.

In figure 4.2 and figure 4.3 measured (dimensionless) values of velocity profiles are
presented from internally supercritical density currents and internally subcritical density
currents respectively. The vertical coordinate ri represents the dimensionless depthz/a2,
and the horizontal coordinate represents the dimensionless velocity function ~" defined
as ~u=u(z)/umax where umax is the maximum flow velocity in the turbidity current. Data
of turbidity currents in a laboratory flume from Garcia (1985, 1990, 1993), Parker et
al. (1987), and Garcia and Parker (1993) are plotted. Additionally data from turbidity
currents in the field presented by Chikita (1989) and Fan (1986, 1991) are plotted.
Furthermore data from saline currents presented by Ellison and Turner (1959) and
Garcia (1990), and measurements of turbulent wall jets in a wind tunnel by Schwarz
and Cosart (1959) and Irwin (1973) are added to show the similarity in shape between
density current and wall jet velocities. The figures show a tendency of similarity
between different profiles measured at different conditions and different locations. The
drawn lines in the figures are the best fits to the data.

Scatter in these figures is partially caused by inaccuracy of the measurement devices


and by differences in instability and stratification of the currents. For example Stacey
and Bowen (1988a) concluded that the measurements of Ellison and Turner (1959) are
at an experimental scale which is being affected by molecular processes.

29
o Ellison & Turner (1959)
0 Garcia et al (1986)
0.8 - Garcia & Parker (1993)
11 x Irwin (1973)
□ Parker et al.(1987)
A Schwarz & Cosart (1961)

0.6

0.4-

.
'it

C
0

o.,i__-.x.-..;~~_,.,.:0=,'"""'~~~~
0 0.4 0.8 1.2
~u

Figure 4.2 Velocity profiles of supercritical density currents.

1.2 o Garcia(1990)
\
◊ \
• Garcia(1990/1993)
, Garcia(1990/1993)

\ , Chlklta(1989)
x Chlklta(1989)

TJ
~" ;"' ° Fan(1986)
a Fan(1986)
A ~•
o Fan(1991)

"<;~;\
" Fan(1991)
0.8 ;~ C

• 0.6

0.4 x 1
fa
,., .../
"/:
0.2

o~==::====--~-~-~-~-~
0 0.4 0.8 1.2 ~
u

Figure 4.3 Velocity profiles of subcritical density currents.

30
The level of maximum velocity is located relatively close to the bed. Now we may
define regions in the current: the lower sub-layer of the turbidity current (below the
velocity maximum) is dominated by boundary-generated turbulence, while above the
velocity maximum (up to the top of the pycnocline) a shear layer exists where
turbulence is suppressed by stratification. A measure for the stabilizing influence of
stratification on turbulence in a shear flow is the gradient Richardson number:

Ri =
(4.2)

Analysis of linear stability show that such a flow is stable if everywhere in the fluid
holds that Rz>0.25 {Miles, 1961). A larger density gradient or a smaller velocity
gradient makes the flow more stable. In real flows this linear threshold may be
somewhat larger than 0.25, e.g. 0.4 (Geyer 1988) or 0.6 (Garcia 1993) due to the
presence of boundary mixing (erosion of a stable layer from boundary-generated
turbulence). Apparently the velocity structure above the velocity maximum is governed
by Ri. Two situations may now occur:
Ri is larger than the stability threshold and the upper-layer regime is laminar.
This situation has been analyzed by Geyer (1988) for an advancing saline wedge
in an estuary. Cross-isopycnal exchange of fluid is small and the velocity
maximum is very sharp.
Ri is smaller than the stability threshold and the upper-layer regime is instable.
In general the interfacial mixing layer is formed in this situation. Unstable
internal waves occur which loose their stability firstly in the region of wave
crests and troughs generating mixing by local turbulence (Turner, 1973).
Eventually this instability and mixing lead to a decrease of velocity gradients
and a subsequent development of a stable interfacial shear flow where
turbulence is again suppressed by stratification (Thorpe, 1971, Gartrell, 1980,
Chu and Baddour, 1984). The instability process is called Kelvin-Helmholtz (K-
H) instability which is associated with the formation and eventually the collapse
of billows.

The first situation logically follows after the latter when a stable interfacial layer is
formed. An intruding density current has an unstable interface at first, but gradually
it can develop into a stable one where turbulence and mixing can eventually disappear.
Turbulence will not receive energy from internal waves any more. However, bottom
generated turbulence can erode the interfacial layer by which its gradient Richardson
number can fall below its threshold again. Once more production of turbulence is
induced causing enhanced mixing until stability is achieved again. This continuously
repeating process causes additional thickening of the interfacial layer (Piat and
Hopfinger, 1981). Also Turner (1986) emphasized that turbulence outside the
interfacial layer is more significant in the mixing processes at the edge of the
interfacial layer, since mixing by K-H billows is relatively small. Note that a in a

31
stable interfacial layer large rotational motions can be observed remaining from the
collapsed billows. More details on these entrainment processes are given in chapter 5.

In its basic form the flow structure of the turbidity current is governed by the 3-D
conservation equations presented in appendix A. Firstly rewrite these basic equations
for a turbulent stratified flow by applying the Boussinesq approximation (density
differences only affect body forces), and neglecting viscous shear stresses:

Momentum in x-direction:

au au 2 auv auw 1 ap I as:a p (4.3)


-+-+-+--+-- = ----+-g
at ax ay az Ptax Pt az Pi X

Momentum in y-direction:

av
-+-+-+--+--
auv av 2 avw 1 ap 1 asyz P
----+-g (4.4)
at ax ay az p ay
1 Pt az Pt y

Momentum in z-direction:

(4.5)

Continuity equation (neglecting sediment mass):


au
-+-+-
av aw = 0 (4.6)
ax ay az

where Sxz, Syz are the local turbulent shear stresses in x,y-direction respectively, p is the
hydrostatic pressure composed of the pressure caused by clear water Pc and a part
caused by excess density p'. If we define an interface level z; = zb + a2 and the water
surface level h,, and if we write
Z;

p(z) = Ps+p 1
= P/8z(hs-z) + P/a'gzfc/z)dz (4.7)
z

then the following boundary conditions can be defined:


At bed level (z=zb): u = v = w = 0 ; s:a = - 'xb

32
At interface (z=z;): p = p s<z) ; sxz "' --r x.i.' syz "' --r.
y,

At water surface (z=h,): p = 0 Syz = 0

In the equations above we may neglect the convective terms and we will assume a
rigid-lid approach in which we express the water-surface gradient as oh/ox= tan(a.,,')
because we used a tilted coordinate system (x-axis has an angle a./ to the horizontal).
Furthermore we assume the underflow to be nearly uniform so that oa/ox = 0 (i.e. the
underflow depth remains constant). The latter assumption is only justified when
pressure variations due to entrainment are small compared to shear stresses. Also
Stacey and Bowen (1988a) assumed pressure gradients to be negligible.

The 3-D model represented by equations (4.3) to (4.6) can be solved numerically for
the time-dependent development of velocity and concentration profiles, especially when
variations iny-direction are negligible (e.g., see Stacey and Bowen, 1988a). However,
in our depth-averaged model we will account for these solutions in a more
approximative way as will be shown here.

Assuming a steady uniform turbidity current and a rigid water surface we can write
equations (4.3) and (4.4) as:

(4.8)

(4.9)

Integration of equation (4.8) over the depth gives:


Z;

Sxz(z,) = -(-rxi+ P1U2W;e) - pfgxaJcs(z)dz (4.10)


zb

And integration of equation (4.9) gives:

33
Z;

Syz(z,) -(-ryi+ p1 v2wie) - p1 gya'fcs(z)dz (4.11)


zb

These equations express the relation between turbulent shear stress and density
distribution.

To derive a simple formulation for flow velocity and sediment transport we will
assume that the x-direction corresponds to the flow-direction where transversal
variations are negligible (in y-direction). This means we eliminate the possibility of
secondary currents. We continue the analysis for equations in x and take 'tx; + Pi½w;,
= 'tx/·
Next we adopt the 'Reynolds stress'-Boussinesq approximation (different to the
Boussinesq approximation mentioned before), and apply the Prandtl mixing length
concept:

-S = p (1-a 1c)e -au acs


p a 1 e u - "'p / 2 ·F(Ri)·-- 1au1 au (4.12)
xz f ft az +
f sx az f m az az
in which F(Ri) is a damping function which expresses the effect of stratification on
turbulence (i.e., on the mixing length), efx is the fluid diffusion coefficient and lm is the
mixing length for momentum. Note that we neglected the terms with c and e,xu(ocjoz)
in this equation, assuming that the concentrations are small. Analysis of velocity and
concentration profiles of Garcia's (1985) experiments showed that the error in S,z
introduced by neglecting this term is of the order of 1% or less.

These relations are used in the following sections to express shear stress, velocity,
Richardson number and diffusion coefficient as function of the depth.

4.3 Flow structure of the lower part of a turbidity current

Usually the velocity structure below the velocity maximum is assumed to be


logarithmic and to be dominated by bed-generated turbulence. This has been observed
in many laboratory and field measurements, e.g. by Ashida and Egashira (1975),
Chikita (1989) and others. Stratification effects in the major part of this sub-layer will
be neglected as is usual in open channel flow. Results of Stacey and Bowen (1988a)
show that only close to the level of velocity maximum the gradient Richardson number
increases largely, inducing a damping effect on turbulence. By assuming an appropriate
shear stress profile this effect is indirectly account for. Prandtl's mixing length concept
for turbulent flow, leading to the law of the wall, can now be applied for the turbid

34
underflow also. The further elaboration of this method corresponds in broad outline
with open channel flow.

We will assume that the shear stress decreases parabolically from the bottom value 'h
to zero at the velocity maximum in agreement with the zero velocity gradient at that
level and the turbulence damping due to an infinite increase of Ri (although the
Reynolds stress may not be zero at this level in reality). The parabolic stress results in
a velocity profile physically superior to that resulting from a linear shear-stress
distribution. Stress and mixing length are:

(4.13)

so that ouloz can be solved from equation 12 (with F(Ri)=l) and after integration the
velocity profile can be written in the following logarithmic form:

au
az (4.14)

+ ✓1 -("'"·)' _ ✓1 l
-("''"·)'

where u. is the shear velocity = ('xh lp1 ) 112


x is the von Karman constant (,::,,0.4).
ri is the dimensionless elevation = (z - zb)la2
1'1m is the elevation at which u(z) is maximal
rio is the elevation at which u(z)=O

For hydraulically smooth boundaries rio ·a2 is usually set as 8/117 with 8= 11.6 ·vlu.
which is the laminar sub-layer thickness. Here v is the kinematic viscosity.
For hydraulic rough boundaries the individual roughness elements have heights greater
than the viscous sub-layer; therefore v becomes irrelevant because the stress is
transmitted by pressure forces in the wakes of the roughness elements. The integration
constant rio ·a2 is then set to k,/32 in which kn is the Nikuradse sand roughness
(kn>6·8).

The level rim at which the velocity maximum has to be determined experimentally. The
similarity collapse of various observed velocity profiles as presented in figures 4.2 and
4.3 indicate that the following values for rim must be chosen:
'llm = 0.15 for internal supercritical turbidity currents

35
11m = 0.40 for internal subcritical turbidity currents

Using equations (4.13) and (4.14) we can write the diffusion coefficient as

efx = t;1:1 = u.1ca211Jl-(11/11m)2 (4.15)

The diffusion coefficient varies from zero at the bed to zero at the velocity maximum.
The maximum value of e1 is located at level ri=V2/211m· A plot of equations (4.14)
and (4.15) is given in figure 4.4.

11

0
u(11)/u* ¾
Figure 4.4 Velocity and diffusion coefficient below velocity maximum.

After choosing the right value for Yim the maximum velocity can be written as:

(4.16)

4.4 Flow structure of the upper part of a turbidity current

The formulation of the velocity distribution in the sub-layer above the velocity
maximum (approximately corresponding to the interfacial mixing layer) is more
complicated due to the necessity of including the damping effect of the density
gradient. This effect expressed by a function F(Ri) requires us to determine the
gradient Richardson number over this sub-layer. To find a general analytical
formulation for the flow in equilibrium state we determined empirical relations for the
velocity and density distribution from laboratory measurements reported in literature.
The resulting stress, Richardson number and diffusion coefficient profiles (using again

36
the mixing length theory) have been compared to the profiles presented by Stacey and
Bowen (1988a).

A fit of the velocity data as presented in figures 4.2 and 4.3 yields the following
relation:
(4.17)

where um = maximum velocity (at TJ91m) described by equation (4.16). A fit of


concentration data from Garcia (1985), and Parker et al (1987) yields:

(4.18)

where cm = maximum concentration at level TJm· These profiles are not applicable to
laminar stratified flows.

Substitution of the empirical profiles into equation (4.2) gives for the gradient
Richardson number profile:

gz -apJaz = 0.1 . Rima. ( al Cm+ 1). exp(6. {11 -11m}3/2)


Ri - -
(4.19)
(11 -11m)(l -11m+ma' +exp(2 · ~ ~;:)]
p(au1az)2

I
where Rim0 = maximum flow overall Ri-number =
gza2 a cm
u,;(1 + a cm)
1

The general form of the damping function F(Ri) may be expressed as


(4.20)

where y' and n1 are calibration constants. In the following we used Stacey and Bowen's
(1988a) calibrated coefficients y'=3.5 or 6.5 and nr4.

The turbulent length scale or mixing length is given by (Stacey and Bowen, 1988a):
(4.21)

(4.22)

By substitution of the relations above into equation (4.12) we can express gradient

37
Richardson number, shear-stress and diffusion coefficient profiles. In figure 4.5 we
plotted these relevant parameters qualitative.

11

F(Ri)

Figure 4.5 Relevant flow parameters plotted for equilibrium flow above the velocity
maximum.

These results greatly correspond to those of Stacey and Bowen's. The plots show us
that the gradient Richardson number increases rapidly near the velocity maximum and
the upper interface, which implies that the turbulence suppressing influence of density
gradients is largest at these boundaries. This explains the shape of the shear stress and
diffusion coefficient profile, where a maximum occurs at the following level:

'llum "' 11m + 0.4·(1 - 11m)


At this level the gradient Richardson number is approximately minimal. According to
section 4.2 {equation 4.2) this implies that there the flow is locally most instable, i.e.
when the mixing layer stabilizes (Ri increases) this is the last level where the flow
remains instable. The local minimal value of Ri can be written approximately (if 11m
= 0.15: ~1=1.137; if 'llm = 0.4: ~1=1.406):

(4.23)

Due to instabilities in the mixing layer entrainment occurs at the interface causing the
depth a2 to increase and the velocity um to decrease (conservation of mass and
momentum). Consequently from equation (4.23) follows that Rima and Rimin increases
until stability is reached and entrainment ceases.

At the interface the turbulent shear stress equals •x/ (the interfacial shear stress and a
pseudo shear stress induced by entrainment of stagnant water from the upper layer).
From figure 4.5 it follows that this value is often very small compared to the bed-shear
stress and may be neglected (with respect to momentum transfer) for stable turbidity
currents. This assumption is in contradiction to the approach of Abraham et al. (1980)
because they assumed the location of the interface in the middle of the mixing layer

38
(e.g., 17:::::!0.7 for a subcritical density current) where shear stresses are of the same order
of magnitude as the bed-shear stress. Their analysis did not include turbidity currents.
It is clear that the definition of the upper boundary of the dense underflow is
representative for the magnitude of this interfacial shear stress.

For the bed-shear stress 'txb follows from equation (4.10):


Z;

-r;xb pfgxo'fcs(z)dz + -c~ = P1gxo'Csa2 + -c~ (4.24)


Zb

or in terms of the shear velocity defined in section 4.3:

u, = J-cxbf Pt= ✓gxo'Csa2 + -r;~fPt"' Jgxo'Csa2


(4.25)

Since the shear velocity expresses the bed-shear stress which is essential for the
computation of velocity and sediment transport this latter equation is very important
for gravity driven turbidity currents on a sloping bed.

In the following section the derived velocity profiles for the flow below and above the
velocity maximum (at 11m) are analyzed, and combined to compute the depth-averaged
flow velocity in terms of roughness length and shear velocity.

4.5 Analysis of the velocity profiles for the full turbidity current

Now we have obtained equations for the velocity profiles of the turbidity current we
can analyze the influence of the different parameters by which these profiles are
expressed. From the profile functions we can derive an equation for the depth-averaged
velocity ½ by means of integration from ri=O to 17= 1. This type of relation is important
for the depth-integrated approach. Depth integration of equations (4.14) and (4.17), for
the velocity below and above its maximum respectively, yields (approximately):

(4.26)

where Iu and fu are functions expressed by


1

Iu = Jexp[-3(11 -11m)3 ]d11


12
"' -0.3511!-0.01411!-0.06911m +0.425 (4.27)
'1m

39
(4.28)

Equation (4.26) gives a relation for the depth-averaged flow velocity u2 as a function
of the roughness length 17 0 and relative depth of maximum velocity.

In figures 4.6 and 4.7 is shown how u2 and um (equation 4.16) vary with the roughness
length 11o for fixed 11m, a2 , and u•. With dashed lines the range of variation in the
plotted parameters is given if a, respectively, positive and negative relative error of 10
% is introduced in the value of l"Jo· Hence for each value of 11o the closed lines indicate
the actual velocities, while the dashed lines indicate the velocities computed with
0.911 0 and l.11'] 0 respectively. Clearly the sensitivity of u2 to the roughness scale is
much larger than the sensitivity of um. In a similar way as in figures 4.6 and 4.7 the
sensitivity of the results to the choice of 11m is illustrated in figure 4.8 where l"Jo is
taken equal to 0.0001.

