Nonuniform sediment transport in alluvial rivers

Transport de sédiments non uniformes en rivière alluviale

WEIMING WU, SAM S.Y. WANG and YAFEI JIA, National Center for Computational Hydroscience and Engineering
School of Engineering, University of Mississippi, MS 38677, USA

ABSTRACT
A correction factor has been developed in this paper to account for the hiding and exposure mechanism of nonuniform sediment transport. This fac-
tor is assumed to be a function of the hidden and exposed probabilities, which are stochastically related to the size and gradation of bed materials.
Based on this concept, the formulas to calculate the critical shear stress of incipient motion and the fractional bed-load and suspended load transport
rates of nonuniform sediment have been established. These formulas have been tested against a wide range of laboratory and field data and com-
pared with several other existing empirical methods. The predictions by these newly proposed formulas are very good.

RÉSUMÉ
Dans ce papier un facteur de correction a été développé pour tenir compte de l’effet de masquage et d’exposition de sédiments non uniformes en mouve-
ment. Ce facteur est supposé être une fonction des probabilités de masquage et d’exposition, elles-mêmes liées de façon stochastique à la taille des grains et
à la distribution granulométrique des matériaux du lit. A partir de ce concept, on a établi les formules pour calculer la tension tangentielle critique de mise en
mouvement et les débits solides en charriage et suspension des classes granulométriques de sédiments non uniformes. Ces formules ont été testées sur un
large éventail de données de laboratoire et de terrain, et comparées avec plusieurs autres méthodes empiriques. Les prédictions des nouvelles formules pro-
posées sont très bonnes.

1 Introduction the hiding and exposure effect in the modeling of nonuniform

sediment transport. Until now, most of the studies on the nonu-
Determining the critical condition for sediment incipient motion
niform sediment transport are based on introducing some kind
and the sediment transport rate is very important in hydraulic
of correction factors to account for this hiding and exposure
engineering. After DuBoys published his research on the bed-
effect and use these factors to modify the existing formulas of
load transport rate in 1879 and Shields (1936) proposed the
uniform sediment transport. In the methods for determining the
curve for the prediction of the critical bed shear stress of incipi-
incipient motion of nonuniform sediment developed by Egia-
ent motion, the uniform sediment movement has been exten-
zaroff (1965), Ashida and Michiue (1971), Hayashi et al.
sively investigated and the transport mechanism is reasonably
(1980), Parker et al. (1982) and Andrews (1983), as well as in
well understood. However, the state-of-art for estimating the
the methods for the fractional sediment transport rate developed
nonuniform sediment transport is still inadequate. The pioneer-
by Proffit and Sutherland (1983), Bridge and Bennett (1992),
ing research to fractionally calculate the nonuniform bed-mate-
Fang and Yu (1998) and Karim (1998), the correction factors
rial load transport rate was attributed to Einstein (1950).
were related to bed-material size by
Afterwards, Egiazaroff (1965), Ashida and Michiue (1971),
Hayashi et al. (1980), Qin (1980), Xie and Chen (1982; see
η i = f 1  -----i ; or ------i-
d d
(1)
Zhang and Xie, 1993), Parker et al. (1982) and Andrews (1983)  dm d 50
developed several formulas to determine the incipient motion of
nonuniform sediment mixtures. Ashida and Michiue (1971), where di is the diameter of the ith fraction of sediment; dm and d50
Parker et al. (1982), Proffit and Sutherland (1983), Misri et al. are the arithmetic mean and 50% sieve diameters of bed materials.
(1984), Samaga et al. (1986a), Mittal et al. (1990), Bridge and Einstein (1950) proposed a comprehensive hiding factor in his
Bennett (1992), Patel and Ranga Raju (1996), and Fang and Yu bed-load function to account for the interaction effects between
(1998) proposed several methods for calculating the fractional coarse and fine particles. After having pointed out the inaccu-
transport rate of nonuniform bed-load. Hsu and Holly (1992) racy of Einstain’s method in the case of discontinuous grada-
also proposed a method to predict the gradation of nonuniform tion, Misri et al. (1984) assumed that the motion of fine
bed-load by considering the probability and availability of mov- particles is dominated by the lift force while the motion of
ing sediment. Samaga et al. (1986b) and Karim (1998) estab- coarse particles is by the drag force, and proposed a semi-theo-
lished empirical functions for estimating the fractional transport retical hiding-exposure correction factor. This correction factor
rates of suspended load and bed-material load. was revised subsequently by Samaga et al. (1986a), Mittal et al.
In the processes of nonuniform sediment movement, the coarse (1990) and Patel and Range Raju (1996), and in general, it can
particles on the bed are easier to be entrained than the uniform be expressed as
sediment of equivalent sizes, because they have higher chance
of exposure to the flow. The situation is reversed for the fine τ'
η i = f 2  M , -----b, τ *i (2)
particles on the bed due to the fact that they are more likely  τc 
sheltered by coarse particles. Therefore, it is needed to consider

Revision received July 20, 2000. Open for discussion till June 30, 2001.

