You are on page 1of 13

Shear Stress and Hydrodynamic Recovery over Bedforms

of Different Lengths in a Straight Channel


Bruce MacVicar 1 and Lana Obach 2

Abstract: Pools and riffles are common morphological features in rivers that are frequently used but poorly specified analogs in restoration
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

design. Here, straight two-dimensional (2D) bedforms are conceptualized as perturbations and flow recovery is measured in a laboratory
flume with an array of ultrasonic Doppler velocity profilers (UDVPs). The objectives are to (1) assess the variation of skin friction, turbulent
stresses, and total stress; (2) assess the role of topographical feedback on flow recovery; and (3) compare flow recovery in isolated and
bedforms in series. The results show that the total shear stress and near-bed turbulence greatly exceed the skin friction in decelerating flow
and the pool and that hydrodynamic recovery tends to occur at length scales similar to geophysical scales despite potential negative feedback
from the bed. Repeating short bedforms can push the flow to a more turbulent and laterally concentrated equilibrium condition. Implications
for sediment entrainment thresholds, existing models of riffle-pool hydrodynamics, and the stability of constructed riffle pools are discussed.
DOI: 10.1061/(ASCE)HY.1943-7900.0001043. © 2015 American Society of Civil Engineers.

Introduction found in gravel-bed rivers of less than 2% in slope (Montgomery


and Buffington 1997), riffles and pools are defined as the high and
Pools and riffles are formed in gravel-bed channels by local varia- low points, respectively, on an undulating bed. They can occur in
tions in boundary shear stress acting on heterogeneous bed-surface straight and meandering channels and have been classically attrib-
particles during varying stream discharges (Lisle et al. 2000). This uted a modal scaling ratio of one bedform unit every five to seven
general principle is well understood. Recent numerical simulations, channel widths (Keller and Melhorn 1978; Leopold and Wolman
for example, show that an unsteady one-dimensional (1D) coupled 1957). Modern geospatial techniques have put this idea of a classic
flow and morphology model with bed-sorting is sufficient to main- riffle-pool couplet into doubt (Wyrick and Pasternack 2014), but
tain pool depth downstream of riffles (de Almeida and Rodríguez pool and riffle subunits do appear to be spaced at a low multiple
2011). In practice, however, engineers report widespread ambiguity of the channel width, even when forced by nonalluvial elements
in the design, construction, and maintenance of instream structures such as boulders, trees, and valley constrictions (Montgomery et al.
such as riffles and pools due to a lack of specification standards 1995; White et al. 2010; Wyrick and Pasternack 2014). Lateral side
(Miller and Kochel 2009; Radspinner et al. 2010). Despite a cost or point bars are typically associated with pools (Church and Jones
of over a billion dollars (US) spent per year in the United States 1982; Wyrick and Pasternack 2014), as is variation in channel
alone on stream restoration activities, with over 25% of this total width (White et al. 2010; Jackson et al. 2015) and size sorting
on instream habitat improvement and channel reconfiguration of sediment (Sear 1996). Large particles tend to imbricate (Clifford
(Bernhardt et al. 2005), the practice of stream restoration has in 1993) and help to fix the location of the riffle during floods (Church
many ways preceded the science (Lave 2008). Successful examples and Jones 1982; de Almeida and Rodríguez 2011). Some restora-
show that instream structures can improve the stability and ecologi- tion approaches attempt to work with the alluvial nature of gravel
cal integrity of river channels that have been negatively affected by bed rivers (Wheaton et al. 2010), but structured approaches to
human activities such as urbanization and channel straightening riffle-pool design assign a relatively fixed geometry to the channel
(Newbury 2013; Pasternack et al. 2008; Rhoads et al. 2008; Rosgen (Newbury 2013; Pasternack et al. 2008; Rhoads et al. 2008; Rosgen
2001), but expensive failures may result if they are poorly con- 2001). Structured riffle pools typically are straighter and steeper
ceived, designed, or constructed (Kondolf et al. 2007; Miller than natural channels (Newbury 2013) and the suppression of bank
and Kochel 2009). Hydraulic studies are needed that demonstrate erosion is often an explicit criterion (Newbury 2013; Rhoads et al.
how basic design parameters such as bedform length affect the 2008). Structured riffles typically comprise very coarse material so
hydro and sediment dynamics of open channels. that they act as permanent grade-control structures (Newbury 2013;
Unlike other hydraulic structures such as broad crested weirs Rhoads et al. 2008; Rosgen 2001), while the lack of aggrading lat-
(Hager and Schwalt 1994), constructed riffle pools are intended eral bars means that sediment supplied from upstream must be
to mimic at least some aspects of a natural analog. Commonly routed through the pool.
Hydraulically, it is not clear how to design riffle pools to meet
1
Assistant Professor, Dept. of Civil and Environmental Engineering, stability and sediment routing criteria because the topography cre-
Univ. of Waterloo, 200 University Ave. West, Waterloo, ON, Canada ates pressure gradients that change the shape of velocity and tur-
N2L 3G (corresponding author). E-mail: bmacvicar@uwaterloo.ca bulence profiles. In the downstream portion or tail of a pool, for
2
Dept. of Civil and Environmental Engineering, Univ. of Waterloo, instance, the reduction in flow depth leads to convectively accel-
200 University Ave. West, Waterloo, ON, Canada N2L 3G.
erating flow (CAF) due to a favorable pressure gradient. In such
Note. This manuscript was submitted on January 1, 2014; approved on
April 2, 2015; published online on June 3, 2015. Discussion period open environments, relatively high velocities occur near the channel
until November 3, 2015; separate discussions must be submitted for indi- bed and turbulence is reduced (Whiting 1997; Piomelli et al. 2000;
vidual papers. This paper is part of the Journal of Hydraulic Engineering, MacVicar and Roy 2007; Yang and Chow 2008; MacVicar and
© ASCE, ISSN 0733-9429/04015025(13)/$25.00. Rennie 2012; MacVicar and Best 2013). In the upstream portion

© ASCE 04015025-1 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


or head of the pool, on the other hand, an increase in flow depth linked to geophysical macrobedform scales. Flow patterns in the
leads to convectively decelerating flow (CDF) due to an adverse straight channel, including near-bed acceleration during CAF and
pressure gradient. In such environments, relatively low velocities the lateral convergence of flow during CDF, shared some character-
occur near the channel bed but turbulence is increased (Clifford istics with observations in natural pool riffle systems used to
and French 1993; Kironoto and Graf 1995; Krogstad and Skare postulate the velocity reversal (Keller 1971) and lateral flow con-
1995; Yang and Chow 2008; Lee and Sung 2009; Nagano et al. vergence hypotheses (MacWilliams et al. 2006), both of which are
1998; Monty et al. 2011; MacVicar and Rennie 2012; MacVicar central to discussions of pool formation. MacVicar and Best (2013)
and Best 2013). Because of the altered near-bed hydraulics, the concluded that the different hypotheses likely described linked as-
Shields (1936) criterion, which was developed for uniform flows, pects of the same hydrodynamic process. Their experiments, how-
may be unreliable over pools and riffles. The addition of form ever, were carried out in a smooth channel over isolated and
roughness in riffles and pools also means that the skin friction must relatively long bedforms. In an alluvial system, feedback from the
be distinguished from the total resistance to flow for sediment bed could accelerate the rate of hydrodynamic recovery or force the
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

transport calculations (Einstein and Barbarossa 1952; Millar 1999; flow to a new equilibrium condition. Riffle-pool bedforms are also
Papanicolaou et al. 2012). The generation of turbulence away from characterized by rough beds and it is not clear that results from
the bed can result in the advection of powerful coherent turbulent smooth bed channels are applicable.
structures toward the bed (Krogstad and Skare 1995; MacVicar The current study continues investigations into the hydrody-
et al. 2013), which can affect scour in straight channels (Sumer et al. namics of riffle pools by exploring feedback and roughness effects.
2003; Dwivedi et al. 2010) and in pools with width constrictions Specific objectives are to (1) assess skin friction, turbulent stresses,
(Thompson 2006; Thompson and Wohl 2009). Shear stress is fre- and total stress over two-dimensional (2D) bedforms with smooth
quently modeled in riffle pools using a time averaged estimate of and rough beds, (2) assess the role of feedback from the bed on the
velocity (e.g., Brown and Pasternack 2009; MacWilliams et al. recovery of shear stress and velocity distributions, and (3) assess
2006, 2010; Caamano et al. 2009; Jackson et al. 2015), but this the effect of repeating vs isolated bedforms on hydrodynamic re-
approach may underestimate shear stress in areas of high covery. The scope of the work is limited to a straight laboratory
turbulence like the head of the pool. Despite recent advances in flume with a fixed bed at a constant discharge. Implications for
physics-based relations that would account for turbulence (Celik sediment transport estimates and riffle-pool design are discussed
et al. 2013; Diplas et al. 2008), not enough is known about based on the shear stress results.
riffle-pool hydrodynamics for the role of turbulence to be properly
accounted for.
The question of shear stress is also linked with the question of Methods
bedform scale. How far apart and how long should the bedforms
be? Designers tend to use empirical geophysical scales (Miller and
Kochel 2009; Newbury 2013), the physical justification for which Experimental Apparatus
is not clear. Bedform scale is investigated in the current study using Experiments were conducted in a 17 m long, 0.6 m wide recircu-
the concept of flow recovery. Flow recovery is defined as the ten- lating flume at a slope of 0.001 m=m. Modular bedforms were con-
dency of flow parameters to return toward uniform flow values structed from PVC sheets. Uniform depth modules 0.4 m long and
when perturbations such as pressure gradients are removed (Smits either low (0.025 m) or high (0.085 m) were added or removed
and Wood 1985). Inner zone parameters such as the shear velocity to create deep or shallow uniform sections of different lengths
tend to recover quickly while outer zone parameters such as the (Table 1). The nonuniform depth modules were 0.51 m long and
velocity defect relative to uniform flow require longer distances fixed at an angle of 7.2° from the horizontal (Fig. 1). This slope
(Smits and Wood 1985). In a study of flow recovery downstream ensures that permanent flow separation does not occur (Simpson
of straight bedforms, MacVicar and Best (2013) found that recov- 1981), and is in the range of typical leeside angles in macrobed-
ery distances for the near-bed Reynolds stress exhibited a modal forms (Best and Kostaschuk 2002; Carling and Orr 2000). The non-
scaling relation with the channel width, which supported Yalin uniform modules could be turned 180° to create either CAF or CDF
and da Silva’s (2001) hypothesis that turbulent length scales are sections. The bedforms were installed downstream of a uniform

