You are on page 1of 83

Welcome to the second module on the Dams and Dam Failure – St.

Venant
Equations, Modeling and a Case Study. My name is Dr. Richard Koehler.

1
This module is intended for river forecast center personnel, service hydrologists and
other National Weather Service staff. It would also be useful for other water
resource professionals, emergency managers as well as academic audiences
interested in dynamic flow routing, and dam failure contingency planning.

The information covered in this two part series will benefit you by providing a
scientific foundation for advanced course work needed to run dam break simulations
and to conduct hydraulic modeling as a part of dynamic wave forecasting.

It is assumed that the person taking this module has already completed module 1,
“Dams and Dam Failure - Module 1: Terminology and Open Channel Hydraulics”.
This module should take you about an hour to complete.

2
This information builds on the previous module and provides an introduction to the
Saint-Venant equations for dynamic wave flow, and flood wave characteristics. The
general dam failure process along with advantages and limitations to modeling of
dam failure, model stability, accuracy, and sensitivity issues are presented.
Additionally, a case study is included.

3
Dam Failure Modeling Process section. This section provides a brief overview of
the modeling process used in dam failure analysis.

4
The idealized failure of a dam involves several different events. The initial causative
action is the weakening of the structure through piping, overtopping or foundation
failure. This process can occur over a period of minutes to days to weeks.

Breach formation and development follow where uncontrolled high volumes of water
are released downstream. This may take place as quickly as a matter of minutes to
several hours. Once the breach has formed, the reservoir contents are released and
the dewatering of the pool occurs.

The water that has flowed through the breach moves downstream as a flood wave.
The routed wave travel time will take minutes to days depending on the distance
downstream from the dam.

5
The steps involved in dam failure modeling can be stated in five general categories.
These are defining the objectives, data collection, flood hydrograph generation,
flood wave routing, and display of the results.

In this module, we will examine aspects of data collection, the mathematics behind
flood hydrograph generation and flood wave routing techniques appropriate for
various conditions.

A link to the European Union Environment and Climate Program report, Dambreak
Modeling Guidelines and Best Practices, is shown at the bottom of the page.

6
The Saint-Venant Equations - This section describes the partial differential
equations used in dynamic flow simulation.

7
Discussed here are the continuity of mass and the continuity of momentum
equations, commonly called the Saint-Venant equations after the French
mathematician Jean Claude Saint-Venant. He lived from 1797 to 1886 and studied
the fields of mechanics, elasticity, hydrostatics and hydrodynamics.

These equations are the basis of dam failure and flood wave modeling and most
any type of dynamic flow simulation. Modeling assumes a one-dimensional
gradually varied flow in open channels. This means computing the average cross
section water depth at any point along the river. Specifically, the continuity equation is
based on the principle of conservation of mass while the momentum equation describes the
equilibrium of forces within a body of water. Such forces include: hydrostatic pressure,
convective and local accelerations, gravity, and friction.

8
Various assumptions are needed to use the Saint-Venant equations. These include uniform
velocity within the cross-section and that the water level is horizontal across the channel,
gradually varied flow with hydrostatic pressure dominant at all points within the flow, that
vertical accelerations are ignored along with lateral and secondary circulations, that cross
section areas remain the same, that the water in the channel has uniform density, the
manning’s formula adequately describes the channel bottom friction. Flow is generally
subcritical. However in the real world, many of these assumptions are not present in a
perfect form. The column “Typical Areas of Exception” shows some actual conditions that
may exist.
i t

Some exceptions include the following:


-Water velocity is never uniform as flow varies across a floodplain. Also, velocity varies
within river reaches with constrictions or with sharp river bends.
- The second listed exception are regions where gradually varying flow is not present such
as near a dam failure site itself where rapidly
p y changingg g flow conditions exist.
-Another exception is the remove of material, called scour and the deposition of these
material occur within rivers from lateral circulation.
-Scour also changes the cross section area, increases the sediment load of the river
thereby changing the water density.
-Finally steep river channel of dam spillways can produce supercritical flow.

Despite
D it th
these exceptions,
ti th
the S
Saint-Venant
i tV t equations
ti are routinely
ti l used
d ffor estimating
ti ti
one-dimensional water levels under dynamic flow situations and adequately model a river
system under a variety of conditions.

9
The continuity equation accounts for change in water volume.

The basis for the equation is to examine the inflow and outflow discharges, cross
section area and any lateral inflows that may be present.

In the figure, a certain cross section is shown at low water within a river reach, Δx
in length
length. There is Qinflow flowing into the reach and Qoutflow leaving the reach
reach. Flows
that enter the channel within a reach of the river are called lateral inflows, sometime
as tributary streams.

