Professional Documents
Culture Documents
Downloaded from ascelibrary.org by UFRJ - Universidade Federal Do Rio De Janeiro on 05/30/23. Copyright ASCE. For personal use only; all rights reserved.
TRANSPORT O F SEDIMENT
By Patricia L. Wiberg 1 and J. Dungan Smith2
ABSTRACT: A bed load transport model based on the mechanics of sediment mov-
ing by saltation yields predicted values of bed load flux as a function of boundary
shear stress, grain diameter and density. The parameters required to calculate bed
load transport (particle velocity, bed load sediment concentration, and the height
of the bed load layer) can all be determined from our model, which computes
sequences of trajectories of individual saltating grains as well as the concentration
of moving grains that the flow can support. The latter is related to the momentum
the accelerating grains extract from the flow. Predicted curves of bed load transport
vs. boundary shear stress agree well with transport data collected by Gilbert, Meyer-
Peter et al., and the Waterways Experimental Station; measured shear stress was
corrected for pressure drag when bed forms were present. Comparison of predicted
bed load transport with common bed load equations reveals considerable similarity
among the relationships. The best agreement with the data is produced by rela-
tionships in which the transport rate vanishes as the shear stress approaches the
critical Shields' number, as in our model and Yalin's equation.
INTRODUCTION
101
particle trajectories, however, is not very extensive and the ultimate test of
any saltation model comes from using it to predict bed load transport rates
for which considerably more data are available. The flux of bed load sedi-
ment is given by the product of the downstream particle velocity and the
sediment concentration integrated over the height of the bed load layer. Ex-
panding the saltation model to produce a bed load transport model thus re-
quires a method for calculating the concentration; this is described in the
first section of the paper. All other parameters needed to calculate bed load
transport rates are available directly from the saltation model. In the follow-
ing sections, we compare the results of our model with existing flume mea-
surements of bed load transport rates, as well as with several commonly used
bed load transport equations.
The volume discharge of bed load sediment per unit width of flow, qs, is
given by the equation,
/•(8«) m
qs = csu4z (1)
J-n
where cs = cs(z) is the concentration of sediment in the bed load layer, z is
height above the bed, us = us{z) is the downstream (bed-parallel) component
of sediment velocity, -n is the height of the nonmoving bed surface, and
(8B),„ is the height of the bed load layer (maximum saltation height). The
saltation model described by Wiberg and Smith (1985) yields the trajectories
of single grains over a sediment bed by solving particle equations of motion
in which the acceleration of the grain is balanced by the gravitational and
fluid mechanical forces acting on it (lift, drag, and relative acceleration).
The governing equation, in vector form, is
d(us) d(Vf) CVAD . 3<uA)
psV —— = pV —-— + p —— uA uA + pVC,„ ——
dt dt 2 at
C A
+ 9-y1 l(ul)T- (ul)B]eL - (p, - p)VQ (2)
where uA is the difference between the fluid and particle velocity, uf — us,
( ) indicates an average over the grain, and the subscripts T and B refer to
the top and bottom of the grain; V is particle volume, AD and AL are particle
cross-sectional area normal to the drag and lift forces, respectively, p and
ps are fluid and sediment density, g is gravitational acceleration, and eL is
a unit vector in a bed-normal direction. The drag coefficient, CD, and the
coefficient of relative acceleration (virtual mass), C,„, are taken as those for
spheres, and the lift coefficient CL is set equal to 0.2 (see Wiberg and Smith
1985). In the derivation of Eq. 2, both RD =? |uA|D/v and R = ufh/v have
been assumed to be significantly greater than one; here, D is the nominal
sediment diameter, v is the kinematic viscosity of the fluid and h is the flow
depth. Neither of these constraints pose significant restrictions on the in-
tended use of Eq. 2. (For details, see Wiberg and Smith 1985.)
102
4r
3-
D *•
0
20 40 60 80 100 120 140
X/D
103
While the absolute magnitude of the concentration is not known yet, the
sediment concentration in the bed load layer can be written as cs = (cs)c*,
where (cs) is the vertically averaged value of the concentration.