Table 1. Channel Geometry and Flow Parameters for Experimental Runs


Geometry Flow parameters upstream of bedform
Shear velocity (cm s−1 )
Run identifier X s (m) X s =Y X d (m) X d =Y X b (m) X b =Y u1o u2o u3o u4o Πo Ψo
R240P240 2.4 4.0 2.4 4.0 5.8 9.7 1.18 0.79 0.69 0.60 −0.40 1.01
R160P240 1.6 2.7 2.4 4.0 5.0 8.4 1.16 0.81 0.74 0.65 −0.26 1.03
R80P240 0.8 1.3 2.4 4.0 4.2 7.0 1.16 0.82 0.68 0.65 −0.19 1.02
R40P240 0.4 0.67 2.4 4.0 3.8 6.4 1.16 0.75 0.74 0.59 −0.35 1.01
R40P160 0.4 0.67 1.6 2.7 3.0 5.0 1.16 0.80 0.58 0.65 −0.23 1.03
R40P80 0.4 0.67 0.8 1.3 2.2 3.7 1.16 0.67 0.49 0.66 −0.26 1.05
R40P40 0.4 0.67 0.4 0.67 1.8 3.0 1.17 0.65 0.79 0.50 −0.38 1.00
R240P240 R 2.4 4.0 2.4 4.0 5.8 9.7 0.68 0.46 0.93 0.53 −0.04 1.01
Note: Channel geometry includes length of shallow section (X s ), length of deep section (X d ), and total bedform length including one deep, one shallow, and
two transitional sections (X b ). Section lengths are normalized by the channel width (Y ¼ 0.60 m) for comparison with other studies. All flow parameters are
measured immediately upstream of the bedform at the channel centerline. Shear velocity is estimated with the Clauser method with a fixed intercept (u1o ),
Clauser method with a floating intercept (u2o ), Reynolds stress method (u3o ), and boundary conditions method (u4o ). Outer zone parameters include the wake
parameter (Πo ), and the centerline lateral concentration of flow (Ψo ).

© ASCE 04015025-2 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Q Y

7.2°
7.2°

Xs Xd
Xb
Uniform Approach Convectively Accelerating Shallow Uniform Convectively Decelerating Deep Uniform
Flow Flow (CAF) (Riffle) Flow (CDF) (Pool)

Fig. 1. Perspective view of the generalized 2D bedform configuration used in the current study; the lengths of the CAF and CDF sections were fixed
while those of the riffle (X s ) and pool (X d ) sections were varied to change the overall bedform length
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

reach just over 6 m in length. Additional bedforms were installed and skewness profiles. Poor quality sampling bins were found
downstream, as space permitted, to the end of the flume. A final to occur due to reflectance from the channel bed and insufficient
bedform was placed at the downstream end of all runs and a curtain seeding concentration. Raw time series and power spectra were re-
gate at the end of the flume was adjusted to ensure the tailwater viewed manually to confirm data quality assessments. Poor quality
elevations were within 0.002 m for all runs. time series were not included in subsequent analyses.
An array of 4 MHz ultrasonic Doppler velocity profilers
(UDVPs) manufactured by Metflow (Switzerland) was used to
measure flow velocities. The UDVPs measure the Doppler shift Determination of Velocity and Reynolds Stress from
in sound velocity from the return signal reflected from neutrally Beam Velocities
buoyant tracer particles. They are single-beam profilers that sample The streamwise and vertical components of flow and the principal
velocity quasisimultaneously in a series of bins along the beam Reynolds stress were calculated from measured beam velocities.
axis, with the recorded velocity vectors oriented in the direction Beam velocities (ui ) were recorded at orientations of 30° and −30°
of the beam. Multiple beams cannot be measured simultaneously. to the vertical (referred to using subscripts 1 and 3, respectively).
The UDVPs are suited to laboratory measurements due to their high Reynolds decomposition was used to calculate the mean (U i ) and
sampling frequency (up to 100 Hz), small size (0.008 m diameter), fluctuating parts (ui0 ) of each beam velocity. The mean and variance
and short blanking distance (∼0.017 m) (Best et al. 2001). The of each beam velocity were linearly interpolated on a normalized
probes were held in fixed holders slightly under the water surface three-dimensional (3D) grid through the sampling volume. Grid
(<0.005 m submergence) to minimize disturbance to the flow. The spacing was set at 1=20 of the channel width (0.03 m), 1=50 of
UDVP measurement parameters were set to record 2-min time the flow depth (variable from 1 to 2 mm depending on local depth),
series at 40 Hz in sampling bins that were 0.001 to 0.002 m long and 1=10 of the length of each morphological subunit in the stream-
in the beam direction and less than 0.005 m in diameter. See wise direction (0.04 to 0.24 m depending on the run and subunit).
MacVicar and Best (2013) for more details on the experimental Following Lhermitte and Lemmin (1994), the streamwise (U) and
apparatus. vertical (W) mean velocities were calculated at each grid node from
the interpolated beam velocities
Quality Control
U1 − U3
Quality control was effectuated using MITT (multi-instrument U¼ ð2Þ
2 sin α
turbulence toolbox) (MacVicar et al. 2014), a set of MATLAB
algorithms designed to clean and classify data quality from veloc-
imeters. Spikes were detected using a skewed velocity threshold U1 þ U3
and the Goring-Nikora (2002) phase-space thresholding algorithm. W¼− ¼ −U 2 ð3Þ
The skewed velocity threshold sets a threshold velocity for good 2 cos α
data as
Again following Lhermitte and Lemmin (1994), the covariance
uT ¼ umedian  Cλsu ð1Þ (−u 0 w 0 ) was calculated from the difference between the variance
terms from the probes pointed up and downstream as
ffi = median velocity; λ = universal threshold
wherepffiffiffiffiffiffiffiffiffiffi
umedian
(λ ¼ 2 ln n, where n is the length of the time series) (Goring  02 
u3 − u102
and Nikora 2002); su = one-sided standard deviation calculated −u 0 w 0 ¼− ð4Þ
from the data on the side of the median that has lower variability; 2 sin 2ϕ
and C = coefficient. This method was found to be effective for data
measured with the UDVP instrument where seeding was insuffi- where ϕ is the beam angle measured from the vertical.
cient (MacVicar and Best 2013). In such a case, a second mode The interpolation method is similar to that of Vermeulen et al.
of velocity samples were found clustered around a null velocity (2014), who decoupled the beam measurements from an acoustic
so that they produced a highly skewed distribution. A high value Doppler current profiler (ADCP) to improve the accuracy of the
of C (∼3.0) was found to reliably detect this type of outlier without mean velocity and covariance estimates. This method reduces
resulting in false positive spike detections. Low-pass filtering was the spatial extent over which flow homogeneity is assumed. The
used to remove high frequency noise following Roy et al. (1997). relative flow depth was used instead of the actual depth following
For each beam of simultaneously measured sampling bins, poor the stream-tube approach, which was found to more accurately ac-
quality data cells were identified from statistical outliers (α ¼ 0.05) count for the bulk flow acceleration and deceleration over nonuni-
to third order polynomial fits of mean velocity, standard deviation form topography (Vermeulen et al. 2014).

