CHAPTER 1

Simplified models for flood flow simulation

Flavio C.B. Mascarenhas & Marcelo G. Miguez

Federal University of Rio de Janeiro, Rio de Janeiro, Brazil.

Abstract

Adequate representation of space dimensions in mathematical models for flood

simulations has been a major challenge for hydrologists, hydraulic engineers and
general scientists worldwide. Indeed, the natural and real phenomena are three-
dimensional and time-dependent; however, there are particular situations where
space dimensional simplifications are valid for practical purposes. For instance,
hydrological models are based on a scalar conservation law; thereby they often
do not consider a particular flow direction but the flow direction itself, even in
a simplified way. By contrast, one-dimensional (complete or simplified) and
pseudo-two-dimensional models can be adequate in most cases, mainly the one-
dimensional models applied to rural areas where the main drainage is carried by
a large river with minor tributaries. Pseudo-two-dimensional models can provide
reliable results when the main flow occurs in a horizontal plane, with compartments
connected to their neighbours through one-dimensional hydraulic laws. Rigorous
two- and three-dimensional models may apply to few special cases, and gener-
ally their boundary conditions are not easily specified. Recent studies have shown
that, in some cases, dimensional reliability has been overestimated, since simpler
models can simulate the major aspects of flood flow. This chapter will present
a brief review of some complete and simplified models, their dimensionality or
negligence and their applicability to achieve reliable results for flood flow and
flood control, both in rural and urban areas, and show that dimensionality should
be considered carefully by engineers and not taken as a rigorous mathematical
entity in all cases.

1.1 Introduction

Mapping and tracking the movement of flood waves is an important task to minimize
risks to population of hazard-prone areas. In hydraulic engineering, this study is

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
doi:10.2495/978-1-84564-560-1/01
2 Flood Prevention and Remediation

known as flood routing. There are many suitable flood-routing methods or models,
from the simplest ones to the more complex. In our opinion, the main difference is
the dimensionality involved in each model. The hydrologists are very familiar with
the general water balance model, where the only dimension involved is the time.
Therefore, this model states the following:

(Inflow rate in a control volume) − (Outerflow rate from a control volume) =

(Time variation of volume inside the control volume) (1.1)

dS
I −O= (1.2)
dt

where I and O are the inflow and outflow rates from a control volume, respectively;
S is the stored volume and t is the time; the term in the right-hand-side of eqn (1.2)
may be a simple relationship to represent a linear reservoir response, or another
relationship for hydrological models such as Muskingum and S.S.A.R.R. Equation
(1.2) is often called as the concentrated form of the continuity equation. It is impor-
tant to mention that this general model should not be confused with rainfall-runoff
models. In that model, without dynamic effects, it is not appropriated to attribute a
main flow direction, with the exception of Muskingum–Cunge model [1].
There are not geometrical (or Cartesian) dimensions in the equation above, i.e.
the dimensionality was theoretically reduced to zero (zero-dimensional model).
A simple way to solve complicated flood flow mathematical models is to reduce
them by one-dimension. This is a common feature in more sophisticated models /
methods such as boundary element method. What is added when the method
performs only discretization on the boundary of the problem domain is that dimen-
sionality is reduced by one, and the mathematics becomes more simple and easier
to deal with because, as is well known, surface integrals are more adequate with
certain approaches than volume integrals. In fact, as we will see, any discretizing
process, such as finite-differences, has a trend to reduce dimensionality. The
rigorous two-dimensional mathematical models are seldom used in flood flow,
although some dam-break mathematical models follow this idea. They are more
used in open water bodies subjected to currents and tides, e.g. in costal engineer-
ing. Dam-break flows have their particularities, despite they can be described
through one-dimensional models. In the next sections, we will briefly present the
physical conservation laws which are the basis for mathematical models of flood
flow and flood control, emphasizing scalar and vector laws. The latter give rise to
Cartesian directions in the corresponding mathematical models for flood routing.

1.2 The scalar law of mass conservation

The general statement of mass conservation is that mass can neither be created
nor be destroyed in a closed system. No mention to any Cartesian dimension

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 3
is done. Some engineers confuse mathematical formulations of this law with
three-dimensions. Assume a general velocity fluid field ν (vector) with compo-
nents on the Cartesian directions x, y and z. Now, introducing the vector gradient
∇, the mass conservation equation is written for incompressible flow, known as
continuity equation:

∇ ⋅n = 0 (1.3)

where

∂ ∂ ∂
∇= i+ j+ k (1.4)
∂x ∂y ∂z

Obviously, the product of eqn (1.3) gives a scalar result, hence confirming that
the mass conservation law is scalar as expected. In eqns (1.3) and (1.4), Cartesian
directions give an illusion of dimensionality, except when the flow is potential
(and non-rotational). Therefore, the equation above becomes the Laplace equation,
which is used to describe standing waves without close solid boundaries (models
2-DH or 2-DV), for circulation problems in shallow water bodies, in two- or three-
dimensions, where it appears in the vortex function. In fact, some researchers state
that close solid boundaries invalidate the non-rotational hypotheses. However,
those are not clearly the case of flood wave movement, which is considered as
quasi-non-rotational, and therefore does not apply to potential flow. Besides, the
flood waves are not standing waves, but progressive ones with net mass transport.
In the first case, besides the Laplace equation, we have

curl x n = 0 (1.5)