10
In this next figure, the cross section area has changed as well as the reach inflow
and outflow. Tracking these changes will allow the development of an equation to
determine the water change within the specific reach of the river.

11
The terms of the continuity equation are the change in discharge over reach
distance plus the change in cross section area over time minus any lateral inflows
equals zero.

Stated another way, the continuity equation takes into account changes in water
volume over time and over channel length.

12
The momentum equation accounts for change in the energy present in the system.

The basis for this more complex equation is to examine the change in the weight of
the water moving downstream, the differences in upstream and downstream
pressure, how the speed of the water changes within the reach, any lateral inflow,
and the associated energies of each.

13
The momentum equation shows the local acceleration term is equal to minus the
convective acceleration term minus the pressure term minus the gravity term minus
the friction term plus any lateral inflow.

Stated another way, the continuity equation takes into account changes in
acceleration and thus changes in force within the river reach.

Except for the lateral inflow term, common variables include water density and
distance. Using the relationship Q=vA and the weight of the flow volume, the
above formula can be simplified.

Using the local acceleration term as an example, the initial equation is rewritten to
sho the change in velocity,
show elocit or acceleration.
acceleration The term is then di
divided
ided b
by the weight
eight
of water for a flow unit, such as cubic feet or meters. Canceling similar terms and
simplifying, we have the alternate form of the local acceleration. This process is
repeated for each term in the momentum equation with the results shown on the
next slide.

14
This more common form of the momentum equation still contains terms
representing the physical elements shown earlier.

Together the two acceleration terms represent the inertia within the system. Note
that the lateral inflow term has been removed to simplify this equation.

15
It is possible to use Manning’s equation to estimate the friction slope of the channel.

The equations show that channel roughness and hydraulic radius along with either
discharge and area, or average velocity can be used to estimate Sf.

16
The full dynamic wave equation includes all of the momentum equation factors
mentioned previously. The dynamic wave assumes unsteady, non-uniform flow, and
that depth, velocity, and acceleration vary only in the up and downstream, or x,
direction.

The equation also assumes gradual varied flow, uniform water density, fixed
channel boundaries, and that no erosion or deposition is taking place.

The full dynamic wave equation is appropriate for all river simulation situations but
extensive physical data is required. Required data include cross-section geometries,
roughness coefficients, off-channel storage areas, and any structures.

Details on the req


required
ired data are presented later in this presentation
presentation.

17
The kinematic wave is a simplified form of the momentum equation that ignores all
terms except for the channel slope. Important assumptions are made if this equation
is used and include steady, uniform flow, gravitational and frictional forces dominate,
that the energy gradeline or friction slope is equal to the channel bed slope and
consistent conveyance.

The river modeling situations where kinematic wave is appropriate are when steep
channel slopes, greater than 0.2 percent, are present. The kinematic wave should
not be used in tidal reaches or when backwater effects are present.

18
The diffuse wave is another simplified form of the momentum equation which
includes only the channel slope and pressure terms. The diffuse wave is also
known by the names “non-inertial” and “zero inertial” wave as there are no inertial
terms in the equation. An important assumption is that flow is steady and may be
either uniform or non-uniform.

The river modeling situations where diffuse wave is appropriate include rivers with
backwater effects and slow to moderate rising flood waves. This equation is not
appropriate for tidal reaches. It is best for large flood-storage basins.

19
The quasi-steady dynamic wave equation has the friction slope equal to the channel
slope, pressure, and the convective acceleration term.

The local acceleration term is ignored. Assumed are steady, non-uniform flow
conditions. The equation is not generally used for flood wave routing but is used to
find water surface profiles.

20
The gravity wave simplification of the dynamic wave equation ignores the friction
slope and the channel slope.

Generally the gravity wave equation is not used for flood wave routing but is
appropriate for conditions where the inertial terms dominate over slope terms.

A modeling situation would be ocean waves or a lake seiche


seiche, also called a standing
wave.

21
Here is a summary of the previous slides on various forms of the dynamic wave
equation. The actual on-the-ground physical conditions will determine which wave
equation may be used in place of the full dynamic wave equation.

Kinematic Wave
Physical setting - Steep channel slopes, No flood attenuation
Flow considerations - Steady, uniform flow, No backwater effects

Diffuse Wave (Non-inertia, Zero inertia)


Slow to moderate rising flood wave
Backwater effects allowed, No tidal reaches

Quasi-steady Wave
Water surface profiles from steady flow conditions
Steady flow

Gravity Wave
Not used for flood routing

Full Dynamic Wave


Any physical setting, any flow condition

22
Here are results from an open channel flow study to show the distribution of
momentum terms within a flood wave. The study used a theoretical river with a
steep channel and rapidly rising hydrograph. Results showed that, under these
conditions, the gravity term, or channel bottom slope, accounted for over 97 percent
of the energy in the system. If only this term is used, the kinematic wave equation
would describe most, but not all, of the energy.

