You are on page 1of 21

Geomorphology 53 (2003) 75 95

www.elsevier.com/locate/geomorph

Evaluation of bed load transport formulae using field evidence


from the Vedder River, British Columbia
Yvonne Martin *
Department of Geography, University of Calgary, Calgary, AB, Canada T2N 1N4
Received 14 August 2001; received in revised form 17 June 2002; accepted 4 October 2002

Abstract
Bed load transport formulae have not been tested extensively over medium spatial and temporal scales. The lack of such
testing is undoubtedly due to a paucity of field data available for testing at these scales. Extensive transport, morphologic, and
discharge data, available for a 10-year period along an 8-km study reach of the Vedder River, British Columbia, are used to test
the predictive capabilities of bed load transport formulae at medium scales. The transport data were obtained using the
morphologic approach, which is based on the premise that changes in channel morphology reflect the transfer of sediment. Such
transport estimates can be made over larger spatial and temporal scales than is possible using traditional measurement techniques.
This study focuses on evaluating the original and revised versions of the Bagnold stream power formula, the Meyer-Peter and
Muller formula and a stream power correlation. Overall, the formulae are found to underpredict gravel transport rates for the
Vedder River. The discrepancy is particularly pronounced in 1982 1983 and 1987 1990, which are periods during which
significant dredging occurred. Dredging activity may have loosened the bed structure or resulted in a disequilibrium morphology,
thereby increasing sediment movement. Analysis is undertaken to assess how the use of mean daily discharge vs. 15-min
increment discharge series affects transport predictions. Results show that only the Meyer-Peter and Muller formula is
particularly sensitive to which discharge series is used in calculations. Modification of the Shields parameter is found to
significantly affect transport predictions, although it cannot explain alone the discrepancies between field data and calculated
results. Predictions of total deposition in the study reach are well within an order-of-magnitude for all equations. The Bagnoldtype formulae and the Meyer-Peter and Muller formula do not capture the relatively even distribution of deposition along the
study reach. Surprisingly, the simple stream power correlation captured this downstream pattern of deposition. Cumulative
distributions of predicted bed load transport are found to be most realistic for the original and revised Bagnold formulae. Results
of this study do not suggest that one particular formula is consistently preferred.
D 2002 Elsevier Science B.V. All rights reserved.
Keywords: Bed load transport; Transport formula; Field data

1. Introduction
Despite the widespread use of bed load transport
formulae in both applied and theoretical studies, critical
* Tel.: +1-403-220-6197; fax: +1-403-282-6561.
E-mail address: ymartin@ucalgary.ca (Y. Martin).

testing of such equations in the literature remains


limited. The most complete evaluation of bed load
formulae to date was undertaken by Gomez and Church
(1989), who noted that there are more bed load formulae in existence than there are reliable data to test
them. They assessed a large number of formulae over
individual flood events and cross sections using field

0169-555X/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(02)00348-3

76

Y. Martin / Geomorphology 53 (2003) 7595

and flume data that met a requirement of equilibrium


transport conditions. Testing the predictive capabilities of bed load formulae over medium temporal scales
(herein defined as approximately decadal) and spatial
scales (herein defined as reach length of at least several
kilometers) remains considerably more difficult due to
a notable lack of field data documenting bed load
transport rates at these scales.
Bed load is defined as the component of the total
load that moves in at least intermittent contact with the
bed. In many rivers, the pattern of entrainment and
deposition for this component of the load largely
determines channel morphology. From an applied
perspective, patterns of bed load can influence flooding (Cazanacli et al., 2002) and are critical in maintaining ecological diversity and habitat in stream
channels (Lisle, 1989). Despite longstanding recognition of the importance of bed load movement, understanding of processes involved in the entrainment,
movement, and deposition of bed load still remains
limited. This is due at least partly to the difficulties
encountered when attempting to observe bed load
entrainment and movement directly in the field. Measurement of bed load transport rates requires a considerable field effort; furthermore, representative measurements are difficult to obtain due to spatial and
temporal variability associated with bed load movement (Hubbell, 1987). As a result, very few data sets
exist that reliably document bed load transport rates.
Moreover, data sets that exist were generally collected
at individual cross sections over individual flood events
(e.g., Jones and Seitz, 1980; Burrows and Harrold,
1983).
Difficulties associated with field measurement of
bed load have resulted in a long history of interest in
developing equations for its prediction (e.g., duBoys,
1879; Schoklitsch, 1934; Bagnold, 1980). Such formulae are usually based on idealized hydraulic principles and attempt to relate the bed load transport rate to
some correlate of the flow (such as discharge, shear
stress or stream power). Bed characteristics are generally incorporated through the inclusion of a grainsize term, usually the median grain size. Research,
however, has indicated that bed structure may significantly affect sediment transport (Church et al., 1998).
Such factors have not generally been incorporated in
bed load formulae and this omission may further limit
their predictive capability.

Despite well-documented limitations in using such


formulae even at relatively small scales (Gomez and
Church, 1989), bed load formulae continue to be
widely used in geomorphological studies over a range
of scales. Applications range from applied studies
making use of the numerical model HEC-6 (HEC,
1991), which predicts erosion and deposition in rivers
and reservoirs, to larger-scale modelling of landscape
evolution in which the channel is treated as a subgrid
process (e.g., Martin, 1998). Nonetheless, research is
required to further evaluate the ability of transport
equations to predict phenomena of interest to geomorphologists, such as trends in aggradation and
degradation over medium to large spatial and temporal
scales.
The development of the morphologic approach to
sediment transport (e.g., Neill, 1987; Martin and
Church, 1995; McLean et al., 1999) has allowed geomorphologists to quantify transport patterns in the
field at medium scales. These same data have allowed
for evaluation of transport equations at these scales.
Nonetheless, testing of bed load formulae at medium
scales has been limited to only a few studies in the
literature (e.g., Carson and Griffiths, 1987; McLean et
al., 1999; Nicholas, 2000). Carson and Griffiths (1987)
evaluated the predictive capability of bed load formulae using time-averaged transport measurements available for the Waimakariri River and other gravel-bed
rivers in New Zealand. In particular, they focused on
the ability of bed load equations, including the Bagnold formula, to estimate transport in braided rivers.
They concluded that bed load formulae often underpredict transport rates by several orders of magnitude.
Nicholas (2000) continued work on the Waimakariri
River and used an extension of sediment transport
theory outlined in Paola (1996), which incorporated
the Meyer-Peter and Muller formula. Results were
found to significantly improve when braid intensity
and spatial variability in flow hydraulics at a cross
section were considered. McLean et al. (1999) used
transport estimates available for a wandering reach of
the Fraser River, based on the morphologic method, to
test several formulae and found that predicted transport
values compared well to field data.
The objective of this study is to evaluate bed load
formulae over decadal time scales using the extensive
transport and morphological field data available for an
8-km reach of the Vedder River, British Columbia.