JOURNAL OF HYDRAULIC RESEARCH, VOL. 38, 2000, NO. 6 427

where M is the Kramer’s uniformity coefficient; τ’b is the grain
shear stress; τc is the critical shear stress for the arithmetic mean
size dm; τ*i = τ’b/[(γs – γ)di]; and γs and γ are the specific weights
of sediment and fluid, respectively.
However, the correction factors given in Eq. (1) only involve
the non-dimensional grain size di /dm or di /d50 and can not effec-
tively account for the influence of bed-material gradation. The
correction factor proposed in Eq. (2) introduces the Kramer’s
uniformity coefficient, but it is more complex and thus less
attractive in applications. In this study, the hiding and exposure
correction factor is related to the bed-material gradation and the Fig. 1. Definition of exposure height of bed material.
hidden and exposed probabilities. Based on this correction fac-
tor, the formulas for the incipient motion and the fractional The total hidden and exposed probabilities of particles di can be
transport rates of nonuniform bed-load and suspended load obtained by summing Eqs. (4) and (5) over all fractions, respec-
have been established. tively,

N
dj
2 New development on the hiding and exposure factor p hi = ∑ pbj --------------
di + d j
- (6)
j=1
The drag and lift forces acting on a particle staying on the bed
depend on how it is situated and surrounded by others. If there
is no other particle on its upstream side, it is exposed com- N
di
pletely to the flow and has maximum upwinding area and expo- p ei = ∑ pbj --------------
di + d j
- (7)
j=1
sure height; otherwise, its upwinding area and exposure height
are reduced. As shown in Fig. 1, we assume that sediment parti-
where N is the total number of particle size fractions of nonuni-
cles are spheres with various diameters and define the exposure
form sediment mixtures; phi and pei are the total hidden and
height ∆i for a particle with size di as the elevation difference
exposed probabilities of particles di. A correlation of phi + pei = 1
between the apexes of this particle and its upstream particle. If
exists between phi and pei. By using them, the hiding and expo-
∆i > 0, the particle di is considered to be at exposed state, and if
sure factor is defined as
∆i < 0, it is at hidden state. Because sediment particles usually
distribute on the bed randomly, ∆i is a random variable. It is m
η i =  -----ei-
assumed to follow a uniform probability distribution f. If the p
(8)
 p hi
upstream particle is dj, f can be expressed as

where m is an empirical parameter.

 1 ⁄ ( d i + d j ), – d j ≤ ∆ i ≤ d i
f =  (3)
 0, otherwise In the situation of uniform sediment, phi = pei = 0.5 and ηi = 1,
which means the hidden and exposed probabilities are equal. In
The probability of particles dj staying in the front of particles di the situation of nonuniform sediment, pei ≥ phi for the coarse
can be assumed to be the percentage of particles dj in the bed particles, and pei ≤ phi for the fine particles. This can be demon-
material, pbj. Therefore, the probabilities of particles di hidden strated with a simple example. For a sediment mixture with two
and exposed by particles dj can be obtained from Eq. (3) as size fractions d1 = 1 mm, pb1 = 0.4 and d2 = 5 mm, pb2 = 0.6, we
can get ph1 = 0.7 > pe1 = 0.3, ph2 = 0.3667 < pe1 = 0.6333. It
dj shows that more coarse particles are exposed and more fine
p hi, j = p bj --------------
- (4)
di + d j ones are hidden.

