VCB

MODELLING ALLUVIAL CHANNEL FORM 511
Copyright 2004 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 29, 511529 (2004)
Earth Surface Processes and Landforms
Earth Surf. Process. Landforms 29, 511529 (2004)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/esp.1062
RATIONAL REGIME MODEL OF ALLUVIAL CHANNEL
MORPHOLOGY AND RESPONSE
BRETT C. EATON,
1
MICHAEL CHURCH
2
* AND ROBERT G. MILLAR
3
1
Department of Geography, The University of British Columbia, Vancouver, British Columbia, V6T 1Z2, Canada
2
Department of Geography and Peter Wall Institute for Advanced Studies, The University of British Columbia, Vancouver, British
Columbia, V6T 1Z2, Canada
3
Department of Civil Engineering, The University of British Columbia, Vancouver, British Columbia, V6T 1Z2, Canada
Received 8 January 2003; Revised 24 June 2003; Accepted 8 September 2003
ABSTRACT
A theoretical model is developed for predicting equilibrium alluvial channel form. The concept of greatest relative stability,
achieved by maximizing resistance to ow in the uvial system, is presented as the basis for an optimization condition for
alluvial systems. Discharge, sediment supply (quantity and calibre) and valley gradient are accepted as independent govern-
ing variates. The model is used to dene a dimensionless alluvial state space characterized by aspect ratio (W/d), relative
roughness (D/d), and dimensionless shear stress (
*) or, equivalently, channel slope (S). Each alluvial state exhibits unique
values of Froude number and sediment concentration. The range of alluvial states for constant values of relative bank
strength (parameterized by an apparent friction angle,
) forms a single plane in the state space (W/d, D/d, * or S). The
scaling relations produced by the model are consistent with laboratory channels exhibiting a range of bank strengths, and
with the behaviour of natural channels. Copyright 2004 John Wiley & Sons, Ltd.
KEY WORDS: alluvial state; bank strength; channel pattern; optimization; river morphology
INTRODUCTION
A quantitative understanding of alluvial channel form and response to changes in governing conditions remains
an important yet elusive goal in uvial geomorphology. Arguably the nearest approach so far is embodied in
the various so-called regime theories constructed to predict the geometry of unlined irrigation canals (e.g.
Lacey, 1929; Blench, 1957; Simons and Albertson, 1960). These developments were essentially empirical, as
is the adaptation to describe river channels in the form of hydraulic geometry (Leopold and Maddock, 1953).
In this paper we present a rational regime approach to predict reach-averaged channel response to the conditions
governing alluvial channel form.
Rational regime theory formalizes the relations amongst governing equations traditionally applied to predict
the geometry of mobile-bed canals. The geometry of a canal is governed by three relations: (1) a bed material
transport equation; (2) a ow resistance equation; and (3) a bank stability criterion (Henderson, 1966, p. 451).
Since the geometry of an alluvial channel is related primarily to the transport and deposition of the sediment
found in it, the sediment transport equation must reect bed material load, rather than the total load. If the
discharge (Q), sediment transport rate (Q
b), and sediment calibre (D) are specied, a canal geometry comprising
gradient (S), hydraulic radius (R) and wetted perimeter (P) can be calculated. However, the bank stability
criterion does not impose a unique value since there is a range of canal geometries that satisfy the bank stability
constraint (Henderson, 1966).
Rivers are not subject to the same controlling conditions as canals. An important difference is the larger temporal
variation in both water and sediment supplies that most rivers experience. In alluvial rivers, moreover, channels are
self-formed and may migrate. The sediment and water supplied to the channel are assumed to be transported with
no net change in average channel form despite sediment exchange and active lateral migration. Valley gradient
(Sv ) remains constant at the timescale upon which channel geometry and alignment adjust and therefore represents
* Correspondence to: M. Church, Department of Geography, The University of British Columbia, Vancouver, British Columbia, V6T 1Z2,
Canada. E-mail: mchurch@geog.ubc.ca
512 B. C. EATON, M. CHURCH AND R. G. MILLAR
an important constraint on the system the largest gradient that the channel may adopt over any signicant distance.
However, as in the case of bank stability, S v merely establishes a limit to the range of possible solutions.
