Hydrological Sciences Journal

ISSN: 0262-6667 (Print) 2150-3435 (Online) Journal homepage: https://www.tandfonline.com/loi/thsj20

Determination of bed shear stress in gravel-bed

rivers using boundary-layer parameters

HOSSEIN AFZALIMEHR & COLIN D. RENNIE

To cite this article: HOSSEIN AFZALIMEHR & COLIN D. RENNIE (2009) Determination of bed
shear stress in gravel-bed rivers using boundary-layer parameters, Hydrological Sciences Journal,
54:1, 147-159, DOI: 10.1623/hysj.54.1.147

To link to this article: https://doi.org/10.1623/hysj.54.1.147

Published online: 21 Dec 2009.

Submit your article to this journal

Article views: 3433

View related articles

Citing articles: 12 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=thsj20
 Hydrological Sciences–Journal–des Sciences Hydrologiques, 54(1) February 2009 147

Determination of bed shear stress in gravel-bed rivers using

boundary-layer parameters

HOSSEIN AFZALIMEHR1 & COLIN D. RENNIE2

1 Water Engineering Department, Isfahan University of Technology, Isfahan, Iran
2 Department of Civil Engineering, University of Ottawa, 161 Louis Pasteur Street, Ottawa, Ontario K1N 6N5, Canada
crennie@genie.uottawa.ca

Abstract The behaviour of velocity profiles and shear velocity for non-uniform flow in gravel-bed rivers is
studied, with the objectives: (a) to test a new method of shear velocity estimation in gravel-bed rivers that is
based on boundary layer parameters, and to compare it with the log law and parabolic law; (b) to consider
the influence of flow non-uniformity on the outer layer region of velocity profiles; and (c) to investigate the
effect of aspect ratio on velocity profiles. For the primary study river, mid-channel velocity profiles were
analysed with relative submergence ranging from 9.7 to 33.3 in channel sections with aspect ratios ranging
between 16.2 and 50. Velocity profiles deviated from the log law in the outer region due to flow non-
uniformity or pressure gradient effects, and the vertical extent of the inner region was variable. Estimates of
shear velocity using the boundary layer parameters (δ* and θ) compared well with estimates from the log
law. In a second study river, boundary-layer parameter estimates of shear velocity compared well to shear
velocity estimates from linear extrapolation of Reynolds stress profiles.
Key words shear velocity; boundary layer; log law; non-uniform flow; aspect ratio; gravel bed

Détermination de la tension de cisaillement du lit en rivières à lit de gravier à l’aide de

paramètres de couche limite
Résumé Le comportement des profils de vitesse et de la vitesse de cisaillement est étudié pour un
écoulement non-uniforme en rivières à lit de gravier, avec les objectifs: (a) de tester une nouvelle méthode
d’estimation de la vitesse de cisaillement en rivières à lit de gravier, basée sur les paramètres de la couche
limite, et de la comparer avec les lois logarithmique et parabolique; (b) de considérer l’influence de la non-
uniformité de l’écoulement sur la couche extérieure des profils de vitesse; et (c) d’étudier l’effet du rapport
hauteur/largeur sur les profils de vitesse. Pour la première étude de rivière, les profils de vitesse en milieu de
section ont été analysés avec une submersion relative allant de 9.7 à 33.3 dans des sections présentant des
rapports hauteur/largeur allant de 16.2 à 50. Les profils de vitesse s’écartent de la loi logarithmique dans la
région extérieure en raison de la non-uniformité de l’écoulement et d’effets liés au gradient de pression, et
l’extension verticale de la région interne est variable. Les estimations de vitesse de cisaillement basées sur
les paramètres de la couche limite (δ* et θ) sont comparables aux estimations fournies par la loi
logarithmique. Pour une deuxième étude de rivière, les estimations de la vitesse de cisaillement basées sur
les paramètres de la couche limite sont proches de celles que l’on obtient par extrapolation linéaire des
profils de tensions de Reynolds.
Mots clefs vitesse de cisaillement; couche limite; loi logarithmique; écoulement non-uniforme; rapport hauteur/largeur;
lit de gravier

INTRODUCTION
Two different flow regions can be distinguished for gravel-bed open-channel flow: the inner
region, near the bed, and the outer region where the velocity profile deviates slightly from the log
law of the wall (Bathurst, 1982; Kironoto & Graf, 1995; Afzalimehr & Anctil, 2001). Kironoto &
Graf (1995) demonstrated that outer region velocities were reduced in accelerating flows (negative
pressure gradient), and increased in decelerating flows (positive pressure gradient). Non-uniform
flow is common, if not typical, in gravel-bed rivers due to longitudinal variation in aspect ratio,
relative submergence, and channel roughness.
The existence of a negative or positive pressure gradient due to flow non-uniformity exerts a
strong effect on the spatial distribution of shear stress (τ) and corresponding sediment transport in
a gravel-bed channel (Kironoto & Graf, 1995; Afzalimehr & Anctil, 2001; Song & Chiew, 2001;
MacVicar & Roy, 2007). Shear velocity (u* = √(τ/ρ), where ρ is fluid density) tends to be less in
non-uniform flows than uniform flows (Kironoto & Graf, 1995; Afzalimehr & Anctil, 2001).
Furthermore, for non-uniform flows, shear velocity for favourable (negative) pressure gradient is

