1656

Intensity and characteristic length of braided

channel patterns1
Peter Ashmore

Abstract: Recent research on braided river morphology shows that the intensity (number of anabranches) of braiding
channels increases with dimensionless discharge and (or) stream power. This variation in intensity reflects the adjustment
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

of total sinuosity of the river to imposed gradient at a given discharge and grain size. Only a subset of channels is active
at a given time and this active braiding intensity reflects the limited number of channels that can sustain bed load transport
as the flow is divided. This is governed mechanistically by the dynamics of bifurcations and avulsions. Braided channel
networks also have a characteristic length scale (or scales) related to the wavelength of the bars from which braiding de-
velops and to the scale of the bars and confluence–bifurcation units within the braided network. The range of scales is lim-
ited by the size (and, therefore, number) of the active channels within the network and the width of the entire river.
Key words: river regime, braiding, fluvial geomorphology, river mechanics, sediment transport, bars, bifurcation, channel
pattern.
Résumé : Les recherches récentes sur la morphologie des rivières anastomosées montrent que l’intensité (le nombre de
bras de la rivière anastomosée) des petits chenaux entrelacés augmente avec la décharge adimensionnelle et/ou la force du
ruisseau. Cette variation d’intensité reflète l’ajustement de la sinuosité totale de la rivière à un gradient imposé pour une
décharge et une granulométrie données. Seul un sous-ensemble de canaux est actif à tout moment donné et cette intensité
de chenaux entrelacés actifs reflète le nombre limité de canaux qui peuvent soutenir le transport de la charge de fond
lorsque l’écoulement est divisé. Ce nombre limité est régi de manière mécaniste par la dynamique des embranchements et
For personal use only.

des avulsions. Les réseaux de canaux anastomosés possèdent aussi une ou plusieurs échelles de longueur caractéristiques
reliées à la longueur d’onde des barres à partir desquelles se développe le réseau anastomosé ainsi qu’à l’échelle des barres
et au nombre de confluences-embranchements dans le réseau de canaux anastomosés. La plage d’échelles est limitée par la
dimension (et, donc, le nombre) des canaux actifs dans le réseau et par la largeur entière de la rivière.
Mots-clés : régime d’écoulement, développement d’un réseau anastomosé, géomorphologie fluviale, mécanique des riviè-
res, transport de sédiments, barres, embranchements, forme de canaux.
[Traduit par la Rédaction]

1. Introduction mon in mountain and piedmont environments, alluvial

plains, and glaciated regions with high rates of supply of
Braided rivers are a distinctive alluvial river morphology
coarse-grained sediment to the river system. For engineering
characterized by multiple intersecting channels separated by
and applied geomorphology, an understanding of braided
ephemeral (and usually unvegetated) exposed bars (Fig. 1).
river morphology and dynamics is needed to effectively an-
Braiding is a ‘‘self-induced form’’ (Yalin 1992) and an ac-
ticipate the consequences of human interference (e.g., flow
tive, ongoing set of processes in the river that is a response,
regime manipulation, river training, encroachment on the
ultimately, to the prevailing discharge, valley gradient, and
channel zone, and gravel extraction) and response to envi-
sediment flux (quantity and particle size). The channel con-
ronmental change (Galay et al. 1998). Increasingly, stream
figuration is extremely unstable and characterized by rapid
channel restoration and natural channel design, and the im-
and abrupt morphological changes, channel avulsion, and
plications of channel pattern change for fluvial habitat and
bar formation or destruction during periods of high flow.
river ecology, require understanding of fluvial morphody-
Braiding occurs in noncohesive sand and gravel and is com-
namics, including that of braided rivers.
Much analysis of braided channel patterns has focused on
Received 7 January 2009. Revision accepted 27 May 2009. the threshold conditions for the onset of braiding in compar-
Published on the NRC Research Press Web site at cjce.nrc.ca on ison to that for single channel rivers (Ferguson 1987) and on
5 November 2009. the fundamental unit processes and morphology of braided
P. Ashmore. Department of Geography, University of Western rivers (e.g., channel confluence dynamics, bar formation
Ontario, London, ON N6A 5C2, Canada and accretion, development of bifurcations) (see Ferguson
(email:[email protected]). 1993). Much less attention has been paid to the larger scale
Written discussion of this article is welcomed and will be characteristics and development of the braided channel net-
received by the Editor until 28 February 2010. work as a regime property of braided rivers and the relation
of this larger scale morphology to the unit channel and bar
1This paper is one of a selection of papers in this Special Issue processes (Ashmore 2001). This paper synthesizes and de-
in honour of Professor M. Selim Yalin (1925–2007). velops ideas and recent analyses of two aspects of gravel-

Can. J. Civ. Eng. 36: 1656–1666 (2009) doi:10.1139/L09-088 Published by NRC Research Press
 Ashmore 1657

Fig. 1. Gravel-bed braided river (Sunwapta River, Alberta). Flow is right to left and the river is approximately 100 m wide. Note the reg-
ularity in spacing of channel confluence bifurcations in the main channel closer to the camera.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

bed braided rivers – the intensity (complexity) of braiding of these indices in a range of experimental conditions using
patterns and the characteristic length scale of braided chan- physical models, and on braided rivers in nature, shows that
nel patterns – and shows the connection between them that braided rivers vary considerably in their braiding intensity
reveals new, and fundamental, aspects of the morphodynam- and therefore there is a continuum of braided river pattern
ics of braided channels. morphology, as there is of river pattern morphology in gen-
eral (Ferguson 1987). Many of these indices are actually
For personal use only.

