Vous êtes sur la page 1sur 19

Journal of Hydraulic Research Vol. 46, No. 5 (2008), pp.

579597
doi:10.3826/jhr.2008.2986
2008 International Association of Hydraulic Engineering and Research

Research Papers

A new integrated, hydro-mechanical model applied to flexible


vegetation in riverbeds
Un nouveau modle hydromcanique intgr appliqu la vgtation
flexible dans les lits de rivires
D. VELASCO, Hydraulic, Maritime and Environmental Engineering Department, Technical University of Catalonia (GITS-UPC),
ETSCCPB, Av/ Gran Capitan s/n, 08034 Barcelona, Spain. Tel.: +34-93-4017013; fax: +34-93-4054194;
e-mail: david.velasco@upc.edu
A. BATEMAN, (IAHR Member), Hydraulic, Maritime and Environmental Engineering Department, Technical University of
Catalonia (GITS-UPC), ETSCCPB, Av/ Gran Capitan s/n, 08034 Barcelona, Spain. Tel.: +34-93-4017064; fax: +34-93-4017064;
e-mail: allen.bateman@upc.edu

V. MEDINA, Hydraulic, Maritime and Environmental Engineering Department, Technical University of Catalonia (GITS-UPC),
ETSCCPB, Av/ Gran Capitan s/n, 08034 Barcelona, Spain. Tel.: +34-93-4017064; fax: +34-93-4017064;
e-mail: vicente.medina@gits.ws

ABSTRACT
This paper suggests a simple new numerical scheme for calculating the vertical velocity profile, turbulent shear stress distribution, and canopy deflection
for flow in vegetated channels. The scheme is derived from the simplified, steady Reynolds equation (momentum balance) and its vertical integration.
This study includes an experiment that compares and calibrates the numerical model to flume data. Fourteen runs were performed in a laboratory
flume, of which nine involved natural grass (cultivated barley) and five involved plastic plants (PVC). Turbulent diffusion coefficient, mixing length,
and a resistance equation (drag coefficient Cd vs. Reynolds number) were the input parameters for the numerical model. Parameters with physical
meaning were calibrated. This scheme offers a low computation time and good estimation of plant deflection and velocity and stress profiles.

RSUM
Cet article suggre un nouveau schma numrique simple pour calculer le profil vertical de vitesse, la distribution de leffort de cisaillement turbulent,
et la dflexion par la couverture vgtale dans les coulements en canaux garnis de vgtation. Le schma est driv de lquation simplifie et
stationnaire de Reynolds (quilibre des quantits de mouvement) et de son intgration verticale. Cette tude inclut une exprience qui compare et
calibre le modle numrique aux donnes du canal. Quatorze essais ont t excuts dans un canal de laboratoire, dont neuf ont impliqu de lherbe
naturelle (orge cultive) et cinq des plants en plastique (PVC). Le coefficient de diffusion turbulente, la longueur de mlange, et une quation de
rsistance (coefficient de trane Cd en fonction du nombre de Reynolds) taient les paramtres dentre pour le modle numrique. Les paramtres
avec une signification physique ont t calibrs. Ce schma prsente un faible temps de calcul et une bonne valuation des profils de dflexion des
plants, des vitesses et des efforts.

Keywords: Drag forces, numerical modelling, riverine vegetationl, turbulent stresses

1 Introduction and natural canopy in their laboratories to obtain empirical rela-


tionships. The complexity of physical model studies and the
When we study fluvial hydrosystems, we must not focus only popularization of numerical schemes in hydraulics have led to
on the various isolated elements involved in the process (water, the development of applications focused on flow through veg-
sediment, vegetation, and fauna). Rather, we must deepen our etation. Three different lines of numerical research have been
knowledge of their common relationships. We aim to study the followed. First, well-known 3D turbulent schemes have been
effect of flexible vegetation on flow. The presence of flexible applied to special geometric boundaries to simulate obstruction
canopy (grass, bushes, and reeds) in the riverbed and along due to vegetation. Lpez and Garca (2001) and Fischer-Antze
the banks changes flow resistance and, in consequence, river et al. (2001) adjusted the two-equation closure model to flow
hydrodynamics. Many authors (Kouwen et al., 1969, 1973, 1980; through vertical rigid cylinders. These simulations are quite well
Petryk et al., 1975; Nepf, 1999) have experimented with plastic adjusted. Cui and Neary (2002) compared the numerical results

Revision received June 27, 2007/Open for discussion until April 30, 2009.

579
580 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 1 Experimental set-up of sandy flume and images of tested plants

of Reynolds-averaged NavierStokes (RANS) and large eddy and inside the vegetation which takes into account not the
simulation (LES) in order to reproduce the particular anisotropy local, but the vertical integrated shear stress balance. A set of
generated in flow through rigid, submerged cylinders. Stoesser integral equations has been proposed to describe the momen-
et al. (2006) also applied large eddy simulation (LES) scheme tum conservation at different depths of the flow. The mechanical
to simulate the non-uniform turbulent shear stress distribution stem-deflection subroutine is able to reproduce large deforma-
with submerged canopy. Finally, Choi and Kang (2004) used tions. Also, a new concept is introduced in this paper that
Reynolds stress modeling (RSM) to improve mean and turbulent clearly improves the previous models, and this is a variable resis-
field results. High-accuracy numerical schemes will be developed tance law. Drag coefficients Cd are totally dependent of obstacle
in the future, but the high computational cost is still a disadvan- Reynolds Number Red and the model evaluates the vertical varia-
tage in terms of their practical application to large-scale hydraulic tions of Cd as a function of the turbulent regime around the stems.
problems. A second group of researchers (Klopstra and Barn- Finally, computational and resistance parameters are calibrated
eveld, 1997; Carollo and Ferro, 2002) has developed analytical and verified with our experimental data, which were measured in
expressions for adjusting velocity profiles U(z) in very specific a flume covered with flexible vegetation.
conditions, but the input parameters involved in the problem are
hard to estimate. The third line of research is very interesting 2 Experimental setup
because it considers the flexural properties of vegetation and cal-
culates the reduction in relative depth h/k (h water depth and Laboratory experiments were conducted in a rectangular flume,
k deflected plant height) due to stem bending. Kutija and Hong where steady, recirculating flow was generated. The flume is 8 m
(1996) developed this line of research based on very simple tur- long, 2.5 m wide, and 0.4 m tall (see Fig. 1). Channel bed consist
bulent models applied to the 1D Reynolds momentum equation. in a D50 = 1.6 mm sand bed. Bed slope was very low (0.04%),
An implicit discretisation of the unsteady Reynolds equation was and nearly uniform regime was finally established. Test series
used and solved to converge toward the steady solution. Two descriptions are presented in Table 1 and are next commented.
different turbulence closure models were combined: eddy vis- Series group T3. Plastic plants (PVC strips) were tested in
cosity equation and mixing length model. The vertical dominion sand flume. The shape of the plant is similar to a bush or a palm.
of application for every turbulent model was defined as an input. Average plant height was h = 15 cm, and frontal area Ao =
The estimation of the resulting deflected plant height k was solved 20.7 cm2 . Dimensions of plastic strips cross-section was 2
using a function that reproduces the elastic behavior of a vertical 0.4 mm2 . Staggered pattern was used and the spacing was set up
beam loaded by drag forces. Unfortunately, no data was avail- in a = 7 cm. The plastic bushes were fixed in the sand bed in a
able to verify the model with flexible vegetation, and the only length of 2.5 m at the central area of the flume. Water levels were
data used in the calibration were from Tsujimotos test with rigid controlled by a downstream movable gate and measured at 12
cylinders. Kutijas work is a well defined and very efficient code control sections along the flume. No strictly uniform regime was
that was applied later in a Quasi 3D hydraulic model (Erduran established, but the analysis of convective term (U. U/x) in the
and Kutija, 2000). runs shows that it is not dominant and then, it can be neglected
This paper tries to get deeper into this third line of research. in these particular conditions. Discharges of 32, 77, 90, 110, and
We develop a new explicit scheme to balance momentum above 136 l/s registered submergence ratios from h/k = 1 to 2.7.
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 581