Y. Martin / Geomorphology 53 (2003) 7595

This study is of particular significance as it uses an


independent set of decadal-scale transport measurements (Martin and Church, 1995), based on the morphologic method, to test several bed load formulae.
Annual values of bed load transport for the period
1981 1990 are available at regular intervals along the
Vedder River. Having bed load data available with
such extensive and regular coverage in both space and
time is unusual.
The present study contributes to the relatively few
studies that have tested the predictive capabilities of
bed load formulae at medium scales. It remains necessary to obtain further results documenting the relative
success and limitations encountered when applying
bed load transport formulae at these scales. Such
studies are important as they help provide a basis for
further identifying the degree of complexity that is
necessary to incorporate when using bed load formulae
to predict transport rates and associated channel
changes over extended time periods. This study does
not consider the situation of selective transport and
associated sediment sorting along the channel length
(e.g., Borah et al., 1982; Hoey and Ferguson, 1994;
Ferguson et al., 2001), which necessarily involves
greater computational complexity than the approach
adopted herein.
A particular focus of this study is to compare the
predictive capability of the original Bagnold formula,
one of the most widely used transport equations, with
several revised versions of this formula that were
presented in Martin and Church (2000). The revised
versions of the Bagnold formula were calibrated using
a larger and carefully compiled database than was the
case for the original formula and, therefore, may lead to
better transport predictions. In addition, the MeyerPeter and Muller formula is evaluated. These bed load
formulae were developed to depict equilibrium transport conditions (Gomez and Church, 1989), a situation
not often met in natural rivers and which may limit their
predictive capabilities. In contrast to the Bagnold and
Meyer-Peter and Muller formulae, which were derived
on the basis of extensive investigation, a simple stream
power correlation is defined and tested herein.
The formulae tested in the present study are straightforward to implement and, hence, remain tractable for
studies considering changes in fluvial systems at
medium and even large scales. In such cases, the
implementation of complex formulae may not be

77

practical or feasible due to computational constraints


and/or a lack of detailed information about channel
characteristics. For these same reasons, simple rules are
used to define values of input variables for calculations.
Results of this study provide baseline information on
the relative success and limitations of the formulae
when applied in this simple manner over medium times
scales and, moreover, provide a basis for determining
which aspects of the approach require further elaboration in future studies. For each particular application,
a balance has to be met between the degree of detail
necessary to obtain reasonable estimates of bed load
transport and the ease of implementation.

2. Field data for the Vedder River


2.1. Study reach
The Vedder River is the most distal reach of the
Chilliwack River and is located about 160 km east of
Vancouver (Fig. 1). The river drains an area of 1230
km2 in the Cascade Mountains and emerges from the
mountains onto an aggrading alluvial fan at Vedder
Crossing. Below this point, it is referred to as the
Vedder River. About 8 km downstream of Vedder
Crossing, the river has been channelized and is known
as the Vedder Canal.
The flood regime of the Vedder River consists of
storm-generated high flows in autumn and winter,
including rain-on-snow events, and snowmelt floods
in the spring and summer. Winter floods typically have
a duration of about 1 2 days, whereas summer floods
may persist for several more days. The mean annual
flood is about 350 m3 s 1. Measures to control flooding include the construction of dykes along a significant portion of the river and episodic dredging to
maintain the flood capacity of the river.
The Vedder River is a cobble gravel channel with
values of D50 ranging from 35 mm at Vedder Crossing
to 10 mm at the entrance to the Vedder Canal. Channel
slope is about 0.0046 at Vedder Crossing and decreases
in a downstream direction, with the most distal study
reach having a value of 0.00035.
The Vedder River consists of four distinct morphological reaches with bank-to-bank widths varying from
as low as 50 m to several hundred meters in the braided
reaches. In the most downstream reach, the river dis-

78

Y. Martin / Geomorphology 53 (2003) 7595

Fig. 1. Location of the Vedder River. The 10 study reaches along the Vedder River are shown.

plays a wandering pattern, with some bar development


and intermediate values of width. Upstream of this
location, the river is confined and becomes narrow and
deep. Widths in this reach appear to approximate the
regime width and, therefore, the river appears to be
stable (Church, University of British Columbia, personal communication, 2001). Upstream of this reach, a
braided pattern with large widths and extensive bar
development is found. Finally, the river becomes narrow and deep once again at Vedder Crossing.
Exceptional transport and morphological field data
exist for the Vedder River as several extensive studies
of this river have previously been undertaken (McLean,
1980; Martin and Church, 1995; Ferguson et al., in
review). Repeat surveys for 49 cross sections are
available along an 8175-m reach of the Vedder River
for the years 1981, 1982, 1983, 1984, 1987, and 1990.
Martin and Church (1995) used these cross-sectional
surveys to estimate volume changes for 10 condensed
reaches along the Vedder River. Two of the time
periods analyzed in this study cover 3-year intervals
(1984 1987 and 1987 1990) as annual surveys did
not occur in some years. Grain-size measurements are
available for the Vedder River as are other miscellaneous field data, which are introduced in this paper as
required.

2.2. Volume changes and transport rates


Successive surveys were analyzed to estimate changes in area for each of the 49 cross sections (Martin and
Church, 1995). Volume changes for the 10 reaches
were then determined by summing the volume changes
associated with cross sections contained in each particular reach. Martin and Church (1995) calculated transport rates using a sediment budget approach, whereby
the volume change in a reach reflects the difference
between the sediment input and output associated with
that reach. If volume changes are known along the
entire study reach, then one transport rate must be
known or be reasonably estimated to extend calculations along the river. Minimal gravel (material >2 mm)
was transported past the most downstream cross section and, therefore, a 0-gravel transport rate was
assumed at that location and calculations were extended upstream. Hence, the transport data of Martin and
Church (1995) represented only the gravel portion of
the sediment load. Grain-size sampling was undertaken
along the study reach to adjust all transport rates to the
portion >2 mm. The transport data were, in fact, gravel
transport rates. In addition, a porosity of 0.25 was
assumed for the volume changes and transport values
were adjusted accordingly. In the event that some

Y. Martin / Geomorphology 53 (2003) 7595

gravel was actually transported into the canal, the


reported field values represented lower-bound estimates. In fact, the assumption of a 0-gravel transport
rate lead to negative transport rates in the most distal
reaches during several time periods, and minor adjustments to the data were required.
Results of Martin and Church (1995) were found to
be consistent with an independent study undertaken by
Ham and Church (2000), in which aerial photography
was used in conjunction with a sediment budget
approach to estimate transport rates along the Chilliwack River upstream of Vedder Crossing. Over the
period 1983 1991, they estimated that about 54,600
m3 year 1 of gravel was transported past Vedder Crossing, which corresponds reasonably well with the
average annual transport of 40,000 m3 year 1 found
by Martin and Church for the period 1981 1989 (the
results of Ham and Church included the large flood in
November 1990 of about 776 m3 year 1, which is not
included in the present study). Error analysis undertaken in Martin and Church (1995) also suggests the
strength of their results. Details on data collection and
methodology are provided in Martin and Church
(1995).