3 Threshold for incipient motion of nonuniform sediment

di
p ei, j = p bj --------------
- (5) Introducing the hiding and exposure factor defined by Eq. (8) to
di + d j
modify the criterion for sediment incipient motion proposed by
Shields (1936), we obtain the formula for determining the criti-
cal bed shear stress for the incipient motion of nonuniform sed-
iment,

τ ci p m
- = θ c  -----ei-
---------------------- (9)
( γ s – γ )d i  p hi

428 JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 38, 2000, NO. 6

where τci is the critical shear stress for particle di in nonuniform
sediment mixtures; and θc can be interpreted as the non-dimen-
sional critical shear stress for the corresponding uniform sedi-
ment or the mean size of bed materials. θc = 0.03 and m = –0.6
are determined by using laboratory and field data in this study.
To verify Eq. (9), it is necessary to know the threshold at which
sediment particles start moving. Theoretically, the threshold
should be defined as zero bed-load transport rate, but it is not
meaningful in practical situation. Many experiments show that
even if the flow strength is much weaker than the critical condi-
tion proposed by Shields, there are still some sediment particles
moving on the bed. Kramer (1935) divided the sediment move-
ment into four stages but his criteria are only qualitative and
difficult to apply. Therefore, several low levels of bed-load
transport rate were suggested as the quantitative critical condi-
tion for incipient motion, for instance, qb =14cm3/m/min by
U.S. Army Corps of Engineers, Waterways Experiment Station,
qb/(ρsdω) = 0.000317 by Han and He (1984). Yalin (1972) also
proposed a quantitative criterion related to the number of mov-
ing particles on the bed. For nonuniform sediment, the thresh-
old for incipient motion is more complex because of the hiding
and exposure mechanism, etc. Parker et al. (1982) suggested the
following threshold for the incipient motion of nonuniform
sediment,

* q bi ( ρ s ⁄ ρ – 1 )g
W ri = -----------------------------------
3
- = 0.002 (10)
p bi u *

where Wri* is a reference transport parameter; qbi is the volume-

tric transport rate per unit width for the ith fraction of bed-load;
ρs and ρ are the specific densities of sediment and fluid; g is the
gravitational acceleration; pbi is the gradation of the ith fraction Fig. 2. Measured and calculated critical shear stresses: (a) Newly
of bed material; and u* is the bed shear velocity. proposed eq.(9); (b) Egiazaroff’s formula; (c) Hayashi et al’s
We adopt Eq. (10) as the reference transport threshold to deter- formula.
mine the critical shear stress. From the collected data shown in
Table 1, we can determine the size fractions at incipient motion In addition, the present formula (9) is compared with Egia-
by using Eq. (10). The corresponding measured bed shear stress zaroff’s (1965) and Hayashi et al’s (1980) formulas by using
is the critical shear stress of these incipient fractions. Fig. 2a the same data set. Egiazaroff’s formula can be written as
shows the comparison of the critical bed shear stresses meas-
ured and calculated with Eq. (9). The agreement is very encour- θ ci log 19 2
------ = --------------------------------- (11)
aging. The coefficient θc is determined as 0.03, rather than in θc log ( 19d i ⁄ d m )
the range of 0.03~0.06 suggested by Shields for uniform
sediment. and Hayashi et al’s as

Table 1. Flow and sediment parameters of bed-load data.  log 8 2

θ ci  --------------------------------- di ⁄ dm ≥ 1
------ =  log ( 19d i ⁄ d m ) (12)
θc 
 dm ⁄ di di ⁄ dm < 1

where θci = τci /[(γs – γ)di].

θc = 0.06 was used by Egiazaroff (1965), which is too large.

Using measurement data Misri et al. (1983) determined
θc = 0.023~0.0303 in Egiazaroff’s and Hayashi et al’s formulas.
These values are very close to 0.03 used in Eq. (9). For the pur-
pose of comparison in this study, θc = 0.03 and the same refer-

JOURNAL OF HYDRAULIC RESEARCH, VOL. 38, 2000, NO. 6 429

ence criterion Eq. (10) are used in Eqs. (9), (11) and (12). As where Rb’ is the hydraulic radius corresponding to the grain
shown in Figs. 2a-c, both Hayashi et al’s and Egiazaroff’s for- roughness on the bed.
mulas significantly over predict in the range of low bed shear By using the Manning’s formula for uniform flow U = Rb2/3J1/2/n
stress. The discrepancy can be attributed to that Egiazaroff’s and U = Rb’2/3J1/2/n’, one can get Rb’ = Rb(n’/n)3/2. Here U is the
and Hayashi et al’s formulas are only related to the grain size, average flow velocity, n is the Manning’s roughness coefficient
and Hayashi et al’s formula uses the same critical shear stress for channel bed, and n’ is the Manning’s roughness correspond-
for all the particles finer than dm. The correction factor used in ing to grain roughness, calculated with n’ = d501/6/20 in this
Eq. (9) is not only related to the grain size but also to the bed- study. Therefore, from Eqs. (15) and (16), a method to calculate
material gradation. Better prediction can be achieved with the grain shear stress can be obtained
Eq. (9) even when the bed-material gradation is very wide and , 3⁄2

