Hydraulic and sedimentary processes causing anastomosing morphology of the
upper Columbia River, British Columbia, Canada
Bart Makaske a,
, Derald G. Smith b
, Henk J.A. Berendsen a,1
, Arjan G. de Boer a,2
,
Marinka F. van Nielen-Kiezebrink a,3
, Tracey Locking b
a
Department of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, The Netherlands
b
Department of Geography, University of Calgary, Calgary, Alberta, Canada T2N 1N4
a b s t r a c ta r t i c l e i n f o
Article history:
Received 5 January 2009
Received in revised form 22 April 2009
Accepted 23 April 2009
Available online 3 May 2009
Keywords:
Anastomosing river
Avulsion
Bedload
Fluvial geomorphology
Sediment transport
Stream power
The upper Columbia River, British Columbia, Canada, shows typical anastomosing morphology -- multiple
interconnected channels that enclose floodbasins -- and lateral channel stability. We analysed field data on
hydraulic and sedimentary processes and show thatthe anastomosing morphologyof the upperColumbiaRiver is
caused by sediment (bedload) transport inefficiency, in combination with very limited potential for lateral bank
erosion because of very low specific stream power (2.3 W/m2
) and cohesive silty banks. In a diagram of channel
type in relation to flow energy and median grain size of the bed material, data points for the straight upper
Columbia River channels cluster separately from the data points for braided and meandering channels.
Measurements and calculations indicate that bedload transport in the anastomosing reach of the upper Columbia
River decreases downstream. Because of lateral channel stability no lateral storage capacity for bedload is created.
Therefore, the surplus of bedload leads to channel bed aggradation, which outpaces levee accretion and causes
avulsions because of loss of channel flow capacity. This avulsion mechanism applies only to the main channel of
the system, which transports 87% of the water and N90% of the sediment in the cross-valley transect studied.
Because of very low sediment transport capacity, the morphological evolution of most secondarychannels isslow.
Measurements and calculations indicate that much more bedload is sequestered in the relatively steep upper
anastomosing reach of the upper Columbia River than in the relatively gentle lower anastomosing reach. With
anastomosing morphology and related processes (e.g., crevassing) being best developed in the upper reach, this
confirms the notion of upstream rather than downstream control of upper Columbia River anastomosis.
 2009 Elsevier B.V. All rights reserved.
1. Introduction
The anastomosing upper Columbia River, British Columbia, Canada
(Fig. 1), is composed of multiple, roughly "straight" channels that are
laterally stable. The morphodynamics of this fluvial system are
dominated by frequent crevassing -- erosive formation of gaps in the
natural levee by floodwater -- and avulsion (Smith,1983; Makaske et al.,
2002). Straight, laterally stable channels like those of the anastomosing
upper Columbia River are an important category of fluvial channels that
have received relatively little attention in geomorphological and
hydraulic research. Work in this field has strongly focused on the
braided-meandering transition (e.g., Leopold and Wolman, 1957;
Carson, 1984; Van den Berg, 1995; Bledsoe and Watson, 2001), while
straight channels were often supposed be rare. Yet "straight" channels --
i.e., low to moderately sinuous but laterally stable channels (Makaske,
2001, pp. 156­157) -- are very common elements of inland anastomosing
river systems (Smith, 1973, 1983, 1986; Makaske, 1998, 2001) and
river deltas (Kolb, 1963; Morton and Donaldson, 1978; Makaske, 1998).
Frequent crevassing and avulsion are generally believed to be
associated with rapid floodplain aggradation (Makaske, 2001). However,
avulsion is not merelya result of (differential) floodplain aggradation, but
also of in-channel processes that result in a decrease of the ability of a
channel to carry sediment and discharge (see also Jones and Schumm,
1999). Recent papers on the hydraulic and sedimentary processes in the
upper Columbia (Tabata and Hickin, 2003; Abbado et al., 2005) addressed
the matter of the efficiency of anastomosing channels for transporting
water and sediment, and the question of whether anastomosing rivers
are equilibrium forms. These studies are responses to an argument made
by Nanson and Huang (1999) that some anastomosingrivers are basically
fine tuned to transport water and sediment with maximum efficiency
across low-gradient plains. This view was contradicted by results reached
Geomorphology 111 (2009) 194­205
 Corresponding author. Present address: Alterra, Wageningen University and
Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands. Tel.: +31 317
481609; fax: +31 317 419000.
E-mail addresses: bart.makaske@wur.nl (B. Makaske), dgsmit@ucalgary.ca
(D.G. Smith), a.de.boer@archeologie.nl (A.G. de Boer), marinka.van.nielen@rws.nl
(M.F. van Nielen-Kiezebrink).
1
Henk Berendsen passed away on May 14, 2007.
2
Present address: ADC ArcheoProjecten, P.O. Box 1513, 3800 BM Amersfoort, The
Netherlands.
3
Present address: Rijkswaterstaat-Waterdienst, P.O. Box 17, 8200 AA Lelystad, The
Netherlands.
0169-555X/$ ­ see front matter  2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2009.04.019
Contents lists available at ScienceDirect
Geomorphology
journal homepage: www.elsevier.com/locate/geomorph
by Tabata and Hickin (2003) and Abbado et al. (2005), which suggested
inefficiency of the anastomosing upper Columbia River channels. Points
not addressed in these studies, however, are the lateral channel stability
of the upper Columbia channels and its role in the formation of the
anastomosing system, whereas detailed analysis of in-channel sedimentary
processes contributing to anastomosis is also still lacking. Previous
work (Smith, 1983; Makaske et al., 2002) demonstrated the long-term
lateral stability of the upper Columbia channels and suggested significant
bed aggradation in the anastomosing channels.
In this study, in-channel hydraulic and sedimentary processes of the
upper Columbia River are examined to better understand its lateral
channel stability and mechanisms of crevassing and avulsion. More
specifically, in this paper we (i) present stream power and sediment
transport data for the straight upper Columbia River channels;
(ii) analyse the position of these channels, relative to meandering and
braided channels, in an energy-grain size plot; and (iii) carry out a
sediment (bedload) budget analysis for two different upper Columbia
River reaches to assess bed aggradation rates.