The potential equation, in terms of vorticity, can be derived from eqn (1.5), which
is useful for some water resources problems, but obviously not for flood problems.
At this point, it is very important to mention that continuity equation used alone
to describe flood flows is essentially scalar, and any attempt to attribute some
Cartesian direction to it is a simplification in flood flow problems and in such cases
is false and a great misunderstanding.
In the next section, we will present the vector momentum conservation law and
its main forms, which may explain dimensionality, although the three Cartesian
directions do not need to be applied to water resources problems, mainly for flood
flow problems.
For river hydraulics representation, there is a preferential flow direction, which
is simply the length of the control volume used for the mass conservation law
application. Applying this law to a control volume in the x direction (Fig. 1.1), the
following distributed continuity equation for rivers is obtained:

∂Q ∂A
+ =0 (1.6)
∂x ∂t

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
4 Flood Prevention and Remediation

Figure 1.1: Control volume used for transforming eqn (1.2).

where Q is the liquid discharge and A is the wetted area. Instead of introducing a
Cartesian direction in a scalar conservation law, the above equation indicates
the main direction of the flow, or more generally the length of the control
volume.
Indeed, if one tries to express flow velocity instead of discharge, the velocity is
the scalar component in the x direction, called u, and therefore all the above equa-
tions are scalar as cited before.
Assuming Z, also a scalar quantity, the water surface elevation, an alternate
form for the continuity equation for floods in rivers becomes

∂Q ∂Z
+B =0 (1.7)
∂x ∂t

where B is the water surface flow width.

For fixed bottom problems, it is easy to verify that the above equation is equiv-
alent to

∂Q ∂h
+B =0 (1.8)
∂x ∂t
where h (also scalar) is the mean flow depth. Now introducing q, the discharge per
unit width, eqn (1.8) turns into

∂q ∂h
+ =0 (1.9)
∂x ∂t
This is sometimes useful for simulating floods in rectangular channels.
An alternative form for eqn (1.8) can be obtained expanding the term:

∂Q ∂ ( Au )
= (1.10)
∂x ∂x

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 5
Leading to the special and still scalar form

∂u ∂A ∂h
A +u +B =0 (1.11)
∂x ∂x ∂t

The three terms above are known as prism storage, wedge storage and rate of
rise, respectively, and have an important significance in the so-called Muskingum
(hydrological) method for flood flow simulation in rivers. The model’s name
comes from the name of river where it was first applied. The general sketch of this
model is presented in Figs 1.2 and 1.3. This method has been applied with success
in some cases of flood propagation in rivers [1].

Figure 1.2: Prism and wedge storages in a river reach.

I(t)
O (t)

Control volume or element

Upstream section
Downstream section

Figure 1.3: Control element for the Muskingum model.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
6 Flood Prevention and Remediation

1.3 Momentum vector conservation law

1.3.1 Some generalities about systems governed by vector laws

Physical systems, such as the ones figuring floods in a basin, are obviously
time-dependent and three-dimensional. After applying the relevant physical
conservation laws, it is possible to establish a mathematical translation for the phys-
ical law, relating dependent and independent variables. The referred translation is
basically what differs mathematical from physical reduced scale models (despite
the fact that physical models often are based on dimensionless scale numbers, such
as Froude number), where there is no need of vector representation, since they are
always scale of the prototype, and the phenomenon itself is reproduced almost like
reality.
In mathematical models, we can observe two types of simplifications: reduc-
tion of dimensionality, as mentioned before, and neglecting some terms of the
mathematical translation of the vector conservation law, what can be done for
engineering convenience. The difference in some cases may be subtle; however,
the reader will be able to clearly identify each one, as we will treat in the next
section. A useful hint to distinguish the simplifications is that dimensionality
reduction is done before the application of the physical law. Engineering conve-
nience is a simplification performed directly on the equation resultant of the
physical law application, e.g. after the application of the physical law. Generally,
these two kinds of simplifications are related to vector physical conservation law.
However, a special care should be taken in the choice of some approximations;
e.g. some important features of the physical phenomenon may be ignored if the
simplifications adopted are inconsistent. In other words, the engineer or the mod-
eller must have a hydraulics expertise support in order to avoid the development
and employment of weak mathematical black boxes, e.g. false mathematical
models for a particular case.
For example, the use of one-dimensional Saint-Venant equations to describe a
flood flow comes clearly from a dimensionality reduction, while the choice of a
kinematic wave or diffusion analogy model is obviously an engineering conve-
nience. Engineering convenience, however, should never be seen as a choice
because of lack of data or necessity of a fast result. These simplifications may lead
to inconsistent models. Simplifications for engineering convenience are related to
physical identification of non-relevant parts of the process.