In order to evaluate the magnitude of the sediment concentration, it is
necessary to consider what limits the number of grains that can be trans-
ported in the bed load layer. When a particle leaves the bed and is accel-
erated by the downstream component of the fluid flow, it extracts momentum
from the fluid, locally reducing the shear stress of the fluid. At some con-
centration, the sediment will extract enough momentum near the bed to re-
duce the boundary shear stress of the flow to the critical value and no ad-
ditional net erosion of particles will occur. If enough of the saltating grains
lodge in the bed to allow the local boundary shear stress to rise above the
critical value, additional particles will be mobilized until the boundary shear
stress again decreases back to the appropriate value. Owen (1964) pointed
out the importance of this self-equilibrating process and equated the critical
shear stress for this process to that for initial motion. He then used this
condition to provide a functional relationship between sediment concentra-
tion and boundary shear stress in a study of eolian saltation.
The first step in applying Owen's approach to the problem of determining
sediment concentration is to partition the total stress in the bed load layer
into the stress the fluid exerts on other fluid parcels, iy, and the stress the
sediment exerts on the fluid, T„ i.e.:
Ttotal = T/ + Ts (3)
If the thickness of the saltation layer is much less than the flow depth, the
total stress in this thin layer above the bed is approximately constant and
equal to the shear stress of the fluid, ib, just above the layer of saltating
grains. The sediment component of stress depends on the amount of mo-
mentum transferred from the fluid to sediment during each trajectory and
the concentration of saltating grains. The transfer of momentum from the
fluid to the sediment, resulting in downstream particle acceleration, is pri-
marily accomplished by the horizontal component of the drag force, FD =
(p/2)CDAD\uA\uA. This drag force component has a minimum value near the
top of the bed load layer where the sediment velocity approaches the fluid
velocity, and is maximum near the bed at the beginning of the trajectory
where the difference between the particle and fluid velocities is the greatest.
The sediment stress is equated to the downstream component of drag force
per unit area, where the appropriate area is the grain diameter times the
average downstream distance between two grains. This distance can be found
from the definition of sediment concentration as the ratio of sediment volume
to total volume (fluid plus sediment). Taking the sediment volume equal to
the volume of a single grain yields the average fluid volume between two
grains, V/cs. The grains can be anywhere in the volume, so two grains will
not, in general, lie in the same cross-stream plane or be at the same level
above the bed. If we wish to calculate the average distance between two
grains "lined-up" in the downstream direction, then this volume must be
divided by D2, yielding a distance of /(z) = V/[D2cs(z)], an area of ID, and
a stress of
104
l(z)D V AD
DAD
The geometric factor aD = ADD/V in this equation is 1.5 for spheres. Eval-
uating Eq. 3 at the top of the nonmoving grains of the bed surface, -n, gives
ftotai — T 6 = T>(T|) + T S (T|). If the fluid stress at the bed is equal to the critical
shear stress, as argued above, the sediment stress at the bed can be written
as T/TI) = T 6 — Tcr. Substituting Eq. 4 for T, produces an equation in which
everything is known except the vertically averaged sediment concentration.
Solving for this yields,
Jb
<c.) = ~ ] \ (5)
c*(iq) :
AD
All of the terms in Eq. 1 necessary to calculate bed load sediment flux
can now be evaluated using the saltation model. Figs. 2(a)-(c) show the
average saltation height, ( 8 ^ ) ^ / 1 ) , the average downstream particle velocity,
(Wj)avg/"*> a n < i m e vertically averaged bed load concentration, (c s ), as a func-
tion of transport stage, T* = T 4 /T„., for five sediment sizes that will be the
focus of the next section: D = 0.035 cm, 0.050 cm, 0.080 cm, 0.33 cm,
and 2.86 cm. The sediment density is taken to be ps = 2.65 g/cm 3 in all
cases. Fig. 2(d) shows the calculated volume flux of bed load sediment, qs
(in units of cm 3 /cm-j), as a function of transport stage for the same five
cases.
Verification of the bed load transport model described above must be made
by comparing bed load flux calculations to carefully made measurements of
bed load transport rates obtained under appropriate flow conditions. In par-
ticular, the flow should be steady, two-dimensional, and horizontally uni-
form. When ripples are present on the bed, the upper part of the flow should
be uniform with the bed roughness scaled by the bed form size; the inner
flow will not be locally uniform, but the spatial average of the inner-flow-
induced boundary shear stress can be related to the sediment transport rate
(Smith and McLean 1977). The parameters required for the model are grain
size and density, water temperature and density, and boundary shear stress.