To close the regime description of alluvial streams, researchers have resorted to applying some sort of
optimality criterion to the solution curves generated by the three governing relations. Accordingly, extremal
hypotheses have been invoked, including minimum stream power (Chang, 1979), minimum unit stream power
Page 1 of 9 Rational regime model of alluvial channel morphology and response
24/07/2014 http://dc344.4shared.com/doc/gpwYSFSj/preview.html
(Yang, 1976), maximum sediment transport efciency (Kirkby, 1977) or capacity (White et al., 1982), and
maximum friction factor (Davies and Sutherland, 1983). Under appropriately dened circumstances, some of
these conditions turn out to be equivalent (White et al., 1982; Davies and Sutherland, 1983). In a variation on
this theme, Huang and Nanson (2000) employed the principle of least action to demonstrate the equivalence of
minimum stream power and maximum sediment transport capacity. A slightly different approach was presented
by Cao (1996), wherein a rational regime theory is based on the assumption that for a system at equilibrium
the probability of width adjustment is the same as that for depth adjustment.
In the following sections we: (1) formulate a rational regime model to predict the behaviour of alluvial
channels; (2) consider the validity of the assumptions implicit in the model; (3) explore a rationale for applying
the model to alluvial channels, emphasizing a physically based optimality criterion; (4) present a dimensionless
framework for describing the state of alluvial systems based on the optimized solutions from the model; and (5)
test the model against existing data from both natural and laboratory streams.
THE THEORETICAL MODEL
Our model is based on that presented by Millar and Quick (1993), which explicitly considers the role of bank
strength. It includes equations for evaluating ow resistance, sediment transport and bank stability. The model
assumes a trapezoidal cross-section with banks at an angle
from the horizontal (Figure 1), the dimensions of
which are iteratively varied to solve the equations subject to the appropriate constraints (see Millar and Quick,
1993). It is designed to describe single-thread channels only, and is less apt to describe wandering or braided
channels adequately, due to the geometry upon which the model is based.
In practice, the model is operated by specifying the formative discharge (Q), gradient (S), and characteristic
grain size (D). The bottom width (P bed ), water depth (Y o) and side slope (
) are then iteratively varied to produce
channels wherein continuity is maintained (Q
= uA, where u is the mean velocity and A is the area of the
trapezoidal cross-section) and the banks are critically stable. The stability of the banks is assessed by determining
the mean shear stress exerted on them and comparing that to the threshold for bank entrainment. Solutions can
be generated for a range of bottom widths but for a given value of P bed the solution is associated with unique
Figure 1. Denition diagrams for trapezoidal and triangular geometries considered in this paper illustrated with constant cross-sectional area
and wetted perimeter, and representing an aspect ratio of approximately 25 for bank strength
= 50: (A) four times vertical exaggeration;
(B) true scale
Figure 2. Solution curves for the regime model subject to the constraints Q
= 100 m
3
s
1
, S
= 0003. Bedload concentration predicted using
the Meyer-Peter and Muller (1948) equation for constant D 50 (32 mm) and variable
(40, 50 60 and 70) is plotted against (A) bed width
of the solution cross-section, and (B) aspect ratio (width/average depth). Bedload concentration for constant
(= 50) and variable D
50
(22 mm, 32 mm, 45 mm) is plotted against (C) bed width, and (D) aspect ratio
values of Y o and
. This yields a solution curve dening channels over a range of bottom widths and with a range
of sediment transport capacities. Details of the equations and computations are given in an appendix on the
Wiley InterScience website (http://www.interscience.wiley.com).
Examples of typical solution curves are presented in Figure 2. The curves in Figure 2A and B were generated
by holding Q, S and D constant and systematically varying P bed for four different values of bank strength (
).