© ASCE 04015025-3 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Determination of Shear Velocity (u  ) ðδ d − δ θ ÞU c
u4 ¼ ð7Þ
Shear velocity and shear stress are notoriously difficult parameters Kδ d
to calculate, particularly over nonuniform beds. To better describe
the hydraulic environment over the bedforms, four different esti- where U c is the maximum velocity in the profile at elevation zc ,
mates of u were made based on established techniques that use K is a constant (∼2=κ) in uniform flow and nonuniform flow
inner and outer zone parameters, mean velocity and Reynolds in equilibrium (Afzalimehr and Anctil 2000, using data from
stress. The first method, commonly referred to as the Clauser Kironoto and Graf 1995), and δ d and δ θ are the boundary layer
method, finds u by fitting the law of the wall to measured velocity displacement and momentum thickness, respectively, defined as
values in the near-bed (inner) zone, written as Z z 
c U
δd ¼ 1− dz ð8Þ
1 0 Uc
uþ ¼ ln zþ þ A ð5Þ
κ
Z  
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

where uþ ¼ U=u and zþ are the normalized time-averaged veloc-


zc U U
δθ ¼ 1− dz ð9Þ
ity and height above the bed, respectively, κ is the von Karman 0 Uc Uc
constant and A is a constant of integration. Over smooth uniform
beds zþ ¼ zu =ν, where ν is the kinematic viscosity, while over The boundary characteristics method has been found to accu-
rough uniform beds zþ ¼ z=ks , where ks ∼ D50 for a uniform rately represent the total shear stress (i.e., skin friction and form
sediment. roughness) over spatially varying boundaries such as a boulder
The Clauser method requires the determination of the vertical array (Papanicolaou et al. 2012). Note that a constant value was
extent of the near-bed log-linear (inner) zone and specification assumed for K following Papanicolaou et al. (2012) despite the
of values for κ and A. It was assumed that κ ∼ 0.41 for all runs nonuniform and nonequilibrium conditions over the bedforms.
(Nezu and Rodi 1986). The inner zone can be approximated as
the lower 20% of the flow depth in uniform flow (Nezu and Rodi
Determination of Outer Zone Parameters
1986; Kironoto and Graf 1994) and in some cases of nonuniform
flow (Kironoto and Graf 1995; Onitsuka et al. 2009; Afzalimehr Coles wake law is commonly used to describe the velocity defect or
and Anctil 2000), but has been found to vary in other cases deviation from the log-law [Eq. (5)] (Coles 1956; Kironoto and
(Afzalimehr and Rennie 2009; Monty et al. 2011; MacVicar and Graf 1995; Song and Chiew 2001; Papanicolaou et al. 2012)
Rennie 2012). For the current study, a set of two-segment log-  
linear relations were fit to the relation between U and z following uc − ū 1 þ 2Π 2 π z
¼ ln z þ sin ð10Þ
MacVicar and Rennie (2012), where the breakpoint between the u κ κ 2 zc
segments was allowed to vary between 10 and 20% of the flow
depth, and the relation with the minimum RMS error was assumed where Π is defined as Coles wake parameter, uc is the maximum
to correctly delineate the inner zone. A is also considered a constant velocity in the profile at elevation zc . Π is approximately 0 in
in uniform flow over smooth (A ∼ 5.3, Nezu and Rodi 1986) and uniform flows, greater than 0 in CDF, and less than 0 in CAF.
rough beds (A ∼ 8.5, Kironoto and Graf 1994) and in some cases (Kironoto and Graf 1995; Song and Chiew 2001). Two estimates
of nonuniform flow (Song and Chiew 2001; Onitsuka et al. 2009; (Π and Π2 ) were determined by fitting the velocity profiles to
Afzalimehr and Anctil 2000) but has been found to vary in other Eq. (10) with u ¼ u1 and u2 , respectively.
cases (Kironoto and Graf 1995; Nagano et al. 1998; Piomelli et al. Another metric tracked in this study is the lateral concentration
2000; Monty et al. 2011). In the current study two estimates of u of flow (Ψ), defined as
were determined using a constant value for A (u1 ) and allowing A to
float as part of the curve fitting exercise (u2 and A2 ). Ψ ¼ qp =qb ð11Þ
Shear velocity was also estimated by extrapolating the principle
Reynolds stress to the bed where qp = specific discharge measured at a lateral location (y) and
qb = average specific discharge across the whole channel. qp is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
determined as the product of U p and Z, where U p is the depth-
u3 ¼ −u 0 w 0 z→0 ð6Þ averaged velocity at y and Z is the corresponding flow depth.
qb is determined as Q=Y, where Q is the total discharge and Y
Many studies have shown u1 ≈ u3 over smooth and rough beds is the channel width. This metric is a measure of the lateral velocity
(e.g., Nezu and Rodi 1986 and Kironoto and Graf 1994). Over non- defect, as Ψ ¼ 1 if the flow has a uniform lateral distribution. In the
uniform beds it can be more difficult to obtain accurate measure- existing flume under similar flow conditions, Ψ ∼ 1.05 along the
ments, but the accuracy of u3 may be superior to u1 over complex centerline of the channel upstream of the bedform and increased
geometry due to a reduced reliance on near-bed measurements to 1.21 in a straight pool with a smooth bed (MacVicar and Best
(Biron et al. 2004). Nonuniform flows disrupt the linear relation 2013). Higher values were observed at the channel centerline
between relative depth and Reynolds stress, but reasonable esti- for narrower channels and in a straight pool with a rough bed
mates are possible if the extrapolation includes only measurements (MacVicar and Rennie 2012).
from the inner zone (Kironoto and Graf 1995). u3 was therefore A dimensionless velocity profile shape factor (H) was used to
calculated in the current study using the inner zone as defined compare results with other studies of nonuniform flow
in the Clauser method.
The boundary characteristics method was also utilized in this δd
H¼ ð12Þ
study (u4 ). This method is derived from Prandtl’s concept of a δθ
velocity defect arising from viscous shear at a solid boundary
and has been successfully applied to gradually (Afzalimehr and H ∼ 1.3 in uniform flows, with lower values in CAF and higher
Anctil 2000) and rapidly varying flows (Papanicolaou et al. 2012). values in CDF (Monty et al. 2011). Flow separation occurs at
u4 was calculated as H ∼ 2.4 in turbulent flow (White 2009).

© ASCE 04015025-4 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Experimental Runs constant included the width (Y ¼ 0.60 m), the discharge
(Q ¼ 0.0154 m3 =s), the upstream flow depth (Zd ¼ 0.12 m), bulk
Nine runs on seven bedform configurations were investigated for
velocity (U b ¼ 0.21 m=s), and Reynolds number (R ¼ 2.5 × 104 ).
the current study (Table 1). The length of the shallow uniform riffle
Given the resulting water depths between 0.06 and 0.12 m and the
(X s ) was tested in the first four runs (R240P240, R160P240,
UDVP parameters, velocity was thus measured quasisimultane-
R80P240, and R40P240), and the length of the deep uniform pool
ously in 50 to 100 bins defined along the beam axis. Velocities near
(X d ) was tested in the fourth through the seventh runs (R40P240,
to the water surface could not be measured due to the UDVP blank-
R40P160, R40P80, and R40P40). The effect of a roughened bed
ing distance. Four to seven profiles were measured across one half
was tested with the R240P240 bed configuration covered with
of the channel width, and cross sections were spaced at 0.15 m in
TrueGrip Traction Tape (Incom, Ancaster, Ontario) (0.36 mm alu-
the streamwise direction. Calculated from velocity and Reynolds
minum oxide particles). For each of the first eight runs, the test
stress profiles immediately upstream of the initial bedform on
section began in the uniform flow approach reach and extended
the channel centerline, the shear velocity varied between 0.5 and
over one complete bedform (consisting of CAF, shallow uniform,
1.2 cm=s, depending on the run and the calculation method. Differ-
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

CDF, and deep uniform subunits (Fig. 1). A ninth run was com-
ences between methods are discussed as part of the results. Coles
pleted using the R40P40 bed configuration by measuring in a sec- wake parameter (Π ∼ −0.40 to −0.04) and the lateral concentration
ond test section over the fourth of five identical bedforms. The total of flow (Ψ ∼ 1.00 to 1.05) were relatively consistent for all runs and
bedform length-width ratio (X b =Y) for the runs ranged from 3.0 to similar to those in the companion experiments (MacVicar and Best
9.7, which straddles the commonly cited scaling relation of 5 to 7 2013). Negative Π values indicate that flow acceleration was affect-
found in natural riffle pools. ing velocity profiles upstream of the bedforms.
The flow was fully turbulent and scaled using the geometry and
Froude number from a field example of a riffle pool (MacVicar
and Roy 2007). For the geometry, the aspect ratio in the shallow Results
uniform section (1∶10) and the ratio of flow depths in the deep
and shallow uniform sections (∼2∶1) were set to match the field
example. For the flow, the Froude number for the experiments Velocity and Reynolds Stress Profiles
(0.54) was comparable to that in the field example at bankfull dis- Velocity and Reynolds stress profiles over the bedforms exhibit a
charge (0.59). Flow measurements at this stage were found to be number of common characteristics that can be illustrated with select
correlated with patterns of gravel entrainment (MacVicar and Roy profiles from run R240P240 (Fig. 2). Velocity profiles tend to vary
2011) and for this reason were thought to be relevant for a discus- in gradient as flow accelerates and decelerates over the bedforms
sion that includes sediment transport. Parameters that were held [Fig. 2(a)]. In CAF, for instance, velocity profiles are flatter than in