Most of the sediment transport data recorded under suitable flow conditions
comes from flume studies; for the present purpose, the most useful of these
were conducted by Gilbert (1914), Meyer-Peter et al. [(1934); the Meyer-
Peter et al. data employed in this paper were taken from Brownlie (1981)]
and the U.S. Waterways Experimental Station (1935). There are many other
bed load transport studies in the literature, but for the most part they are
limited in scope and add little to these three data sets, or are inappropriate
for some other reason. For example, Guy et aL (1966) conducted an exten-
sive study that focused primarily on bed form evolution, but it is, in large
measure, unsuitable for the present purpose because of the three-dimensional
105
FIG. 2. Bed Load Transport Parameters As Function of Transport Stage, r*, for
Five Sizes of Quartz-Density Sediment: (a) Mean Saltation Height; (b) Average
Downstream Particle Velocity; (c) Vertically Averaged Concentration; (d) Volume
Flux of Sediment
nature of the flow. One study that does add to these comprehensive inves-
tigations is that of Fernandez Luque and van Beek (1976), who examined
lower plane-bed transport for which little other data is available.
A number of sets of measurements have been chosen from the Gilbert,
Meyer-Peter et al., U.S. Waterways Experimental Station (WES) and Fer-
nandez Luque and van Beek data sets for the purpose of testing the sediment
flux calculations described above. The selection was made to include a wide
range of sediment sizes with each size represented by a reasonable number
of bed load transport measurements covering as large a range of shear stresses
as possible. In some cases this was accomplished by merging measurements
for the same sediment size made by several investigators; good agreement
106
ments made either by one or several investigators, grain sizes that were roughly
the same, i.e., within about 10% of each other, were grouped together pro-
vided that the differences in their nondimensional critical shear stress values,
(T*)cr = T„/[(ps — p)gD], were also less than 10%.
Nondimensional sediment transport cj> = qs/[(ps/p — l)gD 3 ] 1/2 is plotted
as a function of nondimensional shear stress, T* = ib/[{ps — p)gD] in Fig.
3(a) for the combined data of WES and Gilbert for D = 0.05 cm, p, = 2.65
g/cm 3 . These measurements cover the whole range of bed load transport
from initial motion to upper plane-bed transport. The data fall into two dis-
tinct ranges of shear stress for relatively low transport rates. The lower-left
set of points corresponds to sediment transported in the lower plane-bed phase
(over a flat bed). As the shear stress on the bed increases, the transport rate
increases until a transport stage T* = T6/Tcr of 2 to 2.5 is reached, at which
point ripples begin to form. The ripples produce a drag on the flow, denoted
here as form drag, that reduces the shear stress on the true bed, denoted
here as skin friction. It is the skin friction, T^, that scales the forces on the
sediment grains in the neighborhood of the bed, and hence, that is related
to the sediment transport rate. While the shear stress of the outer flow, T„,
increases as the ripples form, the stress available to transport sediment de-
creases and there is a commensurate drop in the transport rate. Once ripples
cover the bed, the transport rate again increases with an increase in T„; the
lower-right set of points in Fig. 3(a) correspond to measurements made on
a rippled bed. Skin friction can be determined with reasonable accuracy from
the outer shear stress over rippled beds provided the bed forms are two-
dimensional and their heights and wavelengths are known; the procedure will
be described briefly in the discussion of the WES data below.
In the upper portion of the plot, the curve formed by the bed form data
appears to merge with an extension of the lower plane-bed data. Here, the
bed is essentially planar again (upper plane-bed) and form drag is no longer
important. This transition to upper plane-bed conditions can occur in several
ways depending on the flow conditions. First, as the shear velocity begins
to approach the settling velocity of the saltating particles, turbulent velocity
fluctuations start to affect the particle trajectories, making them significantly
longer. Under these conditions, denoted incipient suspension, the relatively
short wavelength bed forms begin to decay in amplitude and the form drag
on these features is reduced, thereby automatically increasing the effective
shear stress on the bed and causing an increase in the intensity of turbulence
on the upper part of the ripples. Thus, once the bed forms begin to decay
from their equilibrium forms, the transition to a flat bed proceeds rapidly.
The bed can also go through the transition from rippled to flat bed because
of limited flow depth, even if the conditions for incipient suspension have
not been reached. The maximum amplitude of two-dimensional bed forms
is restricted roughly to less than one-fifth of the depth of flow. In shallow
flows where the bed forms quickly reach their maximum heights, increases
in transport stage result in longer wavelengths and thereby can cause the bed
forms eventually to become so long compared to their amplitude that they
are essentially indiscernible. The maintenance of the bed forms by the flow
is also dependent on Froude number, F = (w)/Vgft, which varies inversely
with depth, h. As the Froude number increases toward the critical value F
107
10' I01 : e
j©
• o D = 0.05lcm,0.052cm(WES)
10° I0C r
/i
o 6° /oo
JO" 1 r 10" r
of °
$
„ £o with bedform
10" io- - o o o „2,o> correction
o o 3»j
03
: o
0
o /<B
F!G. 3. Nondimensional Bed Load Transport, 4>, as Function of Nondimensional Shear Stress, T*, f o r D = 0.05 cm: (a) Data with Calculated
Curve; (b) Data after Bed Form Correction.