Open for discussion until 1 August 2009 Copyright © 2009 IAHS Press
148 Hossein Afzalimehr & Colin D. Rennie

smaller than for adverse (positive) pressure gradient (Song & Graf, 1994; Afzalimehr & Anctil,
2001).
These remarks call for sophisticated approaches to determine shear velocity in gravel-bed
rivers based on analytical parameters. Approaches based on vertical profiles of mean downstream
velocity, which are relatively easy to obtain, will be explored herein. Direct measurements of
Reynolds stress may be extrapolated to the channel bed to estimate bed shear stress, but these
require relatively sophisticated high-frequency measurements of turbulent velocity in at least two
dimensions, and such measurements are challenging in field situations.
Typical approaches for estimating the bed shear stress from the vertical profile of downstream
velocity include the log law for the inner region and the parabolic law for the outer region. The
outer region is defined as y/h > x, where y is the elevation above the bed reference level and h is
flow depth, such that x is the elevation above the bed where the velocity profile deviates
systematically from the log law. Application of the log law to non-uniform flows requires an
assumption that pressure gradient due to flow non-uniformity does not have a large influence on
the inner region (White, 2006). Furthermore, u* estimation from the log law is sensitive to the
choice of boundary elevation. The parabolic law has been demonstrated to yield reasonable
estimates of u* in decelerating flows (Afzhalimehr & Anctil, 1999), but power-law scaling for
velocity profiles is sensitive to Reynolds number (Barenblatt, 1993). In this study we evaluate
application of a relatively recent approach to determine shear velocity in non-uniform flow over a
gravel bed based on the boundary layer parameters, i.e. the boundary layer displacement (δ*) and
boundary-layer momentum thicknesses (θ) (Afzhalimehr & Anctil, 2000).
The boundary layer approach developed by Afzhalimehr & Anctil (2000) has thus far been
tested using laboratory scale gravel-bed channel data with small aspect ratio W/h < 5.4, where W is
the channel top width. Therefore, the objectives of this study are: (a) to test the boundary-layer
parameter method for determination of shear velocity in actual gravel-bed rivers over a range of
aspect ratios; (b) to compare the results obtained from (a) with the log-law and parabolic methods;
and (c) to investigate the effect of flow non-uniformity via variations of aspect ratio and relative
submergence values on the velocity distribution along the selected reaches. A large number of
vertical profiles of mean downstream velocity from three gravel-bed rivers in Iran, and a few
vertical profiles of three-dimensional (3-D) velocity and Reynolds stress from one gravel-bed river
in Canada, are used to test the applicability of the boundary-layer method.

BACKGROUND
It is impossible to present a boundary-layer thickness in an unambiguous way because the effect of
viscosity in the boundary layer decreases asymptotically outwards (Schlichting & Gersten, 2000).
In order to avoid utilization of an arbitrary boundary layer thickness, it is necessary to consider the
boundary layer displacement thickness, δ* (Clauser, 1956):
h
u
δ * = ∫ (1 − )dy (1)
0
u max
where umax is the maximum velocity in a velocity profile, u is the mean point velocity at a distance
y measured from the reference level. The parameter δ* indicates the distance by which the external
streamlines are shifted owing to the formation of the boundary layer. Furthermore, the momentum
thickness (θ) indicates the loss of momentum in the boundary layer as compared with potential
flow and is defined as:
h
u u
θ =∫ (1 − )dy (2)
0
u max u max
The exact values of δ* and θ depend upon the distribution of downstream velocity in the cross-

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 149

section normal to the flow. Equations (2) and (3) are valid for any incompressible flow, whether
laminar or turbulent, uniform or non-uniform flows.
In 1949, Ludwieg & Tillmann (see Young, 1989) applied the displacement and momentum
thicknesses over a smooth boundary in the presence of non-uniform pressure distribution and
suggested the following equation:
C f = 0.246(u maxθ /ν ) −0.268 10 0.678 H (3)

where Cf is the friction coefficient, ν is the kinematic viscosity and H = δ*/θ is a form factor.
Their equation using δ* and θ to estimate flow resistance over a smooth plate subsequently found
considerable support in aeronautical engineering (Young, 1989).
Using the laboratory data of Song & Graf (1994), Afzalimehr & Anctil (2000) suggested the
following formulation for determining shear velocity in non-uniform flows over gravel-bed
channels:
(δ − θ )u max
u* = * (4)
Cδ*
where C is an empirical constant that was found to be equal to 4.4 in the laboratory experiments.
Due to the applicability of δ* and θ to non-uniform flows, equation (4) is inherently suitable for
estimation of u* in non-uniform gravel-bed flows.
The parabolic law can be written as follows (Afzalimehr & Anctil, 1999):
⎡ ( y + a d 50 ) ⎤
2
u
= −Ω ⎢1 − ⎥ +Φ (5)
u max ⎣ h ⎦
where a is an empirical constant to scale the roughness (a = 0 in this study), and Ω and Φ are,
respectively, the regression slope and intercept. Afzalimehr & Anctil (1999) showed the shear
velocity can be calculated as u * = Ω u max κ 2 x (1 − x ) , where κ is the von Kármán constant, taken
as κ = 0.4 for gravel-bed channels (Kironoto & Graf, 1995).
The log law has often been used to estimate bed-shear stress in alluvial channels. The velocity
distribution of flow over gravel beds can be presented in the inner region by the log law as
follows:
u 1 ⎛ y + w ks ⎞
= ln⎜⎜ ⎟⎟ + c (6)
u* κ ⎝ k s ⎠
where u* is the shear velocity, the parameter ks is the Nikuradse equivalent grain roughness,
approximated by either d50 or d84 (Bray & Davar, 1987), w is a coefficient of adjustment of the
origin below the top roughness elements, and c is a constant. Here, ks = d84 to reflect the greater
influence of larger grains on surface roughness in gravel-bed rivers (Leopold et al., 1964; Charlton
et al., 1978). The value of w may be calculated using a trial and error process by maximizing the
coefficient of determination (r2) between u and ln[(y + wks)/ks)]. The value of c has been reported
as c = 6.25 (Keulegan, 1938) for large relative submergence (h/ks > 20) and c = 3.25 (Graf, 1984)
for small relative submergence (h/ks < 4). Also, this constant varies slightly with the pressure
gradient (Tennekes & Lumley 1972; Reynolds, 1974). Note that the choice of c does not influence
the estimation of u* from the slope of a regression fit between u and ln[(y + wks)/ks)].
Despite the fact that log law was derived assuming zero-mean vertical velocity, and thus
strictly does not apply to non-uniform flow, it appears that the inner layer is relatively unaffected
by pressure gradient and the log-law over gravel-bed channels has been confirmed experimentally
for non-uniform flows (Afzalimehr & Anctil, 2001; Song & Chiew, 2001; White, 2006). However,
in most field studies, the minimum height of velocity measurement above the bed is relatively
large (>0.02 m) owing to the difficulty of measuring close to the bed (e.g. Ferguson & Ashworth,
1992; Robert et al., 1992; Biron et al., 1998). In fitting a logarithmic profile to rough boundary