Fig. 2. Select velocity and Reynolds stress profiles for run R240P240; profiles are normalized by (a and c) outer zone parameters; (b and d) inner zone
parameters; note that all profiles are from channel centerline location at the downstream end of the sections indicated in the legend

© ASCE 04015025-5 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


the upstream uniform flow and the riffle, which results in a higher and Song and Chiew (2001) observed the opposite effect, noting
near-bed velocity. In contrast, velocity profiles are steeper in the that −u 0 w 0 profiles tend to be convex with a peak at the bed in
outer zone as a result of CDF, which results in a lower near-bed CAF. This discrepancy is again thought to be related to the strong
velocity. The velocity profile tends to recover toward that observed pressure gradient in the current study, which will be discussed more
in uniform flow in the pool. Exhibited trends in CAF and CDF are thoroughly following the presentation of results. Within the CDF
in agreement with laboratory studies of depth-varying flows in zone, −u 0 w 0 is characterized by a strong positive gradient near
equilibrium (Kironoto and Graf 1995; Song and Chiew 2001), a the bed so that maxima occur between z=Z ¼ 0.15 and 0.35.
field study in a forced riffle pool (MacVicar and Roy 2007), and Such changes are expected in flow subjected to positive pressure
the general case of flows subjected to pressure gradients (Monty gradients due to higher inertia and the generation of turbulence due
et al. 2011; Piomelli et al. 2000; Nagano et al. 1998), while the to increased strain rates during the deceleration of flow (Kironoto
trends downstream of depth transitions are in general agreement and Graf 1995; Krogstad and Skare 1995; Lee and Sung 2009). The
with studies of flow recovery (Onitsuka et al. 2009; MacVicar increase in turbulence is well known to result in form roughness in
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

and Best 2013). nonuniform open channels and is distinguished from skin friction at
When normalized by the inner zone parameters, it is possible to the bed (Einstein and Barbarossa 1952; Millar 1999). This turbu-
demark a near-bed log-linear zone in all of the profiles [Fig. 2(b)]. lence tends to advect away from the boundary in the downstream
This zone has been observed in other studies of nonuniform flow, uniform flow section and then dissipate as it moves downstream,
although there is some disagreement on the variation of the inter- although turbulence remains high in the outer region of the down-
cept (A). Kironoto and Graf (1995) identified a weak tendency for stream CAF in comparison with the flow upstream of the bedform
the intercept to be higher in decelerating flow and higher in accel- [Fig. 2(c)].
erating flow. In contrast, Piomelli et al. (2000) noted an increase in To compare shear velocity estimates from the near-bed profiles
the thickness of the laminar sublayer and relaminarization due to of velocity and turbulence, −u 0 w 0 was normalized as by u2 in
CAF, while Nagano et al. (1998) and Monty et al. (2011) observed Fig. 2(d). As expected from other studies (Kironoto and Graf 1995;
pffiffiffiffiffiffiffiffiffiffiffiffiffi
a decrease in A as a result of deceleration. The current study shows Song and Chiew 2001), −u 0 w 0 ≈ u2 for zþ < 100 in the uniform
that A tends to increase due to CAF and decrease in CDF. In many flow upstream of the bedform, over the riffle, and in CDF. In CAF,
cases, however, the range of variability in the current study is p ffiffiffiffiffiffiffiffiffiffiffiffiffi
higher than what was observed in the previous studies. The differ- −u 0 w 0 ≈ u2 only at a lower relative elevation (zþ ∼ 20), but it is
ences are thought to be the result of the confounding of inner zone still reasonable to assume equivalence between the two at the bed.
pffiffiffiffiffiffiffiffiffiffiffiffiffi
parameters u and A with the velocity defect in the outer zone and However, −u 0 w 0 ≠ u2 in the deep uniform section and in the
the relatively high pressure gradients observed in the current study. downstream CAF, where the turbulence is relatively elevated close
These issues will be discussed in the following section as part of a to the bed and throughout the flow depth.
comparison of velocity defect shape factors (Π, Ψ, and H).
Reynolds stress profiles vary in magnitude and near-bed gra- Comparison of Shear Velocity (u  ) Estimates
dient over the bedforms [Fig. 2(c)]. As expected from studies of
uniform flow, −u 0 w 0 increases toward the bed when measured up- A comparison of u estimates shows that they converge in some
pffiffiffiffiffiffiffiffiffiffiffiffiffi
stream of the bedform. During CAF, −u 0 w 0 is lower near the bed locations but diverge in others (Fig. 3). −u 0 w 0 measured at
and the profile becomes slightly concave. Kironoto and Graf (1995) z=Z ¼ 0.10 is also shown for comparison because turbulence

Fig. 3. Shear velocity estimates along the channel centerline over the bedforms for run R240P240, including (a) bed topography and water surface
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
elevation; (b) u and −u 0 w 0 z=Z¼0.1 ; (c) log law intercepts (A) and the shape factor (H)

© ASCE 04015025-6 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


successfully to accelerating flow (Afzalimehr and Anctil 2000) and
over boulders (Papanicolaou et al. 2012), and u4 is calculated using
mean velocity measurements in the outer zone, which are generally
more accurate and insensitive to positioning errors than inner zone
measurements. The increase in u4 and form roughness also corre-
spond with a known local energy loss due to increased turbulence
during CDF (Einstein and Barbarossa 1952; Millar 1999; MacVicar
2013). u4 was therefore considered to be representative of the
total stress. In contrast with u4 , the peak in u3 during CDF is rel-
atively short lived, and u3 follows u1 quite closely through the pool
before they diverge again in the downstream CAF zone. u3 is
thought to have a relatively low accuracy due to the difficulty of
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

accurately extrapolating Reynolds stress to the bed in nonuniform


qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

flows (Kironoto and Graf 1995). The peak in −u 0 w 0 z=Z¼0.1 is
more sustained so that it exceeds u1 and u3 while remaining less
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
than u4 through the pool. −u 0 w 0 z=Z¼0.1 is considered to be a
Fig. 4. Scatter plot of Coles wake parameter estimates (Π and Π2 ) ver-
sus the shape factor (H) for all centerline locations in run R240P240; reasonable estimate of the near-bed turbulent stresses based on
values and known thresholds from other studies are shown where the rigorous data quality control and the minimization of the
available assumption of flow homogeneity inherent in the applied interpola-
tion method.

was thought to be relevant for a discussion of pressure fluctuations Tracking Metrics of Velocity and Turbulence
on the bed and their role in sediment transport. The selected eleva- Distribution over Bedforms
tion is close to the bed but within the range that is sampled reliably To demonstrate the effect of the length of bedforms on flow
by the UDVPs. With the exception of u1 , the estimates are in rel- hydraulics, various inner and outer zone parameters are shown
ative agreement upstream of the bedform, through CAF, and over in Fig. 5. Selected inner zone parameters include u1 ,
the riffle [Fig. 3(b)]. u estimates tend to then diverge in CDF qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
before converging in the pool as they relax toward uniform flow −u 0 w 0 z=Z¼0.1 , and u4 , which represent the skin friction, near-
values. Again with the exception of u1 , the estimates all show bed turbulent stress, and total stress, respectively. Consistent with
a local peak in u within the CDF zone. The highest estimates run R240P240, u1 rises and falls with the bed, with the maxima
are predicted using the boundary conditions method (u4 ) and occurring in the riffle (u1 =uo ∼ 2.1) and the minima at the transition
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
from CDF to the pool (u1 =uo ∼ 0.5–0.75). For the runs with differ-
−u 0 w 0 z=Z¼0.1 . u1 follows a unique pattern as it tends to rise
ent riffle lengths, no differences between the runs were noted for
and fall with the bedform, reaching a maximum value in the riffle u1 over the riffles or the pools [Fig. 5(a)]. For the runs with differ-
and a minimum in the downstream end of the CDF section before ent pool lengths, no significant differences were observed over the
recovering in the pool. riffles, but u1 =uo < 1 at the end of the pools where they are rela-
To sift through the predictions of u , the parameters A, A2 , and tively short, which indicates that skin friction is relatively low
H are shown in Fig. 3(c). A2 varies widely from the constant value in these pools and that recovery of this parameter is incomplete
of A, increasing rapidly in CAF, recovering in the riffle, decreasing qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