These figures clarify the errors we make when over- or underestimating the values of
11 0 and l"Jm• In the following table 4.1 these errors are listed, indicating what
approximate error percentage is found if the levels 11o and 11m are over- or
underestimated with 1%.
Tabel 4.1 Sensitivity of the velocity to 11o and 11m·

% error in % error in % error in


um 1<:.lu. Ui'Klu. uju2
1 % error in 0.1 - 0.4 % 0.1- 0.5 % 0 - 0.1 %
11 0 gives
1 % error in 0.1 - 0.15 % 0 - 0.5 % 0 - 0.25 %
11m gives

Clearly all errors are at least smaller than 1%, which implies that the model accuracy
remains within the limitations of accuracy of values l"Jo and 11m· Furthermore we can
see from the table and from figures 4.6 and 4.7 that (for fixed 11m) the influence of 17 0
on the ratio u,ju2 is sometimes negligible (notably for low 11m).

The roughness length-scale 11 0 in the equations quantifies the bed friction. In literature
often a bed-friction coefficient Cv is defined to express this friction, where it is written
for a 1-D flow as
(4.29)

40
30
TJ =0.15
m
[·) 25

i 20

15

10'

5i

um/u2
o.t::-=====================================:
0.000001 0.00001 0.0001 0.001 0.01
~ TJo
Figure 4.6 Relative velocities as function of the roughness height Tio for Tim=0.15.
Dashed lines indicate the 10% error range.

30
TJ m= 0.40

!· I 25 · -

i 20

15j

0.00001 0.0001 0.001 0.01


~ TJo
Figure 4.7 Relative velocities as function of the roughness height Tio for Tim =0.4.
Dashed lines indicate the 10% error range.

In 2-DH flows equation (4.29) is slightly modified for the presence of transversal flow
velocities; see section 4.7.

41
-----:_:_ __ _- - - - - -

0+--------------------
0.1 0.3 0.5
~T]m

Figure 4.8 Relative velocities as function of level Tim for ri 0=0.0001. Dashed lines
indicate the 10% error range.

The relation between CD and TJo from equations (4.26) and (4.29) is plotted in figure
4.9 for the two relevant values of Tim (for sub- and supercritical turbidity currents).

0.02

0.01

0.004

0.002'

0.001--- - - - - - - - - - - - - -
0.00001 0,0001 0.001 0.01
'lo
Figure 4.9 Relation between bed-friction coefficient CD and roughness length scale
TJo for Tim =0.15 and TJm=0.4 respectively.

Again it should by underlined that the equations presented have to be used with care.
They are primarily based on experimental velocity and concentration profiles which
can only approximate the erratic behaviour of a turbidity current. In reality many more
factors slightly affect the actual velocity profile. For instance due to entrainment of
water from above and sediment from below pressure gradients may occur which

42
modify the velocity profile, but are not explicitly included in our approach. The
velocity reduction induced by water entrainment can be incorporated by an entrainment
term in the 2-layer model which acts as interfacial shear, without accounting for a
redistribution of the velocity profile. Note that also the choice of the depth a2 (and
consequently level 11m) is a point of uncertainty for which no agreement exists in
literature.

Since we neglected secondary currents we assume that the presented velocity profiles
also hold for the velocities in y-direction.

4.6 Suspended-sediment transport and equilibrium concentration

The resulting flow profiles and flow parameters can be used to quantify the suspended-
sediment concentration profile in an equilibrium situation. In a steady uniform flow the
concentration profile follows from integration of the convection diffusion equation (see
appendix B, App.B.l, equilibrium profile):

acs (4.30)
W C
ss (1 - Cs)" szaz = 0
+ e -

For particles in the range 50-500 µm we take a:::.5 to account for effects on the fall
velocity, for instance by return flow due to displaced fluid, particle collisions, particle
induced turbulence and modified drag coefficients (Richardson and Zaki, 1954).

For small concentrations 1-c.:::. l (or take a=O, order zero) integration of equation (4.31)
gives the following concentration profile (appendix B, App.B.l):

a -0 : f
cs(TJ) = ca ·exp[- ~dT}
'la es,:
(4.31)

where e.; = e.,)(w. a 2) = normalized sediment diffusion coefficient, w. = the fall


velocity of sediment particles, ca the equilibrium reference concentration, and 'Ila is
dimensionless reference level. For large concentrations, while only taking into account
the vertical return flow due to fluid displacement by the falling particles (a=l), the
following profile is found

0: =1
(4.32)

43
In the following we assume that esz = \Jf' ·efx where \Jf' = ~ -cj> where ~ is a factor
describing the difference in the diffusion of a discrete sediment particle and the
diffusion of a fluid particle, and <I> expresses the damping of the fluid turbulence by the
sediment particles (van Rijn, 1984b). Here the <I> factor reduces von Karman's constant.
Clearly the diffusion coefficients derived in the previous sections by using the mixing
length theory can be used to express sediment concentration as a function of flow
conditions. Especially for the sub-layer below the velocity maximum where the highest
concentrations are found we can define a profile function in a similar way as is
customary in open channel flow. In the upper sub-layer we already assumed the
concentration profile to be expressed by equation (4.18), so that we only have to link
it to the lower sub-layer profile (value cm at 11=rJm).

Analyzing the diffusion coefficient profile we found that locally at the velocity
maximum efx equals zero. Hence an internal boundary against diffusive mixing exists
causing an unrealistic discontinuity in the concentration profile. To overcome this
problem Launder and Spalding (1972) proposed to use a linear bridge connecting the
maxima efwn and eflm in the sub-layer above and below the velocity maximum
respectively (see figure 4.10).

T) [-]

2/31lm . . . . .. ............. \eflm


2
0 7x[m /s]

Figure 4.10 Linear bridge connecting the homogeneous parts of the diffusion
coefficients according to mixing length theory.

For the lower sub-layer we found for the maximum values of the diffusion coefficient
for water the following expression

(4.33)

and for the upper sub-layer we found

44
4 312
efum = 4.50·(TJ 0 +0.5)2-1c2·a2 um[1+6.50Rif J(TJum-'11m)·exp[-3·(TJum-TJm) ]""

ra umJ(num -nm)·
"" 6.7 · 10- 3 2 exp[- 3 · (num -nm)3
12
]

(4.34)

where Riis taken equal to the stability value 0.4 (see later in this section), rim is taken
equal to 0.4 for subcritical and 0.15 for supercritical turbidity currents, and rium is taken
equal to rim+0.4 -(I-rim).

Assume in all sub-layers that esz' = \jJ·eJ(w,a2) where efx is expressed by the equations
given in the previous chapters. Then the profile functions defined by equations (4.31)
and (4.32), with a=O (zero order profile) and a=l (first order profile), can be written
after integration of Iles/ if we define the suspension parameter Z as

ws ws
Z=--=--- (4.35)
\JTKu. P<l>·Ku.
and assume it constant over the depth with cj>= 1 (later in this section we will correct
this latter assumption).

The profiles become for the different sections:

Section ria < ri < rim -(V2)/2:


The diffusion coefficient efx is expressed by equation (4.15) so that integration yields

(4.36)

(4.37)

Section rim t/2)/2 < ri < rim:


The diffusion coefficient is defined as

so that after integration follows

45
-1

l
AT) + B -1/A

,/2/2 •flmA + B}
(4.40)

Here we can approximate the coefficients A and B for practical purposes as follows:
For llm = 0.15:
(4.41a)

(4.42a)

For 11m = 0.4:


A "' (Zt 1 ' [ 6.46 · 10-3 f(Tlo, Tim) - 0.56] "' -0.56 · (Zf 1 (4.41b)

(4.42b)

Function f(llo,TJm) can be written equal to equation (4.27):

Section TJm < 11 < 1:

(4.43)

Analyzing these functions we found that the concentration profile is rather insensitive
to the value of Ri(llum) in the damping function F(Ri) in equation (4.34). We have
chosen a value of 0.4 corresponding to the stability limit. Laboratory measurements
reported by Garcia (1993) show that the value ranges between 0.2 for entraining
supercritical currents to 0.6 for stable subcritical currents. Furthermore small deviations
of the power nr=-4 in this function do not significantly affect the result. Value nr-4

46
gives the most realistic profile function. Calibration parameters in these equations are
Ca,TJo and \jf.

In figure 4.11 is shown how the concentration profile expressed by the equations
presented above, varies with the value of Z for sub- and supercritical turbidity currents.

1.0 \
I

' \
\
----Tim= 0.15
\
\ -T)m=0.40
T) ''
\
0.8 '
i
\
\

''
''
' \

'
\
0.6 \
\

'

0.4

Z'=0.80\
''
0.2 ' \

' \

............ , ...

0
0

-
0

Figure 4.11 Concentration profiles for Z=0.2 and Z=0.8.

For derivation of the profiles for the sections below TJm we assumed the value of Z to
be constant ov.er the depth with the value of~ defined by van Rijn (1984b) as

for 0.1 < (4.44)


u.

However, the <I> damping function appearing in \Jf is dependent on the local
concentration and can be expressed as (van Rijn, 1984b)

0.8 [ ]0.4
[~ ]
<I>= 1 + - 2· ; (4.45)
0

where c0 = 0.65 = maximum volume concentration.

47
Only for very small concentrations the <I> value is approximately constant and equal to
unity. Consequently the concentration profile can only and best be computed by means
of numerical integration of equation (4.30) (using a=S, fifth order profile). The same
problem was encountered by van Rijn (1984b) in open channel flow and he proposed
to a simplified method in which he defined a modified suspension number Z' through
which the first order concentration profile (equation 4.36) can be can be corrected for
additional effects:

Z1 = Z + q> (4.46)

where Zwas defined by equation (4.35) with <j>=l and <pis the overall correction factor
representing all additional effects. Hence Zin equation (4.36) and equations (4.41),
(4.42) is replaced by Z'.

The value of <p is detennined by means of trial and error for the lowest part of the
turbidity current (TJ :s; TJm-V2/2). Multiple fifth-order concentration profiles have been
computed numerically from equation (4.30) (with a=S) and for various sets of
hydraulic conditions (w,, u., ca). Then the <p-values have been determined that yield
zeroth-order concentration profiles (equation 4.36) similar to those numerically
computed. Especially the equality of value c(TJm-V2/2) in the numerical and analytical
approach is used to find appropriate values of <p, since errors in this value significantly
affect the rest of the profile. By means of the method of least squares a regression
equation is found from about 800,000 different <p-values which expresses <p as a
function ofthe hydraulic parameters with inaccuracy of about 5% (0.01 :s; wju. :s; 1,
Tla ::?: 0.01):

~ •[::r·.[Po(::r r P,(::l l + P1 ( :: + + P,
(4.47)

(1) + c<2)
where = c<Ol.,,2
Po k0 a - Cko " 1la k0

(1)
P1 C(O).ri2
kl a - Ckl '11a + c<Zl
kl

(0) 2 (1) + c(2)


P2 Ckz • 11a - Ckz • 11a k2

(0) 2 (1) + c<2l


p3 Ck3 "11a - ck3 " 1la k3

The regression coefficients ck found for Tlm=0.4 and TJm=0.15 respectively are
summarized in table 4.2.

48
Tabel 4.2 Regression coefficients for cp-value.

TJm = 0.15 TJm = 0.4


i 0 I 2 0 I 2
ck<:/ -1994 118 1.38 -1763 111 1.24
ck/ 4263 -278 -2.86 3677 -245 -2.51
CKli -2925 211 1.89 -2445 173 1.61
ck3i -39.4 0.74 0.045 -53.7 2.16 0.052

Van Rijn's much simpler relationship for q> has an inaccuracy of 25% resulting in
similar or larger inaccuracies of the computed concentration. It can easily be shown
(e.g., see van Rijn, 1984b) that for decreasing values of TJa the sensitivity of the
concentration profile to small variations in Z' increases. Van Rijn proposes a minimum
reference level ofria.min = 0.01 at which the concentration profile can be predicted with
an error less than a factor 2 if the error in Z' is less than 20%. We do not want to
violate the basic theoretical model too much, and consider the increase of accuracy
more relevant than the decrease of computation time.

From the equations for the concentration distribution we can derive the depth-averaged
equilibrium concentration C,e by means of depth integration of equations (4.37) for the
lower sub-layer, (4.39) for the intermediate sub-layer, and (4.43) for the upper sub-
layer of the turbid underflow. The resulting equation is rather comprehensive and
unpractical, but by choosing a value of Tim and by using second-order Taylor
approximations for equations (4.36) and (4.39) before integrating, it is possible to
derive a much more manageable equation with good accuracy. The following
(approximative) values for the concentration C,e can be obtained:
For Um= 0.15:

Cse = ca•[ 0·~~ 1 (3- ✓9-40011~)r{o.1SYa[-¾(z)2 Y; + (Z)Ya - 1) + (4.49)

+ exp(-0.67(Z)) · [3.74 · 10-3 (Z)2 + 15.0 • 10-3 (Z) + 0.4114] }

where Ya = (11af 0.15) - (/2/2)


K.QLD.,,,=0.4:

Cse = ca•[ 0 ·~~ 3 (z- ✓4-2511~)r{o.4ra[-¾(z)2 Y; + (Z)Ya - 1) + (4.50)


3 2 3
+ exp(-0.749(Z)) · [16.88 · 10- (Z) + 45.0 · 10- (Z) + 0.3766] }

49
where Ya = (11a I 0.4) - (fl I 2)

An example of a comparison of computed velocity and concentration profiles to those


measured by Garcia (1985, 1990) in a flume is shown in figure 4.12. The only
parameters that are calibrated for this plot are the values of Tio (or Cv) and ca. These
profiles have not been used for the regression of the upper layer profile.

The formulas we have now defined relate the suspended-sediment distribution and the
depth-averaged suspended-sediment concentration to the flow conditions governed by
depth-averaged flow velocity and the boundary shear stresses. However, these formulas
require a value for the reference suspended-sediment concentration ca at level 11 0 , which
has not been described yet. Clearly we have to use an equation which expresses ca as
function of hydraulic parameters also.

In previous studies various existing equations for ca have been used, which were almost
all derived for open-channel flow. The reason therefore is the lack of a specific
equation for turbidity currents and the lack of data to derive one. Furthermore the flow
conditions below the velocity maximum are comparable to those in open channel flow.
Although Parker et al. (1987) tried to fit their data to an empirical relation for open-
channel flow, the inaccuracies in the data and the method of computing the entrainment
rate of sediment resulted in a large scatter. Experiments of a saline current passing over
an erodible bed by Garcia and Parker (1993) yielded more useful results which are
more realistic than those found for open-channel equations. Still large scatter in the
data is due to the method of computing the sediment entrainment (e.g, by
differentiating the concentration and velocities along a section). Garcia and Parker's
relation are summarized in appendix D.

Garcia (1990) and Garcia and Parker (1991) compared the performance of some
equations for computing ca in open-channel flow. One of the equations that performed
best on open-channel data is the one proposed by van Rijn (1984b ). In line with that
conclusion and in line with the theories used for the concentration profile below the
velocity maximum it is preferable to use this equation. Van Rijn (1984b) used its bed-
load concentration function (1984a), with a small modification, to compute the
reference concentration. This function is based on a particle parameter D. and a
transport stage parameter T which expresses the mobility of the particles in terms of
the stage of movement relative to the critical stage for initiation of motion. The
equations involved are given in appendix D for particles in the range 100 µm - 500
µm.

For practical purposes or numerical experiments also a more simple power law can
be used based on the critical shear-stress approach. For instance by considering that
Van Rijn's transport stage parameter is proportional to the square of the underflow
velocity, the reference concentration can be written as some cubic function of this

50
0 0.4 0.8 cfca 1.2 1.6 0 0.4 o.8 cfca 1.2 1.6

I I
\\'
I \\

0.81 0.8

I
TJ TJ

0.4

0 12 16 o 4 8 I 12
0 0.4 0.8 C Ca
I 1.2 1.6
u u*

~ \
'
0.8' '

TJ
\
\\
0.4 \I
\\ /
/
\
8 / 12
u u*

Figure 4.12 Computed velocities and concentrations compared to Garcia's (1985,


1990) experiments of supercritical flows Run 9 (left figure), Run 17
(right figure), and of subcritical flow Depo3 (lower figure).

51
velocity:
(4.51)

or in line with van Rijn ' s approach (cf. appendix D):

(4.52)

where m is a parameter related to sediment characteristics and the reference height Tla
and u.cr is the critical shear velocity for initiation of movement.

4.7 Galappatti's coefficients for the semi-empirical approach

Galappatti's depth-integrated model for suspended-sediment transport is reformulated


for the 2-DH 2-layer approach in chapter 3. The dimensionless adaptation length (La')
and time (Ta') in this equation (3.14) can be obtained from the flow and concentration
profiles for uniform (equilibrium) flow conditions derived in the previous sections.
Starting from the zeroth-order shape factors iio(ri) for the equilibrium concentration and
\JIJr1) for the velocity distribution defined as

\jJ (fl) = U(T)) (4.53)


u u
2

the dimensionless adaptation length and time are written as (section 3.2):

Lai = Yz _ <Xz (4.54)


Yo <Xo

Y1 (XI
Ta 1 = (concentration bed-boundary) (4.55)
Yo ao

Y1+l (Xl
Ta 1 = (gradient bed-boundary) (4.56)
Yo <Xo

where

52
1

a0 = f Wu(TJ)tzo('rJ) dTj
'la

"1 = f Wu(11)a1 (11) d11


1

"2 = f Wu(11)a2(TJ) d11


'la

The inverse operator D· 1 is derived in appendix B (App.B.l). If interfacial mixing is


neglected it can be written as
I I

D- 1[g] = /(11) = ao(ri) f,, ao(TJ)


!(TJ) d11 - Jg(11)d11
,,
+ Bao(TJ) (4.60)

where integration constant B follows from


I

f /(11) d11 = 0 (4.61)


'la

The derived concentration and velocity profiles in this chapter are described by
connecting mathematical fonnulations for the three separate sub-layers of which the
turbid underflow is composed. Analogously the shape factors are split into three parts
in the following way:

For Tia :c;; TJ < Tim v2/2: '-Vu01(TJ) and ii'o1(TJ)


For Tim v2/2 :c;; TJ < TJm: '-Vu0iTJ) and ii'o/TJ)
For TJm:S;TJ < 1: '-Vu0lTJ) and ilol11)

Also the shape factors a'1 and a2 deduced from these zero-order profiles can be
computed for the three separate layers in successive order. From the definition of the
inverse operator given by equation (4.60) it can be shown that
I

ti 1(TJ) = a'0 (11)(l -11) - Ja0 (11)d11 + B1a 0 (TJ) (4.62)


,,

53
I I I

a 2 (11) = f
a 0 (11) Wu(11)d11 - f Wu(11)a (11)d11
0 + B2 a0 (11) Ja (,i)
2
= o (4.63)
'l 11.