Copyright © 2009 IAHS Press

150 Hossein Afzalimehr & Colin D. Rennie

data it is difficult to locate the origin where y = 0. Furthermore, the vertical height of the law of the
wall is not constant and varies with flow conditions, which complicates its application. According
to White (2006), the vertical height of the logarithmic region has been found to decrease when the
pressure gradient increases. Even in uniform flows, many workers have taken a site-specific
approach and the proportion of depth considered to be semi-logarithmic varies considerably from
one study to another. The upper limit of the validity of the log law is often considered as
y/h = x = 0.2 (Bridge & Jarvis, 1976; Song & Graf, 1994; Kironoto & Graf, 1995; Afzalimehr &
Anctil, 1999). However, estimates have ranged from the lower 15% of the flow (e.g. Bridge &
Jarvis, 1977, 1982) to over 50% (Ferguson et al., 1989; Ferguson & Ashworth, 1992; Biron et al.,
1998; Smart, 1999). Afzalimehr & Anctil (2001) indicated that there is not a constant height for
validity of the law of the wall over cobble-bed channel in intermediate scale roughness. Further-
more, in very rough flows, the log-law region can be elevated above a near-bed wake region that is
generated by flow separation behind individual roughness elements (Nowell & Church, 1979;
Biron et al., 1998).
It should be stressed that the log law does not hold in general even for very large Reynolds
numbers and can only be considered an approximate representation of experimental data in the
inner part of the boundary layer (Barenblatt, 1982). However, despite this limitation, many lab-
oratory studies (e.g. Song & Graf, 1994; Kironoto & Graf, 1995; Afzalimehr & Anctil, 1999,
2000, 2001) have shown that estimation of bed shear stress by the log law and Reynolds stress
extrapolation are compatible. Due to the relative difficulty and expense of obtaining high-
frequency turbulent velocity measurements in gravel-bed rivers, it is a worthwhile endeavour to
demonstrate that an approach based only on mean downstream velocity profiles that is
theoretically valid for non-uniform flows (i.e. the boundary-layer approach) can be used to
estimate bed shear stress. In this paper, data from several gravel-bed rivers in Iran and Canada are
utilized to demonstrate the validity of the boundary-layer method for bed shear stress estimation.
Bed shear stress estimates from the boundary-layer method are compared with those from the log
law and the empirical parabolic law. Furthermore, data from the Canadian river include direct
measurements of Reynolds stress profiles, which are used to estimate bed shear stress for
validation and comparison of the various velocity profile methods.

FIELD EXPERIMENTS
Iranian gravel-bed rivers
Data under non-uniform flow conditions were measured in four reaches along 12 km of the gravel-
bed Ghamasiab River in western Iran (Table 1). The selected reaches were straight and were
devoid of in-stream vegetation and bed forms; thus, cross-stream flow components were minimal.
The length of each reach was 75 m, and in each reach six cross-sections with constant separation
distance of 15 m were measured by surveying instruments. Sections are numbered from down-
stream to upstream. All measurements were obtained during periods of constant and low-flow
discharge. The profile from section 12 was eliminated as an apparent outlier.
The primary evidence for the presence of non-uniform flow in all of the selected reaches is the
longitudinal variation of top width (W), centre-line depth (h), grain size (d50 or d84), and velocity
(Table 1). Furthermore, longitudinal variation in the boundary-layer parameters δ* and θ is further
evidence that velocity profiles were changing along each reach. While it is clear that the flow was
non-uniform in each reach, it is difficult to identify sections undergoing decelerating or
accelerating flow due to the complex interaction between W, h, d50 and u in each section and reach
(MacVicar & Roy, 2007).
Velocity profiles were measured in the central axis of all 24 cross-sections of this river using a
micro-current propeller meter (propeller diameter of 5 cm) oriented in the downstream direction.
The lowest point of velocity measurement was located 5 cm above the river bottom. To avoid
water disturbance during the measurements, a mobile bridge was constructed over the water
surface; thus, measurements were carried out without entering into the river. For each velocity

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 151

Table 1 Ghamasiab River hydraulic parameters.