The pressure term accounts for an additional 1.9 percent of the energy present in
the system. If the diffuse wave equation is used, which has both the slope and
pressure terms, 99.2 percent of the energy of the system is described. Likewise, the
convective and local acceleration terms account for 0.6 percent and 0.2 percent of
the system energy, respectively.

Under the stated conditions, a simpler equation in place of the full dynamic wave
equation may be justified as an appropriate trade off between complexity and
description of the system energy.

As the river channel becomes less steep and the hydrograph rise less rapid, the
various momentum terms will have different p
percents.

23
The US Army Corps of Engineers have tabulated the appropriate routing methods
for certain factors and is shown here.

If no calibration hydrograph is available, the dynamic wave, diffuse wave, and


kinematic wave methods are recommended.

If there are backwater influences


influences, the preferred routing technique would be the
dynamic wave, or diffuse wave.

When the flood wave is present within the river channel and overbank areas, any of
the methods that have separate hydraulic properties for the channel and the
overbank areas are appropriate.

Other criteria are available when examining the channel bed slope as a selection
factor. The reference values for depth and velocity should correspond as closely as
possible to the average flow conditions of the hydrograph to be routed.

For steep channels with slopes with a 10 feet elevation decrease per mile, all
methods are valid.

For channels with smaller slopes, less than 10 feet elevation decrease per mile, the
preferred models are the dynamic wave, or diffuse wave.
24
Flood wave characteristics section

25
In a generalized dam failure flood wave, there is a sharp rising limb, a large peak
outflow, and a sharp recession.

There is a very short time interval between the initial water rise and the peak
outflow. Supercritical flow conditions is likely when the channel slope is greater than
or equal to a 10 foot elevation decrease per mile.

26
There are characteristics of a dam failure flood wave that make it very different from
other types of flood waves. The first is the magnitude or peak flow.

Generally, the maximum flow is much greater than runoff-generated flooding and
could be typically many times greater than the flood of record. This means that
there are no observed flows in the range, that no flow statistics are available since
no data exists and that rating curves must be extrapolated to estimate flood stage.

27
The timing is another way the dam failure flood wave differs from other runoff
generated floods.

Typically a runoff flood wave will take some time to reach peak flow as water from
upstream and from contributing streams takes time to move through the river
system. In a similar way, water from a typical runoff event will take time to reach
peak flow, though maybe not a long depending on the storm pattern.

However, the flood wave from a dam failure will peak very quickly as a large volume
of water is released in a very short time period. The result is an extremely sharp rise
in the hydrograph followed by the peak flow then a rapid falling hydrograph.

Sharp rise means less time to warn


arn do
downstream
nstream areas
areas. If at all possible
possible, warnings
arnings
should be issued before the dam failure takes place.

28
Besides the extreme timing and magnitude nature of dam failure floods, problems
exist for routing the flood wave downstream. It is difficult to calibrate since no
comparable flows exists. Additionally, the flood wave is influenced by cross stream
structures such as bridges and can have appreciable peak attenuation in low slope
reaches.

All of these factors should be taking into account when simulating a dam failure
flood.

29
As the flood wave moves downstream, the peak flow typically attenuates, or
decreases especially in low slope reaches. Since observed data are not available
for comparisons, hydraulic modeling is needed to estimate the flood wave stage as
it moves downstream.

Physical channel data must be collected in order to model the flood wave at various
locations within the river system. A key piece of information will be the channel
slope as the flood wave attenuation is highly dependent on the channel bottom
slope.

In this example, physical data was collected to model a flood wave moving
downstream. The simulated peak flow was modeled at 5, 13, 23, 31 and 40 miles
downstream from the dam. As shown, the flood wave rapidly
p y decreases to a more
constant flow downstream. Depending on the physical setting, this general pattern
may be observed for most dam failure floods, especially in regions where channels
have small slopes.

30
In a recent study, Dr. Christina Tsai (Sigh) examined the characteristics of various
flood wave models in rivers and streams. She found that Quasi-steady and the
Dynamic wave models can be considered as a series of continually attenuated
successive waves moving between upstream and downstream boundaries. She
also found that the Diffuse wave model gave a better approximation of the full
dynamic wave than the Quasi-steady dynamic wave model. Characteristics of the
various models are shown in the table.

31
This study found that dam-breach flood waves will attenuate to the same peak
discharge at a certain distance downstream called the ‘‘ultimate discharge’’ and the
‘‘ultimate distance’’ respectively. These values depend on bed slope and the
hydrograph volume.

The “ultimate distance” is strongly related to the bed slope. Flood waves traveling in
steep slopes have kinematic wave-like characteristics while flood waves traveling in
mild slopes have diffusive wave-like characteristics and attenuate very fast.