3. Bed load formulae used in calculations


3.1. Original and revised Bagnold formulae
Various forms of stream power formulae have been
implemented in studies to estimate sediment transport
rates over a range of scales (e.g., Carson and Griffiths,
1987; Hoey and Sutherland, 1991; Tucker and Slingerland, 1994; Martin, 1998). The Bagnold equation,
which is the most well-known derivation of such a
formula, has been found to perform well in comparison to other bed load formulae at relatively small
scales (Gomez and Church, 1989). Bagnold presented
a remarkable graph that suggested the particular
strength of this formula (see Fig. 1 in Bagnold, 1986).
The original Bagnold equation has the form (Bagnold, 1980)


cs
x  xo 3=2
ib
ib ref
d=dref 2=3
cs  c
x  xo ref
D=Dref 1=2

79

where ib is the specific bed load transport rate (kg m 1


s 1), c is the specific gravity of fluid, cs is the specific
gravity of sediment, x is the specific stream power, xo
is the critical specific stream power, d is the depth, and
D is the characteristic particle size (usually denoted by
D50), and the subscript ref refers to some reference
value obtained from a reliable dataset (Bagnold used
data from Williams, 1970). Gomez and Church (1989)
introduced the term cs/(cs  c) to convert the immersed weight to dry weight, the latter being standard
in most fluvial transport formulae. Stream power is
defined as
x qQS=w

where q is the density, Q is the discharge, S is the


energy slope, and w is the channel width. Bagnold did
not include gravitational acceleration in his definition
of stream power. The critical stream power for bed
load transport is calculated from the equation
xo 5:75f0:04cs  cqg3=2 g=q1=2 D3=2
 log12d=D

where g is the gravitational acceleration and 0.04 is the


value chosen for Shields parameter. The threshold
stream power for bed load transport, xo, depends on
depth and grain size and may also depend on bed
structure if the Shields number is allowed to vary with
bed condition (Church, 1978; Church et al., 1998).
Values of the reference variables in Eq. (1) are
ib ref 0:1; x  xo ref 0:5;
dref 0:1; Dref 0:0011

Bagnold (1977, 1980, 1986) based the derivation


and subsequent analysis of this formula on limited
data. Hence, Martin and Church (2000) reanalyzed the
Bagnold formula using a large dataset to obtain the
most consistent empirical representation and to determine if a rationalization of the formula could be found.
On the basis of dimensional analysis, a rational version
of the Bagnold equation was found to have the form
(Martin and Church, 2000)
ib ~x  xo 1:5 d 1 D1=4

Additional terms in the equation collapse to a coefficient of 0.0139, which is placed in front of the right-

80

Y. Martin / Geomorphology 53 (2003) 7595

hand side of the equation. In addition, relations between excess stream power and bed load transport rate
were examined for a fixed slope of 1.5 to empirically
define coefficients. For the empirical trials, the rational
depth and grain-size scalings resulted in a coefficient
of 0.0793, in contrast to the rationally derived coefficient of 0.0139. The version of this equation with the
rationally derived coefficient is herein referred to as
rational formula #1, while the version with the empirically derived coefficient is referred to as rational formula #2 (Table 1).
3.2. Meyer-Peter and Muller formula
The Meyer-Peter and Muller (MPM) formula
(Meyer-Peter and Muller, 1948) has been used frequently to estimate rates of bed load transport (e.g.,
HEC, 1991; Nicholas, 2000) and was found to perform reasonably well by Gomez and Church (1989).
This equation has the form (Gomez and Church,
1989)
ib

cs
1000
cs  c
"
#3=2
QB =QKB =KG 3=2 dS  0:047fcs  c=cgD

0:25=cc=g1=3

as being equal to 1 in wide channels. The term KB/KG


accounts for the presence of form resistance in the
channel, which reduces the shear stress available for
transport.
3.3. Stream power correlation
A simple correlation between stream power and
bed load transport rate is defined for use in this study.
Although not expected to perform as well as more
detailed bed load formulae, in large part due to a lack
of threshold condition, simple stream power correlations are often used in longer-term studies of landscape evolution considering sediment transport at
subgrid scales (e.g., Tucker and Slingerland, 1994).
The ability of such simple relations to predict the bed
load component of total load is assessed herein. Bed
load data found in Gomez and Church (1989) were
analyzed to obtain the best-fit relation used in this
study (R2 = 0.53):
ib 0:0505 x0:89

This equation does not include terms for depth, grain


size, or threshold of transport.

and
4. Input variables
KB u=d

2=3 1=2

KG 26=D90 1=6

where D is the D50 of the material, the term u is the


velocity and QB is the proportion of the discharge that
acts on the bed. Meyer-Peter and Muller defined QB/Q

Table 1
Bagnold-type formulae implemented in calculations (see Martin and
Church (2000) for further details)
Version of Bagnold-type formula

Form of equation

Original Bagnold formula


Rational formula #1a
Rational formula #2b

See Eq. (1) in text


ib = 0.0139(x  xo)1.5d 1D1/4
ib = 0.0793(x  xo)1.5d 1D1/4

a
Version of the formula with rational scalings and coefficient
obtained from dimensional analysis.
b
Version of the formula with rational scalings and empirically
derived coefficient.

4.1. Discharge
Daily mean discharge data are available at Vedder
Crossing for the years 1911 to the present day, with a
gap in the record from 1932 to 1954. As no significant
tributaries enter the Vedder River between Vedder
Crossing and the canal entrance, these values are
representative of the entire study reach. The hydrograph from 1981 to 1990, which is directly used in
transport calculations, is shown in Fig. 2. Two significant flood events occurred during this period. The
flood of January 1984 had a discharge of 524 m3 s 1
while the flood of November 1989 had a value of 647
m3 s 1. The stage recorder at Vedder Crossing was
not functioning during these flood events. Therefore,
these discharge values were estimated using the
results of regression analysis between discharges
measured at Vedder Crossing and an upstream stage
recorder.

Y. Martin / Geomorphology 53 (2003) 7595

81

Fig. 2. Hydrograph for the Vedder River at Vedder Crossing for the study period 1981 1990. Values shown are mean daily discharges.

4.2. Width and depth


Widths at individual cross sections were derived
from the cross-sectional survey data. Bank-to-bank
width varies systematically between the four main
morphological reaches of the Vedder River. Preliminary analysis revealed that the use of bank-to-bank
widths, which include significant bars and islands, and
mean depths based on bank-to-bank widths resulted in
unrealistically low values of calculated stream power
and, hence, transport predictions. A similar finding was
made by Carson and Griffiths (1987) for the original
Bagnold formula. In their study, results improved when
transport in only the main channel of the braided river
was considered, although transport estimates were still
considerably lower than the field data.
During most floods, bed load transport is concentrated in the deepest parts of the channel with reduced
transport on upper bar surfaces. Therefore, widths of
bars and islands within 1 m of the lowest bank are
removed from bank-to-bank widths (Fig. 3A), an
approach similar to that of Nicholas (2000). Carson
and Griffiths (1987) suggested that the predictive
capability of bed load formulae may be improved by
using hydraulic parameters for individual slices across
channel cross sections. Such detailed morphological
information is not generally available for most rivers
and, therefore, this approach is not adopted in the
present study. Channel banks and edges of islands are
generally steep and approximately vertical along the

Vedder River. Therefore, values of width are assumed


to be constant as flow increases.
Distinct patterns of depth are also observed for the
four morphological reaches. As expected, the narrow
reaches have the greatest depths, while the widest
reaches have considerably shallower depths. Similar
to the procedure followed for width and for the same
reasons, the portions of the channel with bars situated
within 1 m of the surface were removed from calculations of mean channel depth (Fig. 3B). However, an
additional complication is encountered when considering depth. Depth, unlike width, must change with
increasing discharge. Fortunately, at-a-station hydraulic geometry equations exist for several locations
along the river. However, these locations are not
necessarily representative of all cross sections. Nonetheless, this information remains the best that is available and, hence, hydraulic geometry equations based
on these data were derived for the present analysis. The
average exponent for the hydraulic geometry equations
has a value of about 0.5 (Church, University of British
Columbia, personal communication, 2001):
d cQ0:5
b

10

where Qb is the bankfull discharge. This equation can


be rearranged to obtain the coefficient, c, for a particular cross section by inserting the bankfull discharge
(i350 m3 s 1) and the mean depth at this discharge at
that particular location.