Table 1 Resume of experimental tests

RUN a h M Ao Vegetation Q h/k Ao /a2


(m) (m) (Plants/m2 ) (cm2 ) model (m3 /s)

T3 0.07 0.15 205 20.7 Artificial PVC 0.1360.032 1.02.7 0.42


TN 0.006 0.09/0.125/ /0.18 22800 1.85/2.8/ /4.77 NaturalBarley 0.1360.026 1.02.89 4.110.6

Table 2 Mechanical properties of the tested plants. h = initial plant height, D = mean diameter of the stem,
B = mean width of leaf, e = thickness of the leaf, I = Inertial modulus of the cross-section of the stem/leaf,
E = stiffness modulus
RUN type of plant h D B e I E E.I
(m) (m) (m) (m) (m4 ) (N/m2 ) (N.m2 )

T3 0.15 0.0025 0.001 2.31E13 7.30E+07 1.69E05


TN 0.090.18 0.003 4.00E12 1.43E+07 5.72E05

Series group TN. The objective of testing with high density, collected at 25 Hz (temporal series of 3000 data) so the series
natural covers was accomplished in this series. Some previous are long enough to guarantee a correct estimation of second
test about hydroponic growth adaptation were done in the sandy order statistics (i.e., turbulent parameters). The ADV could be
substratum for different, common species of grass; the conclusion used without disturbance to the vegetation array, except for TN
was that barley shows the most efficient properties to resist high runs, were barley cover was too dense and stems could enter the
humidity and poor-nutrient soil conditions. After seeding, barley sampling volume. In consequence, some stems were removed
needs a short time of 912 days to grow in the zone of study with to create an empty region (15 8 cm2 ) to introduce the probe.
a very dense, uniform grass cover. An important factor is to get an Local flow accelerations are registered for h/k = 1, but their
elevated superficial uniformity of stems and leaves in the cover importance decrease for h/k > 1.
to avoid inhomogeneous flow conditions along the flume. At the High accuracy results of vertical velocity profiles U(z) over
end of runs TN, 5 control areas (0.4 0.4 m2 ) were detached barley covers (runs TN) are shown in Fig. 2. Measurements
from the cover and the number of stems were counted. An aver- agree with the existence of a piston flow near the bed and the
age density of M = 22500 stems/m2 was measured, so spacing development of a logarithmic law above the extreme of canopy.
a = 0.67 cm was estimated. Growing velocity of stems (2 cm per Also, turbulent shear stresses profiles xz (z) are obtained from the
day) controlled in some sense the test planning. Three different statistic analysis of velocity series. The peak of xz is registered
plant heights (h = 0.09, 0.125, and 0.18 m) were tested and 4 dis- in the extreme of the deflected plants (z = k) and an impor-
charges for each configuration. These stem lengths were chosen tant reduction of turbulent stresses is observed. The development
in order to obtain adequate mechanical properties, because of the of a shearless region is measured below the penetration depth
high sensibility to the maturity degree. Longer and older barley p (Nepf and Vivoni, 2000). Experimental Drag Coefficients Cd
stems presented low stiffness modulus and undesirable flexibil- within canopy are evaluated as Dunn and Lopez (1996), and
ity properties because prone plant conditions were adopted from commented inVelasco (2006). The vertical profile of Drag Coeffi-
the very low discharge condition. Individual frontal areas were cients Cd (z) was calculated using the expression of Eq. (1), where
Ao = 1.85, 2.8, and 4.8 cm2 , respectively and maximum sub- the vertical gradient of xz (Reynolds XZ turbulent stresses), the
mergence ratios of h/k = 2.9 were registered. Barley geometry mean plant width B, and the mean longitudinal velocity U were
presents a unique leaf with a very uniform width distribution. experimentally measured.
Stem geometry was measured and averaged at the moment, as   
well as stiffness modulus E. Deflected plant height k was mea- 2 .g.So + xz (z)
z
sured optically with a thin gauge from water surface. Waving Cd (z) = . . (1)
B(z).M.U 2 (z)
plants were observed in the test, so k measurements present an
estimated error of 1 cm. From the analysis of experimental data (Velasco, 2006) the
Mean (U, V, W ) and turbulent (u , v , w ) velocities corre- vertical distribution of two different turbulent length-scales was
sponding to the stream-wise (x), lateral (y), and vertical (z) calculated: mixing length (l) and Integral Scale (L), above and
directions, respectively, were measured using 3D side-looking within the vegetation (Fig. 3). The mixing length (l) was evaluated
acoustic doppler velocimetry (ADV). Once uniform conditions as the relationship between turbulent stresses xz and vertical
were established in the flume, detailed vertical profiles (z = gradients of velocity U (see Eq. (2)), as described in the mixing
2 mm) were recorded for series groups T3 and TN (see Table 2). length turbulent model.
Vertical profiles were measured in the centerline of the chan-
xz
nel, and just at the middle of the vegetated area, far enough l2 =  .
U  U 
(2)
from the boundary disturbance region. Two-minutes records were . z . z
582 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 2 Dimensionless profiles of velocity U with barley covers (runs TN)

Figure 3 Comparison between experimental profiles of mixing length (dots) and Integral scale (circles) corresponding to run TNh = 0.12 m

On the other hand, the Integral scale (L) was calculated taking steady-uniform conditions, driving forces are expressed in four
into account the frozen whirl Taylor hypothesis, which simpli- terms on the right side of the equation, which are gravity forces,
fies the relationship between the characteristic length-scale and viscous forces, Reynolds XZ turbulent forces and drag forces
time-scale in homogenous turbulence as: L = U.Te . The char- (fcd ) due to the presence of vegetation.
acteristic time-scale (Integral Scale Te ) is evaluated through the  
2 U(z) xz (z)
analysis of autocorrelation functions of the velocity fluctuation 0 = .g.So + .. + fcd (z) (4)
z2 z
u (t) time series (Eq. (3)).
  The next step is the vertical integration of the Reynolds x-equation
u (t).u (t + ) (Eq. (4)) between water surface level and the riverbed level. The
L = U. .d. (3)
0 u2 flow is separated into four different zones, as presented in Fig. 4.
The conclusion is that a linear tendency of l and L is registered The resulting equations are written as Eqs (5)(8):
above the penetration depth p and a constant value below this 0 = .g.So.(h z) xz (z)
point. Also, the values of l and L at z = p are similar. These  