Profile h W d50 d84 Fr Re δ* θ u*BL umax
Section no. (m) (m) (mm) (mm) - - (m) (m) (m/s) (m/s)
1 0.283 8.4 18 30 0.66 310168 0.048 0.029 0.098 1.096
2 0.283 8.1 18 31 0.65 306206 0.042 0.025 0.098 1.082
3 0.323 8 18 36 0.64 365313 0.061 0.036 0.106 1.131
4 0.253 7 21 40 0.57 481206 0.045 0.022 0.104 1.092
5 0.283 5.5 20 32 0.71 335072 0.062 0.035 0.119 1.184
6 0.373 6 22 35 0.66 468488 0.098 0.058 0.118 1.256
7 0.283 7.5 20 40 0.58 275359 0.057 0.036 0.082 0.973
8 0.453 8.5 20 37 0.77 731595 0.119 0.066 0.163 1.615
9 0.183 9 21 38 0.46 113643 0.032 0.017 0.064 0.621
10 0.263 11.3 17 33 0.55 232492 0.046 0.028 0.079 0.884
11 0.383 10 22 33 0.62 232492 0.057 0.035 0.106 1.308
13 0.373 11 18 37 0.67 480051 0.06 0.036 0.115 1.287
14 0.303 12 14 30 0.57 298152 0.051 0.032 0.084 0.984
15 0.403 11 19 37 0.69 553319 0.068 0.041 0.124 1.373
16 0.343 10 16 30 0.57 357749 0.059 0.036 0.089 1.043
17 0.63 10 20 31 0.55 551880 0.04 0.024 0.08 0.876
18 0.333 10.5 16 27 0.6 360639 0.05 0.031 0.095 1.083
19 0.303 8 19 32 0.6 186633 0.045 0.027 0.092 1.026
20 0.233 8.9 23 40 0.53 186633 0.038 0.022 0.075 0.801
21 0.213 9.4 23 36 0.54 167418 0.035 0.02 0.076 0.786
22 0.223 8.9 17 34 0.59 199139 0.04 0.025 0.076 0.893
23 0.263 7.5 18 33 0.54 228021 0.043 0.026 0.079 0.867
24 0.253 8 18 36 0.51 204424 0.033 0.019 0.076 0.808
h: flow depth; W: top width of river; Fr: Froude number defined by um/(gh)0.5; Re: Reynolds number that is umh/v;
δ*: the boundary-layer displacement; θ: the momentum thickness; u*BL: shear velocity using the boundary-layer method;
umax: maximum point velocity at the distance h measured from the reference level.

profile, an average of 14 mean point velocities was measured. The vertical distance between
measuring points near the bed was approximately 1 cm, and in the upper 50% of the flow was
2–3 cm. On average, eight measurements were taken within 25% of the bed surface.
The aspect ratio of the selected sites was in the range: 16.2 < W/h < 50. All cross-sections
were reasonably trapezoidal in section; thus, h is representative of the typical channel depth across
the section. The relative submergence varied as 9.7 < h/d50 < 33.3, where d50 is the median grain
size of the bed surface measured by pebble count. Using the depth-averaged velocity from the
central-axis as a measure of the maximum velocity (um) in the channel, the Froude number was in
the range 0.46 < Fr < 0.77, the Reynolds number was 1.5 × 105 < Re < 7.3 × 105, and the particle
Reynolds number 2770 < Re* < 7600 (see Table 1).
Grain-size distribution was estimated by pebble count (Wolman, 1954). Grain-size distribu-
tions were similar in all reaches (Table 1), with average d50 = 19 mm and d84 = 34 mm, where d84
is the 84% finer particle diameter. Also, the geometric standard deviation of particle distribution
(σ = (d84/d16)0.5) ranged between 1.5 and 2.1 for the 23 cross-sections. Given that grain size
distribution and associated grain roughness varied little between sections, the velocity profiles
permit study of the effects of flow non-uniformity due to variation in aspect ratio and relative flow
depth on the shear velocity in a gravel-bed river.
Similar methods were used to collect vertical profiles of mean downstream velocity in two
other rivers in central Iran: the Kaj River and Behestabad River.

Salmon River, Canada

The Salmon River flows through a forested flood plain in an agricultural catchment in the Fraser
River Valley near Vancouver, Canada. The study site was a relatively straight, gravel-bed reach
with alternating side bars. Mean bankfull width and depth were 14 m and 0.7 m, respectively.
Sinuosity measured along the thalweg was 1.06. Velocity and substrate measurements were made

Copyright © 2009 IAHS Press

152 Hossein Afzalimehr & Colin D. Rennie

(a) (b)

(c)

Fig. 1 Salmon River ADV survey, 23 March 1999: (a) mean downstream velocity at y/h = 0.2, with
velocity contour interval 0.1 m/s, map units in m, bed topography contour interval 0.2 m, and velocity
measurement locations marked by x with labels for f1, d2, c2 and b2; (b) vertical profiles of mean
stream velocity at f1, d2, c2, and b2; and (c) profiles of principal Reynolds stress at f1, d2, c2 and b2.

throughout one channel bend (Fig. 1(a)). The surface D50 and D84 were 7 and 9 cm in the pool
thalweg, 5 and 7 cm in the upstream riffle, and 3 and 5.5 cm on the bar. The flow separated from
the channel boundary on both banks immediately downstream of the bend apex. Vertical profiles
of 3-D velocity and Reynolds stress were measured using a 10 MHz laboratory Acoustic Doppler
Velocimeter (ADV) from a transportable platform at mid-point of the bar, through the bend apex,
and down to the downstream riffle. Thus, measurements were performed from just downstream of
one cross-over point to the next cross-over. River stage did not change appreciably during the
measurements, and the flow was moderate (8% of the 1.5-year return period flow). See Rennie et
al. (1999) for a complete description of the measurements and analysis of the spatial distribution
of turbulence in the bend. In this paper, four vertical profiles along the thalweg that had at least
four points in the vertical are evaluated (locations f1, d2, c2, and b2 in Fig. 1(a)).