Prior investigators have made similar calculations outside the context of a supporting theory (e.g. Pickup and
Warner, 1984). When bed material sediment concentration is plotted against W/d ratio (Figure 2B) all four
curves coincide for wide, shallow channels, since the relative importance of the banks becomes small, but the
maximum of the function shifts towards a smaller aspect ratio with higher bed load concentration as the bank
strength increases. For
= 40, the curve changes direction abruptly at the peak sediment concentration and
never reaches the lower range of W/d. This occurs because the bank angle (
) necessary for maintaining bank
stability (for a given value of P bed ) decreases rapidly as the bottom width of the trapezoid becomes small. As
a result, the total channel width {P bed + 2(Y o /tan
} actually increases as P
bed is reduced. The assumption that the
bed is mobile but the banks are stable breaks down as
becomes small and the distinction between bed and bank
becomes blurred.
Figure 3. Solution curve for regime model subject to the constraints Q
= 100 m
3
s
1
, Q b
= 50 kg s
1
, and D 50
= 32 mm. The slope necessary
to produce an equilibrium channel with the specied Q and Q b for the given D 50 is plotted against aspect ratio (W/d)
Sediment concentration is plotted against bed width (Figure 2C) and aspect ratio (Figure 2D) for three
characteristic grain sizes, bank strength being held constant (
= 50). Q and S are constant, as before. The
general form of the curves is the same, but peak sediment concentration increases and shifts toward larger aspect
ratios as grain size is reduced. For a reduction in grain size from 45 mm to 22 mm, the aspect ratio associated
with the peak sediment concentration approximately doubles, while the peak sediment concentration nearly
triples. As grain size decreases, the lower limb of the solution curve ips, and W/d increases as P bed decreases,
just as it does for the weakest bank strength (40) in Figure 2B.
An alternative way to explore solutions is to vary the channel gradient, holding the sediment transport rate,
Qb , constant. This situation (Q, Q b and D imposed) approximates the circumstances that we expect in nature.
Figure 3 presents examples of this type of solution curve for varying between 40 and 70. No solutions were
found for aspect ratios much smaller than that associated with the minimum gradient, and hence the different
solution curves do not extend over the same range of W/d ratios. As for the solution curves in Figure 2B, these
curves converge with increasing aspect ratio and bank strength limits the minimum W/d that can be attained.
The range of W/d ratios over which rational regime solutions can be found is larger than that usually exhibited
in nature (Schumm, 1963; Pickup and Warner, 1984). Channels in nature seem preferentially to occupy homologous
points on their respective stability curves. This aspect of uvial systems remains poorly understood. Extremal
hypotheses select an unambiguous set of homologous points the extreme values. Finding a physical explana-
tion for this perceived behaviour of alluvial systems remains an important problem in uvial geomorphology
(Davies, 1987).
BASIS FOR APPLICATION TO ALLUVIAL SYSTEMS
In the following we attempt to identify a physical basis for applying rational regime models to characterize
alluvial streams, and explore some evident limitations of this approach to understand alluvial systems.
Time scale
Regime models are equilibrium models, which immediately requires that some time scale be established for
the emergence of the equilibrium state. This time scale must be substantially shorter than the scales of climate
shifts and topographic changes, and longer than that for individual ow events. Unfortunately, the remaining,
intermediate time scale (10
1
to 10
3
a) is likely to be that at which geomorphic history and land use changes
become evident. It is not clear, then, what the appropriate time scale is or even if it exists in nature. If regime
models can be shown to work under conditions that are known to be at equilibrium (e.g. in experimental studies),
then deviations from the predictions might be attributed to variations in Q and Q b, provided the model is properly
parameterized for alluvial streams. But a regime model remains, at best, a severely reduced model of alluvial
river behaviour.
Formative discharge
Regime models are predicated on the assumption that a single discharge and associated equilibrium sediment
transport rate are sufcient to predict the channel form of active stream channels, which may be approximately
true when discharge uctuations are small, but is unlikely to be generally applicable to alluvial streams (Pickup
and Rieger, 1979). Furthermore, the functional relation between discharge and sediment transport capacity
depends not just on the magnitude of sediment supply, but also on its calibre. In gravel-bed channels, the
relations amongst ow, sediment supply and sediment calibre inuence the texture (Dietrich et al., 1989) and
structure (Church et al., 1998) of the streambed in ways that may affect the future capacity of the channel to
adjust, but this does not preclude the application of regime models that admit a range of solution states for some
specied governing conditions.