2. Limits to braiding closely equivalent (Egozi and Ashmore 2008) and this paper
mostly refers generically to braiding intensity or braiding in-
2.1 Regime concepts of braiding intensity dex (see Table 1 for a summary of terminology).
River channels have a predictable morphology that is in The conditions for the development of braided channels
equilibrium with the prevailing water discharge, sediment have been subject to considerable analysis over the past
supply, and topographic (valley) gradient (see Parker et al. 50 years or more, using a variety of observational, experi-
2007 for a recent analysis). In fluvial geomorphology, these mental, and theoretical approaches (Bridge 1993). Explana-
equilibrium relationships are usually called ‘‘hydraulic ge- tions have tended to focus on channel or bar-scale
ometry’’, whereas in engineering this adjustment has usually morphodynamics or on larger scale external controls on the
been referred to as ‘‘river regime’’ (Blench 1969). The es- river regime. At channel-scale there is general agreement
tablished regime relationships for natural channels that pre- that one important control is the width-depth ratio (aspect
dict, for example, equilibrium channel width for the ratio) of the river cross section (where width is the bank-to-
imposed water discharge and bed sediment size, have been bank wetted width of the initial single channel), for which a
developed almost exclusively for single channel (often larger value tends to lead to a greater tendency for braiding.
meandered) rivers. Braided rivers, with their multiple chan- In channels with significant bedload (sand or gravel), prom-
nels and extensive lateral migration, give the superficial ap- inent and regularly spaced depositional bars form with
pearance of being free of these equilibrium regime sharp, downstream margins and associated upstream scour
constraints. In fact, braided rivers have limits to their overall pools from where the bar material is eroded (Fig. 2). These
dimensions, such as channel width (Ashmore 2001), and are sometimes referred to as pool-bar units. The number of
also the complexity of channel pattern, in the same manner bars across the channels is referred to as the mode (Fig. 2).
as single channel rivers. What is the nature of this regime Some have argued that braiding occurs in wide channels be-
adjustment of braiding channel patterns and how does it oc- cause rather than a single row of (alternate) bars (mode 1)
cur? formed in narrow channels, two (mode 2) or more rows oc-
The complexity of a braided channel network is usually cur that develop into bars in the centre of the channel
referred to as ‘‘braiding intensity’’. A number of different around which the braided channel develops (Yalin 1992)
measurement indices of braiding intensity have been used in (Fig. 2). Others propose that braiding can develop by the
fluvial geomorphology in the past 50 years. Most of these widening of an initial alternate bar channel, which allows
are some variation of either the average number of channels chute cutoff of the initial bends, from which braiding propa-
across the braided river along a defined river length gates (Ashmore 1991b; Ferguson 1993). Widening is then
(channel count indices) or the total length of channels per influenced by other channel characteristics, especially bank
unit river length (total sinuosity indices) (Egozi and Ash- resistance (Ferguson 1987), with the consequence that braid-
more 2008). Their purpose is to provide a simple, relative ing may be the ‘default’ channel pattern for alluvial rivers
measure of the extent or complexity of braiding for inter- unless there is a restriction on channel widening due to
comparison of rivers and for analysis and explanation of bank resistance related to bank material internal strength
spatial or temporal changes in channel pattern. Measurement and riparian vegetation (Murray and Paola 1994). The bar-

Published by NRC Research Press

 1658 Can. J. Civ. Eng. Vol. 36, 2009

Table 1. Common terminology used in describing braided river morphology.

Braiding intensity Generic term of the complexity of the braided channel network
Braiding index A defined measurement of braiding intensity. Two are commonly used: (1) The
average number of channels across transects a river reach (2) The total length of
channels per unit river length (sometimes called ‘‘total sinuosity’’).
Active braiding intensity Complexity and associated measurement of only those channels in the braided river
Active braiding index that are transporting bed load and involved in morphological development of the
river bed. (Inactive channels have flow, but no bedload).
Characteristic length Generic term for the dominant length scale of channel and planform morphology.
For example, wavelength of bars or along-channel spacing of confluences and bi-
furcations in a braided river
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

Fig. 2. Bar unit morphology, wavelength, and width (B) in initial scale origins of braiding, and the consequences for channel
straight channels of low (alternate bar) and high (central bar) width pattern morphology, are discussed further in the second part
to depth ratio. Based on diagrams by Ikeda (1984) and Yalin of this paper.
(1992). Explanations of braiding based on regime analysis focus
on the influence of channel (or valley) gradient or upstream
sediment flux (which, it has been argued, is ultimately the
control on channel gradient at a given discharge (Paola
2001)). Early empirical evidence (Leopold and Wolman
1957) pointed to the generally higher gradient of braided
streams for a given discharge and to the effect of valley gra-
dient on the sinuosity of river channels in general (Richards
1982). One approach to the explanation of this effect of gra-
dient on channel pattern is to consider meandering and
For personal use only.

braiding as different modes of adjustment to the imposed

valley gradient (Bettess and White 1983; Ferguson 1987;
Yalin 1992). Assuming that both meandering and braiding
channels begin with a straight channel with gradient in ex-
cess of the (conceptual) stable regime state, Yalin (1992) ar-
gued that meandering is characterized by a reduction of
channel gradient (via increased sinuosity of one channel) to
reach the regime slope, whereas braiding is characterized by
progressive channel subdivision, which effectively increases
the regime channel slope (because regime gradient is inver-
sely related to channel discharge and the discharge of each
anabranch is less than that of the initial undivided channel)
until it is equal to the initial valley gradient. The conse-
quence is that braided rivers have steeper slopes than mean-
ders for the same discharge and bed material grain size. It
should be pointed out that braided anabranches have some
(usually low) sinuosity, so that at least part of the braiding
adjustment is also via slope reduction. Bettess and White
(1983) suggest that braiding is likely when, initially, valley
gradient is at least two times the regime slope and that
braiding may only be stable in the longer term for valley
gradient approximately three times the regime channel
slope. Yalin (1992) suggests instead that the ‘choice’ of
braiding over meandering goes back to channel geometry
(initial channel width or width:depth ratio) and the occur-
rence of single row (meandering) or multiple row (braiding)
bars, which is, in turn, partly a function of initial channel
gradient, as well as other factors.
The implication of these slope-based analyses is that the
braiding intensity (total sinuosity) needed to reduce the ex-
cess slope to zero (i.e., for regime channel gradient to equal
valley gradient) then depends on the initial magnitude of the
excess gradient or excess stream power (the product of dis-
charge, channel gradient, and specific weight of water).
Greater excess power (gradient) would be expected to yield

Published by NRC Research Press

 Ashmore 1659

greater braiding intensity and thus gradient controls the ex- overall channel width), at a given slope and discharge, until
tent to which channels undergo the consecutive bifurcation reaching a stable average value around which instantaneous
which ‘‘perpetuates itself until the equilibrium or regime braiding intensity fluctuates. This ‘equilibrium’ braiding in-
state is reached’’ (Yalin and da Silva 2001). Thus, the inten- tensity increases with stream power or discharge for a given
sity of braiding, as well as the occurrence of braiding, is a grain size. A step increase in discharge, at constant flume
regime adjustment of the river. (valley) gradient in an established stable braided pattern,
causes an equivalent adjustment and increase in braiding in-
2.2 Empirical evidence for braiding intensity regime tensity (Egozi 2006; Egozi and Ashmore 2009). However,
adjustment new channels are added slowly over time so that there is a
Direct testing of the ‘excess gradient’ hypothesis for fairly long adjustment period during which the channel net-
braiding discussed above is frustrated by the empirical scat- work expands to accommodate the increased discharge.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