The table shows that for a reservoir of 27,000 cubic meters and a steep constant
channel slope of 0.01, the ultimate distance is 8,000 km while the same reservoir
conditions with a mild channel slope of 0.0001 has an ultimate distance of 7 km.
The same g general results are seen with a larger
g reservoir simulation. The steep
p
sloped channel has a much larger “ultimate distance” than the mild sloped channel.

Of course in a real world situation, channel bed slopes vary considerably. It is


unlikely that a channel with a steep slope of over 37,000 km even exists. However,
the point that steep sloped river channel attenuates a flood wave more slowly is an
import
p factor to remember.

32
Downstream impacts are another important factor to take into account when
considering dam failure flood waves. Population related problems include possible
loss of life and damaged homes.

Infrastructure issues include the possible cascade failure of additional dams, and
blockages from debris. Transportation disruption are possible and related economic
problems as businesses are damaged or closed.

Environmental impacts may include the spread of hazard waste material, decreased
water quality, erosion and sediment deposition, and river channel changes.

33
Model stability, accuracy, and sensitivity section.

34
Another way to view the difference between flood wave models is to compare a
kinematic wave and dynamic wave profiles over time.

The kinematic wave assumes uniform flow and the water surface rises steadily as
one unit.

In the dynamic wave example


example, unsteady
unsteady, non
non-uniform
uniform flow is assumed.
assumed The water
surface shows a distinct wave moving downstream.

35
Hydraulic modeling involves finding solutions to the St. Venant equations. But
because they are composed of non-linear differential equations, stage and
discharge cannot be found directly.

A different strategy is required. The flood wave changes over time at each location
and changes as it moves down the channel. In this case, the end goal is to find the
stage and discharge at any location and at any moment in time.

Using the previous slide flood wave profiles, a grid can be constructed showing the
relation between water surface, depth, and distance along the river. The goal of
modeling is demonstrated in the diagram. The modeling technique must be able to
find stage at cross sections x1, x2 and x3 and at times t1, t2 and t3.

36
Using the previous figure as a guide, it is possible to represent a matrix of cross
section locations and time. To find solutions, a branch of mathematics called
‘numerical analysis’ uses an implicit finite difference method to fill each matrix node.

However, key pieces of additional data are important to this method. These include
the upstream and downstream boundary conditions, as well as the initial conditions.

The Upstream boundary condition requires either a stage hydrograph, or a


discharge hydrograph and rating curve.

The downstream boundary condition requires the known relationship between


discharge and depth or have the depth and time data.

Finally, the initial conditions consist of known stage or known discharge and rating
curve at every location at the start of the simulation. Only this way can the
numerical analysis method succeed.

37
In order to calculate values for each node, the method tries to balance the various
elements within dynamic wave equation.

To estimate the time and spatial partial differential factors, the numerical analysis
uses a weighting factor called “theta”.

Theta can be viewed as a ratio of a smaller distance within time steps to the
distance between time steps. More detailed information on theta and which values
to use is presented on the upcoming “Model weighting factor” slide.

It is important to note that theta is not related to hydraulics and hydrology per se. It
is technique needed for the numeric modeling of the St. Venant partial differential
eq ations
equations.

38
The selection of theta, a numerical analysis parameter, influences the accuracy and
stability of simulations.

When theta is 1, the model simulation is generally stable but will produce the lowest
accuracy results. The modeling strategy is to start with theta as 1 but then
incrementally decrease theta and see if the simulation remains stable.

If so, subsequent simulations are made with smaller theta values until the value of
0.6 is reached. Results produced with a theta of 0.6 are a balance between
accuracy and stability. Theta values of 0.5 or less produce unstable model
simulations and should be avoided.

39
Whenever finding solutions, it is important to remember that the momentum
equations cannot be solved exactly.

Multiple computations are required and trends noted. An initial estimate is needed
as part of the next computational iteration.

The process repeated until results show smaller and smaller changes
changes. Once the
changes fall within a stated tolerance, or when a set number of iterations has taken
place, the computational cycle is stopped.

Ideally, the results will converge to a particular value . There are two cases where
the results do not converge to a solution. The first is when the results oscillate with
a range bebeyond
ond a set tolerance
tolerance. The other is when
hen a sol
solution
tion “blo
“blowss up”
p” and res
results
lts
becomes useless.

40
Here is a case where multiple computations are made as part of a hydraulic
modeling simulation. The goal is to find the stage at a point in time at a specific
location within a river system.

Each successive iteration takes the results from the previous calculation as part of
the modeling process. In this generalized example, the differences between each
iteration become smaller. We can say the iterations are converging to a solution.