= 1, the factors controlling bed stability are altered and the bed may become
essentially planar before antidunes develop at supercritical Froude numbers.
Downloaded from ascelibrary.org by UFRJ - Universidade Federal Do Rio De Janeiro on 05/30/23. Copyright ASCE. For personal use only; all rights reserved.
The latter two situations are particularly likely to exist in flume channels
where the flume dimensions, pump capacities, and sidewall effects limit the
flow depth that can be used. The decrease in the ratio of bed form height
to wavelength in either of these shallow-flow situations again results in re-
duced form drag and increased shear stress at the bed; however, sediment
can still be moving predominantly in the saltation mode.
Gilbert (1914), Meyer-Peter et al. (1934), and the Waterways Experi-
mental Station (WES) personnel (1935) measured transport rates of sand and
gravel in parallel-sided flumes. Combinations of water depth, surface slope,
flow discharge and flume width were employed to produce a variety of shear
stresses in the experiments. The boundary shear stress of a flow over a flat
bed is given by the product of the flow depth and surface slope, ds/dx, i.e.,
T(, = — pgh(ds/dx). While the hydraulic radius (wetted perimeter divided by
the cross-sectional area of flow) is often used instead of depth when cal-
culating shear stress in natural channels where the roughness of the sides is
comparable to the bottom, the depth is a more appropriate parameter for
flumes with glass sides because the walls are much smoother hydrody-
namically than the bottom. If the flume width is not sufficiently large with
respect to the depth, however, the effect of friction on the sidewalk can
significantly affect the surface slope. This effect of sidewall friction can be
removed from flume data using a sidewall correction such as Williams' (1970)
experimentally determined adjustment,
^measured
^corrected ~ ~. T~> y^)
where b is the flume width; the coefficient 5.5 is consistent with length
measured in centimeters (0.055 in SI units). In the analysis that follows, Eq.
6 was employed to correct all shear stress measurements, but data requiring
more than a 5% correction were discarded to minimize error associated with
uncertainties in adjusting for sidewall effects.
Gilbert (1914) measured transport rates for eight sizes of sand and gravel
with diameters ranging from D = 0.03 c m t o f l = 0.49 cm. These mea-
surements cover a particularly large range of shear stresses, in many cases
including upper plane-bed transport and antidunes. Unfortunately no mea-
surements were made of the height and wavelength of bed forms, so only
the lower and upper plane-bed measurements can be related to the skin fric-
tion on the bed. Many of the flows were quite shallow (~5 cm), so the
ripples and dunes were almost certainly depth limited. For the most part,
the upper plane-bed data correspond to supercritical flow conditions, and
with the possible exception of the D = 0.03 cm case, incipient suspension
was not responsible for the observed transition from dunes to upper plane-
bed, which occurs at a roughly constant transport stage of nine to ten. It
appears, therefore, that the sediment was transported primarily by saltation
in Gilbert's upper plane-bed runs. The grain sizes in the runs used for the
model comparison are D = 0.03 cm, 0.05 cm, 0.08 cm, and 0.32 cm [grades
A, C, D, and F in Gilbert's (1914) notation].
109
FD f CDH\ H ]V'
where \ is the bed form wavelength. The drag coefficient over bed forms is
not well known, but velocity measurements over large dunes by Smith and
McLean (1977) indicate that when separation occurs just downstream of the
bedform crest, the drag coefficient CD is approximately 0.2. The roughness
parameter of the inner flow, (z0)sf, depends on the local roughness length
scale of the bed. Applying this correction to the WES data generally yields
good results. Fig. 3(b) shows the same WES data plotted in Fig. 3(a) after
the bed form correction is made. The correction brings the bed form data
into good agreement with the lower plane-bed data, and at higher transport
stages the data trend up toward the upper plane-bed values.