In agreement with the previously defined shape factors for velocity and concentration
the Galappatti coefficients depend on the parameters Tlo, Tlm, Tla and w/u•. The
integrations which have to be carried out, for instance for application of the inverse
operator, are complex and have to be carried out numerically. Therefore this
dependence gives a convenient opportunity to carry out the computations in advance
and to express the Galappatti coefficients in terms of regression equations. In this way
for all the coefficients Yo to y2, and a 0 to y2, and for La' and T polynomial regression
0
'

equations can be computed for a wide range of values for the dependent parameters.

The numerical regression analysis is carried out for subcritical and supercritical
turbidity currents, both with a gradient and a concentration-type bed-boundary
condition. Integrals are computed with Simpson's rule using about 700 steps for each
sub-layer. Parameter Tlo is varied from about 5·10·6 to 0.01, Tla from 0.01 to 0.05 and
w/u. from 0.01 to 1. For using Galappati's equation we are only interested in the
adaptation scales La' and Ta' which will be expressed by the following polynomial
regression equations:
(4.64)
+ p51

(4.65)
+ p52

where

plj (pl]) 11! + (plJ) 11 + (plJ}


= 0
(4.66)
p2j = (p2J) 11~ + (p2J} 11a + (p2J)

and

(4.67)

The coefficients for these regression equations are summarized in the tables in
appendix C (App.C.1) and can be easily implemented in a numerical model for
modelling turbid underflows.

54
To illustrate the magnitude and sensitivity of the Galappatti coefficients some of the
computed values are plotted, which have been used for the regression analysis. In the
following figures the variations are plotted of La' and Ta' with w ju. for two values of
lla and two values of 17 0 (drawn lines: 17 0 = 0.0067, dashed lines: TJo = 0.000116).
Figures 4.13 and 4.14 represent the adaptation length-scale for super- and subcritical
turbidity currents respectively. Clearly La'-values for llm = 0.4 are lower than those for
llm = 0.15. Also a smaller roughness scale yields lower adaptation lengths. These
results correspond qualitatively to those derived by Galappatti (1983) for open-channel
flow. The adaptation time-scales for the concentration-type bed boundary approach are
plotted in figures 4.15 and 4.16 for super- and subcritical turbidity currents
respectively. The dependence of these time-scales on the relevant parameters is almost
identical to the dependence of the length scales in figures 4.13 and 4.14. However, in
figures 4.17 and 4.18 is shown that Ta' values for the gradient-type bed-boundary
approach at low w/u. are significantly larger than those for the concentration-type bed
boundary. This effect becomes pronounced if w/u. ➔ 0.

Sloff (1992) showed that the dimensionless characteristic celerity for 1-D Galappatti's
equation is

L'a
(4.68)
T'a

Hence the propagation speed of disturbances in the depth-averaged suspended-sediment


concentration is governed by the ratio La'/Ta'· In figures 4.19 and 4.20 is shown how
this ratio (for llm = 0.15) depends on the choice of the bed-boundary condition. For
convective transport only the dimensionless celerity must be equal to unity. However,
the figures illustrate that for larger values of wju. the dimensionless celerities are
slightly larger than unity, while at lower values (notably for the gradient-type bed
boundary) the celerities dip below unity. For subcritical turbidity currents (TJm = 0.4)
similar plots can be found, although the ratio La'/Ta' is somewhat closer to unity.

In addition to the figures presented here, similar figures which express the behaviour
of y/y0 , y/y0, a/a0, and a/a0, are presented in appendix C. Again the similarities
between these results and those presented by Galappatti (1983) for open channel flow
are notable.

From the results in this section can be concluded that behaviour of Galappatti's model
for turbidity currents is similar to that for open-channel flow. The adaptation length-
scale La' and adaptation time-scale Ta' depend on the parameters 17 0, TJm, lla, w/u., and
the choice of a concentration or gradient-type bed-boundary condition. In agreement
with Wang and Ribberink (1986) the model can only be used if suspended-sediment
is the main mode of transport, i.e. w/u. ~ 1, and flow variations are gradual, i.e.
time-scale of flow ► a/u. and length scale of flow ► q/u•.

55
L'a
0.5

o - 1 - - - - - - - - - - - - - - - - - - - - - < w8 /u*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.13 Adaptation lengths for TJm=0.15 (drawn lines: TJo = 0.0067, dashed lines:
TJo = 0.000116)

0.5
L'a Tla=0.01 111110.40

0.4

0.3
----
- -
'
'
0.2 -----------------
----- ------

0.1

0 wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.14 Adaptation lengths for TJm=0.40 (drawn lines: TJo = 0.0067, dashed lines:
TJo = 0.000116)

56
0.5
T'
a

0.4

0.3

0.2

0.1

0 + - - - - - - - - - - - - - - - - - - - - - - - i wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.15 Adaptation times for llm=0.15 and concentration-type bed boundary
(drawn: TJo = 0.0067, dashed: llo = 0.000116)

0.5
T'
a

0.4
'lla=0.01

0.3

0.2

0.1

o.,.....__________________----< wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.16 Adaptation times for llm=0.40 and concentration-type bed boundary
(drawn: TJo = 0.0067, dashed: TJo = 0.000116)

57
0.7
T'
a
0.6

0.5
'lla=0.05
0.4

11a=O.Ol
0.3

0.2

0.1

0 + - - - - - - - - - - - - - - - - - - - - - 1 wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.17 Adaptation times for rim=0.15 and gradient-type bed boundary (drawn:
llo = 0.0067, dashed: llo = 0.000116)

0.7
T'
a
0.6

0.5

0.4

0.3
'lla=0.05
0.2
- - -- - ___________ 11a=O.O_l__________ _

0.1

0 + - - - - - - - - - - - - - - - - - - - - - 1 wsfl4
0 ~ M M M 1n
Figure 4.18 Adaptation times for rim=0.40 and gradient-type bed boundary (drawn:
llo = 0.0067, dashed: llo = 0.000116)

58
1.6,--------------------,
L'a
1"
a
- -- - - - - - - - - - ---- -- - --- -- - -- -- - - - - - - - - - - - - -

0.4

o + - - - - - - - - - - - - - - - - - - - - - - 1 wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.19 Dimensionless characteristic celerity for llm=0.15 and concentration-type
bed boundary.

L'a
Ta' 1.2

0.8

0.4

11m=0.15
o - + - - - - - - - - - - - - - - - - - - - - - 1 wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure 4.20 Dimensionless characteristic celerity for llm=0.15 and gradient-type bed
boundary.

59
4.8 Alluvial roughness

The roughness length-scale 110 is found to be one of the primary calibration parameters
of the presented model, representing the grain and bed-form induced bed-shear stress.
It is essential for the model to define an appropriate roughness scale or a relation
between roughness scale and hydraulic conditions.

The bed-shear stresses in our model can be expressed as a function of the a bed-
friction coefficient Cn or a Chezy value. In section 4.5 the Cn-value is expressed in
terms of the shear velocity and the average flow velocity (equation 4.29), and is shown
to be a function of the roughness length scale TJo = z0 ·a2• In a 2-dimensional model we
assume similarity in flow profiles in x and y-direction and we note that for the bed-
shear stresses holds that 'th =V(-rbx2+-rh/). Then we find:
Bed-shear stress in x-direction (with u. = shear velocity in x-direction)

(4.69)

Bed-shear stress in y-direction (with v. = shear velocity in y-direction)

'tb
-1'....
.~
= V • = CD V2 V~
2 2 2 1
U2 -r V2 = g y a Cs a2
(
+
't;
-2'..
~
I l z gy a 1Cs a2 (4.70)

Corresponding we can define a Chezy value for this approach which expresses the
relation between shear stress coefficients and roughness length:

(4.71)

where Iu is an integral function expressed by equation (4.27), and f(TJo,TJm) is a function


related to the maximum velocity expressed by equation (4.28).

Usually a measure for the roughness scale used in literature is the effective or
equivalent roughness height kn, i.e. the Nikuradse sand roughness. Furthermore, in open
channel flow it is common to define the thickness of the viscous sub-layer at the bed
as
o = Il.6·vfu. (4.72)

Note that the agreement of logarithmic flow-velocity profile of open-channel flow and

60
density current near the bed allows us to use the same equations.

We may now define the roughness length in our model as follows (see also section
4.3):
For hydraulic smooth boundaries (u.kjv<5) use TJo ·a2 = o/117;
For hydraulic rough boundaries (u.kjv>70) use T] 0 ·a2 = kj32.

In most experimental studies to turbidity currents on a mobile bed, bed forms are
reported leading to hydraulic rough boundaries. To include the bed-form roughness into
the model a reliable roughness prediction method has to be defined. Again the
literature does not present us many tools to account for these bed forms in relation
with turbidity currents. Garcia and Parker (1993) used a method developed by Nelson
and Smith (1989) for the removal of form drag due to bed forms for open channel
flow. But this method can only be used if bed-form dimensions are known.

The lack of bed-form roughness predictors for density currents forces us to either
calibrate the friction coefficient for each situation or to choose from predictors for
open-channel flow. However, these prediction methods are highly empirical and usually
have a strong dependence on the flow depth. For a turbidity current we should replace
this depth either by the underflow depth a2 or the level of the velocity maximum, and
then we should calibrate the proposed empirical relations again. In that we can
reformulate for instance Engelund's (1966) or van Rijn's (1984b) predictors for
turbidity currents. Essential for these calibrations is the availability of a large number
of field data.

Until now detailed data for calibrating a roughness predictor are not available and we
have to calibrate the roughness for each simulation, in which the bed-form roughness
is incorporated in the kn-value for the respective turbidity current. Provided that
primary part of the turbidity current has a rather constant densimetric Froude number
we may expect the bed-form regime to be similar along the current (assuming the
Froude number to be the flow criterion for dividing different regimes).

4.10 Summary and conclusions

The flow velocity profile and concentration profile in a steady uniform turbidity
current are analyzed and described by a semi-empirical model. By using this approach
relations are obtained to describe flow velocities and concentrations as functions of
flow and sediment parameters. Thereupon these relations are used to quantify the
adaptation length and adaptation time for the Galappatti model for the depth-averaged
suspended-sediment concentration.

61
From laboratory experiments reported in literature can be concluded that it is possible
to define a standard velocity and concentration profile for equilibrium turbidity
currents. Below the velocity maximum (which is located at about 15% of the
underflow depth for internally supercritical turbidity currents and at 40 % for internal
subcritical currents) the turbidity current has an approximately logarithmic profile.
Therefore the elaboration of the mixing length theory for open channel flow is adopted
for this layer. In the layer above the velocity maximum a regression equation is
defined for the velocity profile based on laboratory experiments. From both profiles
the distribution of the fluid-diffusion coefficient has been derived from the mixing
length theory accounting for turbulence damping effects by density gradients. The
profiles derived in this way are assumed to be representative for all directions (x,y)
provided that secondary currents are negligible.

The derived fluid-diffusion coefficients express the turbulence structure of the turbidity
current and consequently can be used to describe the concentration profile of suspended
sediment. However, locally at the velocity maximum this coefficient becomes zero
causing zero diffusive mixing across this level. To prevent zero diffusion the maximum
values of the diffusion coefficients in the layers above and below the velocity
maximum are connected by means of a linear bridge.

Firstly in the layer below the velocity maximum and below the diffusion maximum
these diffusion coefficients are used in a similar way as proposed by Van Rijn (1984b)
for open channel flow to express the concentration distribution. Therefore the sediment
mass-balance equation (i.e., convection-diffusion equation for steady uniform flow) is
integrated, using a reference concentration ca at a reference level TJa near the bed. The
profile function found for low concentrations is adopted for flows with larger
concentrations by modifying the exponent Z' in this function.

Secondly, in the layer above the velocity maximum, a regression equation for the
concentration profile is derived from laboratory measurements. This profile is
connected to the profiles for the lower layer, and by means of depth-integration an
equation is defined for the depth-averaged equilibrium concentration as a function of
flow and sediment properties, and of the reference concentration.

The concentrations, which can be computed with the derived equations, are highly
dependent on the reference concentration near the bed. A transport formula has to be
adopted which relates this concentration to the flow conditions. For practical
applications this relation can be sought in terms of the shear velocity and sediment
properties. Furthermore the computation of roughness, which is essential for the
velocity and concentration computation, is complicated due to the presence of bed
forms. Existing roughness predictors for open channel flow cannot be used and
consequently the roughness length scale has to be chosen and calibrated for
computations with the model.

62
Comparison of the semi-empirical model and measured velocity and concentration
profiles show a good agreement within the accuracy of measurements. However, one
must bear in mind that this model is largely based on laboratory experiments especially
for the upper part of the turbidity current, and that its performance is primarily
determined by the accuracy of the predictions for roughness and reference
concentration. Furthermore the lack of agreement between different authors in
definition of the depth of the turbidity current makes it rather laborious to verify the
results (verification is only possible by analyzing the full data sets and not by using
averaged parameters which are defined by the authors).

63
64
Chapter 5

Interfacial mixing

5.1 Introduction

A characteristic property of a two-layer flow of miscible fluids is the exchange of fluid


at the interface due to turbulence and shear instabilities. The understanding of the
natural appearance of this interfacial mixing process, and its quantification is an
element of particular importance for modelling stratified flows. In this chapter both
these aspects are discussed on the basis of an extensive literature survey.

The basic mechanisms of mixing are complex and still only a poor empirical
quantification is possible. The present knowledge on basic mixing mechanisms has
mainly been provided from laboratory experiments. Usually these experiments are very
specific and simplified, and their results cannot be applied for general field situations
which differ much from laboratory circumstances. For instance it is shown that vortex-
generated mixing behind the front may be much greater in a gravity current which
enters a diverging channel or a widening reservoir, than the mixing in a prismatic
laboratory flume. It is also possible that different basic entrainment mechanisms can
occur simultaneously in the field which interact. It is therefore not surprising that the
published empirical relations often show large deviations from the mixing rates in the
field.

Mixing between the layers and the governing formation of the m1xmg layer is
determined by shear and turbulence generated interfacial perturbations at the head of
the current (e.g. Kelvin-Helmholtz instabilities). In the following we consider, just like
Bo Pedersen (1980), all mixing as entrainment irrespective whether it is a one-way
(pure entrainment), an equal two-way process (pure diffusion), or somewhere in
between. Hence across any density interface the following processes can occur:
Pure entrainment occurs if one of the layers is laminar: Mixing only occurs
from the non-turbulent into the turbulent layer.
Net entrainment occurs if both layers are turbulent: Mixing occurs in both
directions, so that the density interface can either rise or sink.
Diffusion mixing occurs if net entrainment is zero and the interface is
stationary.
In all these cases an interfacial layer is present with a density gradient which
suppresses turbulent mixing. In the proposed two-layer model the net entrainment at
the upper edge of the interfacial mixing layer is expressed in terms of an entrainment
velocity w;,•

65
Furthermore we distinguish between two basic mechanisms: cusp and vortex-
entrainment (Bo Pedersen, 1980).
Cusp-generated mixing occurs at high overall Richardson numbers Ri0 (e.g., R;o
= edfszCiiu/ > 1 : moderate gravity currents) due to the action of turbulent eddies
striking the interface. Then undulations in the form of Holmboe waves form sporadic
cusps which may be detached by eddies present within the turbulent layer.
More intens is vortex entrainment which occurs at lower overall Richardson
numbers (e.g., 10-2 < RiO < 1 : internal supercritical currents). Vortex entrainment is
characterized by the growth and coalescence of vortices in the interfacial layer
occurring in the critical or supercritical state of the interfacial layer, i.e. interfacial
wave instability and breaking. In flows with very high Richardson numbers molecular
diffusive effects become important, while in flows with very low Richardson numbers
the entrainment rate equals that of turbulent mixing in a homogeneous fluid. Due to
the formation of a stable interfacial mixing layer it is often justified to neglect the net
entrainment in internally subcritical flows (high Ri0 ).

Since perturbations on the interface originate at the front, the front is the controlling
feature of mixing process and of the flow itself. However, none of the proposed
empirical relations for interfacial mixing do account for that (Turner, 1986). Locally
energy losses and entrainment are associated with the flow in the head, i.e., increased
turbulence and shear generated instabilities.

(1)

Figure 5.1 Instabilities of a gravity current head: (1) billows; (2) lobes and clefts.

The two major forms of instability at the front of a fully turbulent gravity current,
which cause energy loss and entrainment, are:
Billows associated to Kelvin-Helmholtz instabilities forming at the front (see
Figure 5.1, part (1)). This type of billows is characteristic for instabilities at a
density interface for fluids moving relative to each other. The formation and
collapsing of the billows are the main processes in which ambient fluid is mixed
into the gravity current: interfacial mixing. The height of the head wave and the
following mixing layer is identical to the breakdown size of the K-H billows
(Simpson and Britter, 1979, Simpson, 1987).

66
Gravitational instabilities in the head caused by rising less dense fluid which is
overrun by the head. This type of instability only occurs due to the effect of
bottom friction. Due to the no-slip condition at the bottom the foremost point
of the head, the nose, is raised, and small amount of less dense water passes
underneath the nose. The flux of this overrun lighter fluid is order 0.01 of the
flux of lighter fluid involved in mixing at the top, but it is responsible for the
non-steady lobe and cleft structure illustrated in Figure 5.1, part (2). The
Kelvin-Helmholtz billows on the head are then broken up in a complicated 3-D
form, indicating that gravitational instabilities significantly affect the
entrainment at the interface (Simpson and Britter, 1979, Simpson, 1987).