ter and range of theoretical solutions of regime gradient These additional channels are created by successive avul-
(Ferguson 1987). However, there is general empirical sup- sions and by occasional bifurcations of existing channels.
port for the idea that steeper valley gradient (or flume gra- Bertoldi et al. (2009b) propose a dimensionless discharge
dient, in the case of experimental analysis) increases and dimensionless stream power as a means of comparing
braiding intensity. For example, Robertson-Rintoul and Ri- data from experiments (and field data) with a range of abso-
chards (1993) analyzed gravel bed braided and meandering lute discharge, channel gradient, and grain size.
rivers in terms of total sinuosity — the length of all chan- ðQSv Þ
nels per unit valley length — within a given valley length. ½2 U ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Channel pattern and valley gradient were measured directly L ðgDd 3 Þ
from maps of the rivers and data on flow and grain size
were extracted from a catalog of alluvial channel data Q
½3 Q ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(Church et al. 1981). Data cover a range of river discharge L ðgDd 3 Þ
from 2 to 5500 m3s–1. The results show that total sinuosity
(P) increases with increasing stream power index (product where U* is the dimensionless stream power index, Q* is
of Q, the mean annual flood, and Sv, the valley gradient) the dimensionless total discharge, L is Q2/5/g1/5, Q is the to-
For personal use only.

with a secondary (and inverse) effect of bed material grain tal discharge, Sv is the valley or flume gradient, g is the ac-
size (D84): celeration due to gravity, D is the submerged relative
density of sediment, 1.63, d is the median grain diameter of
pffiffiffiffiffiffiffiffiffiffiffi
½1 P ¼ 1 þ 5:52ðQSv Þ0:4 D84 0:14 the bed material, and gDd 3 is the Einstein scale for ana-
lysis of particle mobility.) The dimensionless length, L, was
The generality of the predictions is unknown, but the rela- first proposed by Bray (1982) for analysis of cross section
tionship demonstrates some important points. There is a dimensions (width and depth) of single channels and has
clear, continuous trend of increasing total sinuosity (braiding been used recently by Parker et al. (2007) for this purpose.
intensity) with increasing stream power index and that braid- Bertoldi et al. (2009b) adopted it for analysis of braided
ing intensity increases with increasing stream energy above river morphology. Channel width is often used to scale dis-
the braiding threshold, with observed maximum total sinuos- charge and stream power by river size for analysis of chan-
ity of about eight (Robertson-Rintoul and Richards 1993). nel morphology and sediment flux, but in the case of
As sinuosity increases, for medium-coarse gravel, the chan- braided rivers it is not clear which of several possible defi-
nel pattern switches to braided (from single channel) at nitions of width should be adopted and can be measured
QSv values above about 0.1 (with some variation due to consistently. Consequently Bertoldi et al. (2009b) used L as
particle size), and total sinuosity greater than two. This is a length scale instead of width for both the response vari-
in accordance with the ‘excess gradient’ model for the oc- ables (channel morphology) and external control of the mor-
currence of braiding and the suggestion of Bettess and phology.
White (1983) that braiding occurs at excess valley gradient Braiding intensity appears to correlate better with Q* than
(i.e., channel sinuosity) greater than two. However, there is with U* (Bertoldi et al. 2009b). In other words, the regime
overlap between single channels and braids over a range of adjustment is primarily related to channel capacity to con-
QSv values (meandering ceases to occur at QSv above vey the imposed discharge at the given gradient and bed
about 0.5). roughness. This is consistent with the observations that,
Experimental studies from physical models also confirm when formative discharge is stepped upward, flow first
the idea that braiding intensity is a regime property of drowns the pre-existing channel network and then the net-
braided rivers. It is notable that, as Murray and Paola work gradually accommodates the increased flow by eroding
(1994) observed, the channel pattern adjustment in noncohe- new channels and (possibly) enlarging some existing ones
sive sand in laboratory experiments is always by braiding (Egozi 2006).
once sinuosity exceeds a very low value (approximately When compared with field data from several gravel-bed
1.1). This suggests that, in the absence of other constraints braided rivers in New Zealand (Mosley 1981) (some of the
(e.g., high bank resistance), braiding is the ‘default’ channel same data used by Robertson-Rintoul and Richards 1993),
pattern adjustment. Recent studies (Egozi 2006; Bertoldi et the field and experimental data do not quite collapse onto
al. 2009b; Egozi and Ashmore 2009) show that, beginning the same overall trend (Fig. 3), although both field and lab-
from an initial single, straight channel, braiding intensity oratory data cover a very similar range of braiding intensity.
gradually increases over time (accompanied by increasing The reason for the offset of the laboratory and field data is

Published by NRC Research Press

 1660 Can. J. Civ. Eng. Vol. 36, 2009

Fig. 3. Braiding intensity as a function of dimensionless discharge sible number of successive channel subdivisions during the
(eq. [2] for physical model experiments (Bertoldi et al. 2009b; development of braiding if bed material flux is to be main-
Egozi and Ashmore 2009) and several gravel-bed braided rivers in tained in all channels.
New Zealand. The average number of channels in a river cross section
that transport bed material and are involved in development
of the channel morphology is referred to as the ‘active
braiding intensity’ (Ashmore 2001) (Table1). The active
channels at a given time are a subset of the total number of
channels, i.e., active braiding intensity is equal to or less
than the braiding intensity. Braiding intensity, although still
a regime characteristic of braided rivers, is formed and sus-
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

tained over time by the migration, avulsion, and (transient)

abandonment of a much smaller number of active channels
(Ashmore 2001). Abandoned channels remain open, and
may carry flow, and so remain part of the total braiding in-
tensity and may be reactivated some time later (Egozi 2006;
Egozi and Ashmore 2009).
Active braiding intensity is, like braiding intensity, a re-
gime property of the river, varying with stream power for a
given bed material grain size. It is difficult to obtain field
data on this point, although, given cross-sectional topogra-
phy, the bed area and number of channels that, theoretically,
have bed shear stress above critical can be calculated (Ber-
toldi et al. 2009a). Laboratory experiments (Ashmore 1991a;
Egozi 2006; Bertoldi et al. 2009b) show that active braiding
not known, although a combination of effects is possible, in- intensity correlates well with stream power (but less clearly
For personal use only.