The Meyer-Peter et al. (1934), see Brownlie (1981), data for sand and
gravel include measurements for a number of grain sizes over a relatively
narrow range of shear stresses, in some cases primarily in the bed form
range; no information about bed form size is given. The data set is notable
in that it was, in part, the basis of the well-known Meyer-Peter and Miiller
(1948) and Einstein bed load equations. The 2.86 cm data of Meyer-Peter
110
I0 1 I0 1
• • D = 0.031cm, 0.037cm upper plane-bed M%%
(Gilbert)
l 0 o L O D = 0.03lcm,0.035cm
10° r
(WES)
io- io-'-
with bedform
io- 10" ° SP
of correction
10" 10"
Ioo o oo O OCfflO O
FIG. 4. Nondimensional Bed Load Transport as Function of Mondimensional Shear Stress for I) = 0.035 cm: (a) Data with Calculated
Curve; (b) Data after Bed Form Correction.
10'
Downloaded from ascelibrary.org by UFRJ - Universidade Federal Do Rio De Janeiro on 05/30/23. Copyright ASCE. For personal use only; all rights reserved.
• D= 0.079cm (Gilbert)
* D = 0.09cm (Fernandez Luque)
C
I0
10" I _
4>
10- 2 _
10 - 3
10"
io- 6 8 10 -I 6 8 I0C
et al., which is all lower plane-bed, will be used in the subsequent com-
parisons. More recently, Fernandez Luque and van Beek (1976) measured
bed load transport rates for coarse sand and gravel at low transport stages.
All of their measurements were made on a planar bed, although in many
cases the bed was tilted between 12° and 20°. Their data for D = 0.09 cm
and 0.33 cm with a bed slope of zero are used to provide additional low
transport rate data for comparison with the bed load model; each of their
values represents an average of at least 20 measurements.
Figs. 3-6 show the combined data of Gilbert, WES, Meyer-Peter et al.,
and Fernandez Luque and van Beek for grain sizes, D = 0.035 cm, 0.05
cm, 0.08 cm, 0.33 cm and 2.86 cm, with density ps = 2.65 g/cm 3 . The
rippled-bed data for 0.035 cm and 0.05 cm sizes, corrected for form drag,
are plotted in Figs. 3(b) and 4(b); the Gilbert bed form data is not included
in these plots because insufficient information is available to make the form-
drag correction. Gilbert's shear stress values forD = 0.08 cm and 0.32 cm
when bed forms were present, which again could not be corrected, are in-
dicated in Figs. 5 and 6(a) by the small symbols. In each figure the solid
112
10' 10'
• D=0.32cm(Gilbert) x D = 2.86 cm (Meyer-Peter et al.)
* D= 0.33cm(Fernandez Luque)/
10° 10°
10" 10" I _
10-2 10"
io- 10- 3
FIG. 6. Data and Caiculated Curve of Mondimensional Bed Load Transport as Function of Nondimensional Shear Stress for: (a) D = 0.33
cm; (h) D = 2.86 cm.
curve shows the calculated relationship between the nondimensional volume
transport, <)> = qs/[(ps/p - l)gD 3 ] 1/2 , and nondimensional shear stress, T#
Downloaded from ascelibrary.org by UFRJ - Universidade Federal Do Rio De Janeiro on 05/30/23. Copyright ASCE. For personal use only; all rights reserved.
= T(,/[ps - p)gD].
In general, the calculated curves are in good agreement with the lower
plane-bed data, and extend upward into the upper plane-bed data. The ap-
parently planar bed values that lie to the right of the curve for D = 0.33
cm border on the bed form values and may have included subtle bed form
features. The calculated transport rates for the upper plane-bed, D = 0.035
case appear to overpredict the measured transport. This might be associated
with two effects not included as yet in the bed load model. First, the number
of particles moving in the bed load layer becomes quite large at the higher
transport stages [Fig. 2(c)] and interparticle collisions may become impor-
tant. Secondly, the extraction of momentum from the flow by the sediment
modifies the velocity profile in the bed load layer to some extent (see the
previous discussion of calculating the concentration). It is not obvious what
effect the intergranular collisions would have on the transport rate; however,
the extraction of momentum would tend to decrease the drag force on the
grains, thereby decreasing the particle velocity and, perhaps, the flux rates.
It is likely that the latter effect significantly outweighs the former, and that
the calculated fluxes for saltating sediment can be considered an upper bound
on the actual flux rates.