Bank stability
Self-formed canals are thought to assume a form for which the average shear stress exerted on the bank is
equal to the critical shear stress for bank erosion (Lane, 1957). For laterally active streams, this cannot be the
case. It can be argued that, if the average channel width does not change over time, then the banks are
statistically stable insofar as the mean shear stress (averaged over a suitable length of channel bank) is equal
to the critical value for entrainment. While bank erosion must occur and the channel must widen whenever the
shear stress exerted on the banks exceeds the critical value, we suppose that this erosion must be compensated
by deposition along adjacent banks where the effect of widening is to reduce near-bank stress below the critical
value for the calibre of sediment being transported. These conditions plainly involve details of the velocity eld
within the channel that are not articulated in the theory.
Channel cross-sectional shape
Rational regime model predictions can be expected to resemble an alluvial channel most closely at or near
the inection point between successive point bars because, there, the cross-section is relatively symmetrical, and
most closely represents the trapezoidal shape employed in the model. The question remains how the mean hydraulic
variables for such trapezoidal sections relate to other, asymmetrical cross-sections. It has been observed that
mean velocity is nearly constant for formative discharges in natural streams (Carling, 1991; Clifford and Richards,
1992). Accordingly, we have estimated the effect of channel shape on the channel width and mean hydraulic
depth for trapezoidal channels with a range of aspect ratios, and for the corresponding triangular channels (meant
to describe the asymmetric cross-sections characteristic of bends: Figure 1) that have the same area (A) and
perimeter (P). The cut bank in a triangular section is steeper than the stable banks in the trapezoidal section,
having an angle ( ) equal to or greater than angle of repose () for the bank material (Figure 1), and the apparent
friction angle (
) is selected as a representative value ( ). The angle representing the accreting surface ()
and the maximum depth of the triangular section (H) are then uniquely specied by the constraints of equal area
(A) and perimeter (P), which implicitly satisfy the constraint of continuity and produce the same average velocity.
For trapezoidal channels with
= 50 and aspect ratios ranging from 10 to 100, the ratio of the triangular
top width to the trapezoidal top width (W r ) varies from 093 to 099 (with increasing W/d), while the ratio of
the mean hydraulic depths (d r) varies from 107 to 100 (subscript r indicates the ratio of corresponding values).
The differences between the average parameters for the different cross-sectional shapes are small enough that
one can assume that they are equivalent. We conclude that our trapezoidal model represents the average cross-
sectional hydraulic parameters for plausible channel shapes.
Formulation of bed material transport
The transport of bed material in rivers is largely associated with bedload, almost entirely so in gravel-bed
channels. Bedload transport exhibits both spatial and temporal non-equilibrium behaviour. Transport is seldom
uniformly distributed across the channel (Carson and Grifths, 1987), and the locus of transport does not
necessarily reect the uid force distribution (Muhlhofer, 1933). Even with constant Q in a recirculating ume,
channel morphology interacts with the sediment transport eld to produce lateral and longitudinal variations in
sediment transport (Ashmore, 1991). These variations are in fact responsible for the characteristically non-
regular morphology of alluvial stream channels and progressive changes in stream position. The development
of spatial disequilibrium in sediment transport results in temporal variation at a single point, so the distinction
between spatial and temporal disequilibrium is not always clear.
Bed material/ bedload transport formulae express the supposed relation between uid force exerted on
the channel boundary and the capacity of the channel to transport sediment. However, we recognize that the
sediment transport rates predicted by formula may differ substantially from the actual rates in alluvial streams.
Some of the departures may be the consequence of local variations in sediment supply that are eliminated if
sediment supply is held constant in an experimental setting. But the bed surface condition also exerts an
important inuence on local transport (Dietrich et al., 1989; Church et al., 1998; Bufngton and Montgomery,
1999). Additionally, when transport formulae are applied using average hydraulic quantities, the effect of
channel geometry on the actual sediment transport capacity is not incorporated, which may result in signicant
underestimation of Q
b (Ferguson, 2003). It is therefore not clear how useful the existing sediment transport
formulae are in describing alluvial systems. However, the choice of sediment transport equation in a regime
model does not strongly inuence the predicted channel dimensions (Millar and Quick, 1993), so perhaps these
issues are not critical, especially if Q b is interpreted as a scale representation of the equilibrium transport
conditions (Church, 1985), rather than a true estimate of the sediment transport.