Usually once the differences are within a range, such as 0.03 feet or 1 millimeter,
the calculations stop.

41
However, there are times when the values do not converge.

In this example the computed value oscillate around a range of stage values but are
not getting smaller despite the continued iterations.

Such an example does not produce useful results. Factors used in a numerical
analysis such as time step and cross section locations
analysis, locations, should be re
re-examined.
examined
These factors are discussed later in this module.

42
Finally, sometime results may “blow up” and give wildly inappropriate stage
estimates.

In extreme cases, the result will go to plus of minus infinity. Here the model is
unstable and produces unusable results.

In order to reduce such problems


problems, the next few slides give broad guidance when
selecting parameters for a model run. The suggestions are not software specific and
would generally apply to any hydraulic modeling application.

43
It is important to select the right time step for your simulation. Here are some
general guidelines to consider. One common guide is for initial time step settings to
be about 5 percent of the expected rising limb time.

If the hydrograph rises in 30 minutes, the initial time step should be about 1.5
minutes. For modeling areas near structures, like bridges, or regions with quick
changes, a time step of 1 to 5 minutes would be a good initial setting.

Finally for areas with extreme changes, such as just below a dam failure, a time
step between 1 to 60 seconds may be appropriate.

44
As mentioned earlier, one of the most important factors in modeling unsteady flow is
selecting the correct time step. In order to have a useable time step, something
known as the Courant Condition should be met.

The condition simply states that the time step should not be more than the reach
length divided by the sum of the wave velocity and the average water velocity.

Problems that can occur if the time step that’s too small include possible model
instabilities, the leading edge of the wave may be computed to be more steep than
is actually the case, and longer computational processing time.

A time step longer than this means the flood wave may have passed through a
reach within
ithin one time step
step.

Other possible problems include numerical instabilities, diffusion of the wave peak,
and the true rise and fall of the hydrograph may not be captured.

45
Along with time steps, the placements of cross section spacing is another important
factor. Typically cross section spacing should be between 100 to 1000 feet.

The exact spacing will depend on physical setting factors such as channel slope,
roughness, discharge, and the extent of the floodplain.

In general,
general shorter cross section spacing is required when bracketing lateral inflow
locations and hydraulic structures, areas with abrupt slope changes, river bends or
where determining the stage is critical.

Situations allowing longer spacing between cross sections include reaches with a
consistent slope, consistent channel roughness and isolated areas were critical flow
e ists
exists.

46
Knowing the causes of instability before modeling a river system can save time and
effort.

Listed here are some common problems that others have found. Take these items
into consideration before starting a project.

If the time interval is too large


large, the result may be rapidly changing hydraulic
properties.

If too few cross sections are used, then the model may have large changes in
hydraulic properties over distance. This can lead to instability.

Whenever critical flow occurs, it may help to enable a mixed flow computational
regime and to increase roughness coefficients when appropriate.

47
Issues such as bad boundary conditions will almost always cause problems.

Be sure to check that rating curve covers full flow range and ensure the channel
bottom isn’t too steep. Another problem area to avoid is bad cross section
properties.

Finally, avoiding sharp changes in hydraulic properties will help with the stability of a
Finally
simulation.

48
The dam breach process begins as the structure begins to fail. The process can be thought
as a series of steps. The first is the breach development.

Breach development and failure time are important parameters affecting the outflow
hydrograph. When routing a dam-break flood wave, the breach development begins when
a dam is compromised and failure is imminent. The breach is considered fully developed
when it stops increasing in size. In some cases this may be the time of peak outflow, but
for smaller reservoirs the peak outflow could occur before the breach is fully developed
developed.

In an overtopped dam, the crest of the dam will erode away and a growing gap will form.
The shape of the gap is approximated by an initial square or rectangle then taking on a
trapezoidal shape.
How long the breach takes from initial formation to a final shape is the time of failure. This
can happen
pp in less than one hour or could take several hours.

Remember that any breach formation and time of failure calculations provide an estimate
only. No matter the equation, the slower a breach forms, the lower the flood wave peak.
This is because a slower formed breach allows water to leave the reservoir at a slower rate.
A breach that forms quickly allows water to pass though the opening much quicker.

The link to the Wahl report shows multiple regression equations have been developed to
estimate breach time of failure.

49
The NWS model “BREACH” uses various inputs such as the size, shape, rate of
formation to estimate the outflow hydrograph from a breached earthen dam.

The physically based calculations take into account basic hydraulic principles,
sediment transport theory, soil mechanics, a dam’s physical shape and material
along with reservoir properties. Below is a link to a 1988 Danny Fread paper on the
NWS BREACH model.

50
Here is a comparison of properties for steady and unsteady flow. The algorithms for
steady flow are applied river section by river section using an iterative convergence
approach. Unsteady flow uses a matrix solution where each time step is solved
simultaneously.