We primarily use field data collected by Makaske (1998) in a crossvalley
study transect located roughly halfway along the anastomosing
reach of the upper Columbia River, between Radium and Golden, 9 km
upstream of Parson, British Columbia (Fig. 1). These data are
supplemented by bedload transport measurements carried out by
Locking (1983) upstream and downstream of the transect (Fig.1) and by
longitudinal river gradient data of the anastomosing reach collected by
Abbado et al. (2005).
2. Geographical setting and characteristics of the upper
Columbia River
The Columbia River has its source in Columbia Lake and flows
northwestward. Major tributaries to the upper Columbia River drain
the Purcell Mountains, which flank the Columbia Valley to the SW
(Fig. 1). Locally, the Purcell Mountains rise above 3000 m amsl (above
mean sea level) and glaciers surround the highest peaks. Minor creeks
drain the lower Kootenay Ranges of the Rocky Mountains on the NE
side of the Columbia Valley. These ranges are free of ice with a number
of summits close to 2700 m amsl.
The upper Columbia Valley has a humid continental climate with
cold winters and warm summers. Mean monthly temperatures on the
valley bottom for January and July are -10 °C and 17 °C, respectively.
Mean annual precipitation is 490.7 mm at Golden (Meteorological
Service of Canada, personal communication, 2002). Higher in the
mountains, precipitation is much greater, with the Purcell Mountains
being known for deep and prolonged snow covers. At the end of most
winters the equivalent of 500 to 750 mm of rain lies over the massif as
snowpack (Hare and Thomas, 1974, p. 111).
Above the Nicholson gauging station (Fig.1), the drainage area of the
upper Columbia River is 6660 km2
, with an annual mean discharge of
107 m3
/s (data from Water Survey of Canada). Water discharge in the
upper Columbia River fluctuates seasonally in response to rainfall and
snowmelt, while glacial meltwater constitutes only a minor component
of the total discharge. Minimum discharge occurs during the winter --
usually in the period from December to March when ice covers the river.
Minimum monthly mean discharge is 23.2 m3
/s in February. In the
spring, snowmelt and rain cause a sharp rise and a peak in discharge in
June or July (Fig. 2). Maximum monthly mean discharge is 318 m3
/s in
July and bankfull level is exceeded almost yearly.
The upper Columbia River has a narrow floodplain flanked by
mountain slopes and tributary valleys from which alluvial fans protrude
into the Columbia Valley. The system is laterally restricted (1.5 km
wide), but longitudinally continuous for at least 100 km between
Radium and Golden (Fig. 1). Over most of its anastomosing reach, the
upper Columbia River consists of a main channel and a variable number
of smaller parallel channels. Within the anastomosing reach, the alluvial
valley floor has a mean elevation of around 790 m amsl and a generally
gentle slope, averaging 11.5 cm/km. The Holocene part of the
unconsolidated valley-fill is probably about 20 m thick (Junck, 2009).
Fig. 1. The location of the cross-valley upper Columbia River study transect in SE British Columbia, Canada. The grey zone indicates the anastomosing reach of the upper Columbia
River. The dark grey zone indicates the reach shown in Fig. 3. Black triangles indicate the temporary upstream (location 1) and downstream (location 2) gauging stations where
hydraulic and bedload transport data were collected by Locking (1983).
195B. Makaske et al. / Geomorphology 111 (2009) 194­205
Abbado et al. (2005) distinguished two contrasting anastomosing
Columbia River reaches (Fig. 3) downstream of Spillimacheen, where
the Spillimacheen River -- an important tributary -- enters the
Columbia Valley. The relatively steep upper reach is characterized by
well-developed anastomosis, with three to five parallel channels. A
significant break in slope marks the transition to the relatively gentle
lower reach in which anastomosis is weakly developed, with typically
two to three parallel channels. Abbado et al. (2005) suggested higher
aggradation rates in the upper reach as a cause of better developed
anastomosis.
3. Research site and methods
Fieldwork was carried out in the Columbia Valley from May to July
1994 to enable measurements at discharges near bankfull. The crossvalley
transect in which the data were collected is the same as the
transect studied sedimentologically and chronologically by Makaske
et al. (2002), who demonstrated that the channels in the transect
formed by avulsion between 30 and 3000 years ago and that the
channels generally evolved through a consistent pattern of different
morphodynamic stages of young widening and deepening channels to
old shallowing and narrowing channels. They also reported high
lateral facies variability, with clayey and silty floodbasin deposits
contrasting with narrow and thick sandy channel fills that indicate
long-term lateral channel stability.
A cross-valley transect study enables us to sum measured and calculated
water and sediment fluxes to obtain total Columbia River estimates
that can be compared with available station data. The studied transect is
situated near the transition from the "highly anastomosed" reach
upstream and the "weakly anastomosed" reach downstream (Fig. 3),
and thereby facilitates a sediment budget calculation for both reaches
because sediment transport data have been collected (Locking, 1983)
upstream near Spillimacheen and downstream near Nicholson (Fig. 1).
The transect extends across the floodplain from the McKeeman
Creek alluvial fan to the opposite valley side, thereby crossing six active
channels of variable dimensions (Fig. 4). All studied channels have a low
sinuosity index (Pind =channel length/length along channel-belt axis;
Brice,1964; Makaske, 2001, p.156) between 1.00 and 1.05 (Table 1) and
can thus be classified as straight channels (cf. Makaske, 2001) that have
experienced no significant lateral migration. In general, the channels are
relatively deep and narrow, with width/depth ratios below 10, but some
channels have greater width/depth ratios of about 20 (Table 1). The wet
perimeter of the channels typically has a trapezoidal shape. Channel 6
was blocked by a large beaver dam allowing no significant discharge, so
hydraulic data were only gathered for channels 1 through 5. In terms of
discharge, channel 3 is by far the most dominant (Table 2).
The floodplain topography of the transect was levelled relative to
an arbitrary datum, and echo soundings provided data on submerged
channel morphology. The echo sounder used provided a vertical
resolution of about 2.5 cm and was mounted on an inflatable boat
with an outboard motor. Echo soundings were related to land-level
data through frequent water stage measurements.