82

Y. Martin / Geomorphology 53 (2003) 7595


Table 2
Grouped widths and depths for the four morphological reaches
(values are averages for each reach; cross sections contained within
each morphological reach are also shown)
Group

Width (m)

Depth (m)

1
2
3
4

103
94
137
56

1.73
2.98
1.51
2.63

(xs
(xs
(xs
(xs

1 4)
5 6)
7 10)
11)

study reach were not considered to be representative of


the morphology for the reach as a whole.
The Vedder River is confined within dikes that are
set back from the bank in most locations. During the
highest flows, some portion of the discharge will
overtop the channel banks, which affects the channeltransporting capacity in the main channel. This situation would have occurred during two events in the
study period, the large floods of 1984 and 1989. Such
overbank floods will lead to complex channel hydraulics, the simulation of which would require coupling the
formulae with sophisticated hydraulic models (Tucker
et al., 1999) beyond the scope of applications of bed
load formulae at medium to large scales. To preserve
simplicity, an assumption is made herein that the rules
defined for width and depth continue to apply at high
flows.
Fig. 3. (A) Widths along the Vedder River. All bars within 1 m of the
height of the lowest bank have been removed from the values of
width. (B) Mean depths along the Vedder River. Similar to the
approach followed for width, all bars within 1 m of the height of the
lowest bank have been removed from values of mean depth. In all
diagrams, the entrance to the Vedder Canal is at 0 m and Vedder
Crossing is at 8175 m.

Bed load equations are particularly sensitive to


rapid changes in energy gradient over space. To avoid
spurious results, a best-fit line is fit to the elevations

In reality, river geometry at a location is expected to


fluctuate over time about some overall average behavior for its morphological reach. To ensure that width
and depth values represent space- and time-averaged
behavior and do not reflect anomalies at a particular
location and time, grouped values of width and depth
are calculated for each of the four major morphological
reaches. The boundaries of the 10 study reaches of
Martin and Church (1995) are the locations used for
transport calculations in this study. Widths and depths
for these 11 cross sections are assigned the values for
the morphological reach in which the cross section is
located (Table 2). A similar approach was adopted by
Nicholas (2000), as individual cross sections in his

Fig. 4. Mean bed elevations along the Vedder River. The derivative of
the best-fit line is used to obtain values of gradient for input in
calculations.

4.3. Gradient and grain size

Y. Martin / Geomorphology 53 (2003) 7595

83

fore, adjustments must be made to compare test results


with the field data. McLean (1980) found that below
225 m3 s 1, the D50 of bed load in transport was in the
range 0.5 2 mm, whereas above this discharge the D50
rapidly increased to 20 mm (Fig. 6). A discharge of
about 225 m3 s 1 appears to be required for the
transport of significant amounts of gravel. Therefore,
only discharges >225 m3 s 1 are used in calculations
of gravel transport.
5.2. Results
Fig. 5. Median grain size (D50) along the Vedder River. Values of
D50 for input in calculations are obtained from the best-fit line.

along the river, and its derivative is used to calculate


slope (Fig. 4). The equation has the form
h 4:93exp0:00024 d

11

where h is elevation and d is distance upstream from


the Vedder Canal.
A set of 13 grain-size samples obtained at upstream
bar surfaces is available for the Vedder River (Martin
and Church, 1995). Grain size decreases systematically along the study reach. A best-fit line is fit through
the data to estimate grain size (Fig. 5)
D 14:8exp0:00013d

Calculated transport rates are compared to field


data in Fig. 7. The maximum flood during the period
1981 1982 was 142 m3 s 1, which is significantly
lower than the threshold discharge for significant
gravel transport. Hence, all transport equations predict
0 transport rates. Despite the low value of maximum
discharge, the field data indicate that some gravel
transport did occur during this period. This unusual
result cannot be explained (Martin and Church, 1995).
The period 1982 1983 experienced overall very
high transport rates in comparison to most other years,

12

where D is median grain size (D50) and d is distance


upstream from the Vedder Canal. The best-fit line is
likely a suitable representation for the overall pattern
of grain size as the samples themselves are expected to
incorporate some degree of sampling error (Church et
al., 1987). A complete set of field measurements for
grain size is available only for 1990. An assumption is
made that grain size is constant over the 10-year study
period, as any changes that occurred over the study
period are likely to be within the error range of the
grain-size estimates themselves.

5. Gravel transport
5.1. Comparing field data to calculated transport
rates
Transport rates reported in Martin and Church
(1995) represent gravel material only (>2 mm). There-

Fig. 6. Median grain size of bed load in transport for the Vedder
River measured near the town of Yarrow (data from McLean, 1980).
Note the large increase in D50 of bed load at about 225 m3 s 1.

84

Y. Martin / Geomorphology 53 (2003) 7595

Fig. 7. Predicted and observed gravel transport rates along the Vedder River. The error ranges shown are those given in Martin and Church (1995).
The maximum discharge for each period is also shown. Predicted values for 1984 1987 and 1987 1990 have been divided by the 3 years each
period covers for comparison with field values that are annual averages over the 3-year period. (A) 1981 1982, (B) 1982 1983, (C) 1983 1984,
(D) 1984 1987, (E) 1987 1990.

despite only having a maximum flood of 313 m3 s 1.


Calculated transport rates are significantly lower than
field transport rates in all cases, except at the most
downstream cross section. The transport rate at Vedder
Crossing (cross section 11) is about 66,000 m3 year 1
for 1982 1983, with all calculated transport rates
being lower by at least 20,000 m3 year 1.
Conversely, the period 1983 1984 experienced a
relatively high maximum discharge of 524 m3 s 1, yet
field results indicate that transport was relatively low in

comparison to other years. While calculated transport


rates are generally lower than field values, the disparity
between test results and field results is much reduced in
comparison to 1982 1983. The gravel transport rate
based on field data at cross section 11 was about 33,000
m3 year 1, with both the rational formula #2 and the
MPM formula providing results only several thousand
cubic meters per year below this value.
During the period 1984 1987, the maximum
annual floods were 273, 344, and 349 m3 s 1. Total

Y. Martin / Geomorphology 53 (2003) 7595

transport rates over this period were divided by the 3


years that this period covered to obtain annual averages
for both field and test results. Predicted values are once
again lower than field transport rates in most cases,
except in the lower reaches, for the stream power
correlation and the MPM formula. The discrepancy
between calculated and field results is relatively low in
comparison to other time periods.
The final period, 1987 1990, experienced a very
high discharge of 647 m3 s 1 in November 1989. The
next largest flood during this period had a value of only
250 m3 s 1. The field data show that transport activity
was very high during this period. Calculated transport
rates were significantly lower than values based on
field data. For example, the field value of transport at
cross section 11 was about 52,000 m3 year 1, with the
highest calculated value for the MPM formula being
only about half of that value.
Overall, the calculated transport rates are lower than
field transport estimates (Table 3). The rational formula #1 underestimates transport by the greatest
amount, with the results for the original Bagnold
formula and the rational version #2 formula being
only somewhat better. Surprisingly, the stream power
correlation is found to perform quite well with an
average C/O (calculated/observed) value of 0.56. In
particular, this formula captures the more gradual
decline in transport rates as displayed in the field data,
in contrast to the Bagnold-type formulae. The MPM
formula resulted in the best C/O value of 0.77 but is
not consistently a good predictor of transport rates as
indicated by the significant variability about the field
transport estimates.
Transport rates calculated for 1982 1983 and
1987 1990 are significantly lower than field values