For boundary conditions, steady flow simply needs the discharge for each section.
Unsteady flow requires an inflow hydrograph for routing the water downstream.

Non-flow or ineffective flow areas, areas where ponded water may occur, are not
included as part of the active flow area for steady flow. For unsteady flow
conditions, all areas containing water must be taken into account whether or not the
water is moving.

The calibration target for steady flow is water surface elevation while the target for
unsteady flow is not only the water surface elevation but also the timing, hydrograph
shape and flow distribution.

Methods used for calibration under steady flow conditions include adjusting the
M
Manning’s
i ’ n channel
h l roughness.
h U
Unsteady
t d flflow iincludes
l d adjusting
dj ti ththe M
Manning’s
i ’ n
channel roughness but could also include adjusting flow volume.

Finally, steady flow has no ∂t term since flow doesn’t change over time while
unsteady flow includes the ∂t term as flow is changing over time.
51
Dam failure applications section

52
This discussion of dambreak applications will focus on three specific models, HEC-
RAS, the Hydrologic Engineering Center’s River Analysis System from the US Army
Corps of Engineers, the NWS Simple Dambreak program, and some commonly
cited Rules of Thumb.

A brief overview, data input, boundary conditions, and some advantages and
disadvantages of each are highlighted.

53
The HEC-RAS program is a full developed, fully support application for the US Army
Corps of Engineers.

It calculates one dimensional steady, unsteady and mixed flows and has the ability
to model channel networks, account for hydraulic structures and scour, perform
sediment transport-mobile bed modeling and water temperature analysis.

For more complex river systems, the program can model dam and levee failure,
effects of pump stations, simulate navigational dam operations as well as
pressurized pipe systems.

A link to the HEC-RAS homepage is listed below.

54
As with any model using numerical analysis, HEC-RAS requires boundary
conditions to be specified. In this case the upstream location needs flow or stage,
the downstream location needs a rating curve, normal depth and flow or stage. Any
internal boundary conditions need to specify lateral inflow, groundwater interflow,
and any off-channel flow control structure operation.

Data input requires extensive physical information such as basic channel geometry
and cross section data, the reach length, estimated energy loss coefficients,
Manning n channel roughness values, any stream junction information, hydraulic
structure information for items such as bridges, spillways, weirs, culverts, or flow
control structures.

55
Advantages of the HEC-RAS application includes robustness, highly accurate water
surface calculations, public domain availability, peer-reviewed status, along with fee-
based full technical support. The program itself is available free of charge, and is
broadly used in the private sector especially in the civil engineering community.
Extensive documentation is also included.

Limitations of this application include restrictions to modeling of one dimensional


flow, steady or gradually varied flow condition where the channel slope 10 ft per 1
mile or less. The river channels must have a stable bank without backwater. All of
the energy losses must be defined and an assumption that most of bridge energy
losses are associated with piers. Any culverts are assumed to have constant shape,
flow rate, and bottom slope.

56
The SIMPLE dam break model is an NWS product. The latest version is 2000.

This is a specialized program that estimated downstream flooding from a dam


failure.
Data input includes the dam crest height, storage volume of the reservoir, the
maximum spillway capacity, the reservoir surface area, estimated time to fail and an
estimated breach width
width.

57
The Simple Dambreak program has a single boundary condition – the reservoir pool
elevation or dam height.

Advantages of this program include Less physical data, less time to produce model
run and less hydraulic expertise required.

However, since this is simplification of a more complex program


However program, there are
limitations. First is the predicted flow and travel time. As compared to more
complex program, the results from Simple Dam Break may have 10 to 20 percent
error in the results. The number of cross sections is limited to 50, the default values
for physical properties may not match the actual “on-the-ground” conditions. Finally,
this model does not take into account backwater effects.

58
There is no source agency for the Rules of Thumb. The rules are strictly an
empirical method to quickly estimate flood stages from a dam failure. Data input in
simply the dam height or pool height and the general setting for the river channel.

Advantages of the method include no need for extensive physical data and that no
expertise is needed. Limitations are that results are not based on a physical model
and that results provide only very general estimates of stage.

59
The Rules of Thumb are a set of generalized guidelines that are used when the
situation does not permit more involved hydraulic modeling. The initial wave height
is assumed to be one half of the pool height or dam height immediately below the
dam failure.

Flood wave speed is dependent on the channel slope. For normal to shallow sloping
channels, the flood wave speed is 3 to 4 miles per hour. In stepper slope areas,
such as in a foothill area, the flood wave speed may be between 5 to 7 miles per
hour. In the steepest slope areas, such as mountain streams, the flood wave speed
is estimated to be between 8 to 10 miles per hour.