RESULTS
Effect of pressure gradient (flow non-uniformity) on velocity distribution
Ghamasiab River Velocity profiles in the Ghamasiab River deviated systematically from the
log law in the outer region, as expected for non-uniform flow. The deviations displayed different
patterns, with outer region velocities either exceeding or less than the log law (Figs 2 and 3).

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 153

(a)

(b)

(c)

(d)

Fig. 2 Velocity profiles in Ghamasiab River: (a) Reach 1; (b) Reach 2; (c) Reach 3; and (d) Reach 4.

Copyright © 2009 IAHS Press

154 Hossein Afzalimehr & Colin D. Rennie

14

12 Section 16

10
Section 8
8
u/u*

0
0.01 0.1 1 10
y/h

Fig. 3 Log law fit to inner region of two typical non-uniform velocity profiles in Ghamasiab River.

However, as discussed above, due to geomorphologic and hydraulic interaction, and the fairly
large spacing between individual profiles, it is not possible to state unequivocally whether a given
profile corresponded to accelerating or decelerating flow. The velocity distributions indicated that
there was not a constant height of validity of the log-law, and the law could often be extended to
greater than 50% flow depth from the bed (Fig. 3). The height of validity (x) ranged from 0.22 to
0.68.
Salmon River Flow at profile f1 (pool head) was decelerating due to increasing depth, and the
vertical profile of mean downstream velocity displayed excess upper region velocity as expected
under flow deceleration (Fig. 1(b)). The velocity profile in the pool (d2) was complicated and
nonlinear, possibly due to strong vortex action in a region of flow separation in the pool. The two
profiles further downstream (c2 and b2) were in zones of flow acceleration, and suitably display
reduced velocity in the upper region. These results confirm the previous laboratory observations
(Song & Graf, 1994; Afzalimehr & Anctil, 2001) of the influence on pressure gradient on velocity
profiles.

Local shear velocity estimates

Ghamasiab River For 23 velocity profiles in the Ghamasiab River, shear velocity (u*log) was
determined from the velocity gradient in the log-linear region by regressing u against
y + 0.25d84 . For all profiles, the coefficient of determination (r2) exceeded 98%. The value of
ln
d84
w = 0.25 was determined using a trial-and-error process to obtain the best coefficient of determina-
tion (r2) between u and ln y + 0.25d84 . It can be shown through integration that the mean surface
d84
elevation in a bed of spheres of uniform diameter D is 0.18D (w = 0.18) below the top of the
spheres, which agrees well with the results of this study (w = 0.25) and others (w = 0.20) (Einstein
& El-Samni, 1949; Kironoto & Graf, 1994). It seems likely that the elevation of zero velocity is
actually the mean bed elevation.
The effect of using the full flow depth versus the near-bed portion of the velocity profile was
assessed by estimating shear stress with progressively fewer points from the top of the measured
profile, while always retaining the measurements near the bed. The profile that yielded the highest
r2 was used to determine u* (Fig. 3). The results indicate that using the full flow depth instead of
the near bottom of the profile would generate an underestimation of the shear stress when flow is

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 155

0.2
0.18
(a)
0.16
0.14
0.12
u * log

0.1
0.08
0.06
0.04
0.02
0
0 0.05 0.1 0.15 0.2
0.2
0.18 (b)
0.16
0.14
0.12
u * par

0.1
0.08
0.06
0.04
0.02
0
0 0.05 0.1 0.15 0.2

u * BL
Fig. 4 Ghamasiab River—comparison of estimated u* using the log law, parabolic law and the
boundary-layer method, with C = 4.4. Plots show line of perfect agreement and ±25% lines.
Regressions for (a) r2 = 0.91; (b) r2 = 0.64.

non-uniform. Biron et al. (1998) reported a similar behaviour under an assumed uniform flow. For
a few cases some deviation from the log law was observed near the bed. The deviations were not
systematic and can thus be attributed to either: (a) measurement errors of elevation (y) or point
mean velocity (u) (Wilcock, 1996), or (b) local roughness irregularities (Nowell & Church, 1979).
The boundary-layer parameter method was also applied to estimate shear velocity (u*BL).
Figure 4 presents the comparison of the boundary-layer characteristics method (u*BL) with the log-
law method (u*log) and the parabolic method (u*par). The boundary-layer parameter empirical
constant C was set to 4.4, as determined for laboratory data (Afzalimehr & Anctil, 2000). All u*BL
were within ±30% of respective u*log. The coefficient of determination between u*log and u*BL was
r2 = 0.91, and the root mean square (RMS) difference between these methods was 0.021 m/s. The
comparison between u*BL and u*par was similar, although u*par appeared to display more scatter.
The general compatibility between u*BL and u*log observed in Fig. 4(a) further justifies the
application of δ* and θ for rough turbulent flows in gravel-bed streams. Random error between
u*log and u*BL can be attributed to systematic error in positioning instruments and flow measuring
conditions in gravel-bed rivers. Calculation of u*log depends on estimation of the local roughness
and the elevation of zero velocity, both of which depend on the bed characteristics in the
immediate vicinity upstream of the measurement point. Deviation from a log profile near the bed
is expected in the presence of heterogeneous roughness elements because of form drag which
reduces near-bed velocities (Wiberg & Smith, 1987). In such cases, positioning of the velocity

Copyright © 2009 IAHS Press

156 Hossein Afzalimehr & Colin D. Rennie

Fig. 5 Comparison of estimated u* using the parabolic law and the boundary-layer method, for the Kaj
and Behestabad rivers. Plots show line of perfect agreement and ±25% lines. Regression r2 = 0.91.