85

when compared to other periods (the period 1981


1982 is omitted from this discussion as the threshold
condition for gravel transport failed). Martin and
Church (1995) noted that the periods in which significant dredging occurred, 1982 1983 and 1987
1990, coincided with the periods in which field transport rates were particularly high. Dredging was
believed to be a major cause of instability in the river
by loosening the bed structure, and thereby contributing significantly to the total mobile sediment volume.
Such disturbance may have inflated field transport
rates for 1982 1983 and 1987 1990 above the rates
that would have occurred if dredging had not taken
place. Average C/O values of transport are 0.16 and
0.13, respectively, which are lower than for other
periods. Dredging may possibly lead to a disequilibrium morphology, with the river expending much
energy and transporting large amounts of material in
subsequent floods as it fills in holes and rebuilds bars
(Ashmore, University of Western Ontario, personal
communication, 2001).
5.3. Sensitivity of gravel transport calculations
5.3.1. Discharge series
The use of mean daily discharges in transport
calculations may result in biased transport rates.
Flows above and below the mean value will occur
throughout the day, with variability expected to be
greatest during flood events. This variability
depends on the flashiness of the stream response
to triggering events, which in turn depends on
drainage area and other basin characteristics. The
drainage area of the Vedder River is 1230 km2 with
the duration of flood events being, at most, several

Table 3
Calculated gravel transport rates divided by field transport measurements (values represent averages of all cross sections)
Bed load equation

1981 1982

1982 1983

1983 1984

1984 1987

1987 1990

Averagea

Original
Rational #1
Rational #2
Stream power
MPM
Average
Material removed by
dredging during
period (m3)

0.0
0.0
0.0
0.0
0.0
0.0
18,000

0.12
0.029
0.19
0.28
0.19
0.16
75,000

0.28
0.064
0.37
0.54
0.99
0.45

0.46
0.10
0.59
1.32
1.61
0.82

0.087
0.022
0.12
0.13
0.30
0.13
187,000

0.24
0.054
0.32
0.56
0.77

Average for periods 1982 1983, 1983 1984, 1984 1987, and 1987 1990.

86

Y. Martin / Geomorphology 53 (2003) 7595

days. Therefore, significant variability in flood


values is expected over the course of a day. If
the discharge/transport rating curve for a formula is
highly nonlinear, then use of mean daily discharges
may lead to transport estimates that are negatively
biased.
Relations between discharge and predicted transport rate are shown in Fig. 8. The rating curves are not
highly nonlinear for any of the formulae, suggesting
that the use of mean daily discharges may not be a
significant factor in producing the negatively biased
transport predictions. Other factors, such as how often
threshold criteria are met for the two discharge series,
may be more important.
Discharge data measured at 15-min intervals are
available for a number of moderate, gravel-transporting floods during the study period. Such data are not
available for the significant transporting events of
January 1984 and November 1989. To test the effect
of using mean daily discharges vs. the 15-min data,
gravel transport rates are calculated for two sample
floods (Fig. 9). The first flood examined is that of July
11 15, 1983, and the second flood is that of February
24 26, 1986. For each flood, gravel transport is
calculated using mean daily discharges and the 15min data. Only discharge values exceeding 225 m3 s 1
are used in calculations.
Results for 1983 indicate that the Bagnold-type formulae and the stream power correlation are not sensitive to which discharge series is used in calculations
(Fig. 10A,B). However, transport estimates for the
MPM formula (Fig. 10B) differ significantly depend-

Fig. 9. Hydrographs for two sample floods. (A) July 11 15, 1983,
and (B) February 24 26, 1986.

ing on which discharge values are implemented. Results for the 1986 flood are similar to those of the 1983
flood for all formulae (Fig. 11).
The reasons underlying the greater sensitivity of the
MPM formula in comparison to other formulae are
now examined. Examination of calculations for the
Bagnold-type formulae reveals that the threshold condition for transport is always met for both discharge
series. Because the rating curves for these relations are
not highly nonlinear, minimal discrepancy is expected

Fig. 8. Relation between discharge and unit transport rate for all equations tested in this study.

Y. Martin / Geomorphology 53 (2003) 7595

Fig. 10. Gravel transport predictions for the July 1983 flood using
the mean daily and 15-min interval discharge series. (A) Results for
the three Bagnold-type formulae, and (B) results for the Meyer-Peter
and Muller formula and the stream power correlation.

between results. Some minor differences occur, for


example, when some of the 15-min data do not exceed
the 225 m3 s 1 threshold criterion for a particular day,
while the mean daily discharge for that day does exceed the criterion. On the other hand, the MPM formula is sensitive to the use of mean daily discharges.
Analysis of calculations reveals significant variability
as to when the threshold criterion is exceeded for the
two discharge series.
Overall, the low values of gravel transport predicted
by the Bagnold-type formulae and the stream power
formula do not appear to result from the use of mean
daily discharges. Some portion of the discrepancy
between results for the MPM formula and the field data
may likely be due to the use of mean daily discharges.
5.3.2. Bed structure
The Shields value is related to critical shear stress
for entrainment in the following manner:
scr cgqs  qD

13

where scr is the critical shear stress for entrainment, c is


the Shields parameter, g is the gravity, qs is the density

87

of sediment, q is the density of water, and D is the


grain size. The value of c changes depending on the
bed structure and configuration. Shields (1936) originally defined a value for this coefficient of 0.06.
Values ranging from 0.03 to 0.06 have since been
reported in the literature, with 0.045 being adopted as a
representative value in many studies (e.g., Komar,
1988). Other studies, however, have reported a much
greater range of values for the Shields parameter when
bed structure is considered (Church, 1978). For example, Williams (1983) found that for material >10 mm,
values can range from 0.01 to 0.25.
Shields parameters of 0.04 and 0.047 are contained
in the Bagnold-type formulae and the MPM formula,
respectively. Sensitivity analyses are undertaken for
these formulae by implementing Shields values ranging from 0.01 to 0.1. Because of similarity in the form
of the relation, the resultant patterns for the Bagnoldtype formulae are expected to be similar. Therefore,
results are shown only for the rational formula #2 and
the MPM formula (Figs. 12 and 13).
Results for the rational formula #2 indicate that for
the years in which dredging occurred, lowering the
Shields parameter decreases the discrepancy between
calculated and field transport rates. However, transport
predictions are still lower than field values. For the
periods in which dredging did not occur, 1983 1984
and 1984 1987, a decrease in the Shields parameter is
found to increase transport rates at several locations to
within the approximate range of field transport rates.
However, no strong basis exists for assuming that the
Shields parameter was substantially lower than 0.045
during these periods. Nonetheless, these results demonstrate that modification of the Shields parameter can
significantly affect transport calculations, although
alone it cannot explain the discrepancies between the
field data and calculated results.
A reduction in the Shields parameter to values
below 0.047 for the MPM formula significantly
improved the performance for the periods in which
dredging occurred. For example, reductions in the
Shields parameter for 1982 1983 increased transport
predictions to values relatively close to the field data at
several locations. However, the formula continues to
underpredict transport rates in the confined reaches.
For the period 1987 1990, a reduction in the Shields
parameter also improves calculated results at several
locations. The equation still significantly underpre-

88

Y. Martin / Geomorphology 53 (2003) 7595

Fig. 11. Gravel transport predictions for the February 1986 flood using the mean daily and 15-min interval discharge series. (A) Results for the
three Bagnold-type formulae, and (B) results for the Meyer-Peter and Muller formula and the stream power correlation.

dicts transport in the confined reaches. For the periods


in which dredging did not occur, 1983 1984 and
1984 1987, modification of the Shields parameter
reduces discrepancies between calculated and field
values in all reaches.