in CDF and recovering in the pool. H tends to exhibit opposite [Fig. 5(b)]. −u 0 w 0 z=Z¼0.1 diverges from u1 as it is less than half
trends to A2 , falling during CAF and rising during CDF. The the value of u1 by the end of the CAF section and more than twice
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
parameter H is thought to be reasonably accurate given the exper-
the value of u1 at the end of the CDF section. −u 0 w 0 z=Z¼0.1 tends
imental apparatus because, U=U c ∼ 1 near the water surface [used
in Eqs. (8) and (9)], while dz → 0 near the channel bed where the to recover toward u1 nearer the beginning of longer riffles
velocity defect is greater. The parameter A2, on the other hand, is (e.g., R240P240) in comparison with short rifflesffi (e.g., R80P240)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sensitive to near-bed positioning errors and the velocity defect in [Fig. 5(a)]. In a similar manner, −u 0 w 0 z=Z¼0.1 tends to recover
the outer zone (Monty et al. 2011). Near-bed positioning is thought
more in longer rather than shorter pools [Fig. 5(a)]. As discussed
to be a minor source of error here because similar variability of A2 in the previous section u4 tends to follow similar patterns as
was observed for all runs (results not shown). To test the effect of qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the outer zone velocity defect, Π and Π2 (calculated using A and −u 0 w 0 z=Z¼0.1 over the bedforms, with the exception that u4 re-
A2 , respectively) were compared with H (Fig. 4). As shown, Π is mains low through CDF and into the pool for runs with short riffles
highly correlated with H (R2 ¼ 0.97) while Π2 is not (R2 ¼ 0.35). [Fig. 5(a)]. This result means that the peak in u4 during CDF is
It thus appears that the variability in A2 and u2 are confounded with moderated by the length of the riffle. Longer pools allow for a
variability in the outer zone and that u1 is a better estimate of the recovery of u4 to values observed upstream of the bedform.
inner-zone shear stress. The agreement of u2 with u3 and u4 Presented outer zone parameters include Π and Ψ, which were
through CAF and over the riffle may reflect changes to the outer assumed to represent the vertical and lateral velocity defects, re-
zone that are recorded by the three estimates. spectively (Fig. 5). A comparison of Π with u4 shows that these
In contrast with u1 , estimates of u3 and u4 exhibit peaks near the parameters are highly correlated, which is expected given that they
transition from CDF to the pool, with the peak in u4 greatly exceed- are different representations of the velocity defect. u4 is low when
ing the other parameters. Uncertainty in u4 is related to the un- velocity profiles are relatively flat (as they are in CAF) and high
known variability of K in Eq. (7) as a result of nonequilibrium where velocity gradients are higher in the outer zone (as they are in
pressure gradients. However, the same assumption has been applied CDF). Π is also highly sensitive to the estimate for u1 so that a

© ASCE 04015025-7 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 5. Normalized flow and turbulence distribution metrics for (a) riffle length and roughness runs (as listed in the legend a); (b) pool length and
bedforms in sequence runs (as listed in legend b); note that the streamwise distance (xþ ) has been normalized by the length of each bedform subunit
for comparison between runs; a color version of this figure is available online

slightly lower minimum in u1 for run R40P160 results in a higher The results from the run with the rough bed (R240P240 R) show
peak value for Π. Due to the linear interpolation method used for that roughness tends to affect the magnitude rather than the spatial
beam velocities, the error is higher around local peaks (as occurs at patterns of the flow and turbulence metrics [Fig. 5(a)]. The estimate
the transition from CDF to the pool) in comparison with areas for u1 was less for this run than the other runs (Table 1), likely
where flow parameters are gradually changing (e.g., in the pool). due to the method used to normalize the inner zone parameters
The parameter Ψ is also correlated with u4 and Π, but shows a over the rough bed, and this low estimate accounts for the high
wider variability as a result of riffles with different lengths peaks observed for the other parameters.
[Fig. 5(a)] and is relatively insensitive to the length of the pool The fourth pool in run R40P40 is an exception to many of the
[Fig. 5(b)]. Some unexplained variability in Ψ occurs over in described trends because the flow entering the test section is more
CAF and over the riffle (xþ < 3) in the runs with different pool turbulent and more laterally and vertically concentrated as a result
lengths and in particular for run R40P80. Backwater effects over of conditioning over the upstream bedforms [Fig. 5(b)]. Relatively
the short (0.40 m) riffle may be causing some of this variability, qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
although it is noted that Ψ is dependent on measurements across high values of −u 0 w 0 z=Z¼0.1 , u4 , Π and Ψ occur at the upstream
the width of the channel and errors or missing data close to the limit of the CAF section and, while they rapidly decrease within
side wall may be causing some of the variability. The sensitivity CAF, these parameters remain high relative to the other runs in
to riffle length shows that flows remain more laterally concentrated the shallow uniform flow section. Within the CDF section, the same
at the end of the pools in runs with longer riffles [Fig. 5(a)]. parameters all increase in a manner similar to what was observed in

© ASCE 04015025-8 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


the other runs, but the increases in Π and Ψ are less than and greater Krogstad and Skare 1995) studied decelerating flows with a sim-
than the increases observed in other runs, respectively. No dip in ilar velocity defect to the current study (H ¼ 2.0, Fig. 4), and they
u1 is observed at the transition from CDF to deep uniform flow found no shift in the log-law intercept. Ultimately this study’s
and u1 follows the bed profile without exception. Also unlike other results are not able to resolve the question, and better quality
test scenarios, all parameter values at the downstream end of the near-bed (i.e. zþ < 100) measurements are required for such an
bedform are similar to those at the upstream end, which suggests assessment.
that a periodic but equilibrium state is reached over repeating bed- Despite the uncertainty, shear velocity calculated with the
forms that does not occur over isolated forms. The high value of Ψ assumption of a universal intercept (u1 ) is a valuable metric for
indicates that this more rapid return to uniform flow conditions is comparison with the total and turbulent stresses and other studies
coincident with a significant lateral redistribution of flow so that where similar assumptions were made. The key 3D numerical mod-
more flow passes through the middle of the channel and velocities eling study of MacWilliams et al. (2006), for instance, calculated
near the side walls of the channel are relatively low. shear stress from a near-bed velocity and an assumption of a near-
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

bed log law. The comparison with MacWilliams et al. (2006) is


instructive because they found that the near-bed velocity in the
Discussion Dry Creek pool is much lower than the near-bed velocity over
the point bar at high discharges. This same site was originally
Key results from this study relate to the divergence between esti- used by Keller (1971) to propose his velocity-reversal hypothesis.
mates of total shear stress, skin friction, and turbulent stresses, and MacWilliams et al. (2006) used the near-bed velocity to estimate
the perturbation to and recovery of shear stress as a result of the shear stress and showed the existence of a high stress corridor over
bedforms. Perhaps most significant for understanding the hydrody- the bar—a result that was used to propose their lateral flow con-
namics is the observation that the total shear stress and near-bed vergence hypothesis for pool maintenance. Other modeling studies
turbulence greatly exceed the skin friction in CDF and the pool. have similarly used the near-bed velocity to estimate shear stress
In terms of bedform scaling, a key result is that the degree to which (MacWilliams et al. 2010; Caamano et al. 2009; Jackson et al.
the different flow parameters recover is in many cases a function of 2015). In contrast, both u4 and −u 0 w 0 , which are thought to be
the length of the pool or riffle. It was also observed that the repeat- reasonably accurate estimates of the total and near-bed turbulent
ing bedforms tend to push the velocity and shear stress distributions stress, show consistent peaks in the pool. The current series of ex-
to a new equilibrium condition that is different than the uniform periments thus builds on the insight from field studies (Clifford and
flow case, and the results over isolated bedforms must therefore French 1993; MacVicar and Roy 2007; Thompson and Wohl 2009)
be interpreted with some care. The behavior of sediment particles and confirmed in lab experiments (Thompson 2006; MacVicar and
over the bedforms was not tested in this study, but it is the authors’ Rennie 2012; MacVicar and Best 2013) that near-bed turbulence
hypothesis that the local peaks in turbulent and total stresses in tends to peak in the pool. The boundary conditions method (u4 )
CDF and the hydraulic length scales associated with flow recovery was added in the current study to demonstrate that the total shear
regardless of bedform length have implications for riffle-pool dy- stress exhibits similar behavior. The key conclusion is that the spa-
namics and the design of constructed form. The authors discuss the tial patterns of shear stress vary with the type of stress (skin friction,
limitations of the experimental design, the hydrodynamic length turbulent stress, or total stress), with the implication that different
scales including lateral flow concentration, the difficulty of mod- types of stress may need to be considered for a full explanation
eling shear stress over the bedforms, and the implications of this bedform maintenance.
variability for sediment transport processes and bedform stability. Length scales of flow recovery also suggest that total and tur-
The limitations of the experimental design primarily relate to the bulent stresses are important metrics for describing shear stress dis-
measurement technology. Near-bed measurements with the UDVPs tribution over bedforms. To better describe the length scales of flow
can be prone to error due to positioning and signal interference recovery, normalized values of key metrics at the end of the pool
from the bed. The apparatus was designed to allow the rapid col- and riffle sections for the different runs are shown in Fig. 6. Error
lection of a large set of time series over multiple configurations of bars are shown where multiple runs were made for a given riffle or
the bedforms, but it was still difficult to ensure 100% accuracy. pool length. Many parameters exhibit asymptotic trends with bed-
Considerable effort was made to assess data quality and eliminate form length so that they tend to stabilize where bedform length (X)
poor data in a transparent and repeatable manner and this effort is approximately 3–4 times the channel width (Y). For instance,
led to the creation of an open-source software for this purpose u1 remains low through short pools but recovers in long pools,
(MacVicar et al. 2014). The shift of the log-law intercept [Eq. (5)] with the value at X=Y ¼ 2.7 within the error bars of the multiple
as a result of the nonuniform flow was not anticipated because runs at X=Y ¼ 4.0 [Fig. 6(a)]. In contrast, u1 is not sensitive to riffle
previous work in open channels had found a near-constant value qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(Kironoto and Graf 1995; Song and Chiew 2001; Onitsuka et al. length. −u 0 w 0 z=Z¼0.1 and u4 tend to increase in longer riffles and
2009). The variation in the log-law intercept may be related to the decrease in longer pools, again stabilizing between X=Y ¼ 2.7
range of pressure gradients. In the current study, the range of the and 4.0 in the pools [Fig. 6(b and c)]. Outer zone parameters Π
velocity defect (Fig. 4) was much greater than in previous nonuni- and Ψ tend to recover toward uniform flow values, although recov-
form open-channel flow studies by Kironoto and Graf (1995) ery is not complete for the longest two bedform runs, and Ψ is
(Π ¼ −0.30 to 0.41), Song and Chiew (2001) (Π ¼ −0.11 to insensitive to the length of the pool. The gradual increase of tur-
0.19), and Onitsuka et al. (2009) (Π ¼ 0.2 to 1.3). The idea that bulence and the velocity defect over riffles is similar to observations
stronger pressure gradients cause a shift in the intercept is sup- of the boundary layer development over broad-crested weirs (Hager
ported by some studies in wind tunnels such as Piomelli et al.’s and Schwalt 1994), whereas the dissipation of turbulence in pools
(2000) work comparing weak and strong accelerations in the range is similar to the qualitative observations of Yalin and da Silva
of H ¼ 1.25 to 1.60. They noted an increase in the intercept and (2001). These latter researchers proposed that hydrodynamic recov-
relaminarization in the stronger case. Monty et al. (2011) and ery processes occur at length scales that are low multiples of the
Nagano et al. (1998) also noted decreases in the intercept in ad- channel width and that they are linked with macro-scale bedform
verse pressure gradients. Of the studies examined, however, only formation such as pools and riffles. Given the similarity of the