τ b =  ----
, n
highly irregular. Certainly, the comparison of the methods for τb (17)
 n
the critical condition of incipient motion is somehow dependent
on the used reference criterion, but the above comparison and Eq. (17) is similar to the method for calculating the grain shear
the application of Eq. (9) in establishing the nonuniform sedi- stress adopted by Meyer-Peter and Mueller (1948).
ment transport formulas in the next sections show that the The non-dimensional excess bed shear stress Ti = τb’/τci – 1 is
newly proposed correction factor has more advantages. used as an independent parameter in the relationship of φbi.

4 Fractional transport rate of nonuniform bed-load 4.2 Laboratory and field data of bed-load
4.1 Characteristic parameters and relationship Four sets of laboratory data for nonuniform bed-load measured
Meyer-Peter and Mueller (1948), Engelund and Fredsøe (1976), by Samaga et al. (1986a), Liu (1986), Kuhnle (1993) and
van Rijn (1984) and others related the bed-load transport rate to Wilcock and McArdell (1993) are used to investigate the bed-
the excess shear stress τb – τc and other parameters. This type of load transport and the incipient motion in this study. The latter
formulas for the uniform bed-load transport rate or the total three sets of experiments were for bed-load. In the experiments
transport rate of nonuniform bed-load can be written as of Samaga et al. (1986a, b) both bed-load and suspended load
existed. The concentration distribution of suspended load along
τ
φ b = f 3  ----b – 1 (13) the depth was measured, and the suspended load transport rate
 τc 
was calculated by integrating the concentration distribution and
where φb is a non-dimensional bed-load transport rate, velocity profile from 2di above the bed to the water surface. At
q b ⁄ ( γ s ⁄ γ – 1 )gd 3 ; qb is the volumetric bed-load transport the same time, the total load transport rate was also measured.
rate per unit width; and τb is the total bed shear stress or the bed The bed-load transport rate was obtained by subtracting the sus-
shear stress due to grain roughness. pended load from the total load transport rate. These data sets
Eq. (13) is extended to establish the relationship for the frac- cover a wide range of flow and sediment conditions, as shown
tional transport rate of nonuniform bed-load. The non-dimen- in Table 1.
sional fractional bed-load transport rate φbi is defined as In addition, some of the field data from five natural rivers (see
Williams and Rosgen, 1989) are also used. Because the field
q bi measurement of bed-load is very difficult and sometimes inade-
φ bi = --------------------------------------------
- (14)
p bi ( γ s ⁄ γ – 1 )gd i
3
quate, these data are carefully selected in order to avoid some
measurement error. First, the flow and sediment parameters
where qbi is the transport rate of the ith fraction of bed-load per must be measured at the same time. These parameters should
unit width (m2/s). include flow discharge, velocity, depth, surface slope, bed-load
The bed shear stress can be calculated with transport rate, bed-load gradation and bed-material gradation.
Secondly, because bed-load may move as strips and at stages,
most of the selected bed-load data are those averaged from at
τb = γ Rb J (15) least 16 samples across the same cross-section and during a
long enough measurement period. Thirdly, the data of bed-
material gradation are also averaged from several simultaneous
where Rb is the hydraulic radius of channel bed and J is the measurement points along the same cross-section to enhance
energy slope. reliability. The ranges of flow and sediment parameters of these
However, when sand ripples and dunes exist on the bed, it is field data are included in Table 1.
usually considered that bed-load transport is related only to the
grain shear stress, τb’, which is defined as
4.3 Regressional function
, , The collected data of nonuniform bed-load shown in Table 1
τb = γ Rb J (16)
are used to establish the relationship φbi ~ Ti. To calculate τb’
with Eq. (17), n is determined with n = Rb2/3J1/2/U in this study.