Beside for the gravitational instabilities, bottom friction is also important for shear
related turbulence generation and associated energy losses in the gravity current
(Abraham and Vreugdenhil, 1971)

Interfacial mixing due to K-H instabilities increases when the bottom slope increases.
The front speed however remains constant for increasing slopes because the larger
gravitational force is counterbalanced by the increased entrainment in the head and the
flow behind it. Interfacial mixing cause momentum to be imparted to less dense
reservoir fluid entrained into the head (Britter and Linden, 1980). Furthermore it causes
(together with dense inflow from behind) a growth of the head volume and head
height. Although for large slopes u1 remains constant, it has been observed that the
front behaviour is extremely sensitive to slope as it approaches the horizontal.

Background turbulence in the ambient fluid, in the continuous layer, significantly


influences the entrainment mechanism (Noh and Fernando, 1992). At higher turbulence
rates of the upper layer the thickness of the current increases while its velocity
decreases. For turbidity currents diffusion of sediment particles in the direction normal
to the slope is enhanced. In extreme situations the front is diffused over the entire
depth. For small ambient turbulence the effect is negligible, but else its effect increases
as eventually, during propagation, the density of the current becomes smaller and the
front velocity is reduced (Simpson, 1987).

The above described 2-D mixing processes can be disturbed by the presence of lateral
spreading if the current is entering a reservoir. For instance, if a turbidity current is not
running through a bottom channel, e.g., the former river channel in a storage reservoir,
then it will spread over the bottom depending on reservoir geometry, friction, bottom
slope, initial buoyancy flux, and interfacial mixing (Alavian, 1986).
A characteristic difference between two-dimensional and axi-symmetric flows is the
increased intensity of rotational motion, and the rate of mixing, of the internal fluid
during the early stages (Simpson, 1987). The majority of the gravity current fluid
becomes concentrated at the front or in multiple fronts, leaving only a thin layer of
heavy fluid near the ground. The 2-D flows are more uniform in depth shortly after
release. The formation of this leading edge vortex, which occupies almost the full

67
depth of the dense fluid, is associated to the formation of Kelvin-Helmholtz vortices
at the head: see Figure 5.2.

c-r:c ~~~~ \
._,I.,_._,,.,, ,___..____,,,_,,-=-
~ \

m~a:i&tiiitW-ffffftr1ftWfff€f#&~
Figure 5.2 Comparison of gravity currents in a sector tank and in a prismatic
channel.

Simpson (1987) explains the increase of vortex intensity as follows: 'Assuming that
the fluid volume in a vortex is approximately conserved, its cross-sectional area must
decrease as it stretches. Conservation of angular momentum about the centre line of
the vortex then implies that its intensity increases'. Clearly this effect of vortex
stretching is largest during rapid expansions, notably near the source. This vortex
formation is also reported by McClimans (1978) for a fresh surface current entering
a saline fjord.

The presence of a discontinuity other than the front, such as an internal hydraulic
jump, can also be a source of severe entrainment (Dracos, 1986). Among others Wood
and Simpson (1984) have presented experiments and theory of entraining jumps. They
developed a two-layer theory with energy conservation in the upper layer to
approximate the value of fluid entrained into the jump. Also at the plunge point, where
the buoyant flow plunges below the ambient water, entrainment can be expected.
Especially the laterally-unbounded free-jet like intrusions are subjected to intense
mixing (Johnson et al., 1989).

Entrainment mechanisms and quantification of the dilution rate is a topic of much


research. Various empirical relations are developed for various different types of
stratified fluids: surface flows, bottom currents, thermally stratified lakes, etc. In the
laboratory we may distinguish between oscillating grid experiments and shear flow
experiments. The oscillating grid experiments are based on introduction of external
turbulent energy into a quiescent fluid (e.g. comparable to wind shear or surface-wave
breaking). In general the interface continues to sink until the basin is fully mixed (and
fully turbulent). In pressure and gravity driven shear flows the mixing energy must be

68
taken from the flow itself. Therefore this latter type of flows, especially if internally
subcritical, often obtains an equilibrium situation with stationary interface (zero net
entrainment) and a counter-directional compensation current (Eidness, 1986).

Empirical entrainment equations usually relate the entrainment rate to the overall flow
conditions which can be expressed by some form of the Richardson number. Its
original form is the gradient Richardson number which is defined as equation 4.1 in
Chapter 4 and can be interpreted as a measure for the shear stability of the interface.
The entrainment relations are based on the following entrainment hypothesis (Turner,
1986): The mean entrainment velocity across the edge of a turbulent flow is assumed
to be proportional to a characteristic velocity, usually the local time-average maximum
mean velocity, or the mean velocity over the cross section at the level of inflow. The
total inflow at any position will depend also on the surface area and the geometry and
dynamics of the flow, whether it is axi-symmetric or 2-D, a continuous jet or plume
or a suddenly released thermal.

The primary differences between the various empirical entrainment relations are the
parametrization and the region of validity. The most basic parametrizations are based
on a form of the Richardson number (Ri) and a velocity difference (~u) or shear
velocity (u.). In the following section is shown that this parametrization is disputable
since it allows spurious correlations. Although many researchers have tried to increase
the accuracy and generality of the relations by introducing more or other parameters,
it will make little difference for prototype calculations to use their relations or the
more basic ones. From the treatment of some of these empirical relations in the
following section we can conclude that the computation of water entrainment is as
problematic as the computation of sediment entrainment from the bed in morphological
models.

The presented relations can only give an approximative quantification of the dilution
rate in our model and must therefore be verified with field data. The present
knowledge of the basic mechanisms and the lack of detailed observations do not allow
for a more sophisticated representation of entrainment into turbidity currents in the
field. In the following section some empirical relations reported in literature are
presented and discussed.

5.2 Discussion of some empirical entrainment relations

Various entrainment relations are available for computing the expected fluid
entrainment at the interface of a gravity current. Those relations derived from
oscillating grid experiments are not considered here since we are primarily concerned
in shear flows. The entrainment relations usually apply for the steady part of a density

69
current. After plunging the current in a flume adjusts itself quickly to this equilibrium
state where entrainment is governed by the overall flow conditions. Then a certain
form of the Richardson number, which expresses these flow conditions, is the most
important parameter in all the available relations. In this section a review of results are
presented which can be used for a two-layer model.

Although the theory does not primarily concerns shear flows we should mention the
following results of Turner (1973,1986) from his previous studies. He proposed (as
being fundamental) that for high Peclet numbers (Pe>200), defined as Pe = u//Km
where Km= molecular diffusivity, the entrainment rate is proportional to a Richardson
number to the power -3/2:

with (5.1)

Where u1 = r.m.s. value of the horizontal component of turbulent velocity near the
interface, and /1 = length scale of turbulence, which are the relevant velocity and length
scales in this law. Turner noted that this law holds for Ri;>7, and found that for lower
Pe the entrainment is affected by molecular diffusion. For many other experiments,
also with other length and velocity scales, this 3/2 power law still holds.

More relevant for us is that Turner (1986) mentioned the following entrainment
relation for a gravity current for the often cited Ellison and Turner's (1959) results (for
stagnant ambient flow and low Richardson numbers, or high internal Froude numbers):

0.08 - 0.1 · Ri0


with (5.2)
1 + 5·Ri0

This relation follows from the assumption that the flow adjust rapidly to a normal state
with a constant Richardson number. Here the lower layer average flow velocity and
the overall Richardson number Ri0=(1/Fr2)2 are used for parametrization. Ellison and
Turner found that for Ri0 larger than a critical Richardson Ric ~ 0.8, i.e. for internally
subcritical gravity currents, the entrainment can be neglected.

Christodoulou (1986a,b) obtained the following entrainment equations after


examination of the data from various experiments for shear flows from various authors:

= 0.07 for Ri0 e.. < 0.01 (5.3)


au
Corresponding roughly to conditions for turbulent entrainment in homogeneous
fluid.

70
for (5.4)

Corresponding to conditions where vortex entrainment occurs.

for 0.1 < RiOll. < 10 (5.5)

Corresponding to a transition region between cusp and vortex entrainment.

for 1 < Ri0 ll. < 100 (5.6)

Corresponding to conditions where cusp entrainment becomes important.


Here the Richardson number is based on the depth a 2 of the density current and
velocity jump tiu = u2 - u1 over the mixing layer:

(5.7)

Deardorff and Willis (1982) performed experiments with a rotating screen in an


annular tank. They used the velocity jump tiu as well as the shear velocity u. for
parametrization to determine their empirical entrainment relation. Their results are
valid over the range I < Ri01i < 10.

Narimousa et al. (1986) derived the mixing rate from experiments of a surface flow.
Although their results are scaled by a shear velocity and a corresponding shear
Richardson number, they are fairly similar to those of Christodoulou (1986a). Eidness
( 1986) derived empirical relations for stationary pressure-driven density currents based
on the overall Richardson number. His entrainment rate is based on an estimation of
interfacial friction by Bo Pedersen (1980) for large Reynolds numbers and a stagnant
ambient fluid:

(5.8)

where K; = 4.2·10- 3Ri0 for Ri0 < IO


K; = 4.2·10 4 for Ri0 > 10

Atkinson and Wolcott (1990) reported on theory and experiments of a combination of


mean shear and an oscillating grid (corresponding to a mean current with external
turbulence generation). Unfortunately this generalization of existing entrainment
relations however could not give satisfactory results yet for large ranges of the

71
Richardson number and requires more measurements for evaluating the empirical
coefficients.

Grubert (1990) gives an extensive review of literature on entrainment relations and


derived equations for fjord and salt-wedge flows in which the lower layer is quiescent.
He related interfacial mixing to interfacial shear to determine empirical relations.

Bo Pedersen (1980) developed a theory for entrainment which is based on the


hypothesis that a constant part of the available energy is used for the processes which
require energy. His basic hypothesis states that a bulk Richardson number Ill/'
is
constant for a given flow situation. This Ill/'
is defined as the ratio of gain in potential
and turbulent kinetic energy due to entrained mass to the energy available for
turbulence, or is defined as the efficiency of recovery (due to entrainment) of the
energy produced. For subcritical currents Ill/'= 0.045 so that the entrainment function
can be written as (for a dense underflow):

(5.9)

where 10 = bottom slope (which must be gentle).


For highly supercritical density currents lfl/ and the entrainment function become more
complex.

According to Pallesen (1983) the entrainment for turbidity currents must be smaller
than for currents created by a salinity or temperature difference, and can even be zero.
Starting from Bo Pedersen's (1980) hypothesis of available energy, these assumptions
were based on the fact that energy must be available for production of turbulent kinetic
energy as well as for loss in potential energy due to deposition.

By fitting experimental data and by back-calculation, the water entrainment in turbidity


currents in laboratory flume can be written according to Parker et al. (1987) as:

0.075
(5.10)

For small values of Ri0 this function ew approaches a value of 0.075. For large values
of Ri0 the function becomes:

ew = (5.11)

72
To illustrate the uncertainty in predicting the entrainment rate we have plotted in
Figure 5.1 the entrainment relations (5.3) to (5.6) of Christodoulou (1986a) and
equations (5.10) and (5 .11) of Parker et al. (1987) together with experimental data
from Ashida and Egashira (1975), Deardorff and Willis (1982), Ellison and Turner
(1959), Fukuoka and Fukushima (1980), Lofquist (1960) and Parker et al. (1987). Note
that both axes have a logarithmic scale, so that the actual scatter is greatly distorted.

0.0000
0.01 0.1 1 10 100
- Parker et al. (1987) • Lofquist (1960) Riot:.
Parker et al. (1987)
x ° Fukuoka & Fukushi (1980)
· · Christodoulou (1986a) • Ellison & Turner (1959)
• Ashida & Egashi (1975) • Deardorff & Willis (1982)

Figure 5.3 Entrainment rate: comparison between theory and data.

The large scatter in this figure is partially due to various uncertainties in measuring and
scaling the entrainment rate. Using other length and velocity scales, such as proposed
by Christodoulou (1986b) have not shown an appreciable improvement of these results.
As an example of a possible source of uncertainty it should be remarked that the values
of ew from Parker' s (1987) experiments were back-calculated from a finite difference
form of the following fluid mass balance between two sections:

(5 .12)

Inevitably an increase in error is involved due to the differentiation of the measured


values for depth and velocity.

73
Another phenomenon that could militate against the use of the formulations proposed
is the possibility of spurious correlations. Benson (1965) has shown that correlating
two parameters, each containing the same stochastical variables, may lead to spurious
correlation. This can simply be shown by adopting the following measured stochastic
variables: x 1 = w;,, x 2 = !).u, x 3 = a 2, x 4 = !).p/p2 = ed. Their means are i;, i;, i;, x.i
respectively, or more general the mean values are expressed in terms of the expected
value as µ 1=E(x 1), ~=E(x2), µ 3=E(x3), µ4=E(x4). The standard deviations are s 1, s 2, s3,
s 4 respectively. Then their relative errors become

(5.13)

As shown before, for the entrainment relation the entrainment parameter and the
overall Richardson number are used defined by a combination of the measured
stochastic variables:

Y2 (5.14)
Au

Similarly to Benson (1965) the two parameters are expanded into Taylor series with
respect to the means of the original stochastical variables. This allows us to express the
mean and variance of y 1 and y 2 in terms of means, variances and covariances of x 1, x2,
x 3, x 4 • We -define correlation coefficients for the correlation between the different
stochastic variables:

i,j=l, ... ,4; i<j (5.15)

Using these results we can define a correlation coefficient for y 1 and y 2 :


. 2
r 1 r3 r13 + r 1 r4 r14 - 2r1r2r 12 - r2 r4 r24 - r 2 r 3 r23 + 2r2
(5.16)

Let us assume that the original variables x 1 to x 2 are non-correlated, i.e., r I2 = r 13 = r I4


= r 23 = r24 = r 34 = 0. Note this is not necessarily true if for instance the back calculation
of Parker (equation 5.10) is used. Then the correlation coefficient between the two
dimensionless parameters can be reduced to

(5.17)

74
This coefficient shows that if certain ratios between r 1 to r4 happen to exist, a large
correlation coefficient can be found, although the original variables are not correlated.
For instance for a combination r 2 = 3 ·r 1 = 3 ·r3 = 3 ·r4 , a correlation coefficient rY 1Y2
= 0.92 is found.

To avoid these spurious oscillations the parameters for scaling the entrainment relations
should be redefined. For instance an alternative form of the entrainment parameter can
be proposed which can be used (correlation free) in combination with the overall
Richardson number:

(5.18)
3
{iv

where vis the kinematic viscosity. Unfortunately we do not have appropriate data sets
to derive a new empirical relation based on the proposed parameters.

Clearly the use of one of the discussed empirical entrainment relations for interfacial
mixing demands caution. Under- or overestimation of the actual entrainment rate with
a factor of the order of two is possible, which does reduce the predictability of a two-
layer model significantly. All these drawbacks are almost ofno significance ifwe are
dealing with internally subcritical flows (high Ri0) for which is commonly accepted
that entrainment is negligible, primarily due to the formation of a stable interfacial
mixing layer. In summary we may conclude that the relations as presented here may
be the most useful tools at this time, while dealing with internally supercritical flows.
However, one should not underrate the limitation of the model accuracy by using one
of these entrainment relations.

75
9l
Chapter 6

Conclusions

In this report a mathematical model is presented to simulate hydrodynamic and


morphological processes due to a turbidity current in a reservoir. The model is two-
dimensional in the horizontal plane, and it consists of two layers: a turbid
underflow and a clear (fresh) upper-layer flow. Hydrostatic pressure and turbulent
flow are assumed, while stratification is caused by the suspended-sediment
concentration in the turbidity current. The asymptotic approach of Galappatti (1983)
is used to model the readjustment, or adaptation, of the depth-integrated
concentration due to erosion or sedimentation. A semi-empirical model is used to
describe the vertical equilibrium distributions of velocity and concentration, which
in turn resulted in closure relations for the 2-layer model.

The following conclusions have been drawn from this study:

Since turbidity currents in reservoirs usually have average sediment


concentrations less than about 5% of volume, it is possible to significantly
simplify the two-layer model by adopting the Boussinesq approximation,
which states that density differences only affect body forces. Concentration
gradients in the equations of motion disappear except for the pressure term.

Dealing with low density differences, the internal flow phenomena (e.g.,
internal waves and fronts) hardly affect external flow phenomena (e.g.,
surface waves). This allows us to further simplify the model by decoupling
equations of internal and external flow. The relative error hence introduced is
of the order of the relative density difference.

The tw.o-layer model can be reduced to a one-layer model by a rigid lid


approach for the water surface, or by assuming the water depth infinitely
large. It is shown that this commonly used simplification is only justified in
very deep water-bodies, such as in submarine canyons. In regular reservoirs
or laboratory flumes, errors to 50% may occur in characteristic celerities and
computational results with respect to the two-layer approach.

A generalization for velocity profiles in turbidity currents is obtained from


laboratory data. Two sub-layers can be defined for the underflow,
distinguished by the level of maximum velocity which is located at about
40% of the underflow depth for internal subcritical, and at about 15% for
internal supercritical turbidity currents. The logarithmic velocity profile of
the lower sub-layer is described using the mixing-length concept. The

77
velocity profile in the upper sub-layer (affected by density gradients and
interfacial mixing) is described by an empirically-fitted relation. The velocity
only depends on the bed-roughness scale and the underflow depth, or in
condensed form it depends on a friction coefficient CD or a Chezy coefficient
C.

Starting from the mixing length concept and the velocity profiles, general
relations are formulated for the concentration profiles. Since the mixing
length theory yields an unrealistic zero vertical diffusion at the velocity
maximum, we connected the maxima diffusivities of the lower and upper
sub-layers by a linear bridge. The resulting concentrations depend on
roughness scale, underflow depth, and a combination of the value and the
level of reference concentration near the bed.