ANALYSIS
The good agreement between the calculated transport curves and the ex-
perimental data illustrated in Figs. 3-6 lends strong support to the validity
of the saltation/bed load transport model described in the first section. As
mentioned previously, Einstein and Meyer-Peter and Miiller used portions
of the data represented in these figures to set the coefficients in their well-
known bed load equations. It is interesting to compare these and several
other expressions for bed load transport to the calculated curves. Fig. 7 shows
a number of bed load transport expressions plotted with the corrected D =
0.05 cm data [from Fig. 3(b)] and the calculated curve. Meyer-Peter and
Miiller (1948) found that the relatively simple relationship equivalent to $
= 8(T* - 0.047)3/2 fit the Meyer-Peter et al. (1934) D = 0.52 cm and D
= 2.86 cm gravel data and Gilbert's (1914) gravel data (D = 0.32 cm, D
= 0.49 cm and D = 0.70 cm). Fernandez Luque and van Beek (1976) used
a similar equation, <> j = 5.7[T* - (T*)„.] 3/2 , to fit their data for low transport
rates (7% < 2.5). Aside from the lower value of the coefficient, the Fer-
nandez Luque and van Beek equation differs from that of Meyer-Peter and
Miiller in that the nondimensional critical shear stress is explicitly included;
the constant 0.047 in the latter equation is essentially an average value for
the nondimensional critical shear stress over the range of sizes employed in
the experiments.
Einstein (1942, 1950) formulated his semi-empirical transport expression
by relating the rate of bed load transport to the number of particles passing
through a unit cross section normal to the flow at the bed. The flux of par-
ticles was expressed in terms of the probability that a grain will experience
sufficient lift to leave the bed and that it will complete a certain number of
hops before it stops moving. This yields a functional relationship between
nondimensional transport, 4>, and nondimensional shear stress, T* with two
114
10°
IO-1
<p
io- 2
IO"3
lw
|Q-2 2 4 6 8 |Q-I 2 4 6 8 |Q0
tionship underestimates the high transport data, but this is not surprising
since the constant in their equation was set for lower plane-bed transport
data. The constant in the Meyer-Peter and Muller equation was set by com-
parison to higher transport rates, and yielded a higher value. Pursuing this
one step further, we note that Wilson (1966) made some transport measure-
ments for upper plane-bed conditions only and fit a Meyer-Peter and Miiller-
type equation to it obtaining a still larger coefficient, § = 12[t* - (T*)cr]3/2.
The variation of the constant for these three experiments suggests that the
coefficient relating 4> and [T# — (T*) C J 3 / 2 is a function that increases with
shear stress, and that a modified version of the Meyer-Peter and Muller equa-
tion, 4> = 7[T* — (T*)CJ 3/2 . where 7 is a function of T# — (T*)cr that varies
roughly between 5 and 15, would be likely to produce good results. It is
also noteworthy that all of these empirical or semi-empirical bed load trans-
port equations fall in the range of the corrected bed form data, so that, at
least implicitly, these expressions yield transport rate as a function of skin
friction rather than the outer shear stress.
Table 1 compares the agreement between the data and the various curves.
To do this, we calculated the root-mean-square difference between the mea-
sured values of T* and the values of T* computed from the various transport
expressions for each measured value of <j>, i.e., {2[log(T*)meos - log(T*)cafc]2/
N}1'2, where N is the number of measurements. Although boundary shear
stress is the independent variable in bed load equations, the largest error in
the measurement is associated with the shear stress rather than the sediment
transport rate. In all of the cited measurements, boundary shear stress is
determined indirectly in terms of surface slope, depth or hydraulic radius,
and flume width. The D = 2.86 cm data displayed in Fig. 6(b) clearly il-
lustrates the uncertainty in the value of T*. Thus, for this comparison of the
equations to the data, the difference is computed in terms of the boundary
shear stress. Corresponding values were calculated in terms of the difference
116
Aside from the measurement error, the use of depth, h, or hydraulic ra-
dius, R, to calculate boundary shear stress can produce significant differ-
ences in the computed stress if the channel width to depth ratio is not large
enough that R = h. In addition, differences in the way each investigator
corrected for side-wall effects and bed form drag also affect the computed
shear stresses in the measurements. The values of the shear stress represented
in Table 1(a) are computed in the manner that most closely agrees with each
investigator's original analysis. For the model values, shear stress was com-
puted in terms of depth, with Williams' sidewall correction and the bed form
correction given by Eq. 7. Einstein (1950) and Meyer-Peter and Miiller (1948)
both used a side-wall-corrected value of R and a bed-roughness adjustment
to calculate shear stresses for the data they used to determine the coefficients
in their respective equations. The corrections used by Einstein and by Meyer-
Peter and Miiller were different, but produce almost identical values of T*.