430 JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 38, 2000, NO. 6

For natural rivers Rb is given as the water depth h because the (τ − τc)/τc. Therefore, the fractional transport rate of nonuniform
width-depth ratio of these rivers studied in this paper is large suspended load has relationship,
and the influence of banks can be neglected; however, for labo-
τ
φ si = f 4  ----- – 1 -----
U
ratory experiments the bed and bank shear stresses should be  τ ci  ω i
divided and Rb is determined with Rb = τb /(γJ). Fig. 3 shows the (19)
empirical relationship of φbi ~ Ti. All the collected laboratory
where φ si = q si /[ p bi ( γ s ⁄ γ – 1 )gd i , and qsi is the transport
3
and field data of Ti > 0 are included. The data distribute along a
straight strip, in which φbi ranges from 10–5 to 102 and Ti from rate of the ith fraction of suspended load per unit width (m2/s).
10–2 to 102. By the least square curve fitting, the following The laboratory data of nonuniform suspended load measured by
formula for the fractional transport rate of nonuniform bed-load Samaga et al. (1986b) and two sets of field data in the Yampa
is obtained, River and the Yellow River are used here to analyze the
, 3⁄2 2.2 relationship in Eq. (19). The flow and sediment parameters of
τ
= 0.0053  ----
n
φ bi ----b- – 1 (18) these data are listed in Table 2. Fig. 4 shows the relationship of
 n τ ci φ si ∼ ( τ/τ ci – 1 )U /ω i . All the data points scatter in a straight
strip, the trend of which can be expressed as
where τci is determined with Eq. (9) which has taken into
account the hiding and exposure mechanism of nonuniform
τ 1.74
φ si = 0.0000262  ----- – 1 -----
sediment transport. U
(20)
 τ ci  ω i

where τci is determined with Eq. (9). The settling velocity of sedi-
ment particles in Eq. (20) is calculated with the Zhang’s formula,
ω i = ( 13.95v ⁄ d i ) 2 + 1.09 ( γ s ⁄ γ – 1 )gd i – 13.95v ⁄ d i (see
Zhang and Xie, 1993). Here ν is the kinematic viscosity.

Table 2. Flow and sediment parameters of suspended load data.

Fig. 3. Relationship for Fractional Transport Rate of Nonuniform

Bed-Load.

In Fig. 3, the total number of data points is 752, among

which 48.4% points lie within 1/2~2 folds of formula (18),
68.9% in 1/3~3 folds and 81.9% in 1/4~4 folds. The data strip
scattering can be attributed to the complexity and stochastic
behavior of nonuniform bed-load transport process and the
measurement error.

5 Fractional transport rate of nonuniform suspended load

The suspended load transport rate is related to the rate of energy
available to the alluvial system, which can be expressed as τU,
as explained by Bagnold (1966). Here τ is the shear stress on
Fig. 4. Relationship for Fractional Transport Rate of Nonuniform
the entire cross-section, τ = γRJ, and R is the hydraulic radius of Suspended Load.
channel. In addition, the suspended load transport rate is also
influenced by the gravity, which can be accounted for by the
settling velocity ω and the critical shear stress τc. According to
dimensional analysis, the independent parameter τU/τcω is
obtained. Physically, it is more meaningful to replace τ/τc with

