Escolar Documentos
Profissional Documentos
Cultura Documentos
Introduction
Conservation of mass and momentum
General types of flow
Laminar vs. turbulent flow
Shear Stress
Reach-average shear stress
Bed roughness and reach average flow velocity
Shear stress partitioning
Local shear stress
Laminar velocity profile
Turbulent velocity profile
Determining u* and zo
Laminar sublayer
Smooth bed
Rough bed
Flow Energy
Forms of stream energy
Bernouilli equation
Navier-Stokes Equation
Derivation
Simplifications
Reynolds number
Froude number
Hydraulic scaling
(note Q = discharge; A = x-sectional area; u = velocity so these equations are in volume terms)
We will use these two basic principles to derive the shear stress that acts on the channel bed
(and that transports sediment), the velocity profile in a river, and the equations governing
channel flow.
Simple mathematical models of flow in channels can be constructed only if flow is uniform and
steady. Although flow in natural rivers is characteristically non-uniform and unsteady, most
models rely upon the steady uniform flow assumption.
Note that water is assumed to be stuck to the boundary (the no-slip assumption).
Natural rivers have local irregularities in bed and bank topography that introduce
significant local convergence and divergence of flow that can impose large local gradients in flow
velocity and shear stress. We use a reach-average view of channels in order to make for a
solvable analytical model.
First, consider the force balance on the volume of water in an entire reach of length L and
slope :
Assume that acceleration of flow in the reach is negligible and that the bed is not moving, so
there must be a balance between (1) the gravitational force accelerating the water downstream
and (2) the frictional resistance to flow caused by the boundary, which slows the fluid velocity to
zero at the bed and banks and therefore causes internal deformation of the flow.
A L g sin (3)
The total boundary resistance (which is also a force, i.e. = stress area) for the reach equals
b L P (4)
where b is the average drag force per unit area (shear) on the boundary.
Equating the force moving the flow (3) with the force resisting flow (4) (since we are assuming
no additional energy inputs), we get
b L P = A L g sin (5)
If we define the hydraulic radius as R (A/P) then this simplifies to the standard expression for
the reach-average shear stress
b = R g sin (7)
Note that for wide channels A/P D; and for small , sin tan = S.
b = g D sin (8)
**The force exerted by flow on the channel bed is proportional to flow depth and slope**
We've now established that the basal shear stress is related to the depth-slope product, but how
do we get at flow velocity?
Chezy (1775) first applied mathematical analysis to the mechanics of uniform flow. He made 2
assumptions:
#1 Exact balance between force driving flow (downslope component of the weight of water)
and the total force of bed resistance (i.e. the same assumption we made in writing
equation 5).
#2 The force resisting the flow per unit bed area (i.e., b) varies as the square of velocity:
b = k u2 (9)
Assuming no acceleration [Chezys assumption #1 above] then these forces balance and
u2 = ( g / k) R S (15)
and hence
u = C (R S)(0.5) (16)
Equation 16 is called the Chezy equation and C is called the Chezy Coefficient.
Hence, if both of Chezys assumptions are correct, the average velocity in a channel should
increase with the square root of the gradient, the square root of the hydraulic radius (which for
wide shallow channels is equal to the average depth), and a coefficient that reflects the
smoothness of the channel (i.e., the inverse of channel roughness).
C R(1/6)
Substitution of this result into the Chezy equation [eqn. 16] produced the famous Manning
Equation:
Mannings n is a roughness coefficient that depends on channel margin irregularity and the grain
size of the bed material. The scaling of n has been chosen so that constant k1 = 1 in SI units and
1.49 in English units (fps). This has been known to cause confusion.
Mannings n reflects the net effect of all the factors causing flow resistance in a fluid of a given
viscosity (because of the temperature effect on viscosity, a channels n varies slightly throughout
the year).
The third common roughness equation is the Darcy-Weisbach equation for frictional losses in
circular pipes, which can be modified for open channel flow:
ff = 8 g R S / u2 (20)
1. Grain (or "skin") resistance (or "roughness") due to the presence of small, distributed
irregularities such as bed-forming material.
2. Form resistance due to the larger-scale internal deformation in the flow field imposed by
channel bed irregularities such as bedforms (e.g., dunes, bars, pools, etc...) and by
variations in the plan form of the river (e.g., meanders).
3. Spill resistance due to surface waves generated by large obstacles protruding from banks,
steps in the channel bed profile, or other obstacles such as logs and boulders.
The reach average basal shear stress (b) is often considered to be composed of linearly additive
where:
g is the grain roughness,
and so ' is the effective shear stress available for sediment transport.