G (z)
observations are general for all the analyzed runs. Discussion
about Integral Scale L in vegetated channels is also presented in ZONE 1. external region, h > z > k (5)
Velasco and Bateman (2003).
0 = .g.So.(h z) xz (z) cd (z)
ZONE 2. internal region, k z > p (6)
3 Governing equations
0 = .g.So.(h z) cd (z)
The numerical model is based on momentum conservation
equations. The Reynolds x-equation (Eq. (4)) shows that in ZONE 3. shearless region, p z > o (7)
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 583

U(z) The shear due to drag forces is defined as:


0 = .g.So.(h z) . cd (z)
z  k  k
1 B(z)
ZONE 4. viscous region, o z (8) cd (z) = fcd (z).dz = ..Cd (z). 2 .U 2 (z).dz (9)
z z 2 a
Two different boundary conditions must be set in the compu-
where is the water density, g is the gravity acceleration, So tational domain: (1) xz = 0 at the water surface (z = h), and
is the channel slope, h is the water depth, xz is the verti- (2) xz = 0 in the shearless region (0 < z < p). The applica-
cal Reynolds turbulent stress, cd is the drag stress (vegetative tion of the latter to the x component Reynolds equation (Eq. (4)),
stress), is the cinematic viscosity, U is the mean longitudinal and neglecting viscous forces in the shearless region, leads to a
velocity. very simple balance between gravity and drag forces, as shown
This set of equation is reformulated here in the integral form. in Eq. (10).
Reynolds equations in x, applied to vegetation, can be found in 1 B
Dunn and Lopez (1996) and Nepf and Vivoni (2000), but in the 0 = .g.So ..Cd . 2 .U 2 . (10)
2 a
present paper these equations are integrated in vertical direction If average values of drag coefficient (Cd,o ) are assumed and
z to obtain shear as a variable of the problem, and describe the using plant width (Bo ) for the entire shearless zone (z < p), then
proposed conceptual model. a constant value of velocity Uo is explicitly estimated as Eq. (11):
According to Fig. 4, Zone 1, the external region, corre-  1/2
sponds to heights z from h to k, above the canopy, and thus 2.g.So.a2
Uo = , (11)
has no drag forces. Integral Eq. (5) shows how the gravity shear Cd,o .Bo
term is uniquely balanced to turbulent shear stress xz . Maxi- where drag coefficient Cdo and plant width Bo are an estimation
mum turbulent stress is located, then, at z = k. Zone 2, the of mean values in zone 3. This velocity boundary condition is
internal region, is located below z = k. It is characterized by important because the general method is based on the hypothesis
the presence of canopy. The vertical integration of drag forces that the shearless region (Zone 3 in Fig. 4) is developed.
fcd from the top of canopy k to height z is the amount of shear In order to relate Reynolds turbulent stresses to velocity
stress that is absorbed by the plant (cd ). This vegetative stress gradients, we used Von Karmans turbulent mixing-length the-
is defined in Eq. (9), where Cd is the drag coefficient, B is the ory (Rouse, 1946). This 1D-equation turbulence closure model
plant width, and a is the stem spacing. From Eq. (6), we can reflects momentum diffusion in a very simple way and could be
deduce that gravity shear stress is balanced by both vegetative considered a low-accuracy method. However, for the purposes
and turbulent stress. Hence, turbulent stress xz is reduced at of this work, the results are quite good as a first approach to the
z < k and, if cd is equal to gravity shear G , then turbulent problem of turbulence. The turbulence mixing-length equation is
stress xz disappears (see Fig. 4). This means that the momentum expressed as follows:
 
absorption capability of the canopy is greater than the hydraulic  
2  U  U
gravity shear. Zone 3 is delimited by height z = p, which is xz = .l .   . , (12)
z z
called the penetration depth (Nepf, 1999). This zone is called the
where is the water density and l is the mixing length. Vari-
shearless region because the turbulent stress xz is negligible
ous theories on vertical profiles of mixing length l(z), with and
(Eq. (7)). In this region, there is no vertical momentum exchange,
without vegetation, are found in the literature. Kutija and Hong
but turbulence is completely developed; the flow is governed by
(1996) used a potential law for l(z) inside and outside vegeta-
drag forces that balance the gravity shear completely. Finally,
tion. Watanabe (1990) proposed an analytical model of l(z) for
Eq. (8) represents the viscous region, where viscous forces are
canopy where an integrated exponential function is defined for
predominant and velocity gradients are very high. Also, the lower
within canopy and a linear model l(z) = .(z d) for outside
obstacle Reynolds number Red is present in this layer (laminar
canopy, where is the Von Karman universal constant and d is the
regime).
zero-plane displacement. The discussion is complemented with
our own experimental work in canopies which is briefly described
in this paper. It was analyzed the vertical distribution of mixing
Z
length l(z) and Integral Scale L(z) from the experimental velocity
profiles (see Fig. 3). The integral scale L concept is analogous to
the mixing-length l of the Von Karman turbulence closure model
1
G = . g.So.(h-z)
h
in the sense that both concepts reflect a characteristic turbulent
scale (eddy scale). Despite the dispersion of data in the vertical
profiles, it was found a clear linear tendency of l(z) and L(z)
functions above z = p, and a constant value below this point.
2
xz Cd
Then, based in laboratory experiences, it has been proposed a
k
different distribution of mixing-length l, expressed in Eq. (13).
p
3
4 X l(z) = lo +  .(z p) for z > p, (13)
Figure 4 Shear balance scheme for high vegetative momentum absorp- where lo is the mixing length in the shearless zone and  is a
tion capability diffusion coefficient. This model differs from Watanabe because
584 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

the linear distribution of l(z) is considered within the canopy, the vertical gradients of xz above the top of the plants. In order
above the penetration point p. Numerical parameters lo (bed mix- to describe secondary currents effect (the fall of xz close to the
ing length) and  (diffusion coefficient) are unknown and must surface and the reduction of peak xz in the extreme of plant) a two
be calibrated. It is suggested in Velasco (2006) that the dimen- parameters conceptual model is developed. The first parameter
sion lo is related to plant width B and  is much lower that the represents the superficial portion of water depth where measured
Von-Karman constant = 0.41. xz is zero, and is the dimensionless parameter (ranges from
The observation of experiments with natural vegetation in 0 to 1) which modifies the slope of a linear distribution for xz
the flume, and the later analysis of measured data, showed above the extreme of plant. A two parameters (, ) model would
the importance of secondary currents and turbulent anisotropy fit the measured xz distribution as shown in Fig. 7.
in the momentum balance. Turbulent structures in partly veg- A zero-level approximation of this conceptual model is devel-
etated channels have been studied and characterized by Ikeda oped in this paper, and there was assumed a low value of , it
(1996), Naot (1996) and Nezu and Onitsuka (2001) with high- is 0. Under this simplification, a one parameter (a) linear
advanced measurement techniques (PIV, LDA) and they observed model is derived (see Fig. 7) to represent xz distribution above
that secondary currents are highly affected by density of vegeta- the extreme of plant, and it is implemented in the present numer-
tion and shear gradients at the boundary between vegetated and ical model. The equation which relates (, ) and (a) is deduced
non-vegetated zones. Some instability in the flow was detected at z = k by Eq. (14):
during our tests: cellular secondary currents created preferential