profile (e.g. over the top of a clast or in a hollow between clasts) is critical because the zero plane
is highly variable.
The boundary-layer characteristics method resulted in consistently higher estimates of u*
(Fig. 4). It is possible that u*BL represents an estimate of local total shear stress imposed on the
bed, including local form drag and losses associated with flow non-uniformity, whereas u*log
derived from the near-bed log layer and ks = d84 is representative of grain shear. Given that the
parabolic law depends on the outer layer, shear stress estimated by the parabolic law (u*par) should
also be indicative of total boundary shear as opposed to local grain shear. However, it was not
observed for the Ghamasiab River data that u*BL and u*par were similarly scaled. Alternatively,
optimizing the boundary-layer method empirical coefficient C to a value of 4.6 minimized the
difference between u*BL and u*log and u*par for the Ghamasiab River data. Similarly, u*BL and u*par
were scaled for the data from the two other Iranian rivers, the Kaj and Behestabad (Fig. 5) if the
boundary-layer constant was taken to equal C = 4.4 and C = 4.8, respectively.
Salmon River Profiles of the principal Reynolds stress are presented for each Salmon River
profile in Fig. 1(c). Shear velocity (u*RS) was estimated by linear extrapolation to the bed of the
portion of the Reynolds stress profile that displayed decreasing stress for increasing elevation
above the bed. Shear velocities estimated by each method for each profile in the Salmon River are
presented in Table 2. The boundary-layer method utilized the empirical constant C = 4.4, as
determined for laboratory non-uniform flows (Afzalimehr & Anctil, 2000). It can be seen that, on
average, the boundary-layer method produced shear-velocity estimates (u*BL) more similar to the

Table 2 Salmon River shear velocity estimation.

Profile u*RS u*BL u*log u*par
f1 0.085 0.097 (13.6%) 0.150 (76.6%) 0.060 (29.8%)
d2 0.088 0.079 (10.2%) 0.071 (19.5%) 0.002 (97.5%)
c2 0.075 0.038 (49.3%) 0.075 (0.5%) 0.134 (78.1%)
b2 0.110 0.087 (21.1%) 0.067 (39.1%) 0.074 (32.7%)
Average error 23.6% 33.9% 59.5%
u*RS, shear velocity from linear extrapolation of Reynolds stress profile; u*BL, shear velocity from boundary-layer
method; u*log, shear velocity from log-law; u*par, shear velocity from parabolic law. Percentages in brackets represent
absolute percentage difference with respect to u*RS.

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 157

estimates of u*RS than either the log law (u*log) or the parabolic law (u*par). This serves as direct
validation of the boundary-layer method for shear-velocity estimation under non-uniform flows in
gravel-bed rivers.

Influence of aspect ratio on velocity profiles

Altinakar & Graf (1998) stated that the aspect ratio can be used to distinguish 2-D wide flows
(W/h > 5) or 3-D narrow flows (W/h < 5). Side walls effects are negligible in wide channels and
flow is considered 2-D. For 2-D flow, the maximum velocity occurs at the free surface and the
thickness of the boundary layer is equal to the flow depth.
All of the Ghamasiab River velocity profiles presented in Fig. 2, other than Section 24,
confirm that for the range of aspect ratio in this study, 16.2< W/h < 50, the dip value was equal to
zero, i.e. the maximum velocity occurred at the water surface. The velocity profiles measured in
this study indicate that for large aspect ratio, W/h > 16, the position of the maximum velocity is
always at the water surface, irrespective of the pressure gradient.

DISCUSSION
The velocity profiles in the Salmon River displayed the expected increase (decrease) in outer layer
velocity for decelerating (accelerating) flow (Fig. 1). While it could not be verified that the
Ghamasiab River velocity profiles corresponded to acceleration or deceleration, the measured
variability in channel section and variation in calculated boundary-layer parameters is strong
evidence for the presence of flow non-uniformity. It is worth noting that uniform flows with zero
pressure gradient also commonly have velocities in the outer layer that exceed the log law.
Turbulent momentum exchange near the free surface is anisotropic, with no high momentum fluid
being received from above, thus mean downstream velocity must increase to balance momentum
exchange from below (Monin & Yaglom, 1971, p. 317). Furthermore, it is commonly observed in
both uniform and non-uniform flows that maximum velocity occurs below the surface, i.e. there is
a “velocity dip” at the surface. The velocity dip gives the appearance of an outer flow with reduced
velocities. The reason for the velocity dip was explained by Nezu & Rodi (1985, 1986) and Yang
et al. (2006). When side wall effect is present, a strong lateral velocity component (w) is directed
near the free surface from the side wall to the channel centre and a down flow (v) occurs from the
free surface. These secondary velocity components produce a strong free surface vortex and the
occurrence of the maximum velocity below the free surface. Yang et al. (2006) argued that the
velocity dip phenomenon occurs whenever there is a negative vertical velocity at the free surface,
including when the water surface has positive slope due to accelerating flow. This was not
observed for the Ghamasiab River data, likely because wide aspect ratios minimized side-wall
influence.
The log law depends strongly on the selected reference level. For example, varying the
reference level from 0.1d84 to 0.25d84 led to 20% difference in determination of shear velocity by
the log law. However, although the boundary-layer method takes into account non-uniformity
effects, there is little influence of reference level on shear velocity estimation by this method.
Also, when the river plan is non-uniform, such as converging and diverging sections in a gravel-
bed river, the position of maximal velocity may be changed. In this case, the boundary-layer
method can better represent the effect of flow condition on shear velocity estimation. It is worth
noting that the depth–slope product (St Venant method) for shear velocity estimation in non-
uniform flow conditions is difficult due to uncertainty in estimation of the friction slope.
Although the boundary-layer parameter approach does not include a specific parameter to
represent flow non-uniformity on velocity and shear velocity, the displacement (δ*) and momen-
tum thickness (θ) respond to flow non-uniformity. Using the velocity defect it can be demonstrated
that C = 2/κ for uniform flow (zero pressure gradient) (Hinze, 1975). Assuming κ = 0.4, one
obtains C = 5 for uniform flow. The coefficient C = 4.4 in equation (4) was originally determined
using non-uniform flow laboratory experimental data over the ranges of 7.5 < h/d50 < 13,