cluding error in measurement of braiding intensity on maps with discharge) in each set of constant discharge experi-
or aerial photographs, influence of flow and channel pattern ments and that the data collapse into a single trend when
history, vegetation and other secondary effects in the field, plotted against dimensionless stream power (Fig. 4) (a least
uncertainty in defining and measuring an index discharge squares regression using a power function to fit these data
for the field data, uncertainty and differences in grain size has an exponent of 0.77, but a larger sample size is needed
distributions, and the effect of constant discharge in the lab- to do a reliable statistical analysis). It is also clear that the
oratory experiments. If bank resistance is an important sec- number of active channels at a given time under channel-
ondary control on braiding (Ferguson 1987; Millar 2000), forming discharge is much less than the total number of
this effect would be expected to be greater in natural rivers, channels. In laboratory experiments, active braiding inten-
where bank sediments, even when mainly noncohesive, have sity seldom exceeds three, and is often less than two. The
inherent strength due to particle packing, structure, and fine- ratio of active to total number of channels is typically about
grained matrix, which is often lacking in laboratory sedi- 0.5, but varies between 0.3 and 0.8, depending on the di-
ments, and due to bank vegetation. This higher bank mensionless stream power (Bertoldi et al. 2009b; Egozi and
strength is one candidate for the systematic bias towards Ashmore 2009). Therefore, the ratio of active to total num-
lower braiding intensity in full-scale rivers compared with ber of channels in the river is also a regime property that
laboratory channels with the same dimensionless discharge. adjusts gradually to changes in channel-forming discharge
and is controlled by dimensionless stream power. The ad-
2.3 Mechanisms controlling braiding intensity justment of this ratio to a step increase in discharge occurs
Another way of considering the limits to braiding inten- by an almost immediate increase in active braiding intensity,
sity is that, without adjustment in channel gradient or a as flow levels increase in existing channels and a gradual in-
change in bed material grain size, as discharge in an individ- crease in the braiding intensity that results in a gradual de-
ual anabranch is reduced by successive bifurcation, the bed crease in the active intensity to braiding intensity ratio
shear stress approaches the critical threshold for motion and (Egozi and Ashmore 2009).
bed material transport, theoretically, ceases at a threshold The mechanism (or mechanisms) that maintains active
braiding intensity. In general, assuming that mean flow braiding intensity at approximately half the braiding inten-
depth in a regime channel is proportional to about the 0.3 sity is still under discussion. It seems likely that bifurcation
power of discharge (typical for many empirical and theoret- dynamics is a significant element of this adjustment. Field
ical analyses), then splitting the channel into two anab- and laboratory observations, theory, and numerical models
ranches with half the discharge in each would reduce shear (Federici and Paola 2003; Zolezzi et al. 2006; Bertoldi and
stress by a factor of about 1.25. If the channel begins with Tubino 2007; Kleinhans et al. 2008) have shown that in
shear stress, say 1.4 times critical, (typical for single gravel- many cases, although a bifurcation (or avulsion) may ini-
bed channels with little bank vegetation (Parker et al. tially have flow and bed load in both branches, this symmet-
2007)), then very few subdivisions can occur before the rical configuration is unstable. Commonly a bifurcation
channel can no longer transport bed material supplied from evolves to a state of asymmetry in which both branches
upstream. This suggests a fairly tight constraint on the pos- carry flow, but one has larger discharge than the other, and

Published by NRC Research Press

 Ashmore 1661

Fig. 4. Active braiding intensity as a function dimensionless stream under other constraints. Murray and Paola (1994) suggest
power (eq. [3]) for three sets of physical model experiments. that braiding is likely in mobile, noncohesive alluvium un-
less there is significant bank resistance, which limits river
channel widening and bend cutoff. This is close to Yalin’s
(1992) conception of the occurrence of braiding in which in-
itial single channel width (aspect ratio) determines the bar
mode and therefore the type of channel pattern formed.
Paola (2001) argues that the ultimate control on aspect ratio
(and channel gradient) is the upstream sediment flux (bed
load) relative to discharge. This is also strongly modulated
by changes in particle size such that finer sediment favours
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

higher braiding intensity, which is consistent with empirical

observations (Carson 1984; Robertson-Rintoul and Richards
1993; Chew and Ashmore 2001). Vegetation is also a major
control on bank resistance and therefore on channel pattern
morphology and dynamics. Recent work has shown that ri-
parian vegetation limits and modifies established braided
in which the larger branch maintains bed load transport patterns by reducing channel instability (rates of lateral mi-
whereas the smaller one does not. If this trajectory is com- gration), reducing (active) braiding intensity (Gran and
mon in braided river bifurcations, then this could account Paola 2001; Tal et al. 2004), and modifying the conditions
for the ratio of total to active branches averaging about 0.5 for the occurrence of braiding (Millar 2000). A complete
at any given time under channel-forming flows. Larger di- theory of the limits to braiding intensity should include
mensionless power (or excess shear stress) would make it these effects.
possible to maintain transport in a larger number of
branches. Egozi and Ashmore (2009) also identify a process 3. Characteristic length of braided channel
of ‘‘partial avulsion’’ in which a large channel switches into patterns
For personal use only.

a new, or pre-existing (previously abandoned), channel with-

out closing off the original channel, so maintaining flow in 3.1 Bars, braiding, and length scales
both the original channel and the avulsion channel, but hav- Single-channel rivers often have a definable morphologi-
ing bed load in only one of them. In these ways, the local cal wavelength defined by the along-channel spacing of
channel-scale processes affect the larger scale regime char- pool-bar units or the wavelength of quasi-regular meanders.
acter of the channel pattern as a whole. It is important to This apparent, dominant morphological length scale is re-
note that all of the analysis to date concerns gravel-bed riv- ferred to here as the ‘characteristic length’ of the channel
ers (or physical models of gravel-bed rivers) and the condi- morphology (Fig. 2). Yalin (1992), along with others (see
tions (dimensionless power) in sand-bed rivers may be quite reviews by Ferguson 1993 and Bridge 1993), places consid-
different. erable emphasis on the role of bars in the development of
meanders and braided channel patterns. Yalin (1992) argues
2.4 Limits to braiding: discussion that single row bars (alternate bars) are the precursor to me-
It is clear that there are limits to braiding intensity im- anders whereas double (or multiple) row bars are the precur-
posed by the discharge and gradient of the channel, which sor to braiding (Figs. 4 and 5). These distinctions are also a
constitute regime properties of the river. Braiding does not prominent component of theoretical analyses of braiding us-
propagate by successive subdivision of channels to any great ing stability analysis of bars and flow-bed topography inter-
extent, if at all. It is controlled by the basic physics of the actions (Parker 1976). Laboratory experiments with erodible
river. If bed load is conveyed in only one or two channels, banks confirm the role of initial bar formation in braided
with the remainder of the network conveying only water and channel pattern development, but show that braiding may
formed over time by migration, avulsion, and bifurcation of originate equally well from single row (alternate) bars as
the active channels, then the ‘excess slope’ model of braid- from multiple row bars (Ferguson 1987, 1993; Bolla et al.
ing intensity (Bettess and White 1983; Yalin 1992) should 2001). These transitions also occur within individual anab-
refer to active braiding intensity. If braiding intensity (total ranches (the defined, separate channels within the braided
sinuosity) averages four, for example, this is likely to be river) of an active braiding river (Ashmore 1991b), but in
equivalent to an active braiding intensity of 1.5–2.0 in most this case multiple row bars are rare because the aspect ratio
gravel-bed rivers, so that braiding occurs at excess slopes of individual anabranches favours the formation of alternate
similar to that of meandering. This is contrary to the argu- bars or single central bars (which are common downstream
ment of Bettess and White (Bettess and White 1983) that from anabranch confluence scour pools). Ashmore (1991b)
braiding requires excess slope two or three times the regime showed that braiding often initiated by chute cutoff (by
slope (which is hydraulically problematic because it implies passing) of lateral or point bars in low sinuosity single chan-
initially very high excess shear stress). The low excess slope nels (with simultaneous channel widening) (Bridge 1993,
for braiding (which may be close to that for meandering) is Figs. 1 and 7). Under these conditions, overall channel
roughly consistent with laboratory models and with field ob- width increases to an equilibrium state as part of the channel
servation. The ‘choice’ (Yalin 1992) of braiding, rather than pattern development, rather than being the initial state lead-
meandering, is then a matter of which patterns are possible ing to multiple row bars, as in Yalin’s (1992) view. In all