flow corridors above the plants. A steady, transversal pattern in = . 1 . (14)
hk
the deflection of plants was clearly observed (regular waves in
As a consequence, is lower than as a function of depth .
the bended vegetative cover). In Fig. 5, a picture of the waves
Obviously, both conceptual models are equivalent in the case of
above the vegetated cover is shown. The existence and mag-
0, and then = . The (, ) conceptual model will be
nitude of secondary currents were measured in terms of the
used in the future, instead of the ( ) linear model in order to
vorticity generation parameter (v2 w2 ). This term (understood
increase the accuracy of secondary currents effect in the general
as an anisotropy between transversal and vertical velocity mean
flow. In consequence, effective gravity stresses in Eqs (5) and
squared v2 and w2 ) is the driving force for the secondary cur-
(6) are modified by the parameter . (namely sc ) and the new,
rents as well as for the flow gradients and the mean velocities
equivalent expressions are Eqs (15) and (16).
(U, V, W), as documented by Nezu and Nakagawa (1993), but not
for the Primary Shear Stresses (xz , yz , yx ). In Fig. 5, the peaks 0 = sc .g.So.(h z) xz (z)
of (v2 w2 ) above and below the top of the plant are shown. ZONE 1. external region, h > z > k (15)
This vertical distribution denotes the presence of convective cells
within the flow. 0 = sc .g.So.(h z) xz (z) cd (z)
The consequence of secondary currents in the flow is the
ZONE 2. internal region, k z > p (16)
reduction of the measured xz values compared to those expected
from Eqs (5) and (6) (convective terms and secondary forces are Additionally, special care must be taken in the estimation of
neglected in the 1D conceptual model). Figure 6 shows the devia- drag coefficients Cd,j because they are closely related to the resis-
tion of the measured xz from the linear, theoretical profiles. Two tance to flow over vegetated covers. The value of Cd is very sensi-
effects are observed, the fall of the shear stress xz close to the tive to changes in the flow turbulence regime, which is governed
surface (xz 0 near the water surface), and the reduction of by the obstacle or stem Reynolds number Red (= U.B/),

Figure 5 Anisotropy of turbulence. (Left) Picture of the steady waves at the barley cover. (Right) Vertical distribution of (v2 w2 ) for Run
TN-h = 0.12 m. h/k = 1.84
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 585

Figure 6 Dimensionless vertical profiles of measured Reynolds xz-turbulent stress. The red line defines the theoretical vertical distribution

100
Z
Cd {Re(1), Cd(1)}


10
xz = .g.So.(h-z)
.So h
xz= g.So.(h--z) 1 {Re(i), Cd(i)}
.So
0.1

0.01 {Re(n), Cd(n)}


k

0.001
X 1 10 100
U .B
1000 10000

xz
Re d =

Figure 7 A two parameters (, ) conceptual model and a one parameter Figure 8 Proposed Cd(Red) law using n degrees of freedom
(a) model to fit the measured xz vertical distribution (dotted line)

degrees of freedom. The calibration process will fit the optimal


where is the water kinematic viscosity. Therefore, a resistance drag coefficients Cd .
function of the type Cd (Red ) is added to the model to take this Each canopy element is modeled as a vertical, isolated beam
important factor into account. The discussion here is what type or cantilever in order to simplify the structural problem. The
of function Cd (Red ) is suitable for canopy resistance. Unfortu- forces on the stem are imported from the hydrodynamic subrou-
nately, the classical resistance laws Cd (Red ) of 2D rigid bodies tine and are applied to every calculation node in order to obtain
that have been widely studied in the past are not applicable to an approximation y(z) of stem deformation. Classical, analytical
the heterogeneous shape of stems and leaves. New relationships expressions of deformation on beams are not valid here, because
must be researched. In order to further the fitting-optimization they are only derived for small deformation conditions. Natural
procedure, the estimation of the Cd (Red ) law was introduced canopy elements bend easily in river floods. Brushes and aquatic
as a linear approximation between a series of generic discrete species also bend easily in normal flows. The deformation of
points {Red (i), Cd (i)}, for i = 1, . . . , n, in a loglog graph. the top of the canopy is often of the same order as the deflected
Figure 8 shows an example of resistance law as a function of n plant height (y(h )
= k). In the simulation, large deformations
586 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

are required to adequately estimate the vertical deflected plant


height k. The classical beam elasticity equation (Timoshenko and
Young, 1975) is applied in the stem (z coordinate is adopted as the
vertical, longitudinal direction of the stem) and an explicit for-
ward finite-difference scheme is computed to solve the transversal
deformation profile y(z), as in Eq. (17):
2 y Mf
= , (17)
z2 E.I
where y is the transversal deformation, Mf is the moment (N.m),
E is the stiffness modulus (N/m2 ) and I is the cross area inertial
modulus (m4 ). In large deformation conditions the total stem
length h must be conserved, so a vertical load (drag forces)
re-distribution algorithm must be used to reach the convergence
(equilibrium) of the system.