Formulation of ow resistance
Flow resistance equations used heretofore in regime models rely on some measure of the roughness length
associated with grain diameter, rather than channel morphology and pattern. They are not generally capable of
predicting the ow resistance associated with channel sinuosity, bars and other bedforms present in alluvial
channels, which may account for up to 90 per cent of the total ow resistance at bankfull conditions (Millar,
1999), and may play an important role in pattern development (Lewin and Brewer, 2001). While we can
understand the various components of ow resistance conceptually within an alluvial system (see Optimization
section), it is not yet possible to predict the magnitude of each component, nor the total ow resistance for a
given river system. Thus, regime models may be able to describe channels that exist and for which the resistance
is known, but may not be able to predict the channel form under different conditions.
OPTIMIZATION
As things stand, then, the regime approach may describe reach-averaged dimensions of laterally active streams
for which discharge and sediment supply are relatively constant, provided that some sort of expected condition
can be specied for sediment transport capacity and/or ow resistance. Typically, this requirement has taken the
form of some sort of optimization, such as minimum gradient, maximum transport capacity or maximum ow
resistance. The optimization is needed to choose a unique solution from the range of possible solutions for
specied Q, S and D (Figure 2) or for specied values of Q, Q
b and D (Figure 3).
The behaviour of a single-thread alluvial channel can be understood physically in terms of a tendency to
maximize ow resistance. The alluvial system must be considered as a whole, wherein the independent variables
are Q, D, Qb and Sv (Figure 4); W, d and S (hence channel length, L c) are free to vary. The key consideration
is how to quantify the ow resistance of the system. Equations that index ow resistance per unit length of
channel (e.g. f
= (8gRS)/u
2
) do not describe the ow resistance per unit length of the system. Once channel
length is allowed to vary, the two are not equivalent. The total resistance to ow in the system can be indexed
by substituting S v for S in a ow resistance equation. Thus:
f
gRS
u
sys
v
=
8
2
(1)
This transforms the friction parameter from a property of the stream channel ( f ) to a property of the all-
uvial system ( f sys ) that incorporates the effects of channel planform as well as grain size and bedforms of
Figure 4. Denition of the uvial system. (A) Denition of the properties of the uvial system that are constant at regime time scales: valley
gradient (S v ), valley length (L v), and total head drop (z 1
z
2 ); as well as those that are variable: channel slope (S) and channel length (L c ).
(B) Illustration, following Figure 3, that for a constant Q, Q b , D50 and
, a solution curve exists, for which a continuum of theoretically stable
channels exist for slopes less than the valley slope (S v ) and greater than the minimum slope (S min ) capable of transporting the sediment load
intermediate size (such as dunes and bars). These features point to three components of ow resistance for the
system:
fsys
= f + f + f
(2)
where f
is the grain resistance assuming a roughness length equal to the particle diametor (after Millar,
1999); f
is the within-channel form resistance due to bars, dunes and other in-channel features ( f f f
according to Parker and Peterson, 1980); and f
is the reach-scale form resistance due to channel sinuosity
( f
f
sys
f ). Ackers and Charlton (1970) applied a similar division of roughness to meandering labora-
tory channels with thalweg sinuosities of up to 176, concluding that some 60% of the head loss in meander-
ing channels [is attributable to] bends and major variations in cross sections (Ackers and Charlton, 1970,
p. 369).
According to regime theory, stable alluvial channels can be formed at all gradients S min
S S
v (provided
Sv
> S
min , and S v is not too large to prevent alluvial material from accumulating at all), where S min is the minimum
gradient at which the imposed uxes Q and Q b can be transported (Figure 4). For less sinuous channels (see (1)
in Figure 4A), most of the ow resistance for the system derives from grain roughness and from within-channel
bedform roughness such as bars, dunes and clast structures. As sinuosity increases, so does the relative impor-
tance of reach-scale form resistance (see (2) and (3) in Figure 4A).