Copyright © 2009 IAHS Press

158 Hossein Afzalimehr & Colin D. Rennie

0.82 < W/h < 3.16, 2.33 × 105 < Re < 4.88 × 105 and 0.18 < Fr < 0.43 (Afzalimehr & Anctil, 2000).
Using the 23 measured velocity profiles along different reaches in Ghamasiab River, the value of
C = 4.4 gave reasonable results, despite field measurements over a much broader range of condi-
tions. However, a better match between u*BL and u*log was observed for C = 4.6. The difference of
C values in the field and the laboratory was less than 5%. It is noteworthy that increasing C
effectively eliminates the observed overprediction of u*BL with respect to u*par discussed above for
the Ghamasiab River data. Good results were obtained for the Salmon River data using C = 4.4.
Best results for the Kaj and Behestabad rivers were obtained using C = 4.4 and C = 4.8, respect-
ively. Our results in different rivers suggest that, for the case of gradually varied flow in gravel-
bed rivers with median grain sizes ranging from 1.89 to 7 cm, C is reasonably constrained near a
value of 4.4, but may range between 4.4 and 4.8. Future research should further assess the sensi-
tivity of C to local conditions.

CONCLUSIONS
The following conclusions can be drawn from this study. First, outer layer velocity deviations
from the log law observed in the laboratory for accelerating and decelerating flows were con-
firmed in a gravel-bed river. Second, the velocity distribution for non-uniform flow conditions in a
gravel-bed river follows the log law in the near-bed region. The height of log-law validity above
the bed is not limited to a constant value of y/h in each velocity profile and changes with flow
conditions. Third, for a range of 16.2 < W/h < 50 and 9.7 < h/d50 < 33.9 under non-uniform flow
conditions in the Ghamasiab River, the maximum flow velocity occurred at the water surface.
Third, application of the boundary-layer displacement thickness (δ*) and the boundary-layer
momentum thickness (θ) can provide a good estimation of shear velocity which is compatible with
the log-law method, and is insensitive to choice of the velocity datum. The empirical constant C
utilized in the boundary-layer method appears to be reasonably constrained near a value of 4.4,
with an observed range of 4.4–4.8 for the four rivers analysed in this study.

RERERENCES
Afzalimehr, H. & Anctil, F. (1999) Velocity distribution and shear velocity behaviour of decelerating flows over a gravel bed.
Can. J. Civil Engng 26(4), 468–475.
Afzalimehr, H. & Anctil, F. (2000) Accelerating shear velocity in gravel-bed channels. Hydrol. Sci. J. 45(1), 113–124.
Afzalimehr, H. & Anctil, F. (2001) Vitesse de frottement associée à un écoulement non-uniforme et une rugosité relative
intermédiaire. J. Hydraul. Res. 39(2), 181–186.
Ashworth, P. J. & Ferguson, R. I. (1986) Interrelationships of channel processes changes and sediments in a proglacial braided
river. Geogr. Ann. 68A(4), 361–371.
Barenblatt, G. I. (1993) Scaling laws for fully developed turbulent shear flows. Part 1. Basic hypotheses and analysis. J. Fluid
Mech. 248, 513–520.
Bathurst, J. C. (1982) Theoretical aspects of flow resistance. In: Gravel-bed Rivers (ed. by R. D. Hey, J. C. Bathurst &
C. R. Thorne), 83–105. John Wiley & Sons Ltd, Chichester, UK.
Bathurst, J. C. (1985) Flow resistance estimation in mountain rivers. J. Hydraul. Engng ASCE. 111, 625–643.
Biron, P. M., Lane, S. N., Roy, A. G., Bradbrook, K. F. & Richards, K. S. (1998) Sensitivity of bed shear stress estimated from
vertical velocity profiles: the problem of sampling resolution. Earth Surf. Processes Landf. 23, 133–139.
Bray, D. I. (1979) Estimation average velocity in gravel-bed rivers. J. Hydraul. Div. ASCE 105(9), 1103–1122.
Bray, D. I. (1982) Flow resistance in gravel-bed rivers. In: Gravel-bed Rivers (ed. by R. D. Hey, J. C. Bathurst & C. R. Thorne),
109–137. John Wiley & Sons Ltd, Chichester, UK.
Bray, D. I. & Davar, K. S. (1987) Resistance to flow in gravel-bed rivers. Can. J. Civil Engng 14(1), 77–86.
Bridge, J. S. & Jarvis, J. (1976) Flow and sedimentary processes in the meandering River South Esk, Glen Clova, Scotland.
Earth Surf. Processes Landf. 1, 303–336.
Bridge, J. S. & Jarvis, J. (1977) Velocity profiles and bed shear stress over various bed configurations in a river bend. Earth
Surf. Processes Landf. 2, 281–294.
Bridge, J. S. & Jarvis, J. (1982) The dynamics of a river bend: a study in flow and sedimentary processes. Sedimentology 29,
499–541.
Charlton, F., Brown, P. M. & Benson, R. W. (1978) The hydraulic geometry of some gravel rivers in Britain. Report IT 180,
Hydraulics Research Station, Wallingford, UK.
Chen, C. L. (1991) Unified theory on power laws for flow resistance. J. Hydrol. Engng ASCE 117(3), 371–389.
Clauser, F. H. (1956) the turbulent boundary layers. Adv. Appl. Mech. 4, 1–15.