4 Description of the numerical procedure


Figure 10 General flowchart
4.1 Computational domain
height ki+1 is incorporated at iteration i+1. The computation fin-
The vertical computational domain is uniformly discretizied with
ishes when a convergence in {ki } is a set (\ki+1 ki \ < tolk ). The
a z step, as shown in Fig. 9, and all of the variables are cal-
flow solution is then considered dynamically and mechanically
culated at these nodes. Water depth (h), undeflected plant height
balanced.
(h ) and deflected plant height (k) are denoted as ih, ihp, and
The input data consists of hydraulic variables: water depth h,
ik indexes, respectively. Finite-difference approximations (for-
energy slope So and the vegetations geomechanical properties
ward differences) are used in the explicit method when vertical
(undeflected plant height h , plant spacing a, vertical distribution
velocity U profiles are computed, and first-order integral meth-
of plant width B(z), thickness of stems e, stiffness modulus E,
ods (Simpsons rule) are applied in integral operations. Because
the resistance law {Re(j), Cd(j)} and three turbulence parameters
the scheme is explicit, the number of nodes must be considerable
[bed mixing-length lo , diffusion coefficient  and secondary cur-
(z/ h < 0.02).
rents factor sc ]). The results of the model (output data) are unit
discharge q, velocity U(z), Reynolds stresses xz (z), deflected
4.2 General iterative scheme plant height k, and stem deformation y(z).
The model can estimate the unit discharge q given an initial
This numerical model aims to create an iterative succession of water depth h and an energy slope So, which is equivalent to
well-balanced flow solutions that can generate the final stabi- solving the flow resistance equation for any kind of vegetation.
lization or convergence of the deflected plant height k. Figure 10 Friction factors (the DarcyWeisbach f or the Manning n) can
shows the computational flowchart of the model. At each iteration easily be deduced from the hydraulic results.
i, the deflected plant height ki is calculated and, as a consequence The general scheme is very simple in its conception, but the
of the geometric change, a new flow solution U i (z), xz i
(z) is most important and critical part is the creation of adequate mod-
obtained. At this point, the plant deformation is recalculated using ules to calculate velocity profiles and stem deformation. The
the aforementioned flow conditions and a new deflected plant hydrodynamic subroutine calculates flow profiles {U(z), xz (z)}
for a particular plant height k and the mechanical subroutine esti-
zih=h mates the deflected plant height k as a function of drag forces
F(z), which are obtained from the computed velocity profile
U(z). The next section briefly explains the numerical methods
zihp
used in these two subroutines.
h
zik 4.3 Description of the hydrodynamic subroutine
k
The numerical strategy used to estimate well-balanced velocity
z
profiles is based on a predictor-corrector technique for Reynolds
zi
stresses xz which converges into a well-balanced solution.
Figure 11 shows a flowchart and a schematic diagram with
predicted xz and corrected xz profiles. A minimization algo-
z1
rithm is applied to converge toward the optimal xz solution.
Figure 9 Vertical discretization scheme In the first minimization level (denoted by the supraindex i) an
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 587

The deformation profile yi (z) is calculated from Eq. (17). The


Moment law Mf (z) is obtained as the sum of moments due to
punctual loads (drag) at every computational node. The total
stem length is measured along the deformed stem to define the
deflected plant height ki . New nodes are defined according to
ki . A new distribution of forces F is used. Then, yi+1 (z) and
ki+1 are calculated. And an iterative scheme is established in
{yi (z), ki } to converge to the large-deformation deflected plant
height k. Figure 12 presents a diagram of the iterative proce-
dure. The convergence into the solution is fast enough (less than
15 iterations, in general). In calculating large deformations, the
additional bending due to second-order moments in the beam
must be taken into account. New moments Mf  appear in the
beam due to non-orthogonality between loads F and the lon-
Figure 11 Flowchart of the hydrodynamic subroutine gitudinal direction of the deformed beam(s). The second-order
moment Mf  is calculated from the load F (drag) as Eq. (19):
approximate solution for the deflected plant height ki is supposed.
Mf = F(z s. cos(2 )), (19)
Under the ki assumption, a second minimization level (denoted
by the supraindex j) is established where an initial condition where geometric angles , are defined in Fig. 13. The pro-
for the penetration depth pi,j is proposed and a linear distribu- cedure is as follows: (1) first-order moments Mf are used and
i,j
tion of turbulent stresses xz,pred (z) is predicted from z = pi,j a deformed profile y(z) is obtained; (2) secondary moments
to z = ki (as presented in Fig. 11). At the third minimization Mf  are calculated from derivatives y (z), vertical distances and
level, Eqs (12) and (13) are used to calculate explicitly the veloc-
ity profile U i,j (z) with a boundary condition U i,j (pi,j ) = Uo
i,j
defined by Eq. (11). Drag stresses cd are calculated using Eq. (6),
where the drag coefficients Cd are obtained from the resistance
law Cdi, = n (Red,i ) as a function of the calculated obstacle
i,j
Reynolds Number Red (showed in Fig. 8). At this point, Eqs (15)
and (16) are used to calculate the new, corrected turbulent shear
i,j
stress profile, which will be called xz,corr (z). The area which
i,j
exists between the predicted and corrected stress profiles, Areaxz
(see Fig. 11) is calculated by Eq. (18).
 k
 i,j 
=  i,j 
xz,corr (z) xz,pred (z) .dz.
i,j
Areaxz (18)
0
i,j
The value of Areaxz is a function of the penetration depth pi,j . Figure 12 Numerical discretization and iteration process for the
An iterative scheme is proposed to find the optimal value of pi,j mechanical subroutine (large deformations)
i,j
which minimizes function (p) = Areaxz , and, in consequence,
a well-balanced solution for xz = xz,corr (z) and U(z) profiles. h y i(h)
An iterative, first-order minimization method is applied to (p)
and the solution (penetration depth pi , velocity U i (z) and stresses
i
xz (z)) is sent to the main program. The algorithm of optimization
T
for (p) is a Quasi-Newton Method, where derivates d/dp are yi
numerically obtained by incremental quotient in p + p. In con- F zi
xi
sequence, the accuracy of the solution basically depends on the F
definition and smoothness of the numerical function (p). The N
analysis of this function shows that the search methods behavior i
and efficiency are sufficient.
si
4.4 Description of the mechanical subroutine
The mechanical subroutine consists of a numerical code that i
reproduces load-deformation processes in a stem. As commented
before, Eq. (17) is included as the elastic load-deformation equa-
tion, but an iterative process is created to reproduce the real, large Figure 13 Diagram of the normal and tangent components of the drag
deformations. force R to estimate secondary moment Mf 
588 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 14 Comparison of computational stem deformation with (solid line) and without (dashed line) the calculation of secondary moments

0 0

-0.01 Load step n6 -0.01 Load step n9

k/h=0.64 k/h=0.50
-0.02 -0.02
Deformationy (m)

Deformation y(m)

-0.03 -0.03

-0.04 -0.04
Calculated
Measured
-0.05 -0.05

-0.06 -0.06
Measured Measured
Calculated Calculated
-0.07 -0.07
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12
X (m) X (m)

Figure 15 Comparison between measured and calculated deformations of a barley stem (length h = 0.125 m) for increasing loads. (Left) Load step
no 6 (k/ h = 0.64). (Right) Load step no 9 (k/ h = 0.50)

angles , ; and (3) a new calculation of the beam is performed between the measured and calculated deformation profiles pre-
for the total moments Mf + Mf  . The final result is a decrease sented the lowest accuracy for larger deformations (k/ h < 0.6),
from the first-order calculations in the deflected plant height k where the error on total deflection k was 12% (as presented in
and an increase in the bending curvature (observed in real elastic Fig. 15 (Right).
materials). Figure 14 demonstrates the importance of including
secondary moments in order to more accurately reproduce the
bending of elastic stems. 5 Model verification
Stress strain tests were prepared and run in the laboratory
under strictly controlled conditions to: (1) estimate the stiffness The numerical results of the integrated model were compared to
modulus E of canopy and (2) to verify the accuracy of the numer- experimental data in order to verify the accuracy of the code.
ical solutions from the new mechanical model. Initially, PVC Figure 16 shows two examples. First, the experimental data
cylinders were tested with well-known geomechanical properties obtained by Tsujimoto and Kitamura (1990) in flow through rigid
to check the deformation hypothesis and numerical algorithm. cylinders was used in the verification. Tsujimotos runs involved
The methodology was adopted to directly estimate the geome- vertical cylinders 4.59 cm tall and 1.5 mm in diameter, arranged
chanical properties of the canopy (PVC palms and natural barley in a staggered pattern with 2 cm spacing. The second example
stems), including length, average width and thickness, stiffness in Fig. 16 used experimental data from one of our tests with
modulus E and inertial modulus I. These properties were intro- barley cover (run TNh = 0.18 m, Q = 0.136 m3 /s). The flex-
duced as input data in the mechanical model. The results agreed ible barley stems were 18 cm tall. The spacing between stems
with the experimental ones at each load step. The adjustment was a = 6.7 mm and the submergence ratio was h/k = 1.83.
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 589