6. Total bed load transport


The field data only lend themselves to estimates of
gravel transport rates and not total bed load transport
rates. Hence, the ability of the equations to predict
total bed load transport cannot be directly assessed
and indirect methods are necessary.
6.1. Total volumes
The Vedder River is situated on an aggrading
alluvial fan, and net deposition occurs along its length.

Martin (1991) calculated values of total deposition,


which are differences between the input at Vedder
Crossing and output at the entrance to the Vedder
Canal. Values of total deposition calculated for the
five formulae are compared to the field data to assess if
total aggradation is constrained adequately. The ability
of the equations to predict the absolute magnitude of
transport remains unknown; only the relative differences in total transport rates are assessed using this
approach.
The field data require adjustment before they can be
compared to calculated results. Volume changes presented in Martin and Church (1995) include porosity
and all grain sizes; they are bulk changes. The porosity
of 0.25 used by Martin and Church (1995) for transport
calculations is used to remove porosity from all
volumes. Furthermore, the field volumes represent
the effects of bed load transport in addition to any
suspended load deposited as interstitial fill in the bed.

Y. Martin / Geomorphology 53 (2003) 7595

89

Fig. 12. Results of gravel transport calculations for the rational #2 formula using Shields values ranging from 0.01 to 0.1. The solid lines are the
field data and error ranges. The dashed lines are the predicted values using the modified Shields parameters. (A) 1982 1983, (B) 1983 1984, (C)
1984 1987, (D) 1987 1990.

Fig. 13. Results of gravel transport calculations for the Meyer-Peter and Muller formula using Shields values ranging from 0.01 to 0.1. The solid
lines are the field data and error ranges. The dashed lines are the predicted values using the modified Shields parameters. (A) 1982 1983, (B)
1983 1984, (C) 1984 1987, (D) 1987 1990.

90

Y. Martin / Geomorphology 53 (2003) 7595

Fig. 14. Predicted and observed values of total deposition. (A) 1981 1982, (B) 1982 1983, (C) 1983 1984, (D) 1984 1987, (E) 1987 1990.

The field data are adjusted in an attempt to remove the


portion of the volume change moved as suspended
load. Material < 0.1 mm generally goes directly into
suspension, while material >1 mm typically moves in
at least intermittent contact with the bed. Material
ranging in size from 0.1 to 1 mm may move either

as bed load or suspended load depending on flow


conditions. On this basis, field estimates of volume
change are adjusted to represent only that portion
which is expected to have moved as bed load. Grainsize distributions for the Vedder River are used to
determine the percentage of material >0.1 and >1 mm,

Table 4
Predicted total deposition divided by field values of total deposition (first value is for >0.1 mm and second value is for >1.0 mm)

Original
Rational #1
Rational #2
Stream power
MPM
Average
a

1981 1982a

1982 1983

1983 1984

1984 1987

1987 1990

Average

3.1 3.4
1.1 1.2
6.4 7.1
10.9 12.1
0
4.3 4.8

2.0 2.4
0.70 0.80
3.9 4.6
3.3 3.9
0.36 0.41
2.1 2.4

1.1 1.3
0.37 0.42
2.1 2.4
2.0 2.3
0.70 0.80
1.3 1.4

6.2 7.0
2.1 2.4
11.9 13.6
10.1 11.5
1.1 1.2
6.3 7.1

1.9 2.2
0.63 0.74
3.6 4.2
3.2 3.7
0.37 0.42
1.9 2.3

2.9 3.3
0.98 1.1
5.6 6.4
5.9 6.7
0.50 0.57

Average for this period does not include the Meyer-Peter and Muller formula.

Y. Martin / Geomorphology 53 (2003) 7595

91

Fig. 15. Downstream patterns of volume change along the Vedder River for the period 1983 84. The pattern for the original and revised Bagnold
formulae are similar so only one set of results is shown to represent the Bagnold-type equations. (A) field data, (B) original Bagnold formula, (C)
Meyer-Peter and Muller formula, (D) stream power correlation.

which are then used to adjust volume changes accordingly.


Calculated total volumes are compared to the field
data in Fig. 14 and Table 4. Values of C/O total volumes
are generally reasonable, with all calculated values well
within an order-of-magnitude of the field data. The
rational formula #1 performs the best with C/O values
close to unity. The rational formula #2 performs most
poorly, having C/O values of about 6.
6.2. Downstream depositional patterns
Analysis is now extended to evaluate amounts and
patterns of volume change occurring in the 10 study
reaches. The results for one time period, 1983 1984,
in which the predicted values of total deposition are
reasonable for all formulae, are displayed (Fig. 15). In
such a situation, total deposition is adequately constrained; the distribution of this material among the
study reaches can be evaluated. Field data show that

Fig. 16. Cumulative distributions of bed load transport for all equations evaluated in this study.

92

Y. Martin / Geomorphology 53 (2003) 7595

Table 5
Calculated gravel transport rates as a percentage of calculated total bed load transport rates (values represent averages of all cross sections)

Original
Rational #1
Rational #2
Stream power
MPM
Average
a

1981 1982

1982 1983

1983 1984

1984 1987

1987 1990

Averagea

0.0
0.0
0.0
0.0
n/a
0.0

56.7
55.2
55.2
6.9
93.6
53.5

89.0
87.7
87.7
13.8
100.0
75.6

46.4
44.8
44.8
5.9
94.9
47.4

59.8
56.3
56.3
3.3
99.1
55.0

63.0
61.0
61.0
7.5
96.9

Averages for periods 1982 1983, 1983 1984, 1984 1987, and 1987 1990.

deposition occurs in most reaches, except for the


middle, confined reaches and the most upstream reach.
Deposition volumes generally range from several
thousand up to 10,000 m3 year 1. The Bagnold-type
relations do not simulate this pattern. In all cases, a
large amount of material is deposited in the upper
reach, with very low values in downstream reaches.
The MPM formula also does not perform well. Deposition volumes are too low for most reaches, with
only several reaches having somewhat higher values.
Amounts and patterns of deposition predicted by the
stream power correlation are the most realistic and
show a more even distribution of deposition compared
to other methods.

ing relatively insignificant amounts of bed load. This


result occurs due to a lack of a threshold condition.
Given that significant morphological change on the
Vedder River is associated with large flows, this result
is a major shortcoming of the formula. In contrast,
results for the MPM formula suggest that only 3% of
the bed load movement occurs at discharges below 225
m3 s 1. Below about 225 m3 s 1, bed load movement
of material in the sand fraction is expected to occur
(refer to Fig. 6). The three Bagnold-type formulae
provide an intermediate result, with gravel transport
making up about 61 63% of the total bed load transport.
6.4. Bed load transport vs. total load

6.3. Cumulative distribution of transport rates


To obtain an estimate of the total sediment load at
Vedder Crossing, bed load transport rates are added to
the average annual value of suspended load for
Vedder Crossing (Church et al., 1989). Bed load
transport is then reported as a percentage of the total
load. Reported values typically range from just a few
percent to values exceeding 50% for rivers in mountainous drainage basins (Walling and Web, 1987).
Values for the original Bagnold formula, the rational
formula #1, and the MPM formula are reasonable
(Table 6). Results for the rational formula #2 and the