Published by NRC Research Press

 1662 Can. J. Civ. Eng. Vol. 36, 2009

Fig. 5. Transitions from single and multiple row bars (upper and along-channel distance between successive alternate bar
lower sketches) to braided channel morphology showing inheritance crests on the same side of the channel. The wavelength of
of bar wavelength and development of a characteristic length scale alternate bars, from flume experiments and observations, ap-
of braided channel pattern (middle sketch). lb is the unit bar wave- pears to involve much greater variability than this. For ex-
length and LB is the ‘braid wavelength’. ample, Ikeda (1984) approximates bar wavelength as being
equal to 9B (Fig. 3) based on fixed-wall flume experiments.
There is considerable scatter in these data, partly because of
experimental errors, but also because bar wavelength is not
solely dependent on channel width. In fixed-width flumes,
wavelength may vary with other flow parameters, including
depth and friction coefficient, and a more recent compilation
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

(Bertoldi 2005) shows a wide range of observed values, with

some wavelength–width ratios exceeding 15. However, the
bulk of the data fall in the range of wavelength equal to 5–
12B (Fig. 6).

3.2 Braid wavelength

There seem to be very few analyses of bar wavelength in
erodible channels, rather than fixed wall flumes, but Ash-
more (1985, 1991b) measured initial alternate bar wave-
length to channel width ratios of 6.0–8.7 with an average of
7.5 in experimental analysis of the development of braiding
from straight channels. This ratio increased slightly to an
average of about 9.0 for the low sinuosity meanders that de-
veloped from the initial alternate bars. Braiding occurred by
chute-cutoff of established lateral bars, so that every second
For personal use only.

alternate (point) bar became a central bar. At the same time,

cases, bars are a fundamental precursor to braiding and a lateral erosion removed the intervening lateral bar to be re-
fundamental unit of braided river morphodynamics and this placed by a confluence zone downstream of the central bar.
unifies them with other river channel pattern types in which At the onset of braiding, the spacing of the initial central
bed-load transport is a significant morphological process bars created by cutoff of the low-sinuosity meander was
(Ferguson 1987, Fig. 6). similar to, but slightly shorter (by a factor of about 0.9)
Given that alternate, central, or multiple-row bars are a than, the immediate pre-cursor meanders. The shortening of
precursor to braiding and a component of the resulting chan- the wavelength is presumably the result of the smaller width
nel pattern dynamics, then it is reasonable to theorize that of the two separate branches formed by the bifurcation. As-
elements of braided pattern length scales are related to the suming that anabranch width scales with the square root of
wavelength of bars and the spatial patterns of deposition anabranch discharge and assuming equal division of the
and bifurcation related to those features. For example, Yalin flow in the bifurcation, then the wavelength of the bars in
(1992) proposed that braiding from double-row bars in- the branch channels would be about 0.7 times that of the up-
volves bifurcation around a central bar (Bridge 1993) from stream single channel and the overall wavelength of the fea-
which he theorizes that the along-channel spacing of up- ture (from upstream confluences to downstream confluence)
stream convergence and downstream divergence around a would be approximately 0.85 of the initial bar wavelength
central bar is three times the initial channel width. The ini- (Ashmore and Gardner 2008).
tial braided pattern then has a characteristic length (bar Bertoldi and Tubino (2005) have shown that bar wave-
wavelength) inherited from the bars from which the pattern length and bank wavelength (the wavelength of bank–line
first develops. Yalin (1992) goes on to argue that the initial plan curvature) coincide very closely, through the initial al-
regular multiple bar pattern and wavelength is then disrupted ternate bar phase, into meander development and then to the
because the two channels on either side of the central bar development of incipient bifurcation and braiding. This pro-
then develop their own bars at a wavelength commensurate vides important experimental verification of the relationship
with their own dimensions and discharge and therefore between initial bar morphodynamics and the river plan–form
shorter than the initial bar pattern. Figure 5 illustrates, con- characteristics. Crucially, incipient bifurcation occurred at a
ceptually, the development of braiding from multiple and threshold wavelength–channel width ratio of between 4.5
single row bars, and the inheritance of initial bar wavelength and 6.3, which is shorter than the initial wavelength–channel
in the braided channel. width ratio of the alternate bars. The gradual decrease in
The same basic argument is true for braiding developed wavelength–channel width ratio during channel pattern de-
from cutoff of initial low sinuosity (single row bars) mean- velopment occurs mainly by channel widening, not by in-
ders (Dietrich 1987). There remains uncertainty about the creasing bar wavelength. Thus the longitudinal spacing of
wavelength of this initial alternate bar pattern. In general, bifurcation points is tied to the wavelength of the initial
the wavelength scales with the channel width and Yalin bars.
(1992) argues that the scaling for single row bars is, theoret- The wavelength of central or multiple-row bars has been
ically, six times the width (B), defining wavelength as the subject to much less analysis. Consequently, it is difficult to