JOURNAL OF HYDRAULIC RESEARCH, VOL. 38, 2000, NO. 6 431

6 Procedure for computing fractional transport rate of shown in Table 4, the new formula (18) provides the best
bed-material load results. Van Rijn’s formula’s results are not very good, because
the used data exceed its applicability size range of 0.2~2mm
suggested by van Rijn (1984).
The fractional transport rate of nonuniform bed-material load
can be obtained by summing the fractional transport rates of
bed-load and suspended load calculated with formulas (18) and Table 3. Comparison of calculated versus measured transport rates of
uniform bed-material load.
(20). The following steps are required for this calculation:
1. Divide the nonuniform sediment mixtures into several
fractions with different size ranges and determine di, ωi and
pbi for each fraction;
2. Calculate the hidden and exposed probabilities phi and pei;
3. Determine the critical shear stress τci for each size fraction
with Eq. (9);
4. Calculate the shear stress τ and the bed shear stress τb from
the known flow velocity, depth and surface slope. For natu-
Table 4. Comparison of calculated versus measured transport rates of
ral rivers, τb = γhJ can be used, and for experimental situa- uniform bed-load
tions, τb should be obtained by eliminating the bank shear
stress;
5. Determine the Manning’s roughness coefficient n for chan-
nel bed (excluding the influence of banks) and the
, 1⁄6
n ( = d 50 ⁄ 20 ) corresponding to the grain shear stress, and
,
then calculate the grain shear stress τ b with Eq. (17);
6. Calculate the non-dimensional excess shear stress
, Eqs. (18) and (20) are also jointly tested against the nonuniform
Ti = τ b /τci – 1, and then the fractional transport rate qbi for
sediment data collected by Toffaleti (1968), including the
nonuniform bed-load with Eq. (18).
experimental data observed by three groups of investigators:
7. Calculate the parameter (τ/τci – 1)U/ωi, and then the frac-
Nomicos, Einstein-Chien and Vanoni-Brooks, and the field data
tional transport rate qsi for nonuniform suspended load with
in the Rio Grande River, the Middle Loup River, the Niobrara
Eq. (20).
River and the Mississippi River. In order to avoid the deficiency
8. Sum qbi and qsi to obtain the fractional transport rate for non-
in the measurement of suspended load close to river bed, the
uniform bed-material load.
used field data are selected by limiting the lowest measurement
point on the depth to be lower than 0.2m (lower than 0.4m in a
7 Testing of the proposed transport formulas few of the Mississippi River data). These data cover the flow
discharges up to 21,600m3/s, the flow depths up to 17.5m, and
Eqs. (18) and (20) are jointly tested against 1859 sets of uni-
the sediment sizes from 0.062mm to 1mm. The tests are also
form bed-material load data selected from Brownlie’s (1981)
conducted on the Proffit and Sutherland’s (1983) modification
data collection by limiting the standard deviation of bed mate-
of Ackers and White’s formula, the modified Zhang’s formula
rial σ < 1.2 and the Shields parameter θ > 0.055. These data
(Wu and Li, 1992) as well as Karim’s formula (1998). The
were observed in several decades by many investigators, which
modified Zhang’s formula is only for the fractional transport
cover the flow discharges of 0.00094~297m3/s, the flow depths
rate of suspended load, and here it is combined with Eq. (18) to
of 0.01~2.56m, the flow velocities of 0.086~2.88m/s, the
obtain the fractional bed-material load transport rate. Table 5
surface slopes of 0.0000735~0.0367, and the sediment sizes of
shows the comparison between the calculated and measured
0.088~28.7mm. None of them was used to establish Eqs. (18)
fractional transport rates of nonuniform bed-material load. It is
and (20) in the previous sections. The comparison of the calcu-
found that the Proffit and Sutherland’s method systematically
lated and measured transport rates is shown in Table 3. Similar
over predicts for these data and provides the worst results. The
tests using the same data set are also conducted on three widely
newly proposed formulas predict the nonuniform bed-material
used bed-material load transport formulas developed by Enge-
load transport very well.
lund and Hanson (1967), Ackers and White (1973), and Yang
(1973, 1984). One can see from Table 3 that the newly pro-
Table 5. Comparison of calculated versus measured fractional
posed formulas provide very good results. In addition, Eq. (18) transport rates of nonuniform bed-material load
is separately tested against 1345 sets of uniform bed-load data
selected from the previous 1859 sets of uniform bed-material
load data by limiting the Rouse number ω/Ku * > 2.5 . A com-
parison is also conducted with four widely adopted bed-load
transport formulas of Meyer-Peter and Mueller (1948), Bagnold
(1966), Engelund and Fredfsøe (1976) and van Rijn (1984). As