Cumulative distributions of bed load transport with


increasing discharge are used to assess the importance
of various discharges in transporting bed load (Fig.
16). In addition, calculated rates of gravel transport as
a percentage of calculated rates of total bed load are
shown in Table 5. For the stream power correlation,
about 95% of the total bed load transport occurs at
discharges < 200 m3 s 1. This formula overestimates
the relative amount of transport occurring at low
discharges, with medium to high discharges transport-

Table 6
Total bed load as a percentage of total load

Original
Rational #1
Rational #2
Stream power
MPM

1981 1982

1982 1983

1983 1984

1984 1987

1987 1990

Average

40.3
19.6
58.2
67.6
0.0

71.9
46.3
83.1
76.9
39.4

41.3
18.4
56.3
50.3
39.2

66.7
40.3
79.4
72.3
33.1

63.8
37.4
77.3
71.3
34.4

56.8
32.4
70.9
80.3
29.2

Y. Martin / Geomorphology 53 (2003) 7595

stream power correlation are higher than might be


expected.

7. Conclusions
7.1. Gravel transport
Gravel transport results for the stream power correlation are quite good in comparison to the other
formulae. This finding is surprising given the extreme
simplicity of this equation; the Bagnold and MPM
equations were derived on the basis of extensive investigation and include a threshold condition. Calculated
transport rates are generally lower than field values for
the Bagnold-type formulae and the MPM formula.
These formulae also fail to capture the relatively gradual decline in transport rates observed in the field
data. Hoey and Sutherland (1991) found that the
original Bagnold equation generally overpredicted
bed load transport rates for braided rivers that were
in equilibrium or aggrading. The Vedder River is situated on an aggrading alluvial fan, yet the original
Bagnold equation underpredicted transport rates in the
braided reaches.
Sensitivity analyses show that only the MPM formula is particularly sensitive to which discharge series
is used in calculations. The fortuitous inclusion of
dredging in the field data for particular time periods
allows for speculation regarding the role of bed structure in gravel transport. The years in which dredging
occurred are exactly the years in which the equations
underestimate transport rates by the greatest amount.
Dredging is expected to loosen the bed structure and,
hence, increase the volume of mobile sediment. Furthermore, natural changes in bed structure over space
and time, which are not accounted for in this study, may
be implicit in the field data. For example, the length of
time between gravel-transporting floods may affect the
strength of the bed structure. Finally, constants in the
equations could be tuned for specific bed conditions in
individual reaches, although no physical basis exists
presently for modifying equations in this manner.
7.2. Total bed load transport
All of the formulae are found to adequately predict
total deposition along the Vedder River, with all

93

results well within an order-of-magnitude of field


values. Surprisingly, the results for the stream power
formulae display the most realistic pattern of deposition along the study reach, showing fairly even
deposition rates. The other formulae do not adequately replicate this pattern, suggesting that greater
subtleties are involved in the entrainment and transport of sediment than are found in these transport
relations.
The stream power correlation predicts that over
95% of bed load transport occurs at discharges < 200
m3 s 1. This correlation contains no threshold condition, so transport occurs even during the lowest
flows. On the other hand, the MPM formula shows
that almost no bed load transport occurs at discharges
< 200 m3 s 1. During smaller flood events, some
amount of sand is likely to be in transport and, hence,
this result may be unrealistic. The Bagnold-type formulae display an intermediate result.
7.3. Final remarks
Results of this study do not unequivocally suggest
that one formula is strongly preferred. Although the
various equations employed in this study provide
reasonable predictions of overall aggradation along
the Vedder River, results suggest that bed load formulae may have a tendency to underpredict transport
rates over medium time scales not only for braided
reaches (Carson and Griffiths, 1987), but also for other
channel pattern types. Patterns of predicted gravel
transport also differ from the field data. Several possibilities may explain discrepancies between observed
and predicted transport rates found in this study: (i)
error associated with the field transport rates, (ii)
inadequate specification of input parameters, and (iii)
shortcomings of the formulae themselves, such as their
inability to account for bed structure.
Field transport estimates used in this study may
represent lower-bound values due to limitations of the
morphologic approach (see Martin and Church, 1995
for details). However, error analysis (Martin and
Church, 1995) and corroboration of transport rates at
Vedder Crossing (Ham and Church, 2000) suggest that
error in the field data cannot alone explain these
discrepancies. Discrepancies may possibly be due in
large part to the inability of generalized formulae in
straightforward applications to capture the true com-

94

Y. Martin / Geomorphology 53 (2003) 7595

plexity of the channel hydraulics and bed conditions.


However, for applications at medium to large scales, a
balance must be found between (i) overall ease of implementation of the approach and (ii) complexity
incorporated in the formulae and the definition of
input variables insofar as they improve transport
predictions. Applications may also be limited by the
availability of input data for the particular river being
investigated. In many applications of bed load formulae, one calculation is undertaken for each cross
section with the mean depth and width being input
into the equation, as was the case in this study. Such a
simple approach may decrease the predictive ability of
a formula. Alternatively, each cross section could be
divided into a series of slices for bed load calculations,
which would then be summed to obtain the total bed
load transport rate (see Carson and Griffiths, 1987).
However, such an approach requires very detailed
information about bed morphology and only in very
exceptional situations is such detailed morphological
data expected to be available.
A surprising result of this study is that the stream
power formula best captured the gradual decline in
gravel transport rates and the relatively even distribution of volume changes found in the field data for the
Vedder River. However, this simple correlation contains no threshold for transport and no term for grain
size. The lack of these terms in the formula may
represent a potential weakness as changes in entrainment thresholds or grain-size distributions for a particular river or between different rivers cannot be
accounted for in calculations. Hence, this simple correlation may be acceptable for providing gross predictions only in certain well-tested cases. The performance of the stream power correlation is likely to
differ for individual rivers and even over time for one
particular river when significant changes in channel
properties occur. In addition, the cumulative distribution results suggest that it may overemphasize the role
of smaller flows in transporting bed load. Hence, the
use of this equation should be approached with caution
as it may significantly overestimate total bed load
transport. This being said, the Bagnold-type formulae
and MPM formula, which were defined on the basis of
careful investigation, do not adequately simulate the
magnitudes and patterns of transport for the Vedder
River. On the basis of this and earlier studies, we
recommend that research be conducted to more thor-

oughly investigate the reasons why bed load formulae


appear to consistently underpredict transport rates over
medium time scales and to determine the degree of
increased complexity required to improve transport
predictions.

Acknowledgements
I would like to thank several individuals who provided valuable assistance at various stages of this project.
Mike Church and Rob Ferguson provided valuable
information and advice regarding field data for the
Vedder River. Peter Ashmore read a draft of the manuscript and provided thoughtful suggestions. I would
like to thank Sean Willett and Christopher Fuller for
participating in initial discussions. Lynne Campo of the
Water Survey Division, Environment Canada, provided assistance in obtaining the high-frequency discharge data for the Vedder River. An NSERC grant and
a URGC grant to the author provided funding for this
project. The manuscript benefited from the comments
of Basil Gomez and Vincent Neary.