Published by NRC Research Press

 Ashmore 1663

Fig. 6. Bar wavelength: channel width ratios for mode 1 (see central bars, then it is expected that they would have a sim-
Fig. 2) alternate bars. From Bertoldi (2005) and based on a compi- ple scaling relationship similar to the width–wavelength
lation of published data. scaling of alternate bars. Direct measurement of the length
(confluence to head-of-bifurcation) and channel width
(average wetted width of the channel linking the confluence
and bifurcation) yields a consistent linear relationship cover-
ing both physical models and full scale rivers. The length
scale of the confluence–bifurcation unit is approximately
four to five times channel width (Hundey 2007; Hundey
and Ashmore 2009). Larger rivers have larger average spac-
ing for a sample of confluence-bifurcation units, suggesting
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

that the scaling applies to the river as a whole, as well as

individual units– larger rivers have wider anabranches with
greater length scales for the basic confluence–bifurcation
morphological units. This scaling seems to hold regardless
of whether the downstream bifurcation developed from
bend cutoff or from division around a central bar.
This regular spacing of nodes and the tendency for this to
assess Yalin’s (1992) suggestion that upstream convergence increase with the overall size (discharge) of the river is also
and downstream divergence around central bars have an ini- apparent in analyzing the downstream spacing of nodes in a
tial length of 3B. Initiation of braiding around multiple braided river channel network (Ashmore 2001). By measur-
mode or central bars certainly occurs in the manner de- ing the number of confluences or diffluences in a unit length
scribed by Yalin (1992) (Fujita 1989; Ashmore 1991b; Jang of river, and dividing by the total length of channels in the
and Shimizu 2005), but the transition to braiding may be reach, the mean downstream node spacing or ‘braid wave-
more complicated than he proposed. In the most comprehen- length’ can be calculated. Measuring this braid wavelength
sive work on the problem, Fujita (1989) found experimen- over a range of scales shows that it scales very strongly
For personal use only.

tally that multiple row bars follow length and amplitude with river discharge in the same way meander wavelength
relationships similar to single mode bars and assumed a ratio does. In fact, the relationship parallels the commonly ob-
of bar length to bar width of five. This ratio is roughly con- served square root relationship between discharge and mean-
sistent with Yalin’s (Yalin 1992) ‘‘3B proposition’’ for the der wavelength, but braid wavelength tends to be smaller, by
case of mode two bars. However, initial higher mode bar a factor of 0.8, than the meander wavelength for a single
patterns simplified progressively, by merging of bars, to- channel with the same total discharge. Thus, this overall
wards more stable, lower mode bar patterns and only when length scaling of braided channel networks appears to reflect
these lower modes were established did a distinct braided the bar-scale development and control of braiding processes
channel pattern develop (Fujita 1989). This suggests that and morphology.
bar wavelength may adjust during the transition from initial
bar geometry to a fully-developed braided channel. Subse- 4. Linking intensity and wavelength of
quent numerical simulations and physical experiments (Jang
and Shimizu 2005) show this same tendency for initial bar braided patterns
mode to be reduced during the transition to braiding from The first section of this paper shows that there are upper
initial double row bar morphology. The number of scour limits to the intensity of braiding networks. There are two
pools along the flume decreased along with the simplifica- primary reasons for this. One is the necessity to maintain
tion of the channel pattern and then stabilized, suggesting a bed load transport as channels subdivide and the other is
stable average spacing of channel network nodes for the the fact that, for a given imposed valley slope, only small
braided pattern. Recent observations of Bertoldi et al. increases in total sinuosity of the active channel network
(2009b) from physical models also show that the spatial are needed to reach regime channel slope. If braided channel
density of nodes, and the length of links in the braided net- networks have three or fewer active anabranches at any
work, approached a stable value over time at a given con- time, and each has to be large enough to have significant
stant discharge. These observations are consistent with bed material flux, then the range of size of these active
Yalin’s (1992) general proposition. Furthermore, this analy- channels must be limited (although, to date, there are no
sis of bar dynamics (e.g., simplification and stabilization at data to verify this). This limit on the size (discharge, width,
low mode) also provides a possible explanation for the lim- and depth) of these active anabranches would place similar
its to braiding intensity discussed in the first part of this pa- constraints on the wavelength of unit pool-bar morphology
per. in the river. This is a way of explaining the well-defined
Length scaling of braided channel patterns can also be di- trend in average braid wavelength of rivers that occurs de-
rectly assessed by looking for morphological relationships spite the fact that a braided river usually has several anab-
within developed braided patterns. An obvious candidate is ranches with differing dimensions. The secondary active
the confluence-bar morphological unit that is a basic compo- anabranches scale in their dimensions with the primary
nent of braided river morphology (see middle diagram in anabranches, so limiting the range of anabranch dimensions
Fig. 5). If these units function in the same way as pool-bar in a given braided river. This scaling is apparent in the
units in single channels, and develop from initial simple frequency distribution of relative lengths of confluence–

Published by NRC Research Press

 1664 Can. J. Civ. Eng. Vol. 36, 2009

bifurcation units (Hundey and Ashmore 2009) in which  The intensity of braiding is a regime condition governed
data from physical models and full-scale rivers both cover by either the total discharge of the river (in the case of
a range of approximately 0.3–3.0 times the mean. A simi- total braiding intensity) or the total stream power (in the
lar distribution exists for link lengths within a braided net- case of active braiding intensity). The proportion of ac-
work (Ashmore 2001) although in this case there is often a tive anabranches relative to the braiding intensity is typi-
long tail of short links, which may represent falling stage cally about 0.3–0.6 and this may be an outcome of bar
dissection of bar fronts and other small-scale features unre- stability and of bifurcation dynamics, which often leads
lated to the overall channel network. If this conceptual to asymmetric bifurcations with both channels conveying
model is correct, then rivers with higher active braiding in- flow, but only one of the two channels actively transport-
tensity may be expected to have a wider range of conflu- ing bed load.
ence–bifurcation and link lengths.  Braided channel patterns have a characteristic length in-
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

Characteristic length scales of braided channel patterns herited from the length of the initial bar morphology
were also found by Zanoni (2008) using Fourier and wavelet from which they develop, which then persists within indi-
analysis of braided river topography (from physical models) vidual units of the established braided channel. The dis-
to identify characteristic scales of morphology and associ- tribution of characteristic lengths of individual units is
ated processes. Fluctuations in bedload transport rate are constrained by the limited number and size of active ana-
tied to temporal variation in active braiding intensity, time branches in a braided river, so that there is a close con-
scale of bifurcation development and bar migration, and nection between unit processes at the channel and bar
large-scale lateral migration of the channel network. At the scale and the overall intensity and length scale of braided
same time there are characteristic spatial scales in planform channel networks.
and topography. For example, both active braiding intensity
and total wetted width (sum of wetted widths of each anab- Acknowledgements
ranch) of the braided channel co-vary along the river and do
not vary with dimensionless stream power. The largest of My fundamental research on braiding is funded primar-
these scales is persistent in time and space and is about two ily by the Natural Sciences and Engineering Research
to three times the braid plain width (bank-to-bank width of Council of Canada. I thank Walter Bertoldi for sharing
his data on braiding intensity and giving permission to
For personal use only.