Figure 5. Friction factors for the channel ( f ) and for the alluvial system ( f sys , arbitrarily assuming S v
= 00035) are plotted against slope,
assuming the same values of Q, Q b and D as in Figure 3, and setting
= 50. The minimum slope is indicated by a vertical dashed line
The friction factor, f, calculated for the channel and for the system ( f sys ) is presented in Figure 5. While f based
on the channel gradient increases monotonically with increasing slope, f sys steadily declines from a maximum
value at the minimum gradient. It appears that the ow resistance per unit length of channel is greatest for the
channels with the least sinuosity (i.e. the highest gradient), assuming that C 50 (which represents both f
and f ;
see the appendix on Wiley InterScience for denitions) does not vary signicantly as sinuosity increases. On
the other hand, the friction factor for the system reaches a maximum for the lowest stream gradient, which
corresponds to the maximum sinuosity.
We now have a framework within which to describe adjustments of alluvial systems toward a condition
of maximum ow resistance. If an alluvial channel is located somewhere on the relevant stability curve but
the ow resistance for the system is not at a maximum (e.g. channels (1) and (2) in Figure 4), then the ow
possesses kinetic energy that can act to further deform the system until it happens by chance to achieve a form
that has a higher resistance to ow and is therefore more stable. Since the potential to deform the system (and
hence reach another conguration) is least when this excess kinetic energy for the system is least (subject
to the constraint that both the imposed uid and sediment supply be transported), a state of maximum
ow resistance is the most statistically probable. This aspect of system behaviour has been ascribed by
Huang and Nanson (2000) to a principle of least action, whereby equilibrium is achieved by minimizing
the potential for further deformation of the system. Extremal hypotheses appear to function simply as for-
malisms that permit onedimensional regime models to describe the outcome of threedimensional geomorphic
processes.
PROPERTIES OF THE RATIONAL REGIME MODEL
We can now specify the model using a number of specic extremal hypotheses, since they are closely related
(Davies and Sutherland, 1983; Davies, 1987; Huang and Nanson, 2000). Given equations for ow continuity,
ow resistance, sediment transport capacity and bank strength, and independent variates Q, S v , Qb , D and
, we
may predict S, W, d and
. In diagnostic mode, it is computationally inconvenient to work with specied Q and
Qb while varying S, since S is generally known for alluvial streams in nature whereas Q b is not. Additionally,
Qb as predicted by the model is simply a scale for the actual sediment transport in nature (Church, 1985; Millar
and Quick, 1993), and thus cannot strictly be related to transport rates measured in the eld. Therefore,
our model is operated diagnostically by specifying Q and S, and optimizing on some other parameter such
as Qb or
.
Table I. Simulated dataset based on random pairs of Q and D 50
Q (m
3
s
1
) D50 (mm) Slope Q (m
3
s
1
) D50 (mm) Slope
155 113 753
10
3
374 560 841
10
3
181 298 225
10
3
383 444 233
10
3
333 141 498
10
3
408 308 984
10
3
417 987 619
10
3
427 519 619
10
3
464 321 143
10
2
467 919 295
10
3
544 957 116
10
2
467 975 654
10
3
650 279 499
10
3
488 423 381
10
3
753 798 218
10
2
512 538 246
10
3
102 191 402
10
3
554 409 910
10
4
2 3
139 877 145 10 597 637 415 10
165 129 157
10
3
705 148 239
10
3
199 167 733
10
4
750 370 225
10
3
220 248 298
10
3
776 416 551
10
4
231 137 257
10
3
807 978 336
10
3
246 225 310
10
3
817 735 660
10
3
257 992 269
10
3
863 467 687
10
4
286 115 208
10
3
885 896 588
10
3
301 520 735
10
3
899 909 752
10
3
305 484 349
10
3
910 420 642
10
4
344 357 298
10
3
952 331 117
10
3
352 775 478
10
3
959 963 868
10
3
356 435 320
10
3
973 835 618
10
3
358 598 820
10
3
976 374 722
10
4
359 158 227
10
3
986 436 244
10
3
372 422 369
10
3
991 826 293
10
3
A random number generator with a uniform distribution was used to select Q(10
1000 m
3
s
1
), D(10 100 mm) and
*(004 to 01). * was used to avoid generating channels
in which the bed material could not be transported. Values of S were generated for the lowest
bank strength (
= 40) that produced the specied * values. The random pairs are pre-
sented in order of increasing Q, and the minimum and maximum values of D 50 and S for the
dataset are highlighted in bold. The minimum values of Q and D 50 correspond entirely by
chance, since all values were produced by a random number generator.