The vertical distributions of velocity and concentration are used to determine


the adaptation length and adaptation scales of Galappatti's equation for the
depth-averaged suspended-sediment concentration. At low values of wju.
(i.e., suspension is dominating transport mode) the choice of a gradient-type
bed-boundary condition for derivation of Galappatti's equation, yields a
significantly higher adaptation time than the choice of a concentration-type
boundary condition. The choice of one of these boundaries for modelling
turbidity currents must be verified with data.

The bed-roughness or alluvial roughness, partially dominated by possible bed


forms, is an important parameter for the model. However, at present there
are no reliable bed form or roughness predictors for turbidity currents, and
there is not sufficient data to recalibrate predictors from open-channel flow.
As the bed-form regime is expected to be rather uniform along the
underflow, it is possible to assume (and calibrate) a constant roughness
parameter along the underflow for each test case.

The value of the reference concentration near the bed determines the
suspended-sediment concentration in turbidity currents. Due to similarity of
flow in the lower sub-layer of these currents with open-channel flow
(logarithmic velocity distribution), existing empirical relations for the latter
can be used to predict the reference concentration in terms of flow conditions
and sediment characteristics.

The semi-empirical model for velocity and concentration shows good


agreement with measured profiles in the laboratory, but one should bear in
mind that in nature various (complex) processes may cause deviations from
this simple approach. Especially flow in the interfacial mixing layer maybe
subject to unsteady and non-uniform phenomena. An ameliorated empirical
relation for the flow and concentration in the upper sub-layer (interfacial

78
mixing layer) may be obtained by accounting for these facts and by using
more and accurate data.

Entrainment of clear water from the upper layer into the turbidity current is
caused by various mixing processes caused by instabilities at the density
interface. The empirical entrainment relations, which quantify the
entrainment rate for the two-layer model, usually relate a dimensionless
entrainment velocity to some form of the Richardson number. It is shown
that besides a considerable scatter, the choice of the scaling parameters may
allow for spurious correlations. An alternative is presented but due to a lack
of data it is not elaborated and requires more research. In subcritical turbidity
currents it is appropriate to neglect entrainment.

The 2-DH 2-layer model as presented here can be used for various types of
reservoirs, and by means of a simple width integration of the model it reduces to a
convenient and practical 1-D 2-layer model. Clearly by taking the concentration and
the bed-level constant the model reduces to a more conventional 2-layer which is
used for saline density currents (e.g., Vreugdenhil, 1979).

79
80
Main Symbols

a total water depth= a 1 + a2 [L]


a; depth of the interfacial layer [L]
a, = depth of suspended load layer = Z; - za [L]
ao(r1) = equilibrium concentration-profile function = c.(r1)/C,, [-]
a1,a2 depth of upper layer and lower layer respectively [L]
C Chezy roughness coefficient [L 112rr1
ca (near-bed) reference concentration [-]
CD bed friction coefficient= (uJu2)2 [-]
c,(T]) local equilibrium suspended-sediment concentration [-]
cm concentration at level of maximum velocity (TJ=rim) [-]
c.(ri) local sediment volume concentration [-]
c. depth-averaged suspended-sediment concentration [-]
c., depth-averaged equilibrium suspended-sediment concentration [-]
ew entrainment parameter= w;/½ [-]
f Coriolis coefficient for earth rotation rr-1
1
Fr 1 internal Froude number upper layer= u/✓(e,g,a 1 ) [-]
Fr2 internal Froude number lower layer= u/✓(e,g,a2 ) [-]
g acceleration of gravity [Lff2]
gx g-sinax' cosay' (See Appendix 1, App.1.5) [Lff2]
gy g-cosa/sinay' (See Appendix 1, App.1.5) [L/P]
gz g-cosa/cosay' (See Appendix 1, App.1.5) [Lff2]
h, water-surface level= zb + a2 + a 1 [L]
lo gentle bottom slope [-]
kn Nikuradse sand roughness [L]
L'a dimensionless adaptation length suspended sediment [-]
lm mixing length [L]
p hydrostatic pressure [M/(LT2)]
Pe = Peclet number [-]
qx,qy external flow: discharge in x,y-direction (qx=q 1x+q2x) [L2ff]
qix,qly discharge per unit width in x,y-direction in upper layer [L2ff]
q2x,q2y discharge per unit width in x,y-direction in lower layer [Uff]
Ri gradient Richardson number [-]
Ric critical overall Richardson number (:::: 0.8) [-]
Rimin gradient Richardson number at level T] = TJwn [-]
Rim0 maximum overall Richardson number [-]
Ri0 overall Richardson number = e,gza/u/ [-]
Riot. overall Richardson number based on !lu [-]
Rf bulk Richardson number [-]
Sbx,Sby bed-load transport in x,y-direction (L2/T]
ssx,ssy suspended-load transport in x,y-direction [UIT]

81
s:XZ,s)'Z local turbulent shear stresses in x,y-direction [M/(LT2)]
t time coordinate [L]
T transport stage parameter [-]
T'a = dimensionless adaptation time suspended sediment [-]
u,v,w local fluid velocity in x,y,z-direction (also ufi) [LIT]
u,v depth-averaged flow velocity suspended-load layer in x,y [LIT]
uma:c;um = maximum flow velocity in the underflow, level TJ=rtm [LIT]
U1,V1 = depth-averaged flow velocity upper layer in x,y-direction [LIT]
½,V2 = depth-averaged flow velocity lower layer in x,y-direction [LIT]
u. = shear velocity = ('txb lp1 )1 12 [LIT]
u•cr = critical shear stress for initiation of motion (Shields) [LIT]
t:.u velocity difference between two layers [LIT]
Wbe = fluid entrainment from the bottom-layer [LIT]
W;, fluid entrainment velocity at the interface [LIT]
w'e normalized fluid entrainment coefficient = w;)ws [-]
ws fall velocity of sediment particles (relative to z-axis) [LIT]
ws,eff effective particle fall velocity= w,(1-c5 )'1; a:::,5 [LIT]
x,y,z spatial coordinates (x,y = horizontal, z = vertical) [L]
z suspension parameter [-]
Z' = corrected suspension parameter = Z + cp [-]
Za reference level for suspended-sediment concentration [L]
zb = bed level [L]
Z; level of density interface = zb + a2 [L]
a/,ay' average bed slope in x,y-direction [-]
ao,a1,Ui coefficients for Galappatti's model [-]
~ coefficient for suspended-sediment concentration [-]
Yo,Y1,Y2 coefficients for Galappatti 's model [-]
0 laminar sub-layer thickness [L]
Ob depth of the bed-load layer [L]
tp porosity of the bed material [-]
ed relative density = 1 - (p1 /p 2) = cr' ·Cs [-]
ejlm,efum = maxima of efx in lower and upper sub-layer respectively [L2/T]
efx,ejj,,efz fluid-mixing coefficients in x,y,z-direction [L2/T]
esx,esy,esz sediment-mixing coefficients in x,y,z-direction [L2IT]
esz' normalized vertical-diffusion coefficient = e,!(w,as) [-]
s dimensionless y-coordinate = (y-y0 )w/(va5 ) [-]
TJ dimensionless vertical coordinate = (z-zb)la2 or (z-za)/as [-]
Tla dimensionless reference level= (za-zb)la 2 [-]
Tim dimensionless level at which u(z) is maximal [-1
'llwn level at which shear stress is maximal :::, llm+0.4-(l-TJm) [-]
Tlo dimensionless roughness scale at which u(z)=O [-]
)C von Karman's constant (:::,0.4) [-]
V kinematic viscosity [L2/T]
~ = dimensionless x-coordinate = (x-x0 )w/(uas) [-]

82
local density of the flow= p_,(cr'c,+1) [M/I})
density of sediment grains, density of clear water [M/L3]
density of upper layer = p1 [M/L3]
density of lower layer= p_,(cr'C,+l) [M/I})
density difference between two layers [M/I-3]
= relative sediment density= (p,-p1)/p1 [-]
't dimensionless time coordinate= (t-t0 )w/a, [-]
'txb, 'tyb bed shear-stress in x,y-direction [M/(LT2)]
'tx;, 'ty; shear stress at density interface in x,y-direction [M/(LT2)]
"tx/, 'ty/ corrected interfacial shear stress: 'tx/ = 't;,; + p1 u2w;e [M/(LT2)]
<l> damping coefficient for suspended-sediment concentration [-]
<p = correction factor for suspension parameter Z' [-]
<J>c characteristic celerity for Galappatti's equation [LIT]
♦c dimensionless characteristic celerity for suspension = <l>l½ [-]
\jf' coefficient for suspended-sediment concentration = ~ -<j> [-]
\jfc(ri) dimensionless shape function for concentration = c/r1)/C, [-]
\jfu(ri) dimensionless shape function for velocity= u(ri)lu [-]
\jf.(ri) dimensionless shape function for velocity= v(ri)/v [-]

83
84
References

Abraham, G., M. Karelse and G. van Os (1980)


On the magnitude of interfacial shear of subcritical stratified flows in relation
with interfacial stability.
J. Hydr. Res., IAHR, Vol.17, No.4, p.273-287.
Abraham, G. and C.B. Vreugdenhil (1971)
Discontinuities in stratified flows.
J. Hydr. Res., IAHR, Vol.9, No.2, p.293-308.
Akiyama, J., Stefan, H.G. (1985)
Turbidity Current with Erosion and Deposition.
J. Hydr. Engrg., ASCE, Vol.Ill, no.12, p.1473-1496.
Akiyama, J. and Y. Fukushima (1986)
Entrainment of non-cohesive bed sediment into suspension.
In: Wang, S.Y. et al. (eds) Proc. Third International Symposium on River
Sedimentation. Univ of Mississippi, p.804-813.
Alavian, V. (1986a)
Behavior of density currents on an incline.
J. Hydr. Engrg., ASCE, Vol.112, no.I, p.27-42.
Alavian, V. (1986b)
Entrainment and mixing in inclined dense plumes.
In: Stefan, H.G. et al. (eds.) Advancements in Aerodynamics, Fluid
Mechanics and Hydraulics. Proc. Conf. ASCE, Minneapolis, Minnesota,
p.555-561.
Altinakar, S., W.H. Graf and E.J. Hopfinger (1990)
Weakly depositing turbidity current on a small slope.
J. Hydr. Res, JAHR, Vol.28, No.l, p.55-80.
Ashida, K. and S. Egashira (1975)
Basic study on turbidity currents.
Trans. Jpn. Soc. Civ. Engrg., Hydr. Sanit. Engrg. Div., No.7, p.83-86.
Benson, M.A. (1965)
Spurious correlation in hydraulics and hydrology.
J. Hydr. Div., ASCE, No. HY4, p. 35-42.
Bo Pedersen, F. (1980)
A monograph on turbulent entrainment and friction in two-layer stratified
flows.
Series Paper No.25, Inst. Hydrodyn. and Hydraulic Engrg., Tech. Univ. of
Denmark, Lyngby, Denmark.
Britter, R.E. and P.F. Linden (1988)
The motion of the front of a gravity current travelling down an incline.
J. Fluid Mech., Vol.99, Part 3, p.531-543.
Chikita, K. (1989)

85
A field study on turbidity currents initiated from spring runoffs.
Water Resources Res., Vol.25, No.2, p.257-271.
Chikita, K. (1990)
Sedimentation by river-induced turbididity currents: field
measurements and interpretation.
Sedimentology, Vol.37, p.891-905.
Chikita, K., Okumura, Y. (1987)
Dynamics of river-induced turbidity currents from field measurements.
Geophys. Bull. Hokkaido Univ., 49, p.291-300.
Chikita, K., Okumura, Y. (1990)
Dynamics of turbidity currents measured in Katsurazawa reservoir, Hokkaido,
Japan.
J. Hydrol., 113.
Christodoulou, G.C. (1986a)
Interfacial mixing in stratified flows.
J. Hydr. Res., IAHR, Vol.24, No.2, p.77-92.
Christodoulou, G.C. (1986b)
Scaling of interfacial mixing in layered flows.
In: Noutsopoulos, G. (ed.) Proc. Int. Symp. on Buoyant Flows, Athens,
Greece, p.335-347.
Chu, V.T. and R.E. Baddour (1984)
Turbulent gravity-stratified shear flows.
J. Fluid Mech., Vol. 138, p.353-378.
Cunge, J.A., F.M. Holly, and A, Verwey (1980)
Practical aspects of computational hydraulics.
Pitman Publishing Ltd., London, 420 pp.
Deardorff, J.W. and G.E. Willis (1982)
Dependence of mixed-layer entrainment on shear stress and velocity jump.
J. Fluid Mech., Vol.115, p.123-149.
Dracos, Th. ( 1986)
Geophysical density currents.
In: Noutsopoulos, G. (ed.) Proc. Int. Symp. on Buoyant Flows, Athens,
Greece, p.269-288.
Eidnes, G. (1986)
Interfacial mixing in stationary, pressure-driven shear flow.
In: Noutsopoulos, G. (ed.) Proc. Int. Symp. on Buoyant Flows, Athens,
Greece, p.322-334.
Eidsvik, K.J. and B. Bmrs (1989)
Self-accelerated turbidity current prediction based upon (k-e) turbulence.
Continental Shelf Res., Vol.9, No.7, p.617-627.
Ellison, T.H. and Turner, J.S. (1959)
Turbulent entrainment in stratified flows.
J. Fluid Mech., Vol.6, Part 3, p.423-448.
Engelund, F. (1966)

86
Hydraulic resistance of alluvial streams.
Proc. ASCE, Vol. 92, HY2, p.315-326.
Fan, J. (1985a)
Methods of preserving reservoir capacity.
In: Bruk, S. (ed.) Methods of computing sedimentation in lakes and
reservoirs. Unesco, Paris, p.65-164.
Fan, J. (1985b)
Experimental studies on the density currents in settling basins.
Scienta Sinica (Series A), Vol.28, No.3, p.319-336.
Fan, J. (1986)
Turbid density currents in reservoirs.
Water International, Vol.I 1, No.3, p.107-116.
Fan, J. (1991)
Density currents in reservoirs.
Workshop on Management of Reservoir Sedimentation, New Delhi, June,
1991, p.3.1.1-3.1.13.
Fukuoka, S. and Y. Fukushima (1980)
Mechanics of gravity currents advancing into stratified reservoir.
Proc. Japan Soc. Civ. Engrg., No.293, p.65-77.
Galappatti, R. (1983)
A depth-integrated model for suspended sediment transport.
Comm. on Hydr., Report No. 83-7, Delft Univ. of Techn., 114 pp.
Galappatti, R. and Vreugdenhil, C.B. (1985)
A depth-integrated model for suspended sediment transport.
J. Hydr. Res., IAHR, Vol.23, No.4, p.359-377.
Garcia, M.H. (1985)
Experimental Study of Turbidity Currents.
M.S. Thesis, Univ. of Minnesota, 138 pp.
Garcia, M.H. (1990)
Depositing and eroding sediment-driven flows: Turbidity currents.
Ph.D. thesis, Project Report No. 306, University of Minnesota, St. Anthony
Falls Hydr. lab., Minneapolis, 179 pp.
Garcia, M.H. (1993)
Hydraulic jumps in sediment-driven bottom currents.
J. Hydr. Engrg., ASCE, Vol.119, No.10, p.1094-1117.
Garcia, M.H. and G. Parker (1993)
Experiments on the entrainment of sediment into suspension by a dense
bottom current.
J. Geophysical Res., Vol.98, No.C3, p.4793-4807.
Gartrell, G. (1980)
Vertical flux measurements in a density-stratified shear flow.
Proc. 2nd. Symp. on Stratified Flows, Trondheim, Vol.I, p.301-314.
Grubert, J.P. (1990)
Interfacial mixing in estuaries and fjords.