the whole braided river) and relates to large scale braided

network morphology as measured by spatial variation in ac- use a diagram from his Ph.D. thesis. My students, Roey
tive braiding index. Very similar characteristic lengths are Egozi, Tobi Gardner, and Beth Hundey, did much of the
apparent in full-scale braided river patterns mapped from experimental work, data analysis, and helped in the devel-
aerial photographs or satellite images. Using braid plain opment of ideas incorporated into this paper. Thanks also
width as the scaling parameter means that there is no direct to Karen van Kerkoerle for her work on the figures. The
relation to bar or other characteristic lengths of single chan- paper benefitted from the comments of two anonymous
nels, but there is clearly a persistent characteristic length of referees. This paper is a contribution to a special issue in
braided networks that scales with the river size. memory of Professor Selim Yalin, who was always inter-
The limited range of active channel dimensions is appa- ested in, and supportive of, my work. He was more influ-
rent in physical models of braiding. Subdivision of branch ential on the development of my understanding of river
channels, after the initial braid formation, is not commonly mechanics than he knew. I am grateful to Ana da Silva
seen. Instead, new channels are initiated by avulsion or bi- and ‘Krish’ Krishnappan for inviting me to contribute to
furcation at the expense of existing active channels, which this volume. Some of the ideas in this paper can be traced
serves to increase the braiding intensity (because abandoned back to my MSC and PhD research projects and I remain
channels do not fill with sediment and may be re-occupied), grateful to John Shaw and Gary Parker for their influence
but not the active braiding intensity (Egozi and Ashmore and for setting me on a career path intertwined with
2009). In many cases the network consists of a single active braiding.
main anabranch with limited activity in secondary anab-
ranches. Active anabranches make the morphology, and
References
therefore the morphology reflects the size of these active Ashmore, P.E. 1985. Process and form in gravel braided streams:
channels. Only at high values of dimensionless stream laboratory modelling and field observations. Ph.D. thesis, Uni-
power (high sediment mobility) is active braiding intensity versity of Alberta, Edmonton, Alta.
greater than two in most cases. Thus the physical limits to Ashmore, P.E. 1991a. Channel morphology and bed-load pulses in
braiding intensity impose regularities in the geometry and braided gravel-bed streams. Geografiska Annaler, 73(1): 37–52.
length scale of the overall braided channel pattern and, at doi:10.2307/521212.
the same time, the unit processes at bar scale are the mech- Ashmore, P.E. 1991b. How do gravel bed rivers braid? Canadian
anisms by which braiding intensity is established; local bar Journal of Earth Sciences, 28(3): 326–341.
dynamics and larger scale channel pattern characteristics are Ashmore, P.E. 2001. Braiding phenomena: statics and kinetics, In
closely connected. Gravel Bed Rivers V. Edited by M.P. Mosley. New Zealand Hy-
drological Society, Christchurch, New Zealand. pp. 95–120.
Ashmore, P., and Gardner, J.T. 2008. Unconfined confluences as
Conclusions
elements of braided river morphology. In River Confluences,
There has been considerable recent progress on under- Tributaries and the Fluvial Network. Edited by S. Rice, A. Roy,
standing the morphodynamics of braided rivers. Two main and B. Rhoads. Wiley, Chichester, UK. pp. 119–143.
conclusions emerge from this work: Bertoldi, W. 2005. River bifurcations. In Monographs of the School

Published by NRC Research Press

 Ashmore 1665

of Doctoral studies in Environmental Engineering, No. 6. Uni- Form, Process and Economic Epplication. Edited by J.L. Best
versity of Trento, Italy. 123 p.. and C. Bristow. Geological Society Special Publication 75. Geo-
Bertoldi, W., and Tubino, M. 2005. Bed and bank evolution of bi- logical Society, London. pp. 73–87.
furcating channels. Water Resources Research, 41(7): W07001. Fujita, Y. 1989. Bar and channel formation in braided streams. In
doi:10.1029/2004WR003333. River Meandering. Edited by S. Ikeda and G. Parker. Mono-
Bertoldi, W., and Tubino, M. 2007. River bifurcations: experimen- graph 12. American Geophysical Union Water Resources, Wa-
tal observations on equilibrium configurations. Water Resources shington, D.C. pp. 417–462.
Research, 43(10): W10437. doi:10.1029/2007WR005907. Galay, V.J., Rood, K.M., and Miller, S. 1998. Human interference
Bertoldi, W., Ashmore, P., and Tubino, M. 2009a. A method for with braided gravel-bed rivers. In Gravel-bed Rivers in the En-
estimating the mean bedload flux in braided rivers. Geomorphol- vironment. Edited by P.C. Klingeman, R.L. Beschta, P.D. Ko-
ogy, 103(3): 330–340. doi:10.1016/j.geomorph.2008.06.014. mar, and J.B. Bradley. Water Resources Publications LLC,
Bertoldi, W., Zanoni, L., and Tubino, M. 2009b. Planform dy- Highlands Ranch, Colo. pp. 471–512.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20

namics of braided streams. Earth Surface Processes and Land- Gran, K., and Paola, C. 2001. Riparian vegetation controls on
forms, 34(4): 547–557. doi:10.1002/esp.1755. braided stream dynamics. Water Resources Research, 37(12):
Bettess, R., and White, W.R. 1983. Meandering and braiding of al- 3275–3283. doi:10.1029/2000WR000203.
luvial channels. Proceedings of the Institute of Civil Engineers, Hundey, E. 2007. Length scales of braided river morphology. Hon-
75(3): 525–538. ours thesis, University of Western Ontario, London, Ont.
Blench, T. 1969. Mobile-bed fluviology. The University of Alberta Hundey, E., and Ashmore, P. 2009. Length scale of braided river
Press, Edmonton, Alta. morphology. Water Resources Research, 45(8): W08409.
Bolla, M., Federici, B., Repetto, R., Paola, C., Seminara, G., and doi:10.1029/2008WR007521.
Tubino, M. 2001. The morphodynamics of braiding rivers: ex- Ikeda, S. 1984. Prediction of alternate bar wavelength and height.
perimental and theoretical results of unit processes. In Gravel Journal of Hydraulic Engineering, 110(4): 371–386. doi:10.
Bed Rivers V. Edited by P.M. Mosley New Zealand Hydrologi- 1061/(ASCE)0733-9429(1984)110:4(371).
cal Society, Christchurch, New Zealand. pp. 143–180. Jang, C.-L., and Shimizu, Y. 2005. Numerical simulation of rela-
Bray, D. 1982. Regime relations for gravel-bed rivers. In Gravel- tively wide, shallow channels with erodible banks. Journal of
Bed Rivers. Edited by R.D. Hey, J.C. Bathurst, and C.R. Thorne. Hydraulic Engineering, 131(7): 565–575. doi:10.1061/(ASCE)
John Wiley, Chichester, UK. pp. 517–542. 0733-9429(2005)131:7(565).
For personal use only.