' = b (g + bf + s) (23)
Bedform roughness (bf ) can account for 10 70% of the total roughness in channels with well-
Accounting for these various forms of roughness is a major challenge for predicting flow velocity
and sediment transport, and it is done in 3 typical ways:
You will have an opportunity to practice all three on the first field trip.
H
z
The shear stress on any surface at height z above the bed is caused by the downslope
gravitational stress of the water above the plane - i.e., by the downslope component of the weight
of the fluid between z and the water surface (at height H).
Hence, the shear stress at any point within the fluid will be given by :
= g (Hz) sin (24)
Equation (24) indicates that the shear stress decreases linearly with height above the bed.
surface
zs
bottom b
Note also that for the case of z = 0 (i.e., at the channel bed), = b and so equation (24)
reduces to :
b = g D sin (25)
Combining (26) with (24) above [shear stress distribution in the flow]
= du/dz = g (Hz) sin (27)
Rearranging yields:
du = ( g sin / )H dz ( g sin / )z dz (28)
Integrating:
u = ( g sin / ) (Hz) ( g sin / ) (z2 / 2) + C (29)
Combining terms and using the boundary condition that u = 0 when z = 0 [which inspection of
(29) shows implies that C = 0] yields:
u = ( g sin / ) [Hz (z2 / 2)] (30)
large scale eddies, which impart an extra "eddy viscosity" term () that can be considered
analogous to momentum transfer by conventional viscosity:
This works because typically >> and hence turbulent flow is slower than laminar flow at the
same shear stress. This is because the drag from the bed is transferred more efficiently into the
body of the flow by eddies than by viscosity alone.
It is extremely difficult to determine the eddy viscosity, but Prandtl proposed that the eddies
would have a length scale (a distance across which they could exchange momentum between
layers in a unit of time) that was proportional to the distance away from the solid/fluid boundary
-- eddying would be suppressed near the boundary. He also proposed that depended on the
velocity gradient (du/dz). Thus he developed an expression for the eddy viscosity
= l2 (du/dz) (32)
where is the density of water and l is Prandtl's mixing length, which depends on proximity to
the boundary and was experimentally determined as
l=z (33)
where = 0.4
Equation (33) can be substituted back into (32) and then (31) to yield
= 2 z2 (du/dz)2 (34)
but has the dimensions of velocity [i.e. L / t]. It is assumed to be constant near the bed, where
was also assumed to be constant and equal to b:
u* = z (du/dz) (36)
u = (u* / ) ln z + C (38)
If we impose the boundary condition that u = zero at some elevation z0, just above the bed,
then:
0 = (u* / ) ln z0 + C (39)
and therefore
C = (u* / ) ln z0 (40)
This is the "Law of the Wall" (i.e., the equation for turbulent velocity distribution away from, but
close to, a fixed boundary such that b).
surface
lnZ
ln (z)
Z0
bottom
z = zo
Uu
The "Law of the Wall" predicts a logarithmic velocity profile that begins at a roughness length
scale that defines the height above the bed of z0. Below this height flow is must be assumed to be
laminar, because it is indeterminate under our turbulent assumptions (since u = 0 at z = z0).
Note that in equations 3342 is called von Karman's constant (and = 0.4).
1 Newton's viscous flow law applies, as modified in (31) to include an eddy viscosity.
Farther from the boundary, b, and perhaps at such points in the interior of the fluid the eddy
viscosity will depend not on the local distance from the bed (z) but rather the on total flow depth
(H). If so, it will be constant across this interior flow. Mathematically, this is equivalent to
equation (30), i.e. a constant viscosity (only in this case its an eddy viscosity). As a result, the
velocity profile in the interior of the flow will also be parabolic (see equation 30), although with a
different viscosity than in the laminar sublayer.
the slope of the velocity profile on a semi-log plot can be used to measure the local shear stress,
particularly near the channel bed, either over bedforms, or (if the velocity profile can be defined
To obtain u* and z0 in equation (42), measure u at various heights, z, above the bed. If you take
the natural logarithm of the z values, then if the points conform to (42) they will plot as a
straight line (where the x-axis is velocity and the y-axis is ln z) because (42) would be written as
u = (u*/) ( ln z ln z0 ) (43)
Hence u* can be calculated from either the best-fit line through paired values of u and ln z data
or by reading pairs of data and using the equation for the slope of a line.