© ASCE 04015025-9 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Riffle length the pool, the shortened bedform does result in higher turbulent and
Pool length total stresses over the upstream portion of the riffle.
2.2 1.1 The reduction in effective width and recovery to a new equilib-
rium over the shortened bedforms is also interesting because it
2.1 1
u 1/u 1o provides a link with theories of pool maintenance that rely on a
2 0.9 width constriction (Thompson et al. 1999) or topographically in-
1.9 0.8 duced lateral flow convergence (MacWilliams et al. 2006; White
(a) 1.8 0.7 et al. 2010; Milan 2013; Jackson et al. 2015). Topographical lateral
flow convergence is also used to guide flow into constructed pools
2.5 2.5 (Rosgen 2001; Newbury 2013) and is central to the design of some
))u’w’(- / u’w’-o

constructed riffle pools (Rhoads et al. 2008). MacVicar and Rennie


2 2
(2012) showed that a hydraulic lateral flow convergence will occur
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

1.5 1.5 in straight pools as a by-product of CDF due to increased drag and
turbulence generation near the channel sidewalls, and that the rate
(b) 1
1 of this convergence is sensitive to the overall channel width. Based
on the results from the fourth pool of run R40P40 it is likely that
4 5
any constriction of the width in the pool would locally increase u1 ,
3 4
u 4/ u 4 o

which supports recent modeling results by Jackson et al. (2015).


2 3 The results thus allow the conjecture that there is a positive feed-
1 2 back between local channel width constrictions and the rates of hy-
(c) 1 drodynamic recovery, but such an experiment is beyond the scope
0
of the current work.
3 0 The combination of the location of the peak in turbulent stress,
the similarity of the observed hydrodynamic scales to geophysical
2 −2
o

scales, and the persistence of the scales despite feedback from the
/

−4
bed offers a convenient explanation for pool scour and predictions
1
of pool dimensions. This explanation is seen as relatively simple
(d) 0 −6 because the above characteristics are evident with a simple 2D bed-
form geometry at a single discharge. Clearly, however, the conven-
1.1 1.2 ience is not sufficient to prove that turbulence is involved with
bedform genesis and maintenance. Including turbulence in explan-
o

1 1.1 ations of alluvial and constructed pools and riffle sedimentology is


/

not straightforward because of an incomplete understanding of how


turbulence affects sediment entrainment, transport, and deposition.
(e) 0.9 1
0 1 2 3 4 5 At a minimum it is known that turbulence will modify sediment
X/Y entrainment (Jackson 1976; Sumer et al. 2003). Recent research
has more accurately quantified the link using a physics-based force
Fig. 6. Normalized flow and turbulence distribution metrics at the impulse approach (Celik et al. 2013; Diplas et al. 2008). The ap-
downstream limit of the riffle and pool versus the normalized length proaches for calculating force impulse from turbulent velocity sig-
of the riffle (X s ) and pool (X d ) sections; all metrics are normalized nals, however, still rely on drag calculations (Celik et al. 2013),
by their respective values upstream of the bedform (Table 1) whereas the generation of turbulent structures above the bed during
CDF is likely to affect the lift force on particles. Nevertheless, the
results strongly suggest that Shields (1936) mean-stress criterion
hydrodynamic scales observed for the longer isolated bedforms in will not accurately represent entrainment thresholds in a decelerat-
the study by MacVicar and Best (2013) and the shorter bedforms in ing flow. Based on the occurrence of intermittent but high magni-
the current study, it appears that recovery cannot be forced by plac- tude peaks in turbulent stresses, it can be inferred that a mean stress
ing bedforms closer together. This conclusion solidifies the link criterion will underestimate sediment entrainment in the riffle and
between bedform formation and hydrodynamic length scales be- upstream portions of pools. Sediment in motion is likely to deposit
cause it means that the hydrodynamic recovery is persistent despite in the areas of pools where the turbulence levels are insufficient to
potentially negative feedback from the bed. maintain sediment transport. Such deposition would occur in the
The run with short bedforms in series is interesting because tail and sides of the pools rather than the head, as is observed in
it illustrates that equilibrium states other than uniform flow are natural pools (Lisle and Hilton 1999).
possible. Because hydrodynamic recovery is not accelerated by A consideration of flow recovery can be used to hypothesize
the bedforms, hydrodynamic recovery is incomplete over short control mechanisms on the length of riffles and pools. It may be
bedforms in series in the sense that flow passed to downstream bed- possible to construct relatively short pools and riffles, for instance,
forms is strongly conditioned by those upstream. A type of equi- but the result would be higher flow through the middle of the chan-
librium is nevertheless achieved over a series of short bedforms, nel and higher turbulence, particularly in the riffle section. Given
which is characterized by a more laterally concentrated distribution the sensitivity of sediment transport to turbulence, adding turbu-
of flow, particularly in the pool (Fig. 5). The increase in Ψ was lence would increase the risk of failure. In riffles, coarse particles
interpreted as a reduction in the effective width of the channel tend to collect and imbricate (Clifford 1993; Sear 1996). Imbricate
due to the relatively low velocity near the channel sidewalls. This structures are more likely to be stable in an environment of unidi-
reduction in effective width results in higher skin friction in the rectional force vectors, much like stones in an arch can maintain
pool, relatively lower turbulent and total stress peaks, and a lower their stability indefinitely, provided that the force is constant and
velocity defect in the pool. Despite lower turbulent stress peaks in unidirectional. Where turbulence is higher, the particles will be