Bridge, J.S. 1993. The interaction between channel geometry, water Kleinhans, M., Jagers, H.R.A., Mosselman, E., and Sloff, C.J.
flow, sediment transport and deposition in braided rivers. In 2008. Bifurcation dynamics and avulsion duration in meandering
Braided Rivers: Form, Process and Economic Application. Edi- rivers by one-dimensional and three-dimensional models. Water
ted by J.L. Best and C. Bristow. Geological Society Special Pub- Resources Research, 44(8): W08454. doi:10.1029/
lication 75. Geological Society, London. pp. 13–71. 2007WR005912.
Carson, M.A. 1984. The meandering-braided river threshold: A re- Leopold, L. B. and Wolman, M. G., 1957. River channel patterns:
appraisal. Journal of Hydrology (Amsterdam), 73(3-4): 315–334. braided, meandering and straight. U.S. Geological Survey Profe-
doi:10.1016/0022-1694(84)90006-4. sional Paper 282B, pp. 39–85.
Chew, L.C., and Ashmore, P.E. 2001. Channel adjustment and a Millar, R.G. 2000. Influence of bank vegetation on alluvial channel
test of rational regime theory in a proglacial braided stream. patterns. Water Resources Research, 36(4): 1109–1118. doi:10.
Geomorphology, 37(1-2): 43–63. doi:10.1016/S0169-555X(00) 1029/1999WR900346.
00062-3. Mosley, P.M. 1981. Semi-determinate hydraulic geometry of river
Church, M., Moore, D. and Rood, K. 1981. Catalogue of alluvial channels, South Island, New Zealand. Earth Surface Processes
channel regime data. Department of Geography, University of and Landforms, 6(2): 127–137. doi:10.1002/esp.3290060206.
British Columbia. Murray, A.B., and Paola, C. 1994. A cellular model of braided riv-
Dietrich, W.E., 1987. Mechanics of flow and sediment transport in ers. Nature, 371(6492): 54–57. doi:10.1038/371054a0.
river bends. In River Channels: environment and process. Edited Paola, C. 2001. Modelling stream braiding over a range of scales.
by K. Richards, The Institute of British Geographers, Blackwell, In Gravel Bed Rivers V. Edited by M.P. Mosley. New Zealand
Oxford, pp. 179–227. Hydrological Society, Christchurch, New Zealand. pp. 11–46.
Egozi, R. 2006. Channel pattern variation in gravel-bed rivers. Parker, G. 1976. On the cause and characteristic scales of meander-
Ph.D. thesis, University of Western Ontario, London, Ont. ing and braiding in rivers. Journal of Fluid Mechanics, 76(3):
Egozi, R., and Ashmore, P.E. 2008. Defining and measuring braid- 457–480. doi:10.1017/S0022112076000748.
ing intensity. Earth Surface Processes and Landforms, 33(14): Parker, G., Wilcock, P., Paola, C., Dietrich, W., and Pitlick, J.
2121–2138. doi:10.1002/esp.1658. 2007. Physical basis for quasi-universal relations for bankfull
Egozi, R., and Ashmore, P. 2009. Experimental analysis of braided hydraulic geometry of single-thread gravel bed rivers. Journal
channel pattern response to increased discharge. Journal of Geo- of Geophysical Research, 112F4: F04005. doi:10.1029/
physical Research, 114(F2): F02012. doi:10.1029/ 2006JF000549.
2008JF001099. Richards, K.S., 1982. Rivers: Form and process in alluvial chan-
Federici, B., and Paola, C. 2003. Dynamics of channel bifurcations nels. Methuen, New York.
in noncohesive sediments. Water Resources Research, 39(6): Robertson-Rintoul, M.S.E., and Richards, K.S. 1993. Braided chan-
1162. doi:10.1029/2002WR001434. nel pattern and paleohydrology using an index of total sinuosity.
Ferguson, R.I. 1987. Hydraulic and sedimentary controls of channel In Braided Rivers: Form, Process and Economic Application.
pattern. In River Channels: environment and process. Edited by Edited by J.L. Best and C.S. Bristow. Geological Society Spe-
K. Richards, The Institute of British Geographers, Blackwell, cial Publication 75. Geological Society, London. pp. 113–118.
Oxford, pp. 129–158. Tal, M., Gran, K., Murray, A.B., Paola, C., and Hicks, D.M. 2004.
Ferguson, R.I. 1993. Understanding braiding processes in gravel- Riparian vegetation as a primary control on channel characteris-
bed rivers: progress and unsolved problems. In Braided Rivers: tics in multi-thread rivers. In Riparian Vegetation and Fluvial

Published by NRC Research Press

 1666 Can. J. Civ. Eng. Vol. 36, 2009

Geomorphology. Edited by S.J. Bennett and A. Simon. Mono- namics in braided rivers. Ph.D. thesis, University of Trento,
graph 8. American Geophysical Union, Water Science and Ap- Italy.
plication, Washington, D.C. pp. 43–58. Zolezzi, G., Bertoldi, W., and Tubino, M. 2006. Morphological
Yalin, M.S. 1992. River Mechanics. Pergamon, Oxford. analysis and prediction of channel bifurcations. In Braided Riv-
Yalin, M.S., and da Silva, A.M.F. 2001. Fluvial Processes. Interna- ers: Process, Deposits, Ecology and Management. Edited by G.
tional Association of Hydraulic Engineering and Research, H. Sambrook-Smith, J. L. Best, C. S. Bristow, and G. E. Petts.
Delft, The Netherlands. Special Publication 36. International Association of Sedimentol-
Zanoni, L. 2008. Field and laboratory evidence of large scale dy- ogists, Blackwell, Chichester, UK. pp. 227–250.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by EPFL BIBLIOTHEQUE USD on 12/02/20
For personal use only.

Published by NRC Research Press