References
Bagnold, R.A., 1977. Bedload transport by natural rivers. Water
Resources Research 13, 303 312.
Bagnold, R.A., 1980. An empirical correlation of bedload transport
rates in flumes and natural rivers. Proceedings of the Royal
Society of London A372, 453 473.
Bagnold, R.A., 1986. Transport of solids by natural water flow:
evidence for a worldwide correlation. Proceedings of the Royal
Society of London A405, 369 374.
Borah, D.K., Alonso, C.V., Prasad, S.N., 1982. Routing graded
sediments in streams: formulations. Journal of the Hydraulics
Division ASCE 108, 1486 1503.
Burrows, R.L., Harrold, P.E., 1983. Sediment transport in the Tanana River near Fairbanks, Alaska, 1980 81. U.S. Geological
Survey Water Resources Investigations Report, 20 81.
Carson, M., Griffiths, G., 1987. Bedload transport in gravel channels.
Journal of Hydrology, New Zealand 26, 1 151 (Special Issue).
Cazanacli, D., Paola, C., Parker, G., 2002. Experimental steep
braided flow: application to flooding risk on fans. Journal of
Hydraulic Engineering 128, 322 330.
Church, M., 1978. Palaeohydrological reconstructions from a Holocene valley fill. In: Miall, A. (Ed.), Fluvial Sedimentology. Canadian Society of Canadian Petroleum Geologists, Memoir, vol. 5.
Calgary, pp. 743 772.
Church, M., McLean, D., Wolcott, J., 1987. River bed gravels: sampling and analysis. In: Thorne, C., Bathurst, J., Hey, R. (Eds.),

Y. Martin / Geomorphology 53 (2003) 7595


Sediment Transport in Gravel-Bed Rivers. Wiley, Chichester,
pp. 43 79.
Church, M., Kellerhals, R., Day, T.J., 1989. Regional clastic sediment yield in British Columbia. Canadian Journal of Earth Sciences 26, 31 45.
Church, M., Hassan, M., Wolcott, J., 1998. Stabilizing self-organized structures in gravel-bed stream channels: field and experimental observations. Water Resources Research 35, 3169 3179.
duBoys, M.P., 1879. Etudes du regime et laction exerce par les
eaux sur un lit a` fond de graviers indefiniment affouibale. Annales des Ponts et Chaussees 5, 141 195.
Ferguson, R., Church, M., Weatherly, H., 2001. Fluvial aggradation
in Vedder River: testing a one-dimensional sedimentation model.
Water Resources Research 37, 3331 3348.
Gomez, B., Church, M., 1989. An assessment of bed load sediment
transport formulae for gravel bed rivers. Water Resources Research 25, 1161 1186.
Ham, D., Church, M., 2000. Bed-material transport estimated from
channel morphodynamics: Chilliwack River, British Columbia.
Earth Surface Processes and Landforms 25, 1123 1142.
Hoey, T.B., Ferguson, R., 1994. Numerical simulation of downstream fining by selective transport in gravel bed rivers: model
development and illustration. Water Resources Research 30,
2251 2260.
Hoey, T., Sutherland, A., 1991. Channel morphology and bedload
pulses in braided rivers: a laboratory study. Earth Surface Processes and Landforms 16, 447 462.
Hubbell, D.W., 1987. Bed load sampling and analysis. In: Thorne, C.,
Bathurst, J., Hey, R. (Eds.), Sediment Transport in Gravel-Bed
Rivers. Wiley, Chichester, pp. 89 106.
Hydrological Engineering Center (HEC), 1991. HEC-6, Scour and
Deposition in Rivers and Reservoirs, Users Manual. U.S. Army
Corps of Engineers, Davis, CA.
Jones, M.L., Seitz, H.R., 1980. Sediment transport in the snake and
clearwater rivers in the vicinity of Lewiston, Idaho. U.S. Geological Survey Open File Report, 80 690.
Komar, P., 1988. Sediment transport by floods. In: Baker, V., Kochel,
R., Patton, P. (Eds.), Flood Geomorphology. Wiley-Interscience,
New York, pp. 97 111.
Lisle, T.E., 1989. Sediment transport and resulting deposition in
spawning gravels. Water Resources Research 25, 1303 1319.
Martin, Y., 1991. Sediment budget from morphology. MSc Thesis,
Department of Geography, University of British Columbia, Vancouver. 184 pp.
Martin, Y., 1998. Modelling geomorphology in landscape evolution. PhD Dissertation, Department of Geography, University of
British Columbia, Vancouver. 253 pp.
Martin, Y., Church, M., 1995. Bed-material transport estimated
from channel surveys: Vedder River, British Columbia. Earth
Surface Processes and Landforms 20, 336 347.

95

Martin, Y., Church, M., 2000. Re-examination of Bagnolds empirical bedload formulae. Earth Surface Processes and Landforms
25, 1011 1024.
McLean, D., 1980. Flood control and sediment transport study of
the Vedder River. MASc Thesis, Department of Civil Engineering, University of British Columbia, Vancouver. 185 pp.
McLean, D., Church, M., Tassone, B., 1999. Sediment transport
along lower Fraser River: 1. Measurements and hydraulic computations. Water Resources Research 35, 2533 2548.
Meyer-Peter, R., Muller, R., 1948. Formulas for bedload transport.
Proceedings 2nd Meeting International Association of Hydraulic
Research, Stockholm, pp. 39 64.
Neill, C.R., 1987. Sediment balance considerations linking long-term
transport and channel processes. In: Thorne, C., Bathurst, J., Hey,
R. (Eds.), Sediment Transport in Gravel-Bed Rivers. Wiley, Chichester, pp. 225 239.
Nicholas, A., 2000. Modelling bedload yield in braided gravel
rivers. Geomorphology 36, 89 106.
Paola, C., 1996. Incoherent structure: turbulence as a metaphor for
stream braiding. In: Ashworth, P., Bennett, S., Best, J., McLelland, S. (Eds.), Coherent Flow Structures in Open Channels.
Wiley, Chichester, pp. 705 723.
Schoklitsch, A., 1934. Der geschiebetrieb und die geschiebefracht.
Wasserkraft Wasserwirtschaft 5, 1 7.
Shields, A., 1936. Applications of similarity principles and turbulence research to bed-load movement. W.M. Keck Laboratory of
Hydraulics and Water Resources, California Institute of Technology, Report No. 167. 43 pp.
Tucker, G.E., Slingerland, R.L., 1994. Erosional dynamics, flexural
isostasy, and long-lived escarpments: a numerical modeling
study. Journal of Geophysical Research 99, 12229 12243.
Tucker, G., Gasparini, N., Lancaster, S., Bras, R., 1999. A 3D
computer simulation model of drainage basin evolution and
floodplain evolution: Theory and applications. Technical Report
prepared for the U.S. Army Corps of Engineers Construction
Engineering Research Laboratory.
Walling, D.E., Web, B.W., 1987. Material transport by the worlds
rivers: evolving perspectives. In: Rodda, J.C., Matalas, N.C.
(Eds.), Water for the Future: Hydrology in Perspective. International Association of Hydrological Sciences, Proceedings of the
Rome Symposium, April 1987. International Association of Hydrological Sciences Publication, vol. 164, pp. 313 329. Rome.
Williams, G.P., 1970. Flume Width and Water Depth Effects in
Sediment Transport Experiments. United States Geological Survey Professional Paper 562-H. 37 pp.
Williams, G.P., 1983. Paleohydrological methods and some examples from Swedish fluvial environments. Geografiska Annaler
65A, 227 244.

You might also like