If you plot the logarithm of flow depth on the y-axis and velocity on the x-axis, then the slope of
Hence, if you take a linear regression of ln z (the natural logarithm of the flow depth at which
each velocity measurement was made) versus the flow velocity (u) then in the slope-intercept
form of the expression (y = mx + b), the slope of that line (m) is given by /u* and the intercept
So z0 = eb
u* = / m
reasonable approximation only close to the boundary and has therefore become known as "the
Farther away from the bed, the mixing length becomes constant at (an empirically determined)
fraction of the total depth and the velocity profile becomes parabolic above that depth. Log and
parabolic profiles predict the same velocity at 0.2H, which is the presumed level of this
transition. However, the difference between the computed logarithmic and upper parabolic
profiles in most streams is negligible, and so for many applications a logarithmic profile can be
assumed throughout.
The thickness of the sublayer (v) depends on the near-bed shear velocity. By
v = 11.6 / u* (45)
ks
So, what is the scale of v for flow in a typical gravel bed river with a depth of 1 m and sin =
0.005? (about 0.05 mm, but work it out yourself!)
What is the scale of v for flow in a typical gravel bed river with a depth of 2 m and sin =
0.035? [high estimates] ( 0.01 mm)
What is the scale of v for flow in a typical gravel bed river with a depth of 0.5 m and sin =
0.001? [low estimates] ] ( 0.2 mm)
Hence, the length scale of v is about the diameter of silt to fine sand grains.
smooth flow where the turbulence doesn't interact with the bed roughness? We can already
expect that ks must be much less than 11.6 / u*.
ks
Note that we can define a dimensionless ratio of the laminar sublayer thickness to the roughness
elements on the bed. This has been termed the Roughness Reynolds number, and for
dimensional homogeneity (and linear dependence of v on and u ):
*
Re* = ks u* / (46)
From (45), we know that this ratio must be much less than 11.6 (because ks must be much
less than v for hydraulically smooth flow to occur), but only experiments can determine just how
For hydraulically smooth flow, measured velocity profiles in the overrunning turbulent flow
indicate an apparent z0 of
z0 v / 100 (48)
Combining (45) and (48) yields: z0 / (9 u*) (i.e. very small!) (49)
the laminar sublayer will rise and fall over the protuberances, and the grains will begin
contributing addition form drag in addition to ordinary surface friction:
ks
Consequently, turbulence interacts directly with the roughness elements causing z0 to be scaled
by their size. We know that ks must be much greater than v and thus that
Re* must be much greater than 11.6, but once again experiments were required to determine
just how much. Nikuradse's experiments for such "hydraulically rough flow" showed that it
occurred when:
ks u* / 100 (50)
He also anticipated that the value of z0 would depend on ks; by further experiment,
z0 = ks / 30 (51)
Field measurements have shown D84 to provide a reasonable measure of ks, although Whiting
and Dietrich (1991) reported field-measured z0 values that were about 3 times larger than
predicted by equation 51.
For the simple case of spatially-uniform rainfall, the potential energy (Ep) in a catchment is equal
to the integral of the product of water mass (m), gravitational acceleration (g), and elevation (z)
Ep = m g dz (53)
Initially, the total energy of the system (E) consists of potential energy (mgz).
Downslope movement of water converts this potential energy into kinetic energy (mu2/ 2),
pressure energy (mgD), and energy dissipated by friction (F) and turbulence. Conservation of
energy implies that E = 0 and hence this dissipative system is charcterized by
The loss of potential energy is compensated by increased flow velocity, increased flow depth,
and/or greater frictional energy dissipation. Thus,
Assuming that change in the downstream flow velocity is small [i.e., (mu2/ 2) 0], then the
rate of frictional energy dissipation is related to the fall in the water surface per unit channel
length (L):
The frictional energy dissipation per unit channel length effectively scales the channel roughness
(R). Noting that H/L is the water surface slope (S), implies that R S.
In general, changes in slope dominate flow depth changes (Leopold et al., 1964). Since channels
tend to be steep in their headwaters and decrease in slope downstream, this implies that channel
roughness generally decreases downstream.
This leads to the rather counter-intuitive result that steep headwater channels flow slower than
their lowland counterparts.
For many years geologists simply asserted that steep headwater channels obviously flowed faster
than their lowland counterparts.
In 1953 Luna Leopold showed that this conventional wisdom was incorrect by having the
audacity to actually go out and measure stream velocity at many points down a channel network.
This effect is due to the greater roughness of steeper channels -- low gradient rivers can be
deceptively fast!
potential energy: gh
pressure energy: g d cos
kinetic energy: u2/ 2
The term in parentheses is the total head (H) and flow is driven from high to low head. Note
that H is now a distance above the datum, not the total flow depth as before:
This is the Bernoulli equation which describes conservation of energy from reach to reach.
Consider two reaches (designated with subscript 1 and 2):
The head loss between the reaches (H) will be equal to H1 H2 and hence
Note that the energy, water surface, and bed slopes are not necessarily parallel.