Figure 16 Comparison between the measured and calculated velocity profiles. Top: run A11 for vertical rigid cylinders from Tsujimoto and Kitamuras
experiments (1990). Bottom: flexible barley run (TN-h = 0.18 m) from our own experiments

The vertical distribution of plant width B(z) was defined as a multi-parameter optimization technique was developed by the
constant for Tsujimotos cylinders and it was variable for barley authors and applied to 14 runs with vegetation (T3 and TN runs)
(calculated from a frontal image). The drag coefficients Cd (z) in which the canopy was deflected and totally submerged. Table 3
were set as a constant for Tsujimotos cylinders (Cd = 1.5), but shows the basic properties of runs used in the calibration process.
for barley, the measured, experimental values of Cd (z) were used The measured data used in the optimization routine was the
in the computation. unit discharge qmea (m2 /s) and the velocity profiles Umea (z).
Computational input parameters such as mixing length lo, Some discrepancies were obtained between the measured unit
momentum diffusion constant  , and the secondary current factor discharge qmea and the vertical integral of Umea (z) due to 3D flow
sc were adjusted for each run manually. For Tsujimotos run characteristics. Both measurements were integrated in the opti-
they were lo = 2.5 mm,  = 0.17, and sc = 1.0. For the bar- mization function. This function  was defined as a eight vari-
ley run with TN-h = 0.18 m, the parameters were lo = 10 mm, ables quadratic function of the difference between the calculated
 = 0.08, and sc = 0.78. Figure 16 shows very good agree- output data and the measured data, as defined in Eqs (20) and (21):
ment between the measured and calculated velocity profiles, and
 = (lo ,  , sc , Cd1 , Cd2 , Cd3 , Cd4 , Cd5 )
the velocity profile curvatures were reproduced quite well. Dif-
ferences in the model parameters between Tsujimotos run and = (X1 , . . . , X8 ) (20)
barley run are probably due to the geometrical properties of the
obstacles: cylinders are rigid and rounded obstacles while bar-
14
(qi qadv
i
)2 i
(qcalc qweir
i
)2
ley leafs are flexible, with non-uniform shape and they become = calc
+
i=1
error(qadv )2 error(qweir )2
stream flow oriented for medium-high discharges. Also, values
of bed mixing length lo is related to plant width B and spacing a,
5
(Cd,calc (j) Cd,exp (j))2
so different vegetative configurations lead to different model + , (21)
j=1
(Cd,exp (j))2
parameters.
In this study, we were able to calibrate and verify the model where qcalc is the calculated
 unit discharge from calculated veloc-
in an accurate, controlled way due to a previous experimental ity profile Ucalc (z)(q = U(z).dz), qadv is the experimental unit
study. Our own data was vitally important. Although the tests discharge calculated from the ADV velocity profile, qweir is the
were limited in number (a greater variety of plant density and experimental unit discharge estimated from the weir, Cd,calc is
shapes would be desirable), we obtained detailed measurements. the calculated drag coefficient, Cd,exp is the experimental drag
Thought it was concluded above that different canopy config- coefficient. The three terms in  function are normalized using
urations require different computational parameters, the aim of the measurement error of ADV discharge (error qadv ), of weir
the next work was to check the case that plastic tests (T3 runs) discharge (error qweir ), and the standard deviation of the experi-
and Barley Tests (TN runs) would share a unique group of model mental drag coefficient. (s(Cd,exp )). The complexity of Eq. (21)
parameters. The main idea was then to group the model parame- is the result of the imposition of known, coherent limits or
ters from particular adjusts to a general fit. The goodness of that boundaries to the search method.
objective is concluded later. A particular, unique group of general A first-order, two steps, conjugate-gradients technique was
computational parameters needed to be searched to fit all different developed and implemented to find the minimum of  (for more
tests. The list of unknown parameters included drag coefficient details, Velasco (2006)). Due to the strictly numerical definition
points, {Re(j), Cd(j)}, j = 1, 5, mixing length lo , momentum of  (there is no analytical expression for ), the search strategy
diffusion constant  , and the secondary current factor sc . A is complex and the unicity of solution is non demonstrated. In
590 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 17 Optimally adjusted resistance law (red, dashed line) compared to experimental data of runs T3 and TN

Table 3 Resume of experimental runs used in the optimization process

RUN Q h U (N/m2 ) Re k p h/k qmea qcalc


(l/s) (m) (m/s) (N/m2 ) 104 (m) (m) (m2 /s) (m2 /s)

T3 M = 205 48 0.136 0.14 2.6 1.59 0.136 0.136 1.00 0.019 0.009
T3 M = 205 77 0.184 0.17 2.8 2.55 0.111 0.072 1.66 0.031 0.031
T3 M = 205 90 0.206 0.17 3.1 2.98 0.110 0.054 1.87 0.036 0.032
T3 M = 205 107 0.230 0.19 2.9 3.54 0.112 0.057 2.06 0.043 0.038
T3 M = 205 136 0.275 0.20 3.8 4.50 0.102 0.051 2.70 0.054 0.077
TNh = 0.09 50 0.136 0.15 4.5 1.67 0.085 0.039 1.60 0.020 0.009
TNh = 0.09 93 0.193 0.19 4.5 3.06 0.092 0.033 2.10 0.037 0.024
TNh = 0.09 135 0.255 0.21 5.9 4.46 0.088 0.025 2.89 0.054 0.052
TNh = 0.12 72 0.171 0.17 9.8 2.37 0.118 0.060 1.45 0.028 0.019
TNh = 0.12 109 0.217 0.20 8.6 3.61 0.118 0.072 1.84 0.043 0.032
TNh = 0.12 136 0.255 0.21 8.7 4.50 0.122 0.063 2.09 0.054 0.064
TNh = 0.18 71 0.173 0.16 12.4 2.35 0.128 0.072 1.35 0.028 0.012
TNh = 0.18 111 0.221 0.20 10.7 3.67 0.127 0.084 1.74 0.044 0.037
TNh = 0.18 136 0.257 0.21 10.4 4.50 0.140 0.086 1.83 0.054 0.069