87
J. Hydr. Engrg., ASCE, Vol.116, No.2, p.176-195.
Hauenstein, W. and Th. Dracos (1984)
Investigation of plunging density currents generated by inflows in lakes.
J. Hydr. Res., JAHR, Vol.22, no.3, p.157-179.
Hinze, J.O. (1960)
On the hydrodynamics of turbidity currents.
Geologie en Mijnbouw, 39e jaargang, p.18-25.
Irwin, H.P.A.H. (1973)
Measurements in a self-preserving plane wall jet in a positive pressure
gradient.
J. Fluid Mech., Vol.61, Part I, p.33-63.
Johnson, T.R., G.J. Farrell, C.R. Ellis, and H.G. Stefan (1987a)
Negatively buoyant flow in a diverging channel. I: Flow regimes.
J. Hydr. Engrg., ASCE, Vol.113, no.6, p.716-730.
Johnson, T.R., C.R. Ellis, G.J. Farrell, and H.G. Stefan (1987b)
Negatively buoyant flow in a diverging channel. II: 3-D flow field
descriptions.
J. Hydr. Engrg., ASCE, Vol.113, no.6, p.731-742.
Johnson, T.R., C.R. Ellis, and H.G. Stefan (1989)
Negatively buoyant flow in diverging channel. IV: Entrainment and dilution.
J. Hydr. Engrg., ASCE, Vol.115, No.4, p.437-456.
Kranenburg, C. (1983)
Internal kinematic waves and the large-scale stability of two-layer stratified
flows.
J.Hydr. Res., IAHR, Vol.21, No.4, p.259-275.
Lambert, A.M., Giovanoli, F. (1988)
Records of riverbome turbidity currents and indications of slope failures in
the Rhone delta of Lake Geneva.
Limnol. Oceanogr., 33, p.458-468.
Launder, B.E. and D.B. Spalding (1972)
Mathematical models of turbulence.
Academic, Orlando, Florida, 169 pp.
Lax, P.D. (1957)
Hyperbolic systems of conservation laws II.
Comm. on Pure and Appl. Math., Vol.10, p.537-566.
Lofquist, K. (1960)
Flow and stress near an interface between stratified liquids.
The Physics of fluids, Vol.3, No.2, p.158-175.
Luthi, S. (1981)
Experiments on non-channelized turbidity currents and their deposits.
Marine Geology, Vol.40, Letter section, p.M59-M68.
McClimans, T.A. (1978)
Fronts in fjords.
Geophys. Astrophys. Fluid Dynamics, Vol. 11, Gordon & Breach Science

88
publ., p.23-34.
Miles, J.W. (1961)
On the stability of heterogeneous shear flows.
J. Fluid Mech., Vol.IO, p.496-508.
Narimousa, S., R.R. Long, and S.A. Kitaigorodskii (1986)
Entrainment due to turbulent shear flow at the interface of a stably stratified
fluid.
Tellus, 38A (1), p.76-87.
Nelson, J.M. and J.D. Smith (1989)
Flow in meandering channels with natural topography.
In: Ikeda, S. and G. Parker (eds.) River Meandering. Water Resources
Monogr. Ser. Vol. 12, AGU, Washington, D.C., p.69-102.
Noh, Y. and H.J.S. Fernando (1992)
The motion of a buoyant cloud along an incline in the presence of boundary
mixing.
J. Fluid Mech., Vol.235, p.557-577.
Parker, G., Fukushima and Y., Pantin, H.M. (1986)
Self-accelerating turbidity currents.
J. Fluid Mech., Vol.171, okt. 1988, p.145-181.
Parker, G., Garcia, M., Fukushima, Y. and Yu, W. (1987)
Experiments on turbidity currents over an erodible bed.
J. Hydr. Res., IAHR, Vol.25, No.I, p.123-147.
Piat, J.F. and E.J. Hopfinger (1981)
A boundary layer topped by a density interface.
J. Fluid Mech., Vol.I 13, p.411-432.
Ribberink, J.S. (1986)
Introduction to a depth-integrated model for suspended transport (Galappatti,
1983).
Report No.6-86, Delft Univ. of Techn., Faculty of Civil Engrg., Fluid Mech.
Group, 21 pp.
Richardson Y.F. and W.N Zaki (1954)
Sedimentation and fluidization, Part 1.
Trans. Inst. Chem. Engrs., Vol.32, p.35-53.
Rijn, L.C. van (1984a)
Sediment transport, Part I: Bed-load transport.
J. Hydr. Engrg., ASCE, Vol.110, No.IO, p.1431-1456.
Rijn, L.C. van (1984b)
Sediment transport, Part II: Suspended-load transport.
J. Hydr. Engrg., ASCE, Vol.110, No.II, p.1613-1641.
Rijn, L.C. van (1984c)
Sediment transport, Part III: Bed forms and alluvial roughness.
J. Hydr. Engrg., ASCE, Vol.110, No.12, p.1733-1754
Rijn, L.C. (1987)
Mathematical modelling of morphological processes in the case of

89
suspended sediment transport.
Doctoral thesis, Delft University of Technology, 208pp.
Schijf, J.B. and J.C. Schonfeld (1953)
Theoretical considerations on the motion of salt and fresh water.
Proc. Minnesota Intern. Hydraulics Conv., IAHR and ASCE, Univ. of
Minnesota, p.321-333.
Schwarz, W.H. and W .P. Cosart (1961)
The two-dimensional turbulent wall jet.
J. Fluid Mech., Vol.10:4, p.481-495.
Shields, A. (1936)
Anwendung der Ahnlichkeitsmechanik und der Turbulenzforschung auf die
Geschiebebewegung.
Mitt. der Preuss. Versuchsanst. fiir Wasserbau und Schiflbau, Heft 26,
Berlin, Germany.
Simpson, J.E. (1987)
Gravity currents: In the environment and the laboratory.
Ellis Horwood, 244 pp.
Simpson, J.E. and R.E. Britter (1979)
The dynamics of the head of a gravity current advancing over a horizontal
surface.
J. Fluid Mech., Vol.94, Part 3, p.477-496.
Sloff, C.J. (1991)
Reservoir sedimentation: a literature survey.
Communications. on Hydr. and Geotechn. Engrg., Report No. 91-2, Delft
Univ. of Techn., 126 pp.
Sloff, C.J. (1992)
The method of characteristics applied to analyse 2DH hydraulic models.
Communications. on Hydr. and Geotechn. Engrg., Report No. 92-4, Delft
Univ. of Techn., 62 pp.
Stacey, M.W. and A.J. Bowen (1988a)
The vertical structure of density and turbidity currents: Theory and
observations.
J. Geophysical Res. , Vol.93, No.C4, p.3528-3542.
Thorpe, S.A. (1971)
Experiments on the stability of stratified shear flows : miscible fluids.
J. Fluid Mech., Vol.46, p.299-319.
Tsihrintzis, V.A. and V. Alavian (1988)
Interfacial instabilities associated with density currents.
In: Abt, S.R and J. Gessler (eds.) Hydraulic Engineering, Proc. Nat. Conf.,
ASCE, Colorado Springs, Colorado, USA, p.189-194.
Turner, J.S. (1973)
Buoyancy effects in fluids.
Cambridge University Press, London, 367pp.
Turner, J.S. (1986)

90
Turbulent entrainment: the development of the entrainment assumption, and
its applications to geophysical flows.
J. Fluid Mech., Vol.173, p.431-471.
Vreugdenhil, C.B. (1979)
Two-layer shallow-water flow in two-dimensions, a numerical study.
J. Computational Physics, Vol.33, p.169-184.
Wang, Z.B. (1989)
Mathematical modelling of morphological processes in estuaries.
Communications on Hydr. and Geotechn. Engrg., Report No.89-1, Delft
Univ. of Techn., Delft, 208 pp (doctoral thesis).
Wang, Z.B. (1992)
Theoretical analysis on depth-integrated modelling of suspended-sediment
transport.
J. Hydr. Res., JAHR, Vol.30, No.3, p.403-421.
Wang, Z.B. and J.S. Ribberink (1986)
The validity of a depth-integrated model for suspended-sediment transport.
J. Hydr. Res., JAHR, Vol. 24, No.I, p.53-67.
Wood, I.R. and J.E. Simpson (1984)
Jumps in layered miscible fluids.
J. Fluid Mech., Vol.40, p.329-342.

91
92
Appendix A

General derivation of basic equations

App.A.1 Basic equations for 3-D sediment-laden flow in a reservoir

If the flow of a fluid-sediment mixture is assumed to be fully turbulent the three


dimensional flow of a fluid can be presented by the Reynolds equations:

apumJ
at +
( t
'1i(Pii,,.,iilm) + ~p + '1; Su +Su
V) - Ki
- = 0
(Al)

Where:
c, local mean sediment volume concentration (suspension)
p pressure
Sij = normal stress if i = j; ""' 0
S/ = effective turbulence related shear stress (if i -:;; j)
S/ = viscous shear stress
uf.i local mean fluid velocity
um,; local mean velocity of mixture = uJ,; - w)\
w, particle fall velocity relative to z
8; Kronecker delta: 8x = 8Y = 0, 8z = I
p density of the fluid-sediment mixture = p;(l - c.) + Ps c.
p1 density of fluid
Ps density of sediment grains
Note that the vertical sediment velocity is assumed to be composed of the vertical fluid
velocity minus the particle fall velocity.

Further:
For #j => turbulent shear stresses= S/ = Reynolds shear stresses. They can be
written as:

where e1 = diffusion coefficient for water, Es = diffusion coefficient for


sediment, and er' = relative density of sediment grains (p.-p1 )lp1
Viscous shear stresses are only negligible in low concentration fully developed

93
Appendix A: General derivation

turbulent flow. They can be written as:

= -µ m 'vu
j m,i

where µm = dynamic viscosity coefficient of the fluid as modified by the


presence of particles, I.lo = dynamic viscosity of the fluid phase, and a 1 and <¼
are coefficients.
The system of equations is derived relative to an inclined coordinate axis. In
section App.A.5 the following gravity vector is derived:

Turbulent and viscous shear stresses (S) are generated by the bottom and the density
interface: derivatives of Sif to x and y are small compared to those to z; therefore take

as"xy as;
= --= 0
ay ax

Momentum-balance equations:

Reynolds equations for large sediment concentrations can be rewritten in the following
set of differential equations:

apu
at.
+
apu 2
ax
+
apuv
ay
+
apuw
az
+
ap
ax
+
as 1
--
az
.Q
_~(µ au) =
az maz (A2)

= pfv + P8x

apv apuv apv 2 apvw ap astyz


at
+
ax
+
ay
+
az
+
ay
+
az
~(µm
az az
av) =
(A3)

-pfu + pgy

apw apuw apvw apw 2 ap (A4)


+ + + + + pgz = 0
at ax ay az az

94
Appendix A: General derivation

In which:
f coefficient of geostrophic acceleration (Coriolis)
u fluid-velocity in x-direction
V fluid-velocity in y-direction
w fluid-velocity in z-direction

The third equation (A4) yields a hydrostatic pressure relation (neglected are the
turbulent stress and viscous stress gradients compared to 8p/8z).

Mass-balance equations:

The 3-D mass-balance equations for a unit volume in the 2-phase flow are, after time-
averaging of turbulence, applying the eddy-viscosity concept, and assuming that
e/8c/8x) ~ e/8c/8z) for gradually varied flow (e.g., see van Rijn, 1987):

For fluid:

(AS)

For sediment:

(A6)

Where:
e1 fluid mixing coefficient
es sediment mixing coefficient
In the following sections we will assume that fluid and sediment mixing coefficients
are approximately equal for fine sediments: e1 ~ es·

Boussinesq equations:

The Reynolds equations can be significantly simplified by adopting the Boussinesq


approximation for density currents which states that density differences only affect
bode forces. This latter approximation affects the basic equations by taking the density
equal to p1 everywhere except when multiplied with g.

95
Appendix A: General derivation

App.A.2 General basic equations for 2-DH flow in a reservoir

The 2-DH partial differential equations for the flow in a reservoir can be derived by
integration of the 3-D equations (A2) to (A6) over the depth of a layer in the flow .
Therefore consider layers for bed-load transport, suspended-load transport (turbidity),
and for clear water transport (we start the derivation from a three-layer model, which
finally reduces to a two-layer model).

The equations are integrated over the layer depth from z 1 (= lower boundary) to z2 (=
upper boundary). The Leibnitz rule for differentiation of integrals is used for this
derivation,

i.e.:

Since this is a rather straightforward integration we will only pay some attention to the
integration of the pressure terms of equations (A2) and (A3) in the following section
(App.A.3). Only for the pressure term of equation (A4) follows directly:

(A4a)

We define in these equations S1 and S2 = ¾jt'1cljr!dz (shape


'l
factors) and u(TJ) = Y°iTJ)U and v(TJ) = \JfuCTJ)V and p(TJ) = l!'cCTJ)p.
Here T] = (z-zi)/a and a= depth of the respective flow-layer; z 1 = lower-boundary
level relative to x and y axis; z2 = upper-boundary level.

Remarks:
The shape of the velocity distributions ~" in x and y direction is assumed to be
equal, hence helical flow is not considered.
The similarity assumption applies, i.e. the shape of the velocity distribution ~"
and the density or concentration distribution \Jfc are assumed to be constant
throughout the reservoir, independent of x and y).

On the boundaries of the layer the following kinematic boundary conditions apply for
sediment and fluid particles on the boundaries (accounting for the distinction in vertical
velocity of fluid and sediment particles):

96
Appendix A: General derivation

az 1 Bz 1 Bz 1 (A7)
+ u - + V -
Zt C1)'
= w z1 -wel
ot Z1 ax

Bzz + u -
Bzz a½ (A8)
ot ¼ ax
+ V -
¼ ay = w¼ -we1
Where
w, 1 Net velocity at which fluid particles are entrained from a lower
layer (below z 1) into the layer under consideration (for sediment
particles this velocity = w, 1-ws but it will not change the right
hand side of these boundary conditions since also the velocity w, 1
should then be expressed as w, 1-ws).
w,2 Net velocity at which fluid particles are entrained from the layer
under consideration into a higher layer (above z2).

After substitution of the kinematic boundary conditions in the depth-integrated


equations, and assuming shallow water flow (w ~ 0) the following set of differential
equations remains (omitting overbars):

Momentum in x-direction:

1
as1pu a
+
ax (A9)

- fS 1 pva -

Momentum in y-direction:

+---
as2puva
ax
+ J:
½

Z1
dz + [-ryz(½) - 'tyz(z 1)] + (AlO)
+ fS 1 pua - pgp + (pv)½we2 - (pv)z,we 1 = 0

Momentum in the vertical direction (hydrostatic pressure relation):


(All)

Mass balance equation for fluid:

97
Appendix A: General derivation

1 aS (1-Cs)ua
+-~ -~-
ax

Mass balance equation for sediment:

aS1 Csua
+---
ox (Al3)

The pressure in the layer is hydrostatic (equation Al 1), but not a linear function of the
depth (because of the non-linear shape of the density distribution). The relation for the
pressure is:
Zz
p(z) =po +pD = P1g/a1 +½ -z) + gzf(P - P1)dz (Al4)
z

In which: a1 depth of the upper layer when present


P1 (constant) density of the upper layer when present

The depth-integrated forms of the pressure gradients for a turbidity current with
suspended-sediment concentration C, are derived in section App.A.3. After substitution
of these results the equations can be written in their complete form.

App.A.3 Integration of pressure terms for a turbidity current

Consider a turbidity current flowing below a layer with a low sediment concentration.
The upper layer will have a density approximately equal to PJ- The turbidity current
has a density p2 determined by the sediment concentration c,, and it has a depth a2 •
Then hydrostatic pressure can be presented as:
¼
p(z) = Ps +p' = P1gz(a1 +2i-z) + gJ(P2 - P)dz =
z (Al5)
1

= pfgz(a1 +2i-z) + P1gza'Csa2J Wc(TJ)dTJ


Tl,

98
Appendix A: General derivation

In which a1 = depth of the upper layer (with a constant density p1)


c,(TJ) = concentration of suspended sediment = \Jfc(TJ)C,
C, = depth-averaged suspended-sediment concentration
z1 = the lower boundary of the layer
z2 = the upper boundary of the layer = z 1 + a2
T] = (z-z/a2)
pi(TJ) = Pf(cr'c,(TJ) + 1)
er' = relative density of sediment
\Jfc = shape factor for the concentration distribution over the layer

'l'c can assumed to be constant for the full area (x,y) = similarity assumption.

To find the depth-integrated pressure gradient we use the Leibnitz rule for integration
again, yielding:

(Al6)

in which
I

p(z 1) = p1 gz(a 1+a2) + p1 gza 1CsazJ Wc(TJ)dT)


0

The depth-averaged pressure in this equation follows from (Al5):


½ ½ ½I

jp(z)dz = pa2 = p1 gzf(a 1 +Zz-z)dz + p1 gza 1CsazJ f 'ljrcdTJdz


~ ~q (Al6)

= pfgzaz(al+ ½az + a1Csa2R1(cs))

Now the depth-integrated pressure gradient in x-direction becomes

(A17)

and similar in y-direction

99
Appendix A: General derivation

(Al8)

The coefficient R 1 is related to the concentration profile:


11

R1(cs) = ff Wc(TJ)dTJ dTJz "' ~ (Al9)

App.A.4 Derivation of basic equations for a two-layer model

For the derivation of a two-layer model for turbidity currents on a mobile bed we have
to consider a three-layer model at first. This approach has been schematized in figure
Al. The following layers have been defined:

Bottom layer: This is the actual sediment bed consisting of uniform sediment
grains and pores filled with stagnant water. The porosity is f.P the upper
boundary is located at the average bed level zb.
Bed-load layer: This layer is part of the turbidity current and is formed as a thin
layer near the bed in which sediment is transport as bed-load. Its depth ob is of
the order of magnitude of bed-form height or the Nikuradse roughness height.
Its contribution to the turbidity-current flow is very small.
Suspended-load layer: This layer is the primary part of the turbidity current in
which sediment is transported in suspension. Its depth is a,, its lower boundary
is located at level (zb + ob), and its upper boundary corresponds to the density
interface which has to be taken somewhere in the interfacial mixing layer.
Upper layer: This layer may be assumed almost free of sediment so that its
density approximately equals that of clear (fresh) water. Flow velocities in this
layer are mainly caused by river inflow or generated as a return flow by
entrainment of water into the lower layer. Hence flow velocities are small
compared to those of the turbidity current and may be assumed rather uniformly
distributed over the depth.