Copyright © 2009 IAHS Press

 Determination of bed shear stress in gravel-bed rivers using boundary-layer parameters 159

Einstein, H. A. & El-Samni, E. A. (1949) Hydrodynamic forces on a rough wall. Rev. Modern Phys. 21, 520–524.
Ferguson, R. I. & Ashworth, P. J. (1992) Spatial patterns of bed load transport and channel change in braided and near-braided
rivers. In: Dynamics of Gravel-bed Rivers (ed. by P. Billi, R. D. Hey, C. R. Thorne & P. Tacconi), 477–496. John Wiley
& Sons Ltd, Chichester, UK.
Ferguson, R. I., Prestegaard, K. L. & Ashworth, P. J. (1989) Influence of sand on hydraulics and gravel transport in a braided
gravel bed river. Water Resour. Res. 25, 635–643.
Graf, W. H. (1984) Hydraulics of Sediment Transport. McGraw-Hill, New York, USA.
Hinze, J. O. (1975) Turbulence (second edn). McGraw-Hill, New York, USA.
Keulegan, G. H. (1938) Laws of turbulent flow in open channels. J. Natl. Bur. Stand. 21, 707–741.
Kironoto, B. A. & Graf, W. H. (1995) Turbulence characteristics in rough non-uniform open channel flow. Proc. Inst. Civil
Engng, Water, Maritime and Energy 112, 336–348.
Leopold L. B., Wolman, W. G. & Miller, J. P. (1964) Fluvial Processes in Geomorphology. Freeman, San Francisco,
California, USA.
MacVicar, B. J. & Roy, A. G. (2007) Hydrodynamics of a forced riffle pool in a gravel bed river: 1. Mean velocity and
turbulence intensity. Water Resour. Res. 43, W12401. doi:10.1029/2006WR005272.
Monin, A. S. & Yaglom, A. M. (1971) Statistical Fluid Mechanics: Theory of Turbulence (ed. by J. L. Lumley), MIT Press,
Cambridge, Massachusetts, USA.
Nezu, I. & Rodi, W. (1985) Experimental study on secondary currents in open channel flow. Proc. 21st IAHR Congress.
(Melbourne), vol. 2, 115–119, International Association of Hydraulic Research.
Nezu, I. & Rodi, W. (1986) Open channel flow measurements with a laser Doppler anemometer. J. Hydraul. Engng 112,
335–355.
Nowell, A. R. M. & Church, M. (1979) Flow in a depth-limited boundary layer. J. Geophys. Res. 84, 4816–4824.
Rennie, C. D., Fisher, T. S. R. & Millar, R. G. (1999) Spatial variability of turbulent velocity structure in a natural river. In:
Proc. 1999 ASCE Int. Water Resources Engng Conf. (ed. by R. Walton & R. E. Nece). CD-ROM, American Society of
Civil Engineers.
Reynolds, A. J. (1974) Turbulent Flows in Engineering. John Wiley & Sons Ltd, Chichester, UK.
Robert, A., Roy, A. G. & De Serres, B. (1992) Changes in velocity profiles at roughness transitions in coarse grained channels.
Sedimentology 39, 725–735.
Schlichting, H. & Gersten, K. (2000) Boundary-layer Theory. Springer-Verlag, Berlin, Germany.
Smart, G. M. (1999) Turbulent velocity profiles and boundary shear in gravel bed rivers. J. Hydraul. Div. ASCE 125(2),
106–116
Song, T. & Chiew, Y. M. (2001) Turbulence measurement in non-uniform open-channel flow using Acoustic Doppler
Velocimeter (ADV). J. Engng Mech. ASCE 127(3), 219–232.
Song, T. & Graf, W. H. (1994) Non-uniform open channel flow over a rough bed. J. Hydrosci. Hydraul. Engng 12(1), 1–25.
Song, T., Graf, W. H. & Lemmin, U. (1994) Uniform flow in open channels with moveable gravel bed. J. Hydraul. Res. 32(6),
861–876.
Tennekes, H. & Lumley, J. L. (1972) A First Course in Turbulence. MIT Press, Cambridge, Massachusetts, USA.
White, F. M. (2006) Viscous Fluid Flow. McGraw-Hill, New York, USA.
Wiberg, P. L. & Smith, J. D. (1987) Calculations of the critical shear stress for motion of uniform and heterogeneous sediments.
Water Resour. Res. 23(8), 1471–1480.
Wilcock P. R. (1996) Estimating local bed shear stress from velocity observations. Water Resour. Res. 32, 3361–3366.
Wolman, M. G. (1954) A method of sampling coarse river bed material. Trans. Am. Geophys. Union 35(6), 951–956.
Yang, S. Q., Xu, W. L. & Yu, G. L. (2006) Velocity distribution in gradually accelerating flow. Adv. Water Resour. 29(12),
1969–1980.
Young, A. D. (1989) Boundary Layers. American Institute of Aeronautics and Astronautics, BSP Professional Books Washington DC,
USA.

Received 1 August 2007; accepted 15 July 2008

Copyright © 2009 IAHS Press