This can be stated in words that the rate of change of momentum of a body is equal to the
force(s) acting on that body (or particle, or infinitesimal element of material, or whatever).
Recall that momentum is equal to mass (m) times velocity (u), and acceleration (a) is the first
derivative of velocity with respect to time (i.e. the rate of change). Because we do not expect
mass to change with time,
This becomes complicated only because we need to address both body forces (gravity is the
most common of these) and surface forces (also called tractions), and because if we are being
complete then we must deal with them in all 3 dimensions.
The notation for Newtons second law in 3 dimensions, with body and surface forces called out
separately, expressed per unit volume, is:
[u ] = [ g ] + [ ]
d
(64)
dt
This is Cauchys first law, and it applies to any material (since we have only made the assumption
that it behaves in accord with Newtons second law). It says, in tensor notation (i.e. vectors in 3
dimensions, indicated by the ~ symbol over the 3-D variables), that the change in momentum
(per unit volume) equals the sum of the body force (gravity, onlyno magnetic fields allowed!)
and the surface tractions (more about them later).
du
= [ p] + [ ] [ g ] (65)
dt
We have separated the surface forces into those that apply a shear () and those that act
isotropically (p), which we normally call the pressure. It is defined as:
11 + 22 + 33
p (66)
3
where ij is the notation whereby the force in question is acting on the face perpendicular to the
ith axis and is applied in the direction parallel to the jth axis.
With equation (66), we are stuck until the non-isotropic (also called the deviatoric) part can
be expanded. To do this, we need a constitutive equation that relates strain (deformation, or
movement) of the material to the applied stress (which, by definition, is a force per unit area).
This requires experimentation. Fortunately, there is a large class of common materials that
behave rather simply: their strain gradient is proportional to the applied shear stress. In (3-D)
tensor notation, this can be written as:
du j
ij (67)
dx i
The proportionality constant? For these materials, called Newtonian fluids, that constant
(which will vary for different substances, but which is the same value in any direction and under
any applied stress regime), is called the viscosity (). Weve done this already, but we came at it
then with a less explicit set of simplifying assumptions (see equation 26). If we add the additional
requirements that the material is incompressible and isotropic, Cauchys first law (equation 65)
becomes:
du
= [ p] + [ 2u ] [ g ] (68)
dt
These simplifications, applied to equation (68), yield two equations (for the x and z directions):
p/x = zx/z (69)
p/z = g (70)
and since the pressure equals 0 (atmospheric) at the water surface (where z = H), we can define
h as the distance down from the level z = H:
p = gh (72)
Note that this holds strictly for steady, 2-dimensional flow. This rules out turbulence! Even the
simple hydrostatic equation was built from these same assumptions, and so strictly speaking it
too applies only for non-turbulent conditions. We can evaluate whether we need to worry about
turbulence, and we can also figure out what to do about it, using two different approaches.
First, we can just pretend that it doesnt matter and make some experimental measurements.
From these, we find that the basic equations derived from the Navier-Stokes equation (i.e.
equations (72) and (74)) work pretty well, virtually all of the time. So well continue to use
them.
For the velocity distribution (76), however, results are not so friendly. We already found that
where turbulence is important,
and the form of the eddy viscosity () leads to a logarithmic (as opposed to a parabolic)
velocity profile wherever that viscosity depends on height above the bed.
Re = u D / (77)
Velocity components of turbulent flow consist at any point of a time-average mean velocity (i.e.,
x) and flucutating velocity components (i.e., ux')
ux = x + ux'
uy = y + uy' (78)
uz = z + uz'
Froude Number
The Froude number, Fr, named for the English engineer William Froude (18101879) is
important because it is the ratio between the velocity of streamflow (u) and that of a shallow
gravity wave [(gd)0.5], or the ratio of inertial forces to gravity forces, as simplified as follows:
Fr = u / (gd)0.5 (80)
Think of the Froude number as a measure of whether flow can outrun its own wake.
The Reynolds and Froude numbers are independent of the scale of the river and hence provide
dimensionless ways to characterize flow. They also have distinct physical manifestations in the
behavior of flow in a channel.
Hence, if the depth ratio is 1/100, then the velocity ratio must be 1/10.
or by rearranging
(dmum / drur) = (mr / rm) (84)
If also subject to Froude scaling we can substitute (82) into (84) to yield
For a length scale of 1/100 (typical for flumes) rearranging (85) yields
(m / r) = .001 (m / r) (86)
Almost all common liquids have densities close to that of waterbut we need something 1000
times less viscous! Hence, it is impossible to achieve both Reynolds and Froude scaling in
experimental work.