Between these layers exchange of fluid and sediment may occur. However, we may
assume that the interfacial layer inhibits sediment exchange in upward direction, while

100
Appendix A: General derivation

water surface

al
r
'
'
upper layer

pf

-interface
P2
as 'I
~nded-load
! ~
layer

bottom layer

Figure Al Three-layer approach

only fluid is entrained from the upper layer into the turbidity current (into the mixing
layer). Furthermore in the chosen approach we must take interfacial shear •x;, •y;, and
bottom shear •xb, 'yb into account as shown in the figure. For completeness we also
assume a shear stress between the bed-load and suspended-load layer: •xr, •yr·

For each layer we can now rewrite the depth-integrated basic equations as derived in
section App.A.2.

a. Bottom and bed-load layer

Assume that the depth-averaged sediment concentration of the bed-load layer is Cb, the
density of flui~-sediment mixture is Pb (assume R! uniform in vertical direction), the
fluid parcel entrainment from the bottom-layer is wb,, the entrainment at the upper
boundary is w,,, and the depth-averaged velocities are ub and vb. Furthermore assume
that the pressure in this layer which is mainly caused by higher layers can be expressed
by
(A20)

so that the pressure gradient becomes (with p 2 = density of suspended-load layer)

(A21)

101
Appendix A: General derivation

Now we can write the basic equations for the bed-load layer as (neglecting shape
factors):

Momentum in x-direction:

(A22)

Momentum in y-direction:

(A23)

Mass balance equation for sediment:

(A25)

And for the bottom-layer (stagnant layer) with a sediment-volume concentration of (1-
c:P) mass conservation gives:

102
Appendix A: General derivation

Mass balance equation for fluid:

aepzb + [£I acbl (A26)


at az zb

Mass balance equation for sediment:

(A27)

b. Suspended-load layer

Assume that the depth-averaged sediment concentration of the suspended-load layer is


C,, the density of fluid-sediment mixture is p2 = p1 (1 +cr'C,), the fluid parcel
entrainment from the bed-load layer is w,,, the entrainment at the interface is w;,, and
the depth-averaged velocities are u, and v,. Now we can write the basic equations for
the suspended-load layer as:

Momentum -in x-direction:

Momentum in y-direction:

(A29)

Mass balance equation for fluid:

103
Appendix A: General derivation

(A30)

(A31)

C. Uwer layer

Assume that the sediment concentration in the upper layer and the shear stress at the
water surface (wind shear) is negligible, the constant density of fluid is pft the fluid
parcel entrainment from the suspended-load layer at the interface is w;., the water-
surface level (relative to the horizontal coordinate axes) is hs, and the depth-averaged
velocities are u1 and v1 (with uniform depth distributions). Furthermore the pressure
in this layer becomes p(z) = p1 g/hs-z) and the pressure gradient becomes

(A32)

Now we can write the basic equations for the upper layer as:
Momentum in x-direction:

(A33)

Momentum in y-direction:

104
Appendix A: General derivation

(A34)

Mass balance equation for fluid:

= 0 (A35)

d. Two-layer model: momentum equation lower layer

The presented multiple-layer approach can be transformed to a two-layer model by


combination of layers. In the two-layer model the upper layer is equal to the upper
layer of the three layer model, but the lower layer is now a combination of the
suspended-load, bed-load and bed. Combination of momentum equations in x-direction
for the lower layer gives:

(A36)

We can now use the assumptions that the velocities and the depth (or convection and
mass transfer) are much smaller in the bed-load layer than those in the suspended-
transport layer. Furthermore we assume that the total depth of the lower layer is a2 =
8b + as and we adopt the non-slip assumption at the bed which implies that the velocity
ub at zb equals zero. Expressing the density of the bed-load layer as Pb= p2 + .1.p we
may assume that terms with .1.p are small compared to those with p 2 (for sediment-
laden flows). We will take the shape factors equal to unity for this moment (this gives
R 1 = 1/2). These assumptions imply that we can write the equations in terms of average
flow with:

105
Appendix A: General derivation

,::: ½(ub + u,) with Ub <11; U,


,::: ½(vb + v,) with Vb <11; V,
=Ob+ a, With Ob <11; a,
,::: ½(L\p + 2p2) with L\p <11; p2
,::: ½(Cb+ C,) with (Cb - C,)/C, <11; 1

Combination of all equations and expressed in terms of averages the basic equations
for the lower layer becomes:
Momentum in x-direction:
2
a P21U2tlz a P2rU2 a2 ap2tu2v2a2 aa2
+ + + g P2ta2- +
at ax ay z ax
aa 1 ozb 1 2 °P2r (A37)
+ Ptgza2-;f; + gzp2ta2 ax + +
2gzlli OX

Momentum in y-direction:

(A38)

e. Two-lqyer model: mass-balance equation lower layer

Similarly to the previous derivation of momentum equations for the lower layer we can
derive mass-balance equations expressed in terms of average flow variables.
Combination of mass-balance equations for the suspended-load layer, the bed-load
layer, and the bottom layer yields:

Mass balance equation for fluid:

106
Appendix A: General derivation

Mass balance equation for sediment:

(A40)

The right hand terms in both equations represent boundary conditions at the density
interface. Before averaging these equations they can be combined to form mass-balance
equations for the fluid-sediment mixture. Therefore we define bed-load transport
discharge per unit of width sbx = Cbul>b and suspended-load discharge per unit of
width ssx = Csu.as (and similarly in y-direction). Firstly add both equations (A39) and
(A40) and average in the same way as proposed for the momentum equations. The
resulting mass-balance equation becomes:

(A41)

Secondly rewrite equation (A40) while assuming that the time derivative of Cl5b is
much smaller than that of C,as. The second equation becomes:

The right hand terms (interfacial boundary conditions) except for w;, can be neglected
if interfacial sediment exchange is negligible.

f Basic equations uwer layer

The basic equations derived for the upper layer of the multiple layer model can be
used also for the two-layer model. The momentum equations divided by p1, and the
mass-balance equation become:

Momentum in x-direction:

(A43)

107
Appendix A: General derivation

Momentum in y-direction:

(A44)

Mass balance equation for fluid:

= 0 (A45) I

App.A.5 Components of gravity force in an inclined coordinate


system

The coordinate axes of the model as presented here are tilted compared to the
horizontal plane. The x-axis has an angle a,; with the horizontal and its projection on
the horizontal plane is x'. The angle between the y-axis and the horizontal is a;__ and
its projection is y'. In figure A2 a schematization is given how the gravity vector -g can
be resolved in components in x,y,z-direction.

Starting from this figure in 3-D space we can dete~e the components gx, gy, g, in
terms of the angles a/ and ay' and the magnitude of g. The following three equations
can be derived:

gx
tan(a;) = (A46)
gz

gy
tan(a;) (A47)
gz

2 2 2 (A48)
gx + gy + gz = 181 = g

Rearranging these equations gives the following results which are simplified assuming
that the angles a/ and ay' are small so that sin a/ -sin ay' ~ 1:

108
Appendix A: General derivation

y
X

Figure A2 Gravity force in an inclined coordinate system

(A49)

(A50)

[cos( a~)cos( a~)]2 (A51)


1 - [sin(a~)sin(a~)]2

109
110
Appendix B

Derivation of a 2-DH Galappatti model for suspended-


sediment modelling in turbidity currents

App.B.1 General derivation of the asymptotic solution

Galappatti (1983) and Wang (1989) derived a depth integrated form of the convection-
diffusion equation (chapter 3) by substituting an asymptotic solution of the depth-
integrated concentration into the equation. Here we will derive a similar result for the
suspended-load layer which is part of a turbidity current. In normalized form the
convection-diffusion equation can be presented as

(Bl)

uas a
-
with
a = as..§_ -
a - - - -a = vas~ a a
= a-
a-r ws at a~ ws ax ' a, ws ay ' a.., s az

w ws w z-za
if 't = -2(t - t0) ; ~ = -=-(x -x0) ; ( = ~(Y-Yo) ; 11 - - -
as uas vas as
and a, = depth of suspended-load layer(;:::: a2 )
u. == average bed-shear velocity
a = 0 in Galappatti's model; =1 in general model; ;::::5 if large cs
e , = e,)(u.as) = normalized diffusion coefficient
sx
esy ' = eJ(u.as) = normalized diffusion coefficient
e,,' = e,)(w,a,) = normalized diffusion coefficient
\jiu = ulu = normalized main flow-velocity profile in x-direction
\jfv = v/v = normalized main flow-velocity profile in y-direction

This equation (Bl) can be written as: L[cs] = D[c,]

Assumed is that the flow is gradually varying (for which w"" 0), and that the variation
of the equilibrium concentration profile ce with time and horizontal coordinates can be
neglected, i.e. similarity of concentration profiles is assumed (oc/o-c= o/o~ = oc/o(,

111
Appendix B: Derivation of a 2-DH Galappatti model

= 0). Furthermore the effective fall velocity w,.ef can be taken approximately equal to
w, by taking a=O, although this assumption should be verified (when used) if large
concentrations occur. Assume Esx = esy =Ae"" + B. Later we will assume that in practice
the horizontal diffusion coefficients are negligible throughout the reservoir.

Neglecting sediment exchange between the density interface, the interfacial boundary
condition becomes:

Where w,' = wjw,


And the reference boundary conditions at z0 (i.e., Tj 0 =0):
* concentration type/ Dirichlet type: c,(z=zJ=c0 (B3)

* gradient type / Neumann type: (B4)

where c.(z) = c.{TJ) is the equilibrium concentration, and c0 = c,(z0 ) is the equilibrium
bed concentration. The gradient type condition assumes that at level z 0 the upward
diffusive flux is only determined by local conditions.

Equilibrium concentration profile c,.:

An equilibrium condition exists if all gradients to t,x,y in equation (Bl) are zero. The
situation remains unchanged in time and space. Integrating the remaining terms to z
and using the reference boundary condition equation (B3) gives:

0: =0

(B5)
0: =1

According to this result we will express c,(TJ) (with C,,=depth-averaged value) as

112
Appendix B: Derivation of a 2-DH Galappatti model

(B6)

Asymptotic solution:

The theory of Galappatti (1983, 1985) states that terms on the left side of equation
(B 1) are an order of magnitude smaller than terms in right side. They are smaller when
the two terms on the right have the same order of magnitude and opposite signs. This
is true if the flow-condition variation is not too rapid, or better when time scale and
horizontal length scale of the variations are relatively large (hence the approach cannot
be used near turbidity current fronts). Then the solution can be presented as an
asymptotic expansion:

(B7)

where cj is one order of magnitude smaller than cj_ 1•

Now:
Substitute equation (B7) in equation (Bl).
Collect terms of same magnitude:
0 for j=O
D[c) (B8)
=
1 L[cj_ 1] for j>O

Rewrite the boundary conditions at the surface and the bed, and collect terms
of the same magnitude.

Galappatti's most important assumption is that only the zero order term contributes to

l
the depth-averaged concentration C,. Thus
1
Cs for j=O
focjd( =
O for j>O
(B9)

This assumption has the advantage that only a bed-boundary condition has to be given
for the zeroth-order solution. So far, without using the bed-boundary condition, a new
variable C, is introduced.

Equations (B7), (B8), (B9) and the boundary conditions together give an asymptotic

113
Appendix B: Derivation of a 2-DH Galappatti model

solution with (unknown) depth-averaged suspended-sediment concentration C, in it as


a parameter. An equation for C, is found by applying the bed-boundary condition as
shown in the following section.

General asymptotic solution:

After substitution of equation (B7) into (Bl), terms of equal order (of the same
magnitude) are collected. The solutions to each of these terms are defined by applying
an inverse differential operation on these terms, starting from the zeroth-order
concentration profile.

Zero order term (D[c 0]=0)


co(,, CC ri) = Cs0<•, ~, () · ao(ri) (BlO)

with ii0(TJ) = normalized equilibrium concentration profile (equation B4).

First order term (D[c 1]=L[c0])

The first-order term follows from (B8) where D[c 1] = L[c0 ]. Therefore Galappatti
defined an inverse operator 0- 1, which is presented for our model hereafter.
Using this inverse operator 0- 1 on equation (B8) the first-order term becomes:

(Bll)

where:
a~1 = n-1[G~0J; G~2 = n-1[\JfuGo~ ] ; G~3 = n-1r\jJ,G~0J; G~4 = n-1[esx,G~0J; G~5 = n-1resy,G~0J
C, 1 = the unknown depth-averaged value of c 1

Higher order terms (D[c;]=L[c;_ 1])

This process of successive approximations can be repeated for all higher-order parts,
after which the complete asymptotic solution c, is obtained by substitution of all these
parts in (B7). However, for practical applications generally only the zeroth and first-
order terin have to be used (Wang, 1992)

114
Appendix B: Derivation of a 2-DH Galappatti model

Inverse operator 0- 1 (definition) :

To determine the higher order terms from equation (B8) Galappatti defined an inverse
operator 0· 1[ ] as
(B12)

if

(B13)

with interfacial boundary condition

(B14)

and
l

j f(TJ) dTJ = 0 (B15)


0

Integration of (B13) using condition (B14) gives


1

-f g(TJ)dTJ = f + if - f(l)w: = G(TJ) (B16)


11 TJ

which implies.that oGl8rJ = g(ri) and G(l) = 0.

Since the normalized equilibrium concentration profile ii0( ri) satisfies the homogeneous
form ofthis equation (i.e., for G(ri)=O) we may write the solution of(B16) as
(B17)

By means of substitution of (B 17) into (B 16) while using the profile ii0 we find

115
Appendix B: Derivation of a 2-DH Galappatti model

1
- [ A(l)iio(l)we +G(TJ) -
] aa0 = 0
a,,
(B18)

After solving _equation (B18) to A(ri) and multiplying with ao(ri) we can express fas

/(ri) = A(l)iio(l)w: + [a0(1)w:-a0 ](s -J[ao<I)w:-ao]2 aaoa,,


f]
-G(ri) d11] (B19)

where Bis a constant(= A(ri=l)). Finally this equation can be written as


I
D -I [g] = /(11) = ( ii0(ri) -d0(1)w:) f_ g(~)
1
d11 +
'1 ao(11) -ao(l)we (B20)
I

-f g(ri)d11 + B iio(TJ)
f]

where B follows from condition (B15).

If the term a'0(1)w,' = 0, then equation (B20) equals the relation for n- 1 as defined by
Galappatti. A result similar to equation (B20) cannot be found if we consider D[/] for
large concentrations (a;;:::l) because then a complex non-homogeneous non-linear
differential equation has to be solved. Numerical integration can be used here as an
alternative but will not be discussed.

If w,' = 0 it can be shown (Galappatti, 1983, Ribberink, 1986) that the following
mathematical property holds for the inverse operator:

(B21)

This property is used later in this appendix in case of a gradient bed-boundary


condition.

116
Appendix B: Derivation of a 2-DH Galappatti model

Generalized assumption:

In the nth (n>O) order solution we have an under-determined system of n+ I unknowns


Cs0 to Csm and only 1 equation present: the bed-boundary condition. According to
Wang (1989,1992) boundary conditions can be defined for each term c;, but there are
alternatives:

w Galappatti solves the problem by using his essential assumption (B9) of zeroth
order terms and the bed-boundary condition (B3) or (B4).
w Wang (1989) solves this problem more generally by choosing a set of test
functions (taking n=I for first-order solution)
(B22)

While for each <l>k it is assumed that

fork= l, ... ,n (B23)

which means that only the first k terms in the asymptotic solution contribute to
the internal product of the concentration vertical and the ~ component of <I>.
Further, a concentration-type bed-boundary condition
n

L c/11 =0)
j;Q
= c0 (B24)

or a gradient type bed-boundary is used to close the set of equations.

# :t,J'ow for each chosen set of functions <I> a particular asymptotic solution
can be constructed.
# The test functions <I> should together form an orthogonal system.
# The test functions should be such that the internal products of them with
cs have significant physical meanings. Logical is to choose <j> 1=\lfu because
\lfu ·cs is the sediment-transport rate.
# For each (j>k a weighted average concentration can be defined:
1 1 n

J </Jkcs d11 J <PkL cj d11


0
=
O j;l
(B25)

It is possible to eliminate the n+ I variables Csj from (B25) with the help

117
Appendix B: Derivation of a 2-DH Galappatti model

of equations (B23) and (B24), resulting in a system of n equations for


the n weighted average concentrations.

The solution of Galappatti is derived using cI>=(l,1,1, ... ).

In the derivations above the general procedure for the derivation of Galappatti's model
is elaborated. In the following section the derivations are extended to application of the
model in a 2-DH situation.

App.B.2 Galappatti's model for the 2-DH case

The theory for derivation of a 2-DH turbidity-current version of Galappatti's model


greatly corresponds to Wang's (1989) estuary version. The set <I> is chosen (orthogonal
system of which the first 2 components of c,<I> represent the convective sediment
transport in x and y directions respectively, Wang 1989):
(B26)

For unsteady 2-DH problems no more than the first order solution can be applied in
practice, which is only influenced by the first component of <I>:

c, = c0 + c 1 with
(B27)

(B28)

The conditions (B23) and (B25) now become

118
Appendix B: Derivation of a 2-DH Galappatti model

1 1
Cs = f c/TJ)lJr/11) dri f co(TJ)iJr/ri) dri
= = a.0Cs0 (B29)
0 0

acs0
a. - -
1 a. (B30)

in which
1

a.j = f "1u(TJ)d;-CTJ) dri


0
(B31)

Concentration-type bed boundary

Applying the bed-boundary condition c.,o(r1=0)+c,i(r1=0)=ca and eq.(B6) for c, gives

+ C
y1 acs0
+ --- + --- + --- +
y2 acs0 y3 acs0
sl yO a-r: yO a, y O a(
(B32)

_~:? !(? a~s0) - ~: i :,(? a~s0)


in which y1 =. ii/0)

Gradient-type bed boundary

Apply the bed boundary condition (with a=O, low concentration)

(B33)

and adopt the mathematical formulation of the differential of the inverse operator
expressed by equation (B21) (if w, is negligible). The latter equation can be used to
reformulate the relation obtained by substitution of c, = c0 + c 1, expressed by equations

119
Appendix B: Derivation of a 2-DH Galappatti model

(B27) and (B28), into the boundary equation (B33). This derivation, which was also
carried out by Ribberink (1986), the following equation is obtained

ce = c~ 1
+l)aC~ +---
+Cs +-----
(y 1 ac~ +---
ac~ + y2 y3
Yoa. a~ a, Yo Yo
(B34)

_~:?!(?a~~)- ~:?a~(?a~~)
Clearly the only difference with equation (B32) is the coefficient containing y1•

Depth-integrated concentration:

If equation (B29) is used to represent Cs0 and Cs 1 by Cs in equation (B30), and if these
relations are substituted in equations (B32) and (B34) respectively then we find

=Cs+ (!.!_-~)acs
Yo «o a.
+(y
Yo
2
_ a2)acs
ao a~
+('!.J..- a )acsa( +
Yo
3

«o
(B35)

-(::-::J? !(?~·l-(::-::P :c(~il


for the concentration-type boundary condition, and

for the gradient-type boundary condition. These are then the equations for Cs in which
Cs,= ai:,.

Thus the equations for the depth-integrated concentration are then (substitute Cs, and
transform to the original coordinate system):

-(Y1_ «1)asaCs + (Y2_ «2)uasaCs + (Y3_ «3)vasaCs+


Yo «o ws at Yo «o ws ax Yo «o ws OJ (B37)
-(Y4_ <X4)asj_(U•asaCs)-(Y4_ a4)asj_(u•asaCs) + Cs
Yo «o ax ws ax Yo ao oy ws oy

120
Appendix B: Derivation of a 2-DH Galappatti model

for the concentration-type boundary condition, and

(B38)

for the gradient-type boundary condition.