Flow velocity measurements were carried out in the five channels to
determine bankfull discharge as a basis for the calculation of stream
power and sediment transport. Bankfull stage was defined by the break
in slope between the steep channel bank and the flatter surface of the
levee. In the case of a difference in elevation between the break in slope
for the two banks, the lowest level was chosen. On the basis of five
discharge measurements near bankfull stage, stage-discharge graphs
were constructed for each channel from which discharge at bankfull
stage was interpolated. For each discharge measurement, vertical
velocity profiles were measured from the boat, which was attached to
a wire guideline stretched across the channel. A small propeller current
meter mounted on a finned probe was lowered with a winch. Velocities
were measured at 0.25, 0.50,1.00, 2.00 and 4.00 m (if applicable) above
the bottom, and often an additional measurement was carried out near
the water surface. Each velocity measurement lasted 100 s. Velocity
profiles were measured at breaks in the bed morphology of the channel
cross sections. Mean velocity in each vertical profile was calculated from
the plotted surface area of the velocity vertical. Spatial integration over
the channel cross sectionyielded total channel discharge. The number of
velocity profiles per channel averaged 12 in the main channel and 9 in
the smaller channels, and represent the maximum resolution that could
be achieved under the ruling field conditions.
Grab samples of bed material were taken with a Van Veen grab
sampler (Oele et al.,1983, p. 357). Series of three to seven grab samples
were taken twice during the field season in each channel cross section.
Sampling locations were roughly evenly distributed over the wet
channel perimeter. Laboratory grain size analyses were carried out on 44
grabsamples by thestandard sieving/pipette method (McManus,1988).
Channel gradients are needed for the calculation of stream power
and sediment transport. The channel gradient -- or energy gradient in a
channel -- can be approximated by the water-surface gradient if steady
uniform flow is assumed. We did not succeed in accurately measuring
the low channel gradients in this study. Abbado et al. (2005, their Fig. 4)
calculated an average channel gradient of 6.8 cm/km for the reach in
Fig. 2. Mean monthly discharges of the Columbia River at Nicholson, BC, (see Fig. 1 for location) for the period 1903­2005 and for 1982 and 1994 (data from Water Survey of Canada,
station 08NA002).
196 B. Makaske et al. / Geomorphology 111 (2009) 194­205
which our transect is located, based on many GPS data points. This value
is considered the best approximation of the channel gradient of channel
3 (the main channel) in our study. The related valley gradient --
accounting for a channel sinuosity of 1.16 -- is 7.9 cm/km. Channel
gradients of the other channels in the transect were calculated from this
valley gradient, taking into account the differences between the
direction of maximum valley slope and the directions of the channels
on the valley bottom.
As an estimate of the energy available for bank erosion, specific
stream power () under bankfull conditions was calculated for each
channel, applying =gQbfSc/w in which  is the density of water, g
is the acceleration of gravity, Qbf is bankfull discharge, Sc is the
channel gradient, and w is the (bankfull) channel width. In order to
compare the flow energy and sedimentary characteristics of the
Columbia River channels with meandering and braided channels from
the literature, the parameter SvQbf for each channel was plotted
against D50 of the bed material in a diagram of Bledsoe and Watson
(2001). They analysed a large data set of meandering and braided
channels previously published by Van den Berg (1995) and proposed a
logistic regression model to describe the probabilistic ("fuzzy") nature
of the transition from meandering to braiding, thereby extending Van
den Berg's (1995) work on the meandering-braided transition.
Sediment transport capacity at bankfull discharge was assessed for
each channel using three different sediment transport equations: (i) the
Van Rijn formula (Van Rijn,1984a,b), (ii) the modified Van Rijn formula
(Van den Berg and Van Gelder,1993), and (iii) the Engelund and Hansen
(1967) formula. In a test of sediment transport equations, the Van Rijn
formula (Van Rijn,1984a,b) proved to be a fairly reliable predictor (Van
Rijn,1993). Van den Berg and Van Gelder (1993), however, showed that
at low flow velocities the Van Rijn formula underestimated the transport
rates. Under these conditions, the Engelund and Hansen (1967) formula
performed better, which is largely a result of the fact that it does not
account for critical bed shear stress. Therefore, Van den Berg and Van
Gelder (1993) proposed some modifications of the Van Rijn formula,
mainly concerning the Shields limit and the near bed reference
concentration, which enhanced its predictive power.
Locking (1983) measured discharge and bedload transport during
the 1982 flood season near Spillimacheen and Nicholson (Fig. 1). For
Fig. 3. Satellite images of the anastomosing upper Columbia River, between Spillimacheen and a point 6 km downstream of Parson, BC, Canada (see Fig. 1 for location). Images from
Google Earth (details in Fig. 4). The upper Columbia floodplain is flanked by the Kootenay Ranges (NE) and the Purcell Mountains (SW). The Canadian Pacific Railway (CPR) runs
along the northeastern edge of the floodplain. Dark colours indicate the forested southwestern valley side. Images taken in the summer flood season with river stage near bankfull
and inundated floodbasins. Panels (A) through (D) show successive reaches downstream from the point where the Spillimacheen River joins the upper Columbia, with panels slightly
overlapping. (A) and (B) show the "highly anastomosed" reach upstream of the studied transect, which is characterized by three to five parallel channels and channel slopes of up to
21.5 cm/km (Abbado et al., 2005). Note the pale colour of the floodbasin waters in this reach because of a large supply of sediment-rich water through crevasses. (C) and (D) show the
"weakly anastomosed" lower reach, which is characterized by two to three parallel channels and channel slopes of 6.8 cm/km (Abbado et al., 2005).
197B. Makaske et al. / Geomorphology 111 (2009) 194­205
the discharge measurements, Locking (1983) broadly followed the
procedure outlined above, using a Price AA current meter with two
velocity measurements taken at 2/10 and 8/10 depth of each vertical
profile (11­15 vertical profiles across the channel). For bedload
transport measurements, a Helley-Smith bedload sampler was used
for two measurements (15 s each) at each measured vertical velocity
profile. At both gauging stations, 10 coupled discharge-bedload
transport determinations were carried out in the period June­August
1982 and were used to calculate total bedload fluxes for the 1982 flood
season. Application of the Van Rijn (1984a) bedload transport formula
to the hydraulic and bedload grain size data collected by Locking
(1983) yielded that the Van Rijn formula predicts bedload transport
rates that are on average 58 (Spillimacheen station) to 157%
(Nicholson station) of the measured rates.