© ASCE 04015025-10 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


subjected to greater degree of unsteady and three dimensional Editor and three anonymous reviewers substantially improved an
stresses. Long riffles would result in increased turbulence and a earlier version of this manuscript. The experimental apparatus was
more laterally concentrated flow as the boundary layer develops. assembled while the first author was the holder of an NSERC Post
At some threshold, the combined effect of the mean and the shaking Doctoral Fellowship. The experiments were conducted with the
forces would break up stabilizing structures, much like tremors can support of an NSERC Discovery Grant.
bring down an otherwise stable stone arch. In the case of the river
bed, such a breakup would result in local scour and initiate positive
feedback between increased turbulence as a result of CDF and con- References
tinued pool scour. In pools, it follows that if the combination of
mean and turbulent stresses are sufficient to move sediment, then Afzalimehr, H., and Anctil, F. (2000). “Accelerating shear velocity in
the pool will enlarge. In longer pools, the decelerating flow and gravel-bed channels.” Hydrol. Sci. J., 45(1), 113–124.
weakening turbulence will eventually lose the capacity to transport Afzalimehr, H., and Rennie, C. D. (2009). “Determination of bed shear
stress in gravel-bed rivers using boundary-layer parameters.” Hydrol.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

sediment, resulting in deposition as has been hypothesized by Yalin


and da Silva (2001). When the sediment deposits, convective flow Sci. J., 54(1), 147–159.
acceleration will occur, accompanied by a near-bed increase in Bernhardt, E. S., et al. (2005). “Synthesizing US river restoration efforts.”
Science, 308(5722), 636–637.
velocity, and initiate the formation of a downstream riffle. Con-
Best, J., and Kostaschuk, R. (2002). “An experimental study of turbulent
structed long pools would thus be subjected to significant deposi-
flow over a low-angle dune.” J. Geophys. Res. C: Oceans, 107(C9),
tion, again possibly leading to failure. 18-1–18-19.
Best, J. L., Kostaschuk, R. A., and Villard, P. V. (2001). “Quantitative
visualization of flow fields associated with alluvial sand dunes: Results
Conclusions from the laboratory and field using ultrasonic and acoustic Doppler
anemometry.” J. Visual., 4(4), 373–381.
The aim of the current study is to investigate the hydrodynamics Biron, P. M., Robson, C., Lapointe, M. F., and Gaskin, S. J. (2004). “Com-
of straight riffle pools as a means improving the confidence in paring different methods of bed shear stress estimates in simple and
their design parameters. Specific conclusions are complex flow fields.” Earth Surf. Proc. Land., 29(11), 1403–1415.
1. Skin friction, turbulent stresses, and total stress diverge over Brown, R. A., and Pasternack, G. B. (2009). “Comparison of methods for
macro-scale bedforms such that total and turbulent stresses ex- analysing salmon habitat rehabilitation designs for regulated rivers.”
ceed skin friction in the head and through the middle of a pool; River Res. Appl., 25(6), 745–772.
2. Hydrodynamic recovery of shear stress and the velocity defect Caamano, D., Goodwin, P., Buffington, J. M., Liou, J. C. P., and Daley-
in the riffle and pool occurs at length scales of a low multiple Laursen, S. (2009). “Unifying criterion for the velocity reversal hypoth-
of channel width, regardless of constructed bedform length; esis in gravel-bed rivers.” J. Hydraul. Eng., 10.1061/(ASCE)0733-9429
and (2009)135:1(66), 66–70.
3. Length scales of hydrodynamic recovery seem to match with Carling, P. A., and Orr, H. G. (2000). “Morphology of riffle-pool sequences
geophysical scales of riffles and pools. in the River Severn, England.” Earth Surf. Proc. Land., 25(4), 369–384.
Implications for the design of straight riffle-pool instream struc- Celik, A. O., Diplas, P., and Dancey, C. L. (2013). “Instantaneous turbulent
forces and impulse on a rough bed: Implications for initiation of bed
tures are that
material movement.” Water Resour. Res., 49(4), 2213–2227.
1. Skin friction and the Shields criterion may not be suitable for the
Church, M., and Jones, D. (1982). “Channel bars in gravel-bed rivers.”
prediction of sediment entrainment thresholds in the head and Gravel-Bed Rivers, Wiley, Chicester, U.K., 291–338.
through the middle of the pool; Clifford, N. J. (1993). “Differential bed sedimentology and the maintenance
2. Construction of short bedforms may be prone to failure because of riffle-pool sequences.” Catena, 20(5), 447–468.
they allow insufficient space for flow recovery, resulting in a Clifford, N. J., and French, J. R. (1993). “Monitoring and analysis of
more turbulent and laterally concentrated flow exported to turbulence in geophysical boundaries: Some analytical and conceptual
downstream bedforms; and issues.” Turbulence: Perspectives on flow and sediment transport,
3. Construction of long bedforms may be prone to failure, also Wiley, Chicester, U.K., 93–119.
because of flow recovery as boundary layer growth, increasing Coles, D. (1956). “The law of the wake in the turbulent boundary layer.”
turbulence and lateral flow concentration will encourage ero- J. Fluid Mech., 1(02), 191–226.
sion in the riffle whereas turbulent dissipation will encourage de Almeida, G. A. M., and Rodríguez, J. F. (2011). “Understanding
deposition in the pool. pool-riffle dynamics through continuous morphological simulations.”
Future work should more accurately assess the effect of accel- Water Resour. Res., 47(1), W01502.
eration on the near-bed velocity profile and include mobile bed Diplas, P., Dancey, C. L., Celik, A. O., Valyrakis, M., Greer, K., and
sediment in a range of size classes to understand how sediment Akar, T. (2008). “The role of impulse on the initiation of particle
movement under turbulent flow conditions.” Science, 322(5902),
entrainment and transport respond to the highly turbulent environ-
717–720.
ment in pools and illustrate feedback mechanisms between the bed Dwivedi, A., Melville, B., and Shamseldin, A. Y. (2010). “Hydrodynamic
and the hydrodynamics. forces generated on a spherical sediment particle during entrainment.”
J. Hydraul. Eng., 10.1061/(ASCE)HY.1943-7900.0000247, 756–769.
Einstein, H. A., and Barbarossa, N. L. (1952). “River channel roughness.”
Acknowledgments Trans. Am. Soc. Civ. Eng., 117(1), 1121–1132.
Goring, D. G., and Nikora, V. I. (2002). “Despiking acoustic Doppler ve-
The authors would like to thank Jim Best, Gianluca Blois, and locimeter data.” J. Hydraul. Eng., 10.1061/(ASCE)0733-9429(2002)
Marcelo Garcia at the University of Illinois in Urbana-Champaign 128:1(117), 117–126.
for hosting Lana Obach during her time there, allowing access to Hager, W. H., and Schwalt, M. (1994). “Broad-crested weir.” J. Irrig.
the Ven Te Chow Hydrosystems Laboratory, and for providing in- Drain. Eng., 10.1061/(ASCE)0733-9437(1994)120:1(13), 13–26.
put and very useful suggestions over the duration of this project. Jackson, J. R., Pasternack, G. B., and Wheaton, J. M. (2015). “Virtual
We appreciate the spirit of collaboration in which this work was manipulation of topography to test potential pool-riffle maintenance
completed. The insight and suggestions of the Editor, Associate mechanisms.” Geomorphology, 228, 617–627.

© ASCE 04015025-11 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Jackson, R. G. (1976). “Sedimentological and fluid-dynamic implications Miller, J. R., and Kochel, R. C. (2009). “Assessment of channel dy-
of the turbulent bursting phenomenon in geophysical flows.” J. Fluid namics, in-stream structures and post-project channel adjustments
Mech., 77(3), 531–560. in North Carolina and its implications to effective stream restoration.”
Keller, E. A. (1971). “Areal sorting of bed material: The hypothesis of Environ. Earth Sci., 59(8), 1681–1692.
velocity reversal.” Geol. Soc. Am. Bull., 82(3), 753–756. Montgomery, D. R., and Buffington, J. M. (1997). “Channel reach mor-
Keller, E. A., and Melhorn, W. N. (1978). “Rhythmic spacing and origin of phology in mountain drainage basins.” Geol. Soc. Am. Bull., 109(5),
pools and riffles.” Geol. Soc. Am. Bull., 89(5), 723–730. 596–611.
Kironoto, B. A., and Graf, W. H. (1994). “Turbulence characteristics in Montgomery, D. R., Buffington, J. M., Smith, R. D., Schmidt, K. M., and
rough uniform open-channel flow.” Proc. Inst. Civ. Eng. Wat. Marit. Pess, G. (1995). “Pool spacing in forest channels.” Water Resour. Res.,
Energy, 106(4), 333–344. 31(4), 1097–1105.
Kironoto, B. A., and Graf, W. H. (1995). “Turbulence characteristics in Monty, J. P., Harun, Z., and Marusic, I. (2011). “A parametric study of
rough non-uniform open-channel flow.” Proc. Inst. Civ. Eng. Wat. adverse pressure gradient turbulent boundary layers.” Int. J. Heat Fluid
Marit. Energy, 112(4), 336–348. Flow, 32(3), 575–585.
Nagano, Y., Tsuji, T., and Houra, T. (1998). “Structure of turbulent boun-
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