Yalin (1963) used depth in the derivation of his equation, but used Einstein's
1942 data, in terms of hydraulic radius, to set his free coefficient; shear
stresses computed from the data by Einstein's method were compared to
Yalin's equation to produce the variances shown for Yalin in Table 1(a).
Fernandez Luque and van Beek (1976) apparently used a corrected hydraulic
radius in the analysis of their plane-bed transport data; the overall variance
is the same whether depth or hydraulic radius is used, the former being
employed to produce the values in Table 1(a). Only plane-bed and bed form
or bed roughness corrected data were used to calculate the variances in Ta-
ble 1.
Table 1(b) compares the agreement between the data and the equations
when boundary shear stress is calculated as T = — pghds/dx in all cases. As
argued previously, shear stress calculated in terms of depth will give the
best estimate of the actual boundary shear stress for channels in which the
sides are significantly smoother than the bottom, as in a glass-walled flume
or a natural channel with bed forms, although it may still be necessary to
make a side-wall correction. In all cases in Table 1(b), the transport rela-
tionships predicted by Yalin and by our model result in the smallest differ-
ence between the data and the curves. This is largely because the transport
rate goes to zero in these relationships as jb approaches the critical shear
stress for initial motion, Tcr, as is observed in the data when shear stress is
computed in the manner described here. In contrast, both the Einstein and
the Meyer-Peter and Miiller equations are invariant when plotted on 4> — T#
axes. The transport rate also goes to zero as T6 approaches J„ in the Fer-
nandez Luque and van Beek expression, but as discussed above, their re-
lationship was based on low-transport-rate data and significantly underesti-
mates the observed transport at higher transport stages. A modified version
of the Meyer-Peter and Miiller equation including the critical shear stress
and a coefficient that is a function of shear stress could be expected to work
well.
The Meyer-Peter and Miiller (1948), Einstein (1942, 1950) and Yalin (1963)
bed load equations all used some subset of the Gilbert (1914) and Meyer-
Peter et al. (1934) data to determine the coefficients in their equation, in-
cluding the D = 0.08 cm and D = 2.86 cm data of the respective studies.
(Einstein used only these two sets of measurements for his 1950 equation.)
117
ioV
10c
10"1 ?_
4> /irnx
/r
11 jit.
10- 2 r //j* 10-2 u
/ / / • D= 0.079cm
/ / ^ * D=2.86cm
10" 3 ' '•' / Model (D= 0 . 0 8 c m ) _
Model(D=2.86cm)
k — Einstein ( 1 9 5 0 )
FIG. 3. Comparison of Einstein and Meyer-Peter and Muller Bed Load Equations with Calculated Curves and Combined Data for D = 0.08
cm and 2.SS cm: (a) ib = -pghds/dx; (h) T„ = -pgR'ds/dx,
These two data sets can be fit together to produce a fairly smooth curve
covering a wide range of transport values as shown in Fig. 8; in doing this,
Downloaded from ascelibrary.org by UFRJ - Universidade Federal Do Rio De Janeiro on 05/30/23. Copyright ASCE. For personal use only; all rights reserved.
however, the grain size dependence is lost. The larger gravel data are all
for lower plane-bed transport while the smaller size includes transport over
bed forms and in upper plane-bed flow. Fig. 8(a) shows the data with T*
calculated in terms of h with Williams' side-wall correction; the values in-
dicated by the small symbols are Gilbert's bed form data, which could not
be corrected using Eq. 7. Fig. 8(b) shows the data with T* calculated in
terms of Einstein's (1950) sidewall- and bed-roughness-corrected R; here the
bed-roughness correction was applied to all values, including plane-bed mea-
surements, as Einstein appears to have done; the results when the Meyer-
Peter and Miiller corrections are used are almost identical.