4. Results
Bankfull discharge of the anastomosing system in the transect is
220 m3
/s (Table 2). Considerable differences in mean flow velocity at
bankfull discharge exist between the channels (Table 3) and are
probably related to differences in hydraulic radius. In channel 4, flow
velocity is remarkably high for the small hydraulic radius. In Fig. 5 the
measured cross-sectional flow velocity patterns are given for each
channel for a stage close to bankfull. The velocity patterns are fairly
symmetrical. Even in channel 5, where the cross section is located just
downstream of a bend apex, the thalweg is located in the middle of
the channel. Vertical velocity gradients over the flat central portion of
the channel beds are much more pronounced than over the sloping
banks. Near the channel margins of channels 1, 3, and 5, velocity at the
surface is significantly lower than at some depths. This indicates
relatively great influence of sidewall roughness because of variable
bank morphology, stuck driftwood, and tree branches hanging in the
water. In channel 2, rushes growing on the left bank retarded the flow
over the full depth.
The low-relief channel bed and the sloping banks are very different
in grain size composition (Fig. 6). The bank material is cohesive, with
low percentages of clay (5%), but very high percentages of silt
(typically 50%). The silty material probably originates from the
terraces of glaciolacustrine deposits, which flank the Columbia River
floodplain in the south. The bed material typically consists of 85­95%
sand, with exception of the crevasse channel (channel 2), where the
silt and sand contents of the bed material are approximately 40 and
60%, respectively. Lumps of slumped mud, clay pebbles, leaves, and
wood were frequently found in the grab samples of the bed material.
The bed material of most channels consists of medium sand (Table 4),
which is moderately sorted on average (scale of Folk and Ward, 1957).
Again, the bed material of channel 2 is different, being characterized
by fine, poorly sorted sand.
Fig. 3 (continued).
198 B. Makaske et al. / Geomorphology 111 (2009) 194­205
In Table 5, gradient and specific stream power data for the channels
in the transect are listed. The channel gradients vary little as a function
of slight differences in sinuosity and channel orientation relative to
the direction of maximum valley gradient. All specific stream power
values in Table 5 can be classified as extremely low, when compared
to data from the literature (e.g., Ferguson, 1981; Nanson and Croke,
1992). Fig. 7 shows that the median grain size of the bed material is
finely tuned to specific stream power, which underscores that specific
stream power determines present-day sedimentary processes in the
upper Columbia channels.
The SvQbf values for the upper Columbia River channels are below
all SvQbf data plotted by Bledsoe and Watson (2001), which represent
predominantly braided and meandering channels (Fig. 8). Seemingly,
the inactive "straight" upper Columbia River channels plot as a separate
class below meandering and braided channels in the diagram, with
channel 2 taking the lowest position and channel 3 plotting closest to the
meandering channel data points. Two additional data points (6 and 7 in
Fig. 8) represent straight channels in the upper Inland Niger Delta in
central Mali (Makaske,1998) -- an arid-zone, but perennial, anastomosing
river system. Other additional data points (or data ranges) represent
interpreted straight (point 8) and confined meandering (point 9)
palaeochannels from the Rhine-Meuse delta (Makaske and Weerts,
2005; Makaske et al., 2007). The discriminant line between straight and
meandering channels in Fig. 8 is necessarily tentative and in reality may
not parallel the meandering-braided transition. Because of the limited
straight channel data points (and limited data range), we refrained from
a logistic regression analysis (cf. Bledsoe and Watson, 2001) to define a
probabilistic discriminator between straight and meandering channels.
The calculated transport rates (Table 6) show that the sediment
transport capacity of channel 3 greatly exceeds the capacity of the four
other channels together. The highest sediment transport rates are
predicted by the modified Van Rijn formula. These are more than
twice as high as the prediction by the original Van Rijn formula. The
Engelund and Hansen formula predicts slightly higher transport rates
for most secondary channels because the critical shear stress is not
taken into account. In the smaller channels, the bed shear stress under
bankfull conditions was only slightly above the critical shear stress
Fig. 4. Satellite image of the upper Columbia River floodplain around the studied transect. Flow is from right to left. Image is from Google Earth and taken in the summer flood season,
with river stage near bankfull and inundated floodbasins.
Table 1
Morphometric characteristics of the studied channels.a
Width Depth w/d ratio Pind
(m) (m) (-) (-)
Channel 1 19.30 2.12 9.1 1.04
Channel 2 24.80 1.08 23.0 1.00
Channel 3 56.03 5.85 9.6 1.04
Channel 4 20.70 1.04 19.9 1.04
Channel 5 18.63 3.17 5.9 1.05
Channel 6 18.00 2.90 6.2 1.04
a
Bankfull width; maximum depth relative to bankfull level; w/d = width/depth;
Pind = sinuosity index.
Table 2
Bankfull discharge.a
Qbf,1 Qbf,2
(m3
/s) (m3
/s) (%)
Channel 1 9.5 8.4 3.8
Channel 2 2.4 2.8 1.3
Channel 3 193.5 192.0 87.1
Channel 4 5.4 6.1 2.8
Channel 5 14.8 11.1 5.0
Total 220.4 100.0
a
Q bf,1 = bankfull discharge for each individual channel; Q bf,2 = discharge when the
entire anastomosing system is approximately at bankfull stage.
199B. Makaske et al. / Geomorphology 111 (2009) 194­205
required to initiate motion of the bed material. In channel 2, shear
stress did not exceed the critical level, which was confirmed by bed
observations before and after peak discharge, showing that the mud
drape on the sandy bed had remained largely intact. Consequently,
sediment transport in the smaller channels of the anastomosing
system is almost negligible (Table 6), and only stages well above
bankfull can bring about substantial sediment transport in channels
other than the main channel.
An important limitation is that all sediment transport formulae only
predict the transport capacity for the locally available sandy bed
material. Great quantities of finer material are transported as washload,
unrelated to the transport capacity as calculated with a sediment
transport formula. The Van Rijn formula and the modified Van Rijn
formula make a distinction between bedload and suspended load.
Predicted bedload forthe main channel is 20 to 43% of thepredicted total
load. Measurements of sediment concentrations by Locking (1983)
upstream and downstream of the present study transect indicate
bedload to be 11% of the total load. In these measurements, however,
measured suspended load included washload. The measured values
therefore were much greater than the predicted quantities.