As the flood wave moves downstream, the estimated wave height attenuation is
about half for every
y ten miles of travel.

60
The HEC-RAS program is a fully developed, robust hydraulic model that is
appropriate for most streamflow situations. However, it requires extensive physical
data.

The Simple Dam Break model is a single purpose tool to estimate dam failure flood
stages. It is used when minimal physical data are available and makes multiple
simplifying assumptions.

The Rules of thumb is also a single purpose tool and roughly estimates dam failure
flood stages when physical data and time to simulation the flood flow are not
available.

The HEC
HEC-RAS
RAS model wouldo ld gi
give
e the most accurate
acc rate results
res lts when
hen the river
ri er system
s stem is
properly calibrated, while the Simple Dam Break and Rules of Thumb will give less
accurate general guidance.

61
Model data section

62
Cross section data is required for most models.

Attributes include the channel geometry, overbank flow areas, any levees, all off
channel storage areas, the location of hydraulic structures such as bridges, any
lateral inflow and outflow sites along with the hydraulic properties of these items.

63
This detailed example of a cross section shows how the main channel and any
secondary channels may be present within the data.

Also the proper placement of channel cross sections is shown in the inset. At no
time should cross sections overlap. Note that cross sections always cross
perpendicular to the flow of the river.

64
Off channel storage areas are locations where river water can be diverted, or can
be sources of water released into a river.

The connections between a river and storage areas or between multiple storage
areas and a river system will depend on the specific situation.

Water movement usually depends on the river stage and the structure elevation
elevation.

The effects on a flood wave would be similar to those to overbank storage where
water is removed from the rising limb of the hydrograph.

65
Multiple sources for hydraulic modeling and model results exist. Dam Emergency
Operations Plans may be available from the U.S. Army Corps of Engineers, Bureau of
Reclamation, various State and county agencies or even dam owners or operators.

The U.S. Geological Survey maintains the National Elevation Dataset as well as river
stages for locations around the country.

NOAA’s National Environmental Satellite, Data and Information Service – National


Geophysical Data Center archives flood information from multiple agencies and
organizations.

Other modeling and flood related information is available from academic sources such as
the University Corporation for Atmospheric Research, UCAR.

66
A simulated flood wave gives an opportunity to compare the results from the fully
developed HEC-RAS model to the Rules of Thumb.

The figure of the Bald Eagle Creek dam failure shows the maximum stage for all
points along the river, regardless of the time peak. The dam is located 16 miles
above the confluence of Bald Eagle Creek with a larger river.

Below the dam, the depths estimated with the Rules of Thumb are consistently
higher than the depths computed with the HEC-RAS model. In this case, the Rules
of Thumb overestimates by 15 to 20 feet for locations near the dam. Further
downstream we find the opposite case. Here the Rules of Thumb depths are
consistently lower than the depths computed by HEC-RAS.

Here the Rules of Thumb underestimate the flood depths by over 20 feet at the
confluence, or at the zero x-coordinate.

The lessons to take away from this example are 1 - that a full hydraulic model of a
dam failure is best and 2 – the Rules of Thumb can significantly over or
underestimate
d ti t flood
fl d wave ddepths.
th

67
Breach formation and flood wave attenuation are affected by breach time. The
slower a breach forms, the lower the peak flood that occurs.

However an important fact to remember is that the degree of flood wave attenuation
decreases as the wave moves downstream. The figure shows three simulations of a
dam failure using various breach formation times. For quicker breach failures, the
higher the peak flow.

Yet all three simulation attenuate to a set discharge downstream. Attenuation near
the failure site is dependent on the channel slope. Mild slopes attenuate the flood
wave faster while steep channel slopes attenuate the flood wave more slowly.

Since the R
Rules
les of Th
Thumb
mb do not take into acco
accountnt channel slope for flood wave
a e
attenuation, it is expected it will overestimate flood stage for mild sloped river
sections and underestimate flood stage for steep sloped river sections.

68
Dam failure case study

69
The Teton River Dam failure is a famous case that has been studied extensively. In
this case study rather than examine multiple simulations from numerous studies, a
timeline of events is presented.

The dam was located in eastern Idaho and was justified on the basis of irrigation,
power production, recreation, mitigation of project-caused loss of fish & wildlife, and
ironically flood control. While the project was authorized in 1964, monies were not
allocated until the 1970s. Actual construction did not begin until 1973.

70
Here we see the location of the dam is in extreme eastern Idaho on the Teton River.

The Teton River which flows into the Henry’s Fork River, which in turns flows in the
Snake River.

The town of Rexburg, population about 17,000, is located along the Teton River
approximately 9 miles downstream of the dam site
site.

71
The dam was sufficiently constructed to begin filling the reservoir in October, 1975.

One month later, the dam was essentially completed at a height of 305 feet. In
January generators were installed.