Kondolf, G. M., Anderson, S., Lave, R., Pagano, L., Merenlender, A., and
Bernhardt, E. S. (2007). “Two decades of river restoration in California: dary layer subjected to adverse pressure gradient.” Int. J. Heat Fluid
What can we learn?” Rest. Ecol., 15(3), 516–523. Flow, 19(5), 563–572.
Krogstad, P. Å., and Skåre, P. E. (1995). “Influence of a strong adverse Newbury, R. W. (2013). “Designing fish-passable riffles as gradient
pressure gradient on the turbulent structure in a boundary layer.” controls in Canadian streams.” Can. Water Resour. J., 38(3),
Phys. Fluids, 7(8), 2014–2024. 232–250.
Lave, R. (2008). “The Rosgen wars and the shifting political economy of Nezu, I., and Rodi, W. (1986). “Open-channel flow measurements with a
expertise.” Ph.D. thesis, UC Berkeley, Berkeley, CA, 257. laser doppler anemometer.” J. Hydraul. Eng., 10.1061/(ASCE)0733
Lee, J. H., and Sung, H. J. (2009). “Structures in turbulent boundary layers -9429(1986)112:5(335), 335–355.
subjected to adverse pressure gradients.” J. Fluid Mech., 639, 101–131. Onitsuka, K., Akiyama, J., and Matsuoka, S. (2009). “Prediction of
Leopold, L. B., and Wolman, M. G. (1957). “River channel patterns: velocity profiles and Reynolds stress distributions in turbulent open-
Braided, meandering, and straight.” Professional Paper 282-B, U.S. channel flows with adverse pressure gradient.” J. Hydraul. Res.,
Geological Survey, Washington, DC, 85. 47(1), 58–65.
Lhermitte, R., and Lemmin, U. (1994). “Open-channel flow and turbulence Papanicolaou, A. N., Kramer, C. M., Tsakiris, A. G., Stoesser, T.,
measurement by high-resolution Doppler sonar.” J. Atmos. Ocean. Bomminayuni, S., and Chen, Z. (2012). “Effects of a fully submerged
boulder within a boulder array on the mean and turbulent flow
Technol., 11(5), 1295–1308.
fields: Implications to bedload transport.” Acta Geophys., 60(6),
Lisle, T. E., and Hilton, S. (1999). “Fine bed material in pools of natural
1502–1546.
gravel bed channels.” Water Resour. Res., 35(4), 1291–1304.
Pasternack, G. B., Bounrisavong, M. K., and Parikh, K. K. (2008).
Lisle, T. E., Nelson, J. M., Pitlick, J., Madej, M. A., and Barkett, B. L.
“Backwater control on riffle-pool hydraulics, fish habitat quality, and
(2000). “Variability of bed mobility in natural, gravel-bed channels
sediment transport regime in gravel-bed rivers.” J. Hydrol., 357(1–2),
and adjustments to sediment load at local and reach scales.” Water
125–139.
Resour. Res., 36(12), 3743–3755.
Piomelli, U., Balaras, E., and Pascarelli, A. (2000). “Turbulent structures in
MacVicar, B. (2013). “Local head loss coefficients of riffle pools in gravel-
accelerating boundary layers.” J. Turb., 1, X1–16.
bed rivers.” J. Hydraul. Eng., 10.1061/(ASCE)HY.1943-7900.0000787,
Radspinner, R. R., Diplas, P., Lightbody, A. F., and Sotiropoulos, F. (2010).
1193–1198.
“River training and ecological enhancement potential using in-
MacVicar, B. J., and Best, J. (2013). “A flume experiment on the effect of stream structures.” J. Hydraul. Eng., 10.1061/(ASCE)HY.1943-7900
channel width on the perturbation and recovery of flow in straight pools .0000260, 967–980.
and riffles with smooth boundaries.” J. Geophys. Res.-Earth Surf.,
Rhoads, B. L., Garcia, M. H., Rodriguez, J., Bombardelli, F., Abad, J. D.,
118(3), 1850–1863. and Daniels, M. D. (2008). “Methods for evaluating the geomorpho-
MacVicar, B. J., Dilling, S., and Lacey, R. W. J. (2014). “Multi-instrument logical performance of naturalized rivers: Examples from the Chicago
turbulence toolbox (MITT): Open-source MATLAB algorithms for the metropolitan area.” River restoration: Managing the uncertainty in
analysis of turbulent flow time series.” Comput. Geosci., 73, 88–98. restoring physical habitat, Wiley, Chicester, U.K., 209–228.
MacVicar, B. J., Obach, L., and Best, J. L. (2013). “Large-scale turbulent Rosgen, D. L. (2001). “The cross-vane, W-weir and J-hook vane structures:
flow structures in alluvial pools.” Coherent flow structures in geophysi- Their description, design, and application for stream stabilization and
cal flows at earth’s surface, Wiley, Chichester, U.K., 243–259. river restoration.” Proc., Wetlands Engineering and River Restoration
MacVicar, B. J., and Rennie, C. D. (2012). “Flow and turbulence redistrib- 2001, ASCE, Reston, VA, 22.
ution in a straight artificial pool.” Water Resour. Res., 48(2), W02503. Roy, A. G., Biron, P. M., and Lapointe, M. F. (1997). “Implications of
MacVicar, B. J., and Roy, A. G. (2007). “Hydrodynamics of a forced riffle low-pass filtering on power spectra and autocorrelation functions of
pool in a gravel bed river. 1: Mean velocity and turbulence intensity.” turbulent velocity signals.” Math. Geol., 29(5), 653–668.
Water Resour. Res., 43(12), W12401. Sear, D. A. (1996). “Sediment transport processes in pool-riffle sequences.”
MacVicar, B. J., and Roy, A. G. (2011). “Sediment mobility in a forced Earth Surf. Proc. Land., 21(3), 241–262.
riffle-pool.” Geomorphology, 125(3), 445–456. Shields, A. (1936). “Application of similarity principles and turbulence re-
MacWilliams, M. L., Jr., Tompkins, M. R., Street, R. L., Kondolf, G. M., search to bed-load movement.” Mitteilungen der Preuss Versuchsanst
and Kitanidis, P. K. (2010). “Assessment of the effectiveness of a con- für Wasserbau und Schiffbau, 26 (in German).
structed compound channel river restoration project on an incised Simpson, R. L. (1981). “Review—A review of some phenomena in turbu-
stream.” J. Hydraul. Eng., 10.1061/(ASCE)HY.1943-7900.0000196, lent flow separation.” J. Fluids Eng., 103(4), 520–533.
1042–1052. Smits, A. J., and Wood, D. H. (1985). “The response of turbulent boun-
MacWilliams, M. L., Jr., Wheaton, J. M., Pasternack, G. B., Street, R. L., dary layers to sudden perturbations.” Ann. Rev. Fluid Mech., 17(1),
and Kitanidis, P. K. (2006). “Flow convergence routing hypothesis for 321–358.
pool-riffle maintenance in alluvial rivers.” Water Resour. Res., 42, Song, T., and Chiew, Y. M. (2001). “Turbulence measurement in non-
W10427. uniform open-channel flow using acoustic Doppler velocimeter
MATLAB [Computer software]. Natick, MA, MathWorks. (ADV).” J. Eng. Mech., 10.1061/(ASCE)0733-9399(2001)127:3(219),
Milan, D. J. (2013). “Sediment routing hypothesis for pool-riffle mainte- 219–232.
nance.” Earth Surf. Proc. Land., 38(14), 1623–1641. Sumer, B. M., Chua, L. H. C., Cheng, N. S., and Fredsoe, J. (2003).
Millar, R. G. (1999). “Grain and form resistance in gravel-bed rivers.” “Influence of turbulence on bed load sediment transport.” J. Hydraul.
J. Hydraul. Res., 37(3), 303–312. Eng., 10.1061/(ASCE)0733-9429(2003)129:8(585), 585–596.

© ASCE 04015025-12 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025


Thompson, D. M. (2006). “The role of vortex shedding in the scour of White, F. M. (2009). Fluid mechanics, McGraw-Hill, New York.
pools.” Adv. Water Res., 29(2), 121–129. White, J. Q., Pasternack, G. B., and Moir, H. J. (2010). “Valley width
Thompson, D. M., and Wohl, E. E. (2009). “The linkage between velocity variation influences riffle-pool location and persistence on a rapidly
patterns and sediment entrainment in a forced-pool and riffle unit.” incising gravel-bed river.” Geomorphology, 121(3–4), 206–221.
Earth Surf. Proc. Land., 34(2), 177–192. Whiting, P. J. (1997). “The effect of stage on flow and components of the
Thompson, D. M., Wohl, E. E., and Jarrett, R. D. (1999). “Velocity rever- local force balance.” Earth Surf. Proc. Land., 22(6), 517–530.
sals and sediment sorting in pools and riffles controlled by channel Wyrick, J. R., and Pasternack, G. B. (2014). “Geospatial organization
constrictions.” Geomorphology, 27(3–4), 229–241. of fluvial landforms in a gravel-cobble river: Beyond the riffle-pool
Vermeulen, B., Sassi, M. G., and Hoitink, A. J. F. (2014). “Improved flow couplet.” Geomorphology, 213, 48–65.
velocity estimates from moving-boat ADCP measurements.” Water Yalin, M. S., and da Silva, A. M. F. (2001). Fluvial processes, International
Resour. Res., 50(5), 4186–4196. Association of Hydraulic Engineering and Research (IAHR), Delft,
Wheaton, J. M., et al. (2010). “Linking geomorphic changes to Netherlands.
salmonid habitat at a scale relevant to fish.” River Res. Appl., 26(4), Yang, S. Q., and Chow, A. T. (2008). “Turbulence structures in non-
469–486. uniform flows.” Adv. Water Resour., 31(10), 1344–1351.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 10/11/15. Copyright ASCE. For personal use only; all rights reserved.

© ASCE 04015025-13 J. Hydraul. Eng.

J. Hydraul. Eng., 04015025

You might also like