432 JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 38, 2000, NO. 6

8 Conclusions ENGELUND, F. and HANSON, E. (1967), A Monograph on Sediment
Transport in Alluvial Streams, Teknisk Vorlag, Copenhagen, Den-
The hiding and exposure effect among the particles of nonuni- mark.
form bed material is proven important in the prediction of non- ENGELUND, F. and FREDSØE, J. (1976), A sediment transport model for
straight alluvial channels, Nordic Hydrology, Vol.7, 293–306.
uniform sediment transport. A probabilistic model for predict- FANG, D. and YU, G.L. (1998), Bedload transport in cobble-bed rivers,
ing this effect has been presented in this paper. The hiding and Proc. Of International Water Resources Engineering Conference,
exposure correction factor developed can account for not only Memphis, USA.
the influence of sediment particle size but also that of bed-mate- HAN, Q.W. and HE, M.M. (1984), Stochastic Theory of Sediment
Movement, Science Press, Beijing. (in Chinese)
rial gradation. A formula to determine the critical shear stress HAYASHI, T.S., OZAKI and T. ICHIBASHI (1980), Study on bed load
for the incipient motion of nonuniform sediment has been transport of sediment mixture, Proc. 24th Japanese Conference on
established based on this correction factor. The comparisons Hydraulics.
HSU, E.M. and HOLLY, F.M. (1992), Conceptual bed-load transport
using the laboratory and field data show that the methodology model and verification for sediment mixtures, J. of Hydraulic Engi-
developed has more advantages than Egiazaroff’s and Hayashi neering, ASCE, Vol.118, No.8, pp.1135–1152.
et al’s approaches which only consider the influence of sedi- KARIM, F. (1998), Bed material discharge prediction for nonuniform
ment particle size. Furthermore, the formulas for calculating the bed sediments, J. of Hydraulic Engineering, ASCE, Vol.124, No.6,
pp.597–604.
fractional transport rates of nonuniform bed-load and suspended KUHNLE, R.A. (1993), Fluvial transport of sand and gravel mixtures
load have also been established and tested by using a large with bimodal size distributions, Sedimentary Geology, Vol.85,
number of laboratory and field data of both uniform and non- pp.17–24.
KRAMER, H. (1935), Sand mixtures and sand movement in fluvial
uniform sediment transport. These newly proposed formulas models, Trans. ASCE, Vol.100, pp.798–838.
provide better predictions than several existing methods. More LIU, X.L. (1986), Nonuniform bed load transport rate and coarsening
research will be conducted in this area and more laboratory and stabilization, Thesis for Master Degree of Chengdu University of
field data will be collected to enhance the accuracy and reliabil- Technology. (in Chinese)
MEYER-PETER, E. and MUELLER, R. (1948), Formulas for bed-load
ity of these formulas and to broaden their validity in application transport, Report on Second Meeting of IAHR, Stockholm, Sweden,
to a wide range of sediment transport studies. 39–64.
MISRI, R.L., GARDE, R.J. and RANGA RAJU, K.G. (1983), Experiments
on bed load transport of nonuniform sands and gravels, Proc. 2nd
9 Acknowledgements Int. Symp. on River Sedimentation, Nanjing.
MISRI, R.L., RANGA RAJU, K.G. and GARDE, R.J. (1984), Bed load
This work is a result of the research project supported in part by transport of coarse nonuniform sediments, J. of Hydraulic Engineer-
the Specific Cooperative Agreement No.58-6408-7-035 ing, ASCE, Vol.110, No.3, pp.312–328.
MITTAL, M.K., POREY, P.D. and RANGA RAJU, K.G. (1990), Bed load
between the National Center for Computational Hydroscience transport of nonuniform sediments, Proc. of the Euromech 262-Col-
and Engineering of the University of Mississippi and USDA loquium on Sand Transport in Rivers, Estuaries and the Sea, Wall-
Agricultural Research Service. Dr. Roger A. Kunhle of the ingford, U.K.
QIN, Y.Y. (1980), Incipient motion of nonuniform sediment, J. of Sed-
USDA-ARS National Sedimentation Laboratory, Oxford, Mis- iment Research, No.1 (Resume Publication). (in Chinese)
sissippi and Mr. Kwonkyu Yu are acknowledged for their con- PARKER, G., KILINGEMAN, P.C. and MCLEAN, D.G. (1982), Bed load
sultation and help. and size distribution in paved gravel-bed streams, J. of Hydraulic
Division, ASCE, Vol.108, No.4, pp.544–571.
PATEL, P.L. and RANGA RAJU K.G. (1996), Fractionwise calculation of
References bed load transport, J. of Hydraulic Research, Vol.34, No.3.
PROFFIT, G.T. and SUTHERLAND., A.J. (1983), Transport of nonuni-
ACKERS, P. and WHITE, W.R. (1973), Sediment transport: A new form sediment, J. of Hydraulic Research, Vol.21, No.1, pp.33–43.
approach and analysis, J. of Hydraulics Division, ASCE, 99(HY11), SAMAGA, B.R., RANGA RAJU, K.G. and GARDE R.J. (1986a), Bed load
pp.2041–2060. transport rate of sediment mixture, J. of Hydraulic Engineering,
ANDREWS, E.D. (1983), Entrainment of gravel from naturally sorted No.11, pp.1003–1018.
riverbed material, Geological Society of America Bulletin, Vol.94, SAMAGA, B.R., RANGA RAJU, K.G. and GARDE R.J. (1986b), Sus-
pp.1225–1231. pended load transport rate of sediment mixture, J. of Hydraulic
ASHIDA, K. and MICHIUE, M. (1971), An investigation of river bed Engineering, No.11, pp.1019–1038.
degradation downstream of a dam, Proc. 14th Congress of the IAHR. SHIELDS, A. (1936), Anwendung der Aechlichkeitsmechanik und der
BAGNOLD, R.A. (1966), An approach to the sediment transport prob- Turbulenz Forschung auf die Geschiebebewegung, Mitteilungen der
lem from general physics, U.S. Geological Survey Professional Pruessischen Versuchsanstalt fuer Wasserbau and Schiffbau, Berlin.
Paper 422-J. TOFFALETI, F.B. (1968), A procedure for computation of the total river
BRIDGE, J.S. and BENNETT, S.J. (1992), A model for the entrainment sand discharge and detailed distribution, bed to surface, Technical
and transport of sediment grains of mixed sizes, shapes and densi- Report No.5, US Army Corps of Engineers, Vicksburg, Miss.
ties, Water Resources Research, Vol.28, No.2, pp.337–363. VAN RIJN, L.C. (1984), Sediment transport, part I: bed load transport,
BROWNLIE, W.R. (1981), Compilation of fluvial channel data: labora- J. Hydr. Engi., ASCE, Vol 110, No. HY10, 1431–1456.
tory and field, Rep. No. KH-R-43B, W.M. Keck Lab. of Hydr. and WILCOCK, P.R. and MCARDELL, B.W. (1993), Surface-based frac-
Water Resources, California Institute of Technology, Pasadena, tional transport rate: mobilization thresholds and partial transport of
Calif. a sand-gravel sediment, Water Resources Research, Vol. 29, No.24,
EGIAZAROFF, I.V. (1965), Calculation of nonuniform sediment concen- pp.1297–1312.
tration, J. of Hyd. Div., Proc. ASCE, No.HY4. WILLIAMS, G.P. and ROSGEN, D.L. (1989), Measured total sediment
EINSTEIN, H.A. (1950), The bed load function for sediment transporta- load (suspended loads and bed loads) for 93 United States streams,
tion in open channel flows, Tech. Bulletin, No.1026, United States USGS Open-File Report 89–67.
Dept. of Agriculture. WU, W. and LI, Y. (1992), One- and two-dimensional nesting mathe-
matical model for river flow and sedimentation, The Fifth Interna-