When the measured bedload transport data from Locking (1983)
are combined with the calculated bedload transport data from the
present study transect, sediment budgets of the upper and lower
anastomosing reaches of the upper Columbia can be estimated. The
upper anastomosing reach between Locking's measurement site near
Spillimacheen (Fig. 1) and the present study site is 15 km long (main
channel distance) and is characterized by an 10-km-long, relatively
steep section with a channel slope of 21.5 cm/km (Abbado et al.,
2005). The lower anastomosing reach between the present study site
and Locking's measurement site near Nicholson (Fig. 1) is 35 km long
(main channel distance) with a fairly constant average channel slope
of 6.8 cm/km. Based on the hydraulic and sediment transport data
from Locking (1983), bedload input into the upper anastomosing
reach during bankfull discharge (220 m3
/s) is estimated at 7.1 kg/s.
At the same discharge, bedload transport at the study site is 1.9 kg/s,
according to our calculations. However, because the Van Rijn formula
predicts upper Columbia River bedload transport rates between 58
and 157% of the measured values, real bedload transport at the study
site may be in the range of 1.2 to 3.3 kg/s. Near Nicholson, bedload
output from the lower anastomosing reach at bankfull discharge
(220 m3
/s) is estimated at 0.6 kg/s (Fig. 9).
Table 3
Flow velocities and flow dimensions at bankfull discharge.a
umean R hmean A
(m/s) (m) (m) (m2
)
Channel 1 0.32 1.4 1.54 29.8
Channel 2 0.15 0.6 0.65 16.0
Channel 3 0.79 4.1 4.37 244.8
Channel 4 0.35 0.8 0.75 15.5
Channel 5 0.40 1.8 1.99 37.0
a
umean = mean flow velocity in flow area; R = hydraulic radius; hmean = mean
channel depth; A = flow area.
Fig. 5. Measured cross-sectional flow velocity (in m/s) patterns near bankfull discharge in (A) channels 1, 2, 4, and 5 and (B) channel 3 (note the different scale of this cross section).
The black dots indicate flow velocity measurements. Flow is away from the observer.
200 B. Makaske et al. / Geomorphology 111 (2009) 194­205
Thus, at bankfull discharge, 58 to 91% of the bedload trapped between
Spillimacheen and Nicholson is sequestered in the upper anastomosing
reach, and 9 to 42% is sequestered in the lower anastomosing reach
(Fig. 9). Assuming that (i) these proportions are representative for the
entire flood season, i.e., bankfull discharge is dominant discharge; (ii)
bedload sequestration in the anastomosing reach largely occurs in the
summer flood season, as indicated by Locking's measurements showing
limited bedload surplus at stages well below bankfull; and (iii) the 1982
flood season roughly represents the average flood season, we can estimate
the total amounts of bedload that are trapped annually in the upper and
lower anastomosing reaches (Fig. 9). Because the 1982 flood was above
average (Fig. 2), we might slightly overestimate bedload sequestration.
However, this overestimation is probably compensated for by the fact that
we neglect bedload sequestration out of the flood season.
5. Lateral channel stabilityandbed aggradation causing anastomosis
The specific stream power of the studied upper Columbia River
channels (Table 5) is within the ranges mentioned in the literature for
straight, laterally stable, channels. Nanson and Croke (1992) suggested
an upper boundary of 10 W/m2
for low-energy river systems with
laterally stable channels. Ferguson (1981) mentioned a range of 1 to
60 W/m2
-- with a median of about 15 W/m2
-- for (laterally) inactive
channels. The low specific stream power of the upper Columbia River
channels indicates that little energy is available for lateral erosion of the
cohesive silty banks (Fig. 6), which explains the straight channel
morphology.
In our analysis of channel morphology, we followed the approach of
Van den Berg (1995), Bledsoe and Watson (2001), and Van den Berg and
Bledsoe (2003). Their work, however, has been criticised by Lewin and
Brewer (2001, 2003) who claimed that the analysis based on energygrain
size plots (such as Fig. 8) insufficiently appreciates a range of
important factors influencing river channel patterns. As an example of
the latter they state that "the limited migration of anastomosing
channels may well in part reflect low bankfull stream power, but bank
cohesion seems to be important,..." (Lewin and Brewer, 2003, p. 341).
Our analysis shows that -- although bank cohesion undoubtedly
contributes to the lateral stability of anastomosing upper Columbia
River channels -- the factors "energy" and "bed grain size" seem
sufficient to discriminate these inactive "straight" channels from meandering
and braided ones.
In meandering rivers, bedload is stored in point bars, and hence,
storage capacity is largely determined by the rate of lateral channel
migration. In the upper Columbia River -- where lateral migration of
channels is negligible and point bars are absent -- storage of bedload
takes place on the channel bed. For the channels in the study transect,
we estimate that channel shallowing occurs if bed aggradation exceeds
the average natural levee growth rate of 1.65 mm/year (Fig. 10). This
study shows negligible sediment transport rates in secondary channels
(Table 6), rendering bed aggradation rates in excess of 1.65 mm/year
highly unlikely for these channels. Calculated sediment transport rates
suggest that only in the main channel (channel 3) can bed aggradation
exceed 1.65 mm/year.
Our sediment budget calculations indicate an annual bed aggradation
in the main channel of 13.3 to 20.9 mm in the upper
anastomosing reach based on Fig. 9, a main channel bed area of
1,500,000 m2
(15 km long, 100 m wide), and a sediment density of
1770 kg/m3
(Locking, 1983). Even if bed aggradation is substantially
overrated and levee accretion is substantially underrated, bed
Fig. 6. Grain size distributions of bed and bank material of the main channel (channel 3).
Table 4
Grain size of the bed material.a
Mean D50 Mean D90 s
(m) (m) (-)
Channel 1 282 680 1.86
Channel 2 107 323 2.29
Channel 3 431 1077 2.01
Channel 4 253 405 1.54
Channel 5 301 970 2.18
a
D50, D90 = grain sizes in a distribution for which 50% and 90% by weight,
respectively, are finer; s =1/2 (D50/D16 +D84/D50) representing the gradation of the
bed material (based on means of D16, D50 and D84) with D50 as given earlier and D16 and
D84 representing grain sizes in a distribution for which 16% and 84% by weight,
respectively, are finer.
Table 5
Channel gradients and stream power.a
Sc 
(cm/km) (W/m2
)
Channel 1 7.8 0.38
Channel 2 7.4 0.07
Channel 3 6.8 2.30
Channel 4 7.6 0.19
Channel 5 7.4 0.58
a
Sc = channel gradient;  = specific stream power at bankfull discharge.