Table 4 Adjusted parameters of the drag coefficient law adjusted parameters Red (j) and Cd(j) (see Table 4), is consistent
and turbulence parameters with the measured data, as shown in Fig. 17. The optimization
lo (m) 0.0125 procedure of parameters Red (j) and Cd(j) proposed in this paper
 0.082 takes into account the minimization error of the measured and
sc 0.45 calculated flow discharge (Fig. 21), that is more complex than a
Red Cd simple fitting procedure.
10 4.0 Extrapolation of drag coefficients out of the range of Red was
90 1.0 necessary because the numerical model evaluates the existence
630 0.011 of solutions even beyond the limits of experimental Red .

consequence, a set of searches with different initial conditions 6 Numerical results


were run to verify that the convergence collapse toward a similar
solution. Table 2 shows the optimal adjusted parameters. This section discusses the results of the numerical model which
These adjusted parameters are consistent with the physical has been run using the optimized, global group of computational
measurements. The resistance law, which is defined by the parameters for all the tests (see Table 4). Figures 18, 19 and 20
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 591

Figure 18 Comparison between calculated (green line) and measured (blue circles) velocity U(z) and turbulent shear stress xz (z) profiles,
corresponding to run T3
592 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 19 Comparison between calculated (green line) and measured (blue circles) velocity U(z) and turbulent shear stress xz (z) profiles
corresponding to run TNh = 0.09 m

show all runs and compares experimental data to computed data. seems to be accomplished in general. The top of the canopy in
Velocity U(z) and Reynolds turbulent stress xz (z) profiles are Figs 1821 corresponds to the peak of calculated stress (xz,max )
presented. In general, the goodness of the results in velocities and and approximately agrees with the peak of measured xz . In the
turbulent stress xz are acceptable, with a limited accuracy. It is external zone (z > k), the slope of calculated xz does not fit the
important to note that the more the discharge and submerged ratio data. The calculated velocity U(z) reproduces the gradients in
increases, the more the calculated profiles are well fitted. Agree- the internal zone quite well, but there is some discrepancy close
ment is observed for the penetration depth p, and the shearless to the free surface (the dumping effect of secondary currents in
zone becomes well defined in terms of velocity and shear. Veloc- shear stresses xz. toward the water surface is modeled in a too
ity Uo in this zone is calculated as a constant profile in the range of simple way).
the measured Uo . We therefore conclude that the proposed veloc- Figure 22 shows the calculated Red and Cd vertical pro-
ity boundary condition (Eq. (11)) is efficient. In Zone 2 (internal files and resistance law for a barley run, and compares them
zone p < z < k), the calculated stress xz shows some deviation to the experimentally measured data. The obstacle Reynolds
from the measured data, but vegetative momentum absorption number shows good agreement (plant shape B(z) controls Re
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 593

Figure 20 Comparison between calculated (green line) and measured (blue circles) velocity U(z) and turbulent shear stress xz (z) profiles
corresponding to run TNh = 0.12 m

distribution), and computed drag coefficients agrees quite well but the estimations of k are generally in a sufficient error range
with the experimental resistance law Cd (Red ). of 7%. More experiments are needed to check that the calcu-
The next step in the results analysis is mass conservation. In lated deflected plant k is adjusted to reality (especially in large
Fig. 23, the unit discharge (an output of the numerical model) is deformation conditions, h /k > 0.6). The global minimization
compared to the experimental unit discharge. For all 14 tests, the procedure consists in adjusting a set of parameters to all the
comparison between the calculated unit discharge qcalc (m2 /s) and experimental data sets (Figs 1821), and it is expected that some
the measured qmea shows certain dispersion. The model is under- experimental data do not agree with the model. Figure 15 shows
estimating discharges for low flow conditions and overestimating the results of a particular run in which it is possible to adjust
for the higher submergence conditions. The average estimation perfectly the parameters using the optimization procedure.
error is 25%, but the experimental error between qmea and the
vertical integral of experimental Umea (z) was set at 10%. We
conclude that the data on discharge q is not much more accu- 7 Limitations of the model
rate than typical laboratory work involving vegetation covers.
Figure 23 also shows the results for deflected plant height k The integrated vegetation model is very useful for reproducing
and penetration depth p. Some dispersion of data is detected, vegetative covers in submerged plant conditions. Due to its low
594 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

Figure 21 Comparison between calculated (green line) and measured (blue circles) velocity U(z) and turbulent shear stress xz (z) profiles
corresponding to run TNh = 0.18 m

Figure 22 Measured and calculated values of Red and Cd


Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 595

Figure 23 Numerical results for unit discharge q (left) and deflected plant height k and penetration depth p (right)

computational cost, this model can easily be incorporated into vegetative covers were fixed on the channel bed. Vertical veloc-
any general 1D or 2D hydraulic code to estimate friction fac- ity profiles U and turbulent shear stress xz (obtained using ADV
tors in riverbeds and floodplains. However, the numerical code measurement techniques) provided valuable experimental data on
has certain limitations. The governing equations are only defined every flow and vegetative configuration. Fourteen experimental
for the existence of the shearless zone (Zone 3), so vegetation runs were performed to verify and calibrate the numerical model.
must be able to absorb the total gravity shear G . This model is Turbulent diffusion coefficients, mixing length and a resistance
therefore recommended for high-to-medium plant density (high equation (drag coefficient Cd vs. Red ) were the input parameters
Ao /a ratio). The model should be modified for flooding scenar- of the model. Parameters with physical meaning were calibrated.
ios in which the canopy is totally prone, friction factors are low A multi-parameter search algorithm was used to adjust the com-
and an effective shear stress is acting on the river bed. Finally, puted profiles to the experimental data. Finally, the parameters
a more intense experimental campaign is needed to verify the were estimated. The use of a global, optimized group of compu-
proposed drag coefficient law Cd (Red ) for new plant shapes and tational parameters to simulate different shapes and configuration
configurations. of canopy shows a sufficient but limited accuracy; we think that
particular model parameters in the calculation of every type of
canopy would better adjust data. In practice, the uniqueness of
8 Summary and conclusions this set of general parameters should be verified in the future
using different species of vegetation.
This paper presents an experimental and numerical study that This integrated model offers a low computation cost and a
provides deeper knowledge of resistance to flow over flexible reasonable estimation of plant deflection, velocity, and stresses.
vegetation. A numerical model was computed to simulate vertical Vertical profiles are obviously less accurate than any 3D-turbulent
velocity profiles and plant deflection. Kutija and Hong (1996) scheme. However, the approximation errors are quite small and
have used in the past the 1D Reynolds equation in uniform would therefore be easily dealt with in a general floodplain calcu-
regime and simple turbulence closure models to create an implicit lation. This quick model is therefore very suitable and fast enough
model to reproduce vertical profiles of velocity and shear over to be included in a future general 1D or 2D hydraulic model as a
canopy covers. With a similar point of view, this paper presents resistance function for vegetation.
a scheme that is derived from the simplified, steady Reynolds
equation, which was vertically integrated above and below the
deflected plant height k. A mechanical module was computed Notation
to estimate large elastic deformation in stems. A new, modified
mixing-length theory was used in a turbulence closure equa- a = Spacing between plants
tion for the hydrodynamic subroutine. Also, drag coefficients Ao = Frontal area of plant
Cd estimation is based on the obstacle Reynolds number Red , b = Channel width
in order to represent properly the resistance to flow due to veg- B = Average plant width
etation. We proposed an iterative convergence scheme for plant Cd = Drag coefficient
deformation k as a function of balanced drag forces. Finally, the E = Stiffness modulus
well-balanced vertical velocity profile U(z) was calculated using h = Water depth
a simple, explicit finite-difference scheme. h = Undeflected plant height
In the experimental chapter, two different experimental flumes I = Second-order geometrical momentum
were used. Both natural (barley) and artificial (PVC strips) k = Deflected plant height
596 D. Velasco et al. Journal of Hydraulic Research Vol. 46, No. 5 (2008)