If we assume similarity of flow and concentration profiles which implies that the
coefficients iff.,fi3 (Galappatti 1983, if\Jfu~\Jf.), and we assume that ii4 ~ii1 (Wang 1989,
p.61) then we can rewrite equation (B35) as follows:

Cse - Cs= (~-~)asacs + (Y2_ o:2)iiasacs + (Y2_ o:2)vasacs (B39)


Yo O:o ws at Yo o:o ws ax Yo o:o ws ay
and similarly for equation (B36).

In terms of dimensionless adaptation time Ta' and dimensionless adaptation length La'
these equations can be formulated in a more general form:

I as acs I uas aCs I Vas aCs


Cse - Cs= T0 - - + L -- + L -- (B40)
ws at a ws ax a ws ay

Complete asymptotic solution:

The complete asymptotic solution c/ri) = co(ri) + ci(ri) can be found by eliminating
Cs0 and C, 1 with equations (B29) and (B30) as follows:

(B41)

Substitute this in eq.(B30) to present C, 1 as a function of the depth-averaged C,.

With Cs0 and C, 1 and equations (B27) and (B28) relations for c0 (= i'i0C/a0 ) and c 1 as
a function of the depth-integrated concentration C, can be determined. Then the
complete first order asymptotic solution can be found by c, = c0 + c 1 (for both types
of bed-boundary conditions):

121
Appendix B: Derivation of a 2-DH Galappatti model

iio
-C
a s
0
(B42)

Also equation (B42) can be simplified if we assume similarity of flow and


concentration profiles which implies that the coefficients {i2-;::a3, and if we assume that
d4~d1:

(B43)

The quantification of the various shape factors in these relations is treated in appendix
C.

122
Appendix C

Galappatti's coefficients for turbidity currents:


tables and figures

App.C.1 Regression coefficients

For using Galappatti's equation we are only interested in the adaptation scales La' and
Ta' which will be expressed by the following polynomial regression equations:

r: = r )r
pl 1[1og( ::) +p21 [log(:: +p3 1 [log(:: Jr+ p4 1 [log( :J + p5 1 (Cl)

L: = )r
p1 2 [log( :: ) ]\ p22 [log(:: +p32 [log(:: Jr+ p42 [log(::)] +p52 (C2)

where

plj = (plJ) 11! + (pl:) Tia + (pl])


(C3)
p2j = (p2J) 11! + (p2J) "Ila + (p2J)

and

(C4)

The coefficients for these regression equations are summarized in the following tables.
We distinguish between internal sub- and supercritical turbidity currents, and
Galappatti coefficients for approaches with concentration- and gradient-type bed-
boundary conditions. Note that the adaptation lengths La' are not dependent on the type
of bed-boundary. The coefficients for La' in the tables for the gradient-type bed
boundary can also be used if a concentration-type boundary is chosen.

123
Appendix C: Tables and figures

Table C 1 Regression coefficients supercritical turbidity currents


11m = 0.15 ; gradient-type bed boundary
i c0; cl; c2;
1 146.3 -318. l 61.48
pl/ 2 -12.46 28.66 -8.360
3 0.2991 -0.7181 0.2051
1 618.6 -1368 304.3
p2/ 2 -54.58 125.5 -36.39
3 1.381 -3.212 0.9289
1 760.4 -1724 458.6
T'a p3/ 2 -71.25 163.3 -46.53
3 1.841 -4.287 1.279
1 98.97 -271.2 151.5
p4/ 2 -12.83 30.80 -11.15
3 0.3947 -0.9145 0.2235
1 -255.0 514.9 -29.80
p5/ 2 22.70 -47.66 6.131
3 -0.7200 1.303 0.1575
1 233.4 -476.2 36.48
pl2; 2 -16.85 38.55 -10.87
3 0.3171 -0.8568 0.4077
1 968 .7 -1999 197.l
p22; 2 -71.15 164.3 -48.72
3 1.525 -3.832 1.601
1 1147 -2418 332.1
L'
a p32i 2 -87.29 204.5 -65.50
3 2.162 -5.126 1.691
1 136.4 -326.7 116.5
p42i 2 -12.36 32.57 -16.44
3 0.4634 -1.044 0.2134
l -305.0 635.9 -74.60
p5/ 2 27.11 -57.15 7.815
3 -0.9316 1.696 0.2002

124
Appendix C: Tables and figures

Table C2 Regression coefficients subcritical turbidity current


Tim= 0.40 ; gradient-type bed boundary
i c0; cl; c2;
1 345.1 -701.6 49.76
pl/ 2 -23.71 48.31 -3.631
3 0.3222 -0.7639 0.2794
1 1538 -3091 152.4
p2/ 2 -109.1 220.4 -12.93
3 1.676 -3.953 1.329
1 2086 4118 62.09
T'a p3/ 2 -154.9 309.0 -10.79
3 2.799 -6.472 1.956
1 581.2 -1092 -90.91
p4/ 2 -47.81 92.65 1.997
3 1.155 -2.536 0.5551
1 -394.5 767.6 10.45
p5/ 2 31.92 -63.28 1.487
3 -0.8584 1.632 0.1172
1 424.9 -833.6 2.148
pl/ 2 -29.43 59.38 -3.346
3 0.2546 -0.7102 0.4341
1 1926 -3741 -64.02
p2z; 2 -133.9 269.1 -13.14
3 1.392 -3.652 1.831
1 2670 -5115 -227.1
L'a p32i 2 -186.8 373.2 -14.33
3 2.473 -5.975 2.232
1 811.8 -1499 -178.6
p42i 2 -58.95 115.4 0.1454
3 1.088 -2.411 0.5673
1 -389.3 768.5 -11.14
p52i 2 30.12 -59.76 1.480
3 -0.7804 1.471 0.1326

125
Appendix C: Tables and figures

Table C3 Regression coefficients sub-/super-critical turbidity current


Tim= 0.15; concentration-type bed boundary
i cO; cl; c2;
I 163.4 -321.0 1.271
pl/ 2 -14.47 28.81 -0.9594
3 0.3681 -0.7163 -0.0630
I 690.6 -1375 42.16
p2I; 2 -62.74 125.5 -4.959
3 1.653 -3.178 -0.1738
I 850.5 -1724 112.6
T'a p3/ 2 -80.91 162.0 -6.813
3 2.123 -4.191 0.006758
l 130.0 -269.3 28.65
p4/ 2 -14.73 29.71 -1.734
3 0.4168 -0.8564 0.0248
l -263.4 519.2 -5.820
p5/ 2 24.36 -48.37 1.176
3 -0.6438 1.290 -0.1083

I Tim= 0.40 ; concentration-type bed boundary I


1 341.0 -705.2 72.70
pl2; 2 -23.82 48.76 -4.059
3 0.3809 -0.7033 -0.06444
1 1507 -3125 336.6
p22i 2 -109.0 222.3 -17.00
3 1.913 -3.646 -0.1801
1 2020 -4196 469.1
T'a p3/ 2 -153.5 311.2 -20.38
3 3.068 -6.000 -0.004996
1 543.9 -1141 147.6
p42i 2 -46.53 93.83 -5.207
3 1.203 -2.359 0.02341
1 -387.5 768.3 -17.92
p52i 2 31.83 -63.44 2.139
3 -0.8231 1.652 -0.05743

126
Appendix C: Tables and figures

App.C.2 Figures

The coefficients which appear in Galappatti's equation for turbidity currents are
computed numerically. To illustrate the dependence of these coefficients to the relevant
parameters, they are plotted in the figures presented in this section. The derivation of
these coefficients is outlined in chapter 4, section 4.7. The different coefficients are
defined as follows:
1

a0 = J1jJJTJ)ii (TJ) dTJ


0
(C5)
'la

al= J1jJu(TJ)ii1(TJ)dTJ (C6)


'la

a2 = Jljlir1)ii (TJ) dTJ


2
(C7)
'la

(C8)

The inverse operator 0- 1 is derived in appendix B (App.B. l ). These coefficients form:

Y2 az
La' = (C9)
Yo ao

Ta 1 =
Y1 <Xl
( concentration bed -boundary) (ClO)
Yo ao

Y1 + 1 (XI
Ta 1 = -- (gradient bed -boundary) (Cll)
Yo ao

127
Appendix C: Tables and figures

Y, / o.s
7Yo Tla=0.05
0.4
Tla=0.01
0.3

0.2

0.1

O + - - - - - - - - - - - - - - - - - ~ ~/~
0 0.2 0.4 0.6 0.8 1.0
Figure Cl y/y0 for TJm=0.15 (drawn: TJo = 0.0067, dashed: TJo = 0.000116)

Y, / o.s
7Yo
0.4

0.3

0.2
Tla=0.05

0.1 Tia=0.01

o + ' - - - - - - - - - - - - - - - - - - - l wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure C2 y/y0 for TJm=0.40 (drawn: TJo = 0.0067, dashed: TJo = 0.000116)

128
Appendix C: Tables and figures

0.6,----------------------,
Yyyo TJa=0.05

0.5

0.4
TJa=0.01

0.3

0.2

0.1
'11m=0.15
o +--------------------< wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure C3 y/y0 for llm=0.15 (drawn: llo = 0.0067, dashed: llo = 0.000116)

Y,/ o.s
7Yo
0.4

0.3

0.2

0.1 na=0.01

Figure C4 y/y0 for llm=0.40 (drawn: llo = 0.0067, dashed: TJo = 0.000116)

129
Appendix C: Tables and figures

0.08 ~ - - - - - - - - - - - - - - - - ~
«1/
7 «o Tla=0.05

0.04
Tla=0.01

0 +----------'---------------!

'l'la=0.01
-0.04
11m=o.1s I
+----------------------1 ws u*
0 0.2 0.4 0.6 0.8 1.0
Figure CS a/a.0 for llm=0.15 (drawn: TJo = 0.0067, dashed: TJo = 0.000116)

0.02
"Ya«o
0

-0.02

-0.04

-0.06

- 0 . 0 8 + - - - - - - - - - - - - - - - - - - - - 1 wsfu*
0 0.2 0.4 0.6 0.8 1.0
Figure C6 a/a0 for TJm=0.40 (drawn: llo = 0.0067, dashed: TJo = 0.000116)

130
Appendix C: Tables and figures

0.08 ~ - - - - - - - - - - - - - - - - ~

7ao 'lla=0.05

0.04

11a=0.0l

Q-1------------------------,

-0.04
'lla=0.01

- 0 . 0 8 - 1 - - - - - - - - - - - - - - - - - - - - - 1 w /u
0 0.2 0.4 0.6 0.8 LO s *
Figure C7 a/a0 for TJm=0.15 (drawn: TJo = 0.0067, dashed: 'llo = 0.000116)

0.02
a¾Uo
0

-0.02

-0.04

-0.06

-0.08

- 0 . l V I - - - - - - - - - - - - - - - - - - W8 /U*
0 0.2 0.4 0.6 0.8 LO
Figure C8 a,_/a0 for rim=0.40 (drawn: TJo = 0.0067, dashed: 'llo = 0.000116)

131
132
Appendix D

Equation list
App.D.1 Vertical distribution of velocity and concentration

A summary is given of the equations derived and used. For internal sub-critical
turbidity currents choose llm=0.40, for internal supercritical turbidity currents choose
rim=0.15. The captions used in this list correspond to the captions used in the Chapter
4 for convenience of reading.

Flow velocity, sub-layer ri~ri<rim:

(4.14)

Flow velocity, sub-layer ri,;:;;ri<l:


(4.17)

with

Depth-averaged flow velocity ½ (and similar for v2)

(4.26)

where Iu and ( are functions expressed by


1

I"= f exp[-3(ri-rim)3 12 )dri "' -0.3511!-0.014ri!-0.069TJm+0.425 (4.27)


'lm

133
Appendix D: Equation list

(4.28)

Shear velocity in x-direction u.

u; = Cvu2 ✓u'f +vi= "xb = (gxa'Csa2 + •~] "'gxa'Csa2 (


4 ,69 )
Pf P1

Shear velocity in y-direction v.

v; • CD v,/"i +vJ; • ~ • (c,a'C,a, + ~] • g,a'C,a, (4.70)

Corresponding Chezy value

If a Cv value is known the corresponding roughness scale follows from

K//c; + 1.755034
log(flo) = if 11m = 0.40
-2.656458

K//c; + 1.903016
log(flo) if 11m = 0.15
-1.937993

Concentration profile, sub-layer Tla < T] < Tlm -(¥2)/2:

(4.36)

134
Appendix D: Equation list

Concentration profile, sub-layer TJm f:/2)/2 < TJ < TJm:

C
s(11)
= C
s
(../2
11m
)·[(/i,/2)·11mA +
A 1l + B
Bll/A (4.39)
2

For TJm = 0.15:


A z (zr 1 • [ 5.61 · 10-3 f(11 0 , 11m) - 0.195] z -0.19 •(zr 1 (4.41a)

(4.42a)

For 11m = 0.4:

A z (zr 1 • [ 6.46 · 10-3 f(11o, 11m) - 0.56) "" -0.56 ·(Zf 1 (4.41b)

(4.42b)

Concentration profile layer TJm < TJ < 1:

(4.43)

Depth-averaged concentration (approximate) following from the relations above:

ToLn,,, = 0.15, with Ya = (11af 0.15) - (/i,/2)

c~ =

+
c.t~~ 1(3-/H)()~;)J'{o.15y.[-¾(z')'~
exp(-0.67(Z)) · [3.74· 10-3 (Z) 2
+ 15.0· 10-3 (Z)
+

+
(Z)Y. -

0.4114] }
1] + (4.49)

ToLn,,, = 0.40. with Ya = (11a / 0.4) - (fj, / 2)

3
Cs•= ca·[0·~~ (2-J4-2511~)r {o.4Ya[-¾(Z) Y~
2
+ (Z)Ya - 1] + ( 4 .SO)

+ exp(-0.749(Z))·[16.88·10-3 (Z)2 + 45.0•10-3 (z) + 0.3766] }

135
Appendix D: Equation list

and the Z'-value as

C(O). TJ2 - c(l). TJ + c<2)


where Po kO a kO a kO

C(O). TJ2 - C (1). TJ + c(2)


P1 kl a kl a kl

C(O). TJ2 (1) + c(2)


P2 k2 a - C1,;2 • TJa k2

(0) 2 - C (1). TJ + c<2)


P3 C1c3 "TJa k3 a k3

with regression coefficients:


Table Dl Regression coefficients for <p-value.

llm = 0.15 llm = 0.4


i 0 1 2 0 1 2
Clo)i -1994 118 1.38 -1763 111 1.24

ck/ 4263 -278 -2.86 3677 -245 -2.51


C i -2925 211 1.89 -2445 173 1.61
k2

C i -39.4 0.74 0.045 -53 .7 2.16 0.052


k3

App.D.2 Empirical relations for the reference concentration

The quantification of the reference concentration can only be retrieved from empirical
relations. Presently there are hardly no general relations available from literature.
Therefore it is usual to apply relations derived for open-channel flow, eventually after
recalibration of the empirical coefficients. Three relations are presented here, which can
be used, although they must be used with care.

136
Appendix D: Equation list

Garcia and Parker's (1991) equation for reference concentration ca

Garcia and Parker (1991,1993) defined an entrainment parameter Es which tends to ca


for steady uniform flow. For uniform material (fitted for coal 100 µm - 180 µm) we
can take:

(Dl)

where ~ is a constant equal to 1.3 ·10-7•


The particle Reynolds number is defined as

Dsg ·JatgDsg (D2)


RP=-~~-~
V

and the shear velocity associated with skin friction is defined as (for the evaluated
experiments)

2
u•s 'tb =
= _s CDs u22 "' [ 124

· 107 Re -2 -75 ] u22 (D3)
Pt
Here Re = Reynolds number of flow (= u2a/v) and Dsg = geometric mean size of
sediment.
Parameter Zu is defined as: for RP ~ 3.5:
u
z = ~Ro.6 (D4)
u w p
s

and for 1 < RP < 3

(D5)

Van Rijn's (1984b) equation for reference concentration ca

The equations involved are derived for open-channel flow, calibarted for particles in
the range 100 µm - 500 µm. The following equations are defined

137
Appendix D: Equation list

a. Reference concentration

ca = 0.015 ·( D50) Tl.5 (D6)


a2Tl a Do.3

in which 11a = 0.5·!::Ja2 or 11a = kja 2 with Tla,min = 0.01


~ = bed-fonn height

b. Particle parameter

D = D I
og ll/3 (D7)
• so [ v2

in which D 50 = median particle size, v = kinematic viscosity (µ/p)

b. Transport stage parameter

T = (u.')2 - (u.,c,) 2 (D8)


(u.,c, )2

in which

u~ = u2 • {c{; = bed-shear velocity related to grains,


C0 ' = drag coefficient related to grains (from eq.4.71 with 1lo"'''3 ·Drx}a2 )

u?,c, = ec,o'gD50 = critical bed-shear velocity after Shields (1936),


where 0c, follows from the Shields (1936) curve:

if D. ::s; 4;

0 er = 0.14 ·D-• 0·64 if 4 < D. ::s; 10;

if 10 < D. ::s; 20;

0 29
0 er = 0.013 ·D • · if 20 < D. s 150;

ec, = 0.055 if D. > 150;

138
Appendix D: Equation list

d. Fall velocity
If particles are smaller than 100 µm (Stokes range):

1 o 1gD; (D21)
ws =
18 V

If particles are in range 100-1000 µm:

0.01 o /gDS3]0.5 (D22)


v2

If particles are larger than 1000 µm:

ws =
1
1.1 ·[ o gDsf
5 (D23)

where Ds is the representative particle diameter defined as


Ds (D24)
= 1 + 0.011 ·(os - l)(T-25)
Dso

or Ds = D50 if T 2 25.
Here as is the geometric standard deviation defined as

(D25)

Simple power law equation for reference concentration c.

A power law equation which can be considered as a summary of van Rijn's (1984b)
equation can be written as

(D26)

where m is a parameter related to sediment characteristics and the reference height 1la·
A more simple power law (deduced from the equation above) is
(D27)

139

You might also like