201B. Makaske et al. / Geomorphology 111 (2009) 194­205
aggradation very likely outpaces levee accretion in this reach (Fig. 11).
Levee accretion in the upper anastomosing reach may exceed the
1.65 mm/year that was inferred for the present study site, but is
unlikely to be much greater than 3.7 to 4.7 mm/year, representing the
maximum measured levee accretion during single flood seasons
(Makaske et al., 2002; Filgueira-Rivera et al., 2007). For the lower
anastomosing reach, main channel bed (35 km long, 100 m wide)
aggradation rates between 0.9 and 4.1 mm/year can be calculated.
With a long-term levee aggradation rate of 1.65 mm/year channel
shallowing is still likely to occur but at a much slower rate than in the
upper anastomosing reach (Fig. 11). Locally, some channel widening
Fig. 8. Diagram of channel type in relation to grain size and the flow energy parameter SvQbf (Sv is valley slope and Qbf is bankfull discharge), with a discriminator based on logistic
regression analysis indicating the probability of braiding (Bledsoe and Watson, 2001 with data from Van den Berg, 1995). Data points B (Barwon River) and Y (Yellow River) added
from Van den Berg's original data set, but not used by Bledsoe and Watson (2001) in the logistic regression analysis. Data on straight channels from the present study are plotted
along with some additional data from the literature. Channels with a sinuosity (P)N1.3 were classified as meandering by Van den Berg (1995) and Bledsoe and Watson (2001). Note
that some of these "meandering" channels are in fact laterally stable, having a low sinuosity index (Pind =channel length/length along channel-belt axis; Brice, 1964; Makaske, 2001,
p. 156). This means that the sinuosity of these channels is not the result of lateral migration. An example is the Barwon River (data point B) (Taylor and Woodyer, 1978), which plots
near the straight channels from this study. A preliminary discriminant line is drawn between the populations of meandering and straight (P1.3 and Pind b1.3) channels.
Fig. 7. Plot of median grain size (mean of multiple samples) of the channel bed material
against the specific stream power at bankfull discharge. Numbers of data points refer to
channel numbers.
Table 6
Total sediment transport (bedload+suspended load) at bankfull discharge according to
three equations.
Eng.-Hansen Van Rijn Mod. Van Rijn
Qt Qt Qt
(g/s) (g/s) (g/s)
Channel 1 126 0 16
Channel 2 24 0 0
Channel 3 5877 4411 9824
Channel 4 75 14 62
Channel 5 236 43 312
202 B. Makaske et al. / Geomorphology 111 (2009) 194­205
by slumping of river banks may partly compensate for the loss of flow
area caused by bed aggradation. Rapid bank retreat, however, will
destroy the levee and induce crevassing and avulsion.
Hence, the main channel is most liable to avulsions, many secondary
channels being virtually inactive morphologically. Exceptions are young
secondary channels formed by avulsion, which may rapidly develop
(widening and deepening) because of significant gradient/energy
advantage. Older secondary channels exist for very long periods, their
morphological evolution in general being determined by trapping of fine
suspended and washload in bank vegetation leading to very slow
channel narrowing (Makaske et al., 2002, their Fig. 14). Log jams may
speed up abandonment of secondary channels. Some secondary
channels receive ample bedload and coarse suspended load and totally
fill up with sand because of specific conditions near the channel
entrance. Channel 4, for example, is very shallow and -- unlike other
secondary channels -- totally fills up with sandy bedload because its
entrance is located on the inside of a bend in the main channel (Fig. 4).
Because of this configuration, this channel receives a relatively large
proportion of the coarse load from the main channel, as shown by
general bifurcation modelling by Kleinhans et al. (2008), which
demonstrated that meandering river bifurcations are inherently
unstable because of asymmetric division of water and sediment
associated with morphological evolution in meander bends. Usually,
within a geologically short period (a few decades to centuries), one
branch will be plugged by sediment, whereas the other branch receives
most of the discharge. Only very special conditions permitted longer
bifurcation stability. This modelling result explains the observation
(Makaske, 2001) that anastomosing rivers that are composed of
individual meandering channels, such as the Ovens and King Rivers
(Schumm et al.,1996), are much less common than anastomosing rivers
consisting of straight channels, like the upper Columbia River.
Anastomosis of the upper Columbia River initially was understood
to result from a downstream control -- i.e., a rising base level caused
by aggrading cross-valley alluvial fans (Smith, 1983). This idea was
challenged by Abbado et al. (2005), who showed that anastomosis is
most pronounced in the upper reach between Spillimacheen and our
study transect, where much sediment is introduced by the Spillimacheen
River (Fig. 3). Our results confirm their idea of an upstream
(excessive bedload supply) rather than a downstream cause (base
level rise) of upper Columbia River anastomosis. That both upstream
and downstream controls can give rise to frequent avulsions in lowenergy,
multiple channel systems was demonstrated in the RhineMeuse
delta by Stouthamer and Berendsen (2001) and Gouw and
Erkens (2007).
6. Conclusions
Results from this study indicate that anastomosing morphology of
the upper Columbia River is generated from the inability of the river to
transport its entire bedload. An important circumstance also, is that
the upper Columbia channels have no ability to migrate laterally
because of low stream power and erosion-resistant bank material. In a
diagram of channel type in relation to flow energy and median grain
size of the bed material, data points for the straight upper Columbia
Fig. 9. Longitudinal profile of the upper Columbia River main channel between locations 1 and 2 in Fig.1 -- showing a relatively steep upper reach and a relatively gentle lower reach -- and
bedload fluxes at bankfull discharge (220 m3
/s). Amounts and proportions of bedload (relative to total bedload stored in anastomosing reach) stored annually in the upper and lower
anastomosing reaches are also given. Light grey bars represent scenario with maximum difference between sediment storage in upper and lower anastomosing reach (based on minimum
calculated sediment throughput in transect), dark grey bars represent scenario with minimum difference between sediment storage in upper and lower anastomosing reach (based on
maximum calculated sediment throughput in transect). The channel gradients indicated were measured by Abbado et al. (2005) over the period 13­15 October 2000 when river discharge
was 55 m3
/s. These low stage water surface gradients may slightly deviate from high stage water surface gradients.