The model then yields predictions of alluvial channel dimensions for various imposed ows and grain
sizes. These predictions can be collapsed onto dimensionless relations among the key variables describing
the state of the alluvial system for specied bank strengths (
). represents a fundamental boundary condition
that controls the development of the system. All channels with the same
are described by a single function
relating
* and W/d. Scaling based on these two dimensionless parameters is consistent with previous argu-
ments about the controls of sediment transport, channel deformation and channel pattern (Ferguson, 1987;
Todd, 1996; Dade and Friend, 1998; Dade, 2000). Once D/d is known (in addition to
* and W/d), the alluvial
state is fully specied, exhibiting unique values of Froude number (Fr), gradient (S) and sediment concentration
(Qb /Q).
A complete description of an alluvial state by the regime model is thus summarized by three dimensionless
quantities: (1) average dimensionless shear stress (
*) or, equivalently, channel gradient (S); (2) channel shape
(W/d); and (3) relative roughness (D/d). Implicit in this description of the alluvial state are constant values for
C50 and
*
c, which are considered to be additional secondary descriptors of the alluvial state.
The state-space (W/d, D/d,
* or S) denes the scaling behaviour of alluvial channels. We investigate the
scaling behaviour using random combinations of Q, D (which together specify the relative roughness), and
S. A dataset of 50 such randomly generated alluvial channel specications (Table I) is used (1) to illustrate the
characteristics of the alluvial scaling implied by the rational regime model; (2) to examine the sensitivity of
the predicted scaling behaviour to the particular optimizations used; (3) to determine the effect of varying the
channel-scale form roughness, C
50 ; and (4) to establish the limits of applicability of the model.
Figure 6. (A) Dimensionless scale relations between
* and W/d for various bank strength values (represented by different values, in
degrees). (B) Dimensionless scale relations between transport rate G (Parker, 1990) and channel shape W/d. The curves were generated using
a regime model (that maximizes shear stress) with the dataset in Table 1. Supplemental data (open symbols) were added for
= 60, 70
and 80 so as to extend the range of W/d up to 150 to conrm the general structure of the scaling relations
Table II. Power relations t to dimensionless scale collapse*
(deg.)
a b n R
2
40 358
10
5
298 50 0998
50 146
10
5
312 50 0998
60 403
10
4
311 49 0997
70 948
10
3
316 43 0996
80 899
10
2
319 24 0998
The relations were t to data where W/d
> 7 only.
* The tted model is W/d
= a( *)
b
Supplemental data were added for the relations for 60, 70 and
80 to extend the range of W/d up to 150 (see Figure 6A).
Dimensionless scaling relations
We obtain most of our predictions by maximizing the shear stress,
, for a given Q and S. This approach is
consistent with the implications of adopting an extremal hypothesis. By setting
equal to 40, 50, 60, 70
and 80, ve curves relating
* and W/d were generated using the dataset (Figure 6A). For W/d greater than
about 7, all the relations are nearly log-log linear and almost parallel. They are well described by power
functions, wherein the coefcient for the relations decreases with increasing bank strength (Table II) such that
stronger banks result in smaller W/d ratios for the same dimensionless shear stress. For W/d less than 7, there
is a pronounced curvature to the functions which arises from the rapidly changing relative importance of the
bank height in the denition of the hydraulic radius, and thus in the shear stress.
These functions capture much of the system behaviour, in dimensionless form. This is particularly evident
when
* is replaced by Parkers (1990) dimensionless sediment transport rate, G (Figure 6B). The same general
structure persists but there is a distortion along the x axis owing to the non-linear nature of the transport relation.
The relations in Figure 6A and 6B may thus be interpreted as a dimensionless relation between the sediment

VCB

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

VCB

Transféré par

Droits d'auteur :

Formats disponibles

MODELLING ALLUVIAL CHANNEL FORM 511

Vous aimerez peut-être aussi