l = Mixing length Erduran, K.S., Kutija, V. (2000). Quasi-three-dimensional


L = Integral scale numerical moldel for flow through flexible, rigid, submerged
M = Density of vegetation (plants/m2 ) and non-submerged vegetation. J. Hydroinformatics 5(3),
Mf = Moment 189202.
p = Penetration depth Fischer-Antze, T., Stoesser, T., Bates, P. (2001). 3d numerical
q = Unit discharge (= Q/b) modelling of open-channel flow with submerged vegetation.
Q = Discharge J. Hydraul. Res. 39(3), 303310.
Re = Reynolds number (= U.h/) Ikeda, S., Kanazawa, M. (1996). Three-dimensional organized
Red = Obstacle Reynolds number (= U.B/) vortices above flexible water plants. ASCE J. Hydraul. Eng.
Rh = Hydraulic radius 122(11), 634640.
Sf = Energy slope Klopstra, D., Barneveld, H.J. (1997). Analytical model for
Somaxshear = Experimental energy slope obtained from hydraulic roughness of submerged vegetation. Proceedings of
the xz, vertical profile the 27th Congress of the IAHR, San Francisco (USA).
So = Geometric slope Kouwen, N., Unny, T.E., Hill, H.M. (1969). Flow retardance
U, V, W = Longitudinal, transversal, and vertical in vegetated channels. ASCE J. Irrigation and Drainage Div.
mean velocities, respectively 95(IR2), 329342.
u , v , w = Longitudinal, transversal, and vertical Kouwen, N., Unny, T.E. (1973). Flexible roughness in open
velocity fluctuations, respectively channels. ASCE J. Hydraul. Div. 99(HY5), 713727.
u = Shear velocity Kouwen, N., Li, R. (1980). Biomechanics of vegetative channel
, sc = Secondary current factor linings. ASCE J. Hydraul. Div. 106(HY6), 10851103.
= Secondary current factor Kutija, V., Minh Hong, H.T. (1996). A numerical model for
d = Secondary current parameter assessing the additional resistance to flow introduced by
do = Viscous layer depth flexible vegetation. J. Hydraul. Res. 34(1), 99114.
= Specific weight of water Lpez, F., Garca, M.H. (2001). Mean flow and turbulence struc-
= Von Karman universal diffusion constant ture of open-channel flow through non-emergent vegetation.
 = Diffusion coefficient ASCE J. Hydraul. Div. 127(HY5), 392402.
= Kinematic viscosity Naot, D., Nezu, I., Nakagawa, H. (1996). Hydrodynamic behav-
= Water density ior of partly vegetated open channels. ASCE J. Hydraul. Eng.
xz , xy , yz = Reynolds stresses 122(11), 625633.
cd = Vegetative drag stress Nepf, H.M. (1999). Drag, turbulence and diffusion in flow
= Bed shear stress through emergent vegetation. Water Resour. Res. 35(2),
479489.
Acknowledgments Nepf, H.M., Vivoni, E. (2000). Flow structure in depth-limited,
vegetated flow. J. Geophys. Res. 105, 2854728557.
This research was supported by a UPC per la recerca grant (Tech- Nezu, I., Nakagawa, H. (1993). Turbulence in open channel
nical University of Catalonia) and projects REN2000-1013/HID flows. In: Balkema (ed.), Monograph Series IAHR.
and BTE2002-0375. We are especially grateful to Prof. J.M. Nezu, I., Onitsuka, K. (2001). Turbulent structures in partly
Redondo, who was very helpful in the initial phase of the vegetated open-channel flows with lda and piv measurements.
experimental work. J. Hydraul. Res., IAHR 39(3), 629642.
Petryk, S., Bosmajian, G. (1975). Analysis of flow through
vegetation. ASCE J. Hydraul. Div. 101, 871884.
References Rouse, H. (1946). Elementary mechanics of fluids. Dover
Publications.
Carollo, F., Ferro, V. (2002). Flow velocity measurements in Stoesser, T., Liang, C., Rodi, W., Jirka, G.H. (2006). Large
vegetated channels. ASCE J. Hydraul. Eng. 128, 664673. eddy simulation of fully-developed turbulent flow through
Cui, J., Neary, V.S. (2002). Large eddy simulation (LES) of submerged vegetation. Proceedings of River Flow 2006 1,
fully developed flow through vegetation. Hydroinformatics 227234.
2002: Proceedings of the 5th International Conference on Timoshenko, S., Young, D.H. (1975). Unin Tipogrfica Edito-
Hydroinformatics, Cardiff, UK. rial Hispanoamericana. In: S. A. Mxico (ed.), Elementos De
Choi, S., Kang, H. (2004). Reynolds stress modeling of vegetated Resistencia De Materiales.
open-channel flows. J. Hydraul. Res. 42(1), 311. Tsujimoto, T., Kitamura, T. (1990). Velocity profile of flow in
Dunn, C., Lpez, F. (1996). Mean flow and turbulence in a lab- vegetated-bed channels. KHL Progressive Report, Hydraulic
oratory channel with simulated vegetation. Civil Engineering Laboratory, Kanazama University, Japan.
Studies, Hydraulic Engineering Series No. 51. University of Velasco, D., Bateman, A. (2003). An open channel flow exper-
Illinois, Urbana-Champaign. imental and theoretical study of resistance and turbulent
Journal of Hydraulic Research Vol. 46, No. 5 (2008) A new integrated, hydro-mechanical model applied to flexible vegetation in riverbeds 597

characterization over flexible vegetated linings. Flow, Turbu- Hidrosistemas Fluviales. PhD Thesis. (in Spanish). Technical
lence and Combustion, Vol. 70. FTAC Special Issue: Fluxes University of Catalonia-UPC.
and Structures in Fluids. Kluwer Publications, 6988. Watanabe, T., Kondo, J. (1990). The influence of canopy struc-
Velasco, D. (2006). Estudio Experimental y Modelacin de la ture and density upon the mixing length within and above
Resistencia al Flujo y Caracterizacin Turbulenta de la Inter- vegetation. J. Meteorol. Soc. Jpn. 68(2), 227235.
accin FlujoVegetacin Flexible. Aplicacin numrica a los

Vous aimerez peut-être aussi