Fig. 10. Plot of natural levee thickness (including basal crevasse splay deposits) against
the age of the associated upper Columbia River channel (data from Makaske et al., 2002,
representing the same cross-valley transect as in the present study). Numbers of data
points refer to channel numbers. This graph indicates that after initial rapid crevasse
splay deposition of 2 m -- associated with avulsive channel formation -- natural levee
aggradation approaches a nearly constant rate of 1.65 mm/year for all channels. This
value may seem low relative to the long-term average upper Columbia floodplain
sedimentation rate of 1.75 mm/year (Makaske et al., 2002), but the latter includes rapid
crevasse splay sedimentation, which importantly contributes to overall long-term
floodplain sedimentation.
203B. Makaske et al. / Geomorphology 111 (2009) 194­205
River channels cluster separately from the data points for braided and
meandering channels. Because of lateral channel stability, no lateral
storage capacity for the surplus of bedload supplied to the Columbia
River main channel is created. Much of it is stored on the channel bed,
leading to gradual decrease in flow capacity, increased overbank
flooding and sedimentation, crevassing, and cutting of new channels
on the floodplain.
In a number of publications (e.g., Slingerland and Smith, 1998;
Törnqvist and Bridge, 2002), avulsion is described as a result of
floodplain accretion, creating steep cross-levee gradients that provide
an energy advantage for floodwaters relative to channel flow. Results
from this study point to bed aggradation as the primary mechanism of
avulsion in the upper Columbia River. This study quantitatively
demonstrates that bed aggradation in the upper Columbia main channel
outpaces levee accretion, whereas the morphological evolution of most
secondary channels is slow because of very low sediment transport
capacity. The development of steep cross-levee gradients is probably not
an important driver of avulsions in the upper Columbia River, because
valley width is limited and floodbasins are small with water levels
usuallyclose to channel waterlevels. In other words, in our case avulsion
seems rather "pushed" by bed aggradation than "pulled" by gradient
advantage.
Our calculations demonstrate that bed aggradation in the upper
anastomosing reach of the upper Columbia River where anastomosis is
particularly well developed, is much faster than in the lower reach, and
therebyconfirm the idea (Abbado et al., 2005) of an upstream (excessive
bedload supply) rather than a downstream cause (base level rise) of
upper Columbia River anastomosis. This study shows that -- next to
ample bedload supply -- anastomosis is favoured by a relatively low
stream power limiting lateral channel migration and creation of lateral
storage capacity for excess bedload. In general, low stream powers may
result from low slopes caused by rising base levels; so in this indirect
sense, rising base levels may contribute to anastomosis -- for example in
near-coastal settings subjected to relative sea level rise. For the upper
Columbia River, the present role of downstream control is uncertain
because aggradation of downstream cross-valley alluvial fans determining
local base level seems to be insignificant today, with incised fan
channels and stable fan surfaces.
Acknowledgements
This paper is dedicated to co-author Henk Berendsen, who initiated
this research many years ago and extensively commented on drafts of
this paper until shortly before he passed away. Maarten Kleinhans is
thanked for sharing his insights into bifurcation morphodynamics. We
are grateful to Stephen Tooth for his useful comments on a draft of this
paper. This paper benefitted from constructive reviews by Nicholas
Pinter and an anonymous referee. Financial support for carrying out
fieldwork and analyses was granted by Utrecht University and the
Natural Sciences and Engineering Research Council (NSERC) of Canada.
References
Abbado, D., Slingerland, R., Smith, N.D., 2005. Origin of anastomosis in the upper
Columbia River, British Columbia, Canada. In: Blum, M.D., Marriot, S.B., Leclair, S.M.
(Eds.), Fluvial Sedimentology VII. Special Publication of the International Association
of Sedimentologists, vol. 35. Blackwell, Oxford, UK, pp. 3­15.
Bledsoe, B.P., Watson, C.C., 2001. Logistic analysis of channel pattern thresholds:
meandering, braiding, and incising. Geomorphology 38, 281­300.
Brice, J.C., 1964. Channel patterns and terraces of the Loup Rivers in Nebraska. U.S.
Geological Survey Professional Paper 422-D, Washington, DC.
Carson, M.A., 1984. The meandering-braided river threshold: a reappraisal. Journal of
Hydrology 73, 315­334.
Engelund, F., Hansen, E., 1967. A Monograph on Sediment Transport in Alluvial Streams.
Teknisk Forlag, Copenhagen, Denmark.
Ferguson, R.I.,1981. Channel form and channel changes. In: Lewin, J. (Ed.), British Rivers.
Allen & Unwin, London, pp. 90­211.
Filgueira-Rivera, M., Smith, N.D., Slingerland, R.L., 2007. Controls on natural levee
development in the Columbia River, British Columbia, Canada. Sedimentology 54,
905­919.
Folk, R.L., Ward, W.C., 1957. Brazos River bar: a study in the significance of grain size
parameters. Journal of Sedimentary Petrology 27, 3­26.
Gouw, M.J.P., Erkens, G., 2007. Architecture of the Holocene Rhine-Meuse delta (the
Netherlands); a result of changing external controls. Netherlands Journal of
Geosciences - Geologie en Mijnbouw 86, 23­54.
Hare, F.K., Thomas, M.K., 1974. Climate Canada. Wiley, Toronto.
Jones, L.S., Schumm, S.A., 1999. Causes of avulsion: an overview. In: Smith, N.D., Rogers,
J. (Eds.), Fluvial Sedimentology VI. Special Publication of the International
Association of Sedimentologists, vol. 28. Blackwell, Oxford, UK, pp. 171­178.
Junck, M.B., 2009. Sub-surface Imaging of Cross-valley Alluvial Fans: using Ground
Penetrating Radar (GPR) and Electric Resistivity Ground Imaging (ERGI) to
determine Holocene Sedimentation Thicknesses in Trunk Valleys. M.Sc. thesis,
Department of Geography, University of Calgary, Alberta, Canada.
Kleinhans, M.G., Jagers, H.R.A., Mosselman, E., Sloff, C.J., 2008. Bifurcation dynamics and
avulsion duration in meandering rivers by one-dimensional and three-dimensional
models. Water Resources Research 44, W08454.
Kolb, C.R., 1963. Sediments forming the bed and banks of the lower Mississippi River.
Sedimentology 2, 227­234.
Leopold, L.B., Wolman, M.G., 1957. River channel patterns: braided, meandering and
straight. U.S. Geological Survey Professional Paper 282-B, Washington, DC.