The curves predicted by the model for both grain sizes shown in Fig. 8(a)
indicate that the transport relationships for these sizes are considerably dif-
ferent. Curves calculated using Yalin's bed load equation for two sediment
sizes agree very well with the model results. As expected, the Einstein curve
lies closer to the data, particularly the 2.86 cm gravel values, in Fig. 8(£>)
where the stress is calculated from the corrected hydraulic radius; the same
is true for the Meyer-Peter and Miiller relationship. This good level of agree-
ment requires that the entire Einstein or Meyer-Peter and Miiller correction
(for side-wall effects and for bed-roughness, even for planar bed data) be
applied to the data. While this can be relatively straightforward for flume
data, the complications of natural systems, e.g., channel curvature and mul-
tiple bed form length scales, makes application of these shear stress correc-
tions problematic for field data. If we consider the Einstein and the Meyer-
Peter and Miiller curves in Fig. 8(a) in the range 4> = 10~4 to 10~3 where
the data for the two sizes approach the respective critical shear stresses, both
curves fall near the middle of the range of nondimensional critical shear
stress values [(T#)cr = 0.03-0.06] for grain sizes of fine sand and larger
(essentially any size that moves as bed load). Thus, these equations basically
represent average transport relationships.
There are two significant respects in which the bed load equations dis-
cussed above and our model do not reflect natural bed load transport con-
ditions. First, all of the bed load transport expressions included in this com-
parison relate bed load sediment transport to skin friction, i.e., the shear
stress at the actual bed. This is straightforward when the bed is flat, but
when ripples or dunes are present on the bed accurate estimates of bed load
flux can only be obtained after a correction is applied to the shear stress on
the mean bed (T = — pQhds/dx) in order to determine a spatially-averaged
estimate of the skin friction. It might seem more appropriate to take the bed
forms into account in the calculations and to relate transport rates directly
to the measured (uncorrected) shear stress. The difficulty with relating bed
load transport to this stress is that a unique relationship must exist betwen
the height and the wavelength of bed forms as a function of grain size and
shear stress, and this relationship must be known accurately. An attempt was
made during this analysis to construct such a relationship using the WES
data for D = 0.02 cm, 0.035 cm, and 0.05 cm, but even for this limited
range of sizes and flume conditions the dependence of ripple height and
wavelength on transport stage could not be generalized. Thus, until we have
a better understanding of the relationship between flow parameters and bed
119
CONCLUSIONS
The bed load transport model presented herein differs from existing meth-
ods of calculating transport in the bed load layer in that the formulation
proceeds directly from the equations of motion for a particle in a fluid and
it contains no coefficients set using bed load transport measurements. The
good agreement between the calculated values and data taken from Gilbert
(1914), Meyer-Peter et al. (1934), and "Studies of river bed materials" (1935)
suggests that the basic model accurately reflects the dominant features of the
mechanics of sediment moving as bed load, at least until the concentration
of sediment is sufficiently high that intergranular collisions become impor-
tant. Not only can sediment flux rates be predicted for a large range of
conditions, but the details of the saltation layer also can be examined with
the model. Accurate knowledge of saltation parameters is important for a
120
ACKNOWLEDGMENTS
This work was supported by the National Science Foundation under Grant
OCE-8310712.
APPENDIX I. REFERENCES
121
Meyer-Peter, E., Favre, H., and Einstein, A. (1934). "Neuere Versuchsresultate uber
den Geschiebetrieb." Schweizerische Bauzeitung, 103(13).
Meyer-Peter, E., and Miiller, R. (1948). "Formulas for bed load transport." Proc,
2nd Meeting Int. Association Hydr. Res., Stockholm, 39-64.
Owen, P. R. (1964). "Saltation of uniform grains in air." J. Fluid Mech., 20(2),
225-242.
Smith, J. D., and McLean, S. R. (1977). "Spatially averaged flow over a wavy
surface." / . Geophysical Res., 82, 1735-1746.
"Studies of river bed materials and their movement, with special reference to the
lower Mississippi River." Paper 17, U.S. Waterways Experimental Station, Vicks-
burg, Miss.
Yalin, M. S. (1963). "An expression for bed load transportation." J. Hydr. Div.,
ASCE, 89(HY3), 221-250.
Wiberg, P. L., and Smith, J. D. (1985). "A theoretical model for saltating grain in
water." J. Geophysical Res., 90, 7341-7354.
Wiberg, P. L., and Smith, J. D. (1987). "Calculations of the critical shear stress for
motion of uniform and heterogeneous sediments." Water Resour. Res., 23(8), 1471-
1480.
Williams, G. P. (1970). "Flume width and water depth effects in sediment-transport
experiments." Prof. Paper 562-H, U.S. Geological Survey, Reston, Va.
Wilson, K. C. (1966). "Bed load transport at high shear stress." J. Hydr. Div.,
ASCE, 92(HY6), 49-59.
Subscripts
avg = average value;
/ = fluid;
s = sediment;
D = drag; and
L = lift.
123