As the spring snowmelt began, permission was received to fill the reservoir at a two
feet per day rate
rate.

In April spillway gate assembly begin and by June, the reservoir contained some
234,000 acre-feet of water. The stage for catastrophic events was now set.

72
On June third, personnel noticed two small springs on the right side of the dam. The
estimate flow was 0.09 cfs. This wasn’t necessarily worrisome as earthen dams can
show wet spots and there were no problems. The next day, June fourth, another
spring was noticed below the bottom, or toe, of the dam. The flow from this new
spring was about 0.05 cfs, than the previous two springs.

The morning of June fifth brought new concerns as at 7:30 a.m. muddy water was
found flowing, at about 20 cfs, from the rock near the right abutment of the dam.
This was significantly more than previously seen. There were two important
differences with this new spring.

First was the flow rate, 20 cfs. The other was that it was muddy water, indicating
active erosion was taking gp
place. An hour and a half later another leak with a flow of
2 cfs was found below the dam. By 9:30 a.m. a downstream spot was found that
began to flow and erode. Bulldozers tried to fill the hole but efforts were abandoned
when one bulldozer was captured by the growing hole.

At 10:43 a.m. the local sheriff’s department was notified that it should “prepare to
evacuate”. Byy 11:30 a.m. the sheriff's department
p was told to begin
g evacuations.
Local radio stations broadcast the alert to downstream residents.

73
At 11:55 am, the dam was breached. The estimate peak outflow was 2.6 million cfs.
The powerhouse was washed away along with approximately 4 million cubic yards
of material.

Residents of the small towns of Sugar City, Teton, and Newdale were evacuated
around 12:30 p.m. At Sugar City, about 7 miles downstream, a 15 foot wave at 1
pm. The flood wave continued downstream to Rexburg, arriving at 2:45 pm.

Because of the volume of water, from June sixth to ninth, the US Bureau of
Reclamation began releasing water from the largest main stem Snake River
reservoir, American Falls, in order to create enough storage to capture to incoming
flood waters. This effort was successful as flood waters slowed on June ninth.

The estimated damage included 14 deaths, 8,000 homes destroyed and


approximately a half a billion dollars in damages.

74
Models of the Teton Dam failure show some interesting characteristics. The graph
shows the distance downstream versus the flood peak travel time. The slope of the
line indicates the travel speed. Comparison slopes of 1, 2, 3, and 10 miles per hour
are shown for comparison.

Initially, the peak flood travel time was very fast as the curve is nearly flat, indicating
a change in distance over very little time. However, as the flood wave moved
downstream, the curve is much closer to the 10 mile per hour slope. For a majority
of the time, the flood peak is traveling between 1 and 2 miles per hour.

Not until the end of the simulation , near 60 miles from the dam, does the speed
change back to a value closer to 10 miles per hour.

75
A series of photographs available from the Bureau of Reclamation are shown here.
The left photo taken about 9 a.m. and shows the resulting of internal piping. The
right photo shows the increased piping and external erosion as time passed. For
scale, bulldozers are circled and look like ants on top of the dam.

76
The left photograph shows continued piping and erosion. Note that the erosion is
progressing towards the top of the dam. The right photograph taken about 11:55 am
shows the moment as the dam is breached. Failure of the dam was assured before
this point. The only question was how fast the entire failure process would take and
how long it would take to empty the reservoir.

77
The left photograph shows the water flowing through the breach in an uncontrolled
manner.
Because of the sediment mixed with the water, flood waters are brown. The right
photograph shows the breach continuing to enlarge and the reservoir surface is
seen through the breach.

78
This photograph shows the fully developed breach and the reservoir waters flowing
through the gap. Initial peak flow was estimated at 2,600,000 cfs. It took several
hours for the reservoir to empty.

79
In summary, this module has presented material and topics related to on the dam
failure process with a generalized failure timeline and steps involved with modeling.

An overview of the Saint-Venant equations was provided that included explanation


of terms within the conservation of mass and momentum equations, assumptions
and physical setting, the various wave equations, and situation where the different
wave equations would be appropriate.

This module also addressed dam failure flood wave characteristics such as
magnitude, timing, and routing along with wave attenuation and downstream effects.

80
Also discussed were sections dealing with hydraulic modeling, numeric analysis,
guidance when selecting a model time step, stability issues, cross section
information, and a generalized breach process.

An overview and comparison of the HEC-RAS model, Simple Dambreak and Rules
of Thumb was provided.

Finally, a review of the Teton Dam Failure was included.

Once both module 1 and module 2 are completed, the student will have the
scientific foundation for advanced course work needed to run dam break simulations
and conduct hydraulic modeling as a part of dynamic wave forecasting.

81
82
83

You might also like