Lewin, J., Brewer, P.A., 2001. Predicting channel patterns. Geomorphology 40, 329­330.
Lewin, J., Brewer, P.A., 2003. Reply to Van den Berg and Bledsoe's comment on Lewin
and Brewer (2001) "Predicting channel patterns," Geomorphology 40, 329­339.
Geomorphology 53, 339­342.
Locking, T., 1983. Hydrology and Sediment Transport in an Anastomosing Reach of the
Upper Colombia River, B.C. M.Sc. thesis, Department of Geography, University of
Calgary, Alberta, Canada.
Makaske, B., 1998. Anastomosing Rivers; Forms, Processes and Sediments. Nederlandse
Geografische Studies 249. Koninklijk Nederlands Aardrijkskundig Genootschap/
Faculteit Ruimtelijke Wetenschappen, Universiteit Utrecht, Utrecht, The Netherlands.
Fig. 11. Bed and levee accretion rates for the upper Columbia River main channel in the upper and lower anastomosing reaches (Figs. 3 and 9). Vertical exaggeration approximately 10×.
204 B. Makaske et al. / Geomorphology 111 (2009) 194­205
Makaske, B., 2001. Anastomosing rivers: a review of their classification, origin and
sedimentary products. Earth-Science Reviews 53, 149­196.
Makaske, B., Weerts, H.J.T., 2005. Muddy lateral accretion and low stream power in a
subrecent confined channel belt, Rhine-Meuse delta, central Netherlands. Sedimentology
52, 651­668.
Makaske, B., Smith, D.G., Berendsen, H.J.A., 2002. Avulsions, channel evolution and
floodplain sedimentation rates of the anastomosing upper Columbia River, British
Columbia, Canada. Sedimentology 49, 1049­1071.
Makaske, B., Berendsen, H.J.A., Van Ree, M.H.M., 2007. Middle Holocene avulsion-belt
deposits in the central Rhine-Meuse delta, The Netherlands. Journal of Sedimentary
Research 77, 110­123.
McManus, J., 1988. Grain size determination and interpretation. In: Tucker, M.E. (Ed.),
Techniques in Sedimentology. Blackwell, Oxford, UK, pp. 63­85.
Morton, R.A., Donaldson, A.C., 1978. Hydrology, morphology, and sedimentology of the
Guadalupe fluvial-deltaic system. Geological Society of America Bulletin 89,1030­1036.
Nanson, G.C., Croke, J.C., 1992. A genetic classification of floodplains. Geomorphology 4,
459­486.
Nanson, G.C., Huang, H.Q., 1999. Anabranching rivers: divided efficiency leading to
fluvial diversity. In: Miller, A., Gupta, A. (Eds.), Varieties of Fluvial Form. Wiley, New
York, pp. 477­494.
Oele, E., Apon, W., Fischer, M.M., Hoogendoorn, R., Mesdag, C.S., De Mulder, E.F.J., Overzee,
B., Sesören, A., Westerhoff, W.E., 1983. Surveying The Netherlands: sampling
techniques, maps and their application. Geologie en Mijnbouw 62, 355­372.
Schumm, S.A., Erskine, W.D., Tilleard, J.W., 1996. Morphology, hydrology, and evolution
of the anastomosing Ovens and King Rivers, Victoria, Australia. Geological Society of
America Bulletin 108, 1212­1224.
Slingerland, R., Smith, N.D., 1998. Necessary conditions for a meandering-river avulsion.
Geology 26, 435­438.
Smith, D.G., 1973. Aggradation of the Alexandra--North Saskatchewan River, Banff Park,
Alberta. In: Morisawa, M. (Ed.), Fluvial Geomorphology. Proceedings of the 4th
Annual Geomorphology Symposia series, Binghamton, NY, September 27-28, 1973.
State University of New York, Binghamton, pp. 201­219.
Smith, D.G., 1983. Anastomosed fluvial deposits: modern examples from western
Canada. In: Collinson, J., Lewin, J. (Eds.), Modern and Ancient Fluvial Systems.
Special Publication of the International Association of Sedimentologists, vol. 6.
Blackwell, Oxford, UK, pp. 155­168.
Smith, D.G., 1986. Anastomosing river deposits, sedimentation rates and basin
subsidence, Magdalena River, northwestern Colombia, South America. Sedimentary
Geology 46, 177­196.
Stouthamer, E., Berendsen, H.J.A., 2001. Avulsion frequency, avulsion duration, and
interavulsion period of Holocene channel belts in the Rhine-Meuse delta, The
Netherlands. Journal of Sedimentary Research 71, 589­598.
Tabata, K.K., Hickin, E.J., 2003. Interchannel hydraulic geometry and hydraulic efficiency
of the anastomosing Columbia River, southeastern British Columbia, Canada. Earth
Surface Processes and Landforms 28, 837­852.
Taylor, G., Woodyer, K.D., 1978. Bank deposition in suspended load streams. In: Miall, A.D.
(Ed.), Fluvial Sedimentology. Canadian Society of Petroleum Geologists Memoir, vol. 5.
Canadian Society of Petroleum Geologists, Calgary, p. 257­275.
Törnqvist, T.E., Bridge, J.S., 2002. Spatial variation of overbank aggradation rate and its
influence on avulsion frequency. Sedimentology 49, 891­905.
Van den Berg, J.H., 1995. Prediction of alluvial channel pattern of perennial rivers.
Geomorphology 12, 259­279.
Van den Berg, J.H., Bledsoe, B.P., 2003. Comment on Lewin and Brewer (2001):
"predicting channel patterns," Geomorphology 40, 329­339. Geomorphology 53,
333­337.
Van den Berg, J.H., Van Gelder, A., 1993. Prediction of suspended bed material transport
in flows over silt and very fine sand. Water Resources Research 29, 1393­1404.
Van Rijn, L.C., 1984a. Sediment transport, part I: bed load transport. Journal of Hydraulic
Engineering 110, 1431­1456.
Van Rijn, L.C., 1984b. Sediment transport, part II: suspended load transport. Journal of
Hydraulic Engineering 110, 1613­1641.
Van Rijn, L.C., 1993. Principles of Sediment Transport in Rivers, Estuaries and Coastal
Seas. Aqua Publications, Amsterdam, The Netherlands.
205B. Makaske et al. / Geomorphology 111 (2009) 194­205