JOURNAL OF HYDRAULIC RESEARCH, VOL. 38, 2000, NO. 6 433

 tional Symposium on River Sedimentation, Karlsruhe, Germany,
Vol. 1, pp.547–554.
YALIN, M.S.(1972), Mechanics of Sediment Transport, Pergamon
Press.
YANG, C.T. (1973), Incipient motion and sediment transport, J. of
Hydraulics Division, ASCE, 99(HY10), pp.1679–1704.
ZHANG R.J. and XIE, J.H. (1993), Sedimentation Research in China,
Systematic Selections, China Water and Power Press, Beijing.

Notations
The following symbols are used in this paper:

di diameter of the ith fraction of sediment

dm arithmetic mean diameter of bed materials
d50 50% sieve diameter of bed materials
f probability distribution function of exposure height
g gravity acceleration
h water depth
i size fraction number
J energy slope
m exponent for correction factor
N total number of particle size fractions
n Manning’s roughness coefficient for channel bed
n’ Manning’s coefficient corresponding to grain
roughness
pbi percentage of the ith fraction of bed material
phi,j, pei,j probabilities of particles di hidden and exposed by
particles dj
phi, pei total hidden and exposed probabilities of particles di
qb bed-load transport rate
qbi, qsi fractional transport rates of bed-load and suspended
load
Rb hydraulic radius of channel bed
,
Rb hydraulic radius corresponding to grain roughness
Ti non-dimensional excess shear stress
U average flow velocity
u* shear velocity
W*ri a reference transport parameter
γ , γs specific weights of water and sediment
ηi hiding and exposure correction factor for nonuniform
sediment
θc non-dimensional critical shear stress of uniform-
sediment
θci non-dimensional critical shear stress of the ith fraction
of sediment
ρ, ρs specific densities of fluid and sediment
,
τb, τ b bed shear stress and grain shear stress
τc, τci critical shear stresses for dm and di
φb non-dimensional bed-load transport rate
φbi, φsi non-dimensional fractional transport rates of bed-load
and suspended load
∆i exposure height of the ith fraction of bed material
ωi settling velocity of the ith fraction of sediment

434 JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 38, 2000, NO. 6