1.3.2 One-dimensional dynamic equation

The dynamic equation for flood flow results from the application of the vector law
for (linear) momentum conservation (Newton’s second law) to a one-dimensional
flow control volume. The general law states that

d
F= P (1.12)
dt

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 7
where F is the resultant force acting on the control volume and P is the linear
momentum.
If we consider only the x dimension, the above equation obviously reduces to

d
Fx = P (1.13)
dt x
where

Px = m u (1.14)

Indeed, m is the mass and u is the x component of the flow velocity.

Considering now the one-dimensional control volume shown in Figs 1.4 and
1.5, the application of eqn (1.14), after some manipulation, results in the one-
dimensional flow equation [2]:

∂Q ∂ ⎛ Q 2 ⎞ ∂h
+ b
∂t ∂x ⎜⎝ A ⎟⎠
+ gA
∂x
= gA S0 −S f ( ) (1.15)

where

∂Z
S0 = − (1.16)
∂x

n2 (Q / A )
2
Sf = (1.17)
( A / P )4 / 3

( Fp1 )x1
u
h
u + ∂u dx
Fg ∂x
δ
(F p1 )x 2
h + ∂h dx
Ff ∂x
δ
x
x1 x = x1 + dx

Figure 1.4: One-dimensional control volume, longitudinal view.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
8 Flood Prevention and Remediation

z
B(ξ )

A
h-ξ
h

Z dA dξ
ξ

Z0
y

Figure 1.5: One-dimensional control volume, transverse view.

where P is the wetted perimeter, β is the Boussinesq coefficient and n is the

Manning roughness coefficient.
The Boussinesq coefficient is introduced to correct the assumption of the mean
velocity u at a cross-flow section in the momentum equation, and is given by

∫A u
2
dA
b= 2
(1.18)
u A

This coefficient must not be confused with the Coriolis coefficient α, which is
adopted to correct the kinetic charge in the energy (Bernoulli) equation

∫A u dA
3

a= (1.19)
u3 A
It must be clear that when one chooses the set of Saint-Venant equations (1.8) and
(1.15) for flood flow, the dimensionality reduction is assumed and it is irreversible.
This is called complete one-dimensional model for flood flow. Other simplifi-
cations, related to engineering conveniences, can still be done and are perfectly
reversible, as they basically consist in neglecting some terms in the dynamic equa-
tion (1.15). We will deal with some of them in the next subsection.
At a first time, the change of dependent variables in the Saint-Venant
equations may be looked as simple algebraic manipulations. However, these
manipulations should be used carefully as some important conservation proper-
ties may be lost when they are carried out. Such manipulations will be done in
the next section, because some properties of the equations appear more clearly
with other dependent variables and simplified Saint-Venant models become
easier to understand.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 9
1.3.3 Simplified one-dimensional models for flood flow

The partial differential equations system is most often used by engineers because it
may be interpreted physically in a simple and natural way. This system corresponds
to Saint-Venant equations with dependent variables Z, u, and A, written for pris-
matic channels [2], and the original system presented by Saint-Venant in 1871 [3]:

∂Z 1 ∂ (uA )
+ =0 (1.20)
∂t B ∂x

1 ∂u u ∂u ∂Z
+ + + Sf = 0 (1.21)
g ∂t g ∂x ∂x

All terms in eqn (1.21) can be considered as a slope. The first two terms, being
‘u’ a uniformly distributed velocity over the cross section, are the inertia terms
or acceleration slopes. The first term represents the slope of the energy grade
line due to the velocity variations in time (acceleration). The second term is the
slope corresponding to the velocity variations (u2/2g) (in steady flow) in space.
The third term is the slope of the water surface. Finally, the fourth represents the
slope due to the resistance to friction forces opposed to the flow, called friction
slope.
The terms in eqn (1.21) are known in hydraulic engineering and have relative
importance in different flow situations. For example, when the only interest is the
global development of a flood in steep rivers, the acceleration terms can be
neglected; and also when ∂h / ∂x is lower than bed slope, the dynamic equation
(1.21) is simplified to Sf – S0 = 0, which leads to the simplest form of the kinematic
wave model. Another kinematic wave model form is obtained as follows:

Q = K (h ) S0 (1.22)

where K is the cross-section conveyance.

Equation (1.22) implies that Q = Q (A) and A = A (Q), or the presence of a
single-valued relationship between the wetted area and the discharge at a given
point x = x0. Therefore, the first term of the continuity equation (1.20) may be
rewritten as follows:

∂A ⎛ dA ⎞ ∂Q
= (1.23)
∂t ⎜⎝ dQ ⎟⎠ x ∂t
0

and then

∂Q ⎛ dQ ⎞ ∂Q
+ =0 (1.24)
∂t ⎜⎝ dA ⎟⎠ x ∂x
0

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
10 Flood Prevention and Remediation

Equation (1.24) is one of the practical forms of the kinematic wave model. This
model gives good results if the hypotheses are valid for a particular flow, which is
decided by engineers and scientists involved with the problem.
Now, it will be considered only the effect of neglecting the inertia terms. When
the first two terms of eqn (1.21) are dropped, the following equation is obtained:

∂h ∂Z 0 Q Q
+ + =0 (1.25)
∂x ∂x K2

Now, assuming that B is constant and differentiating eqns (1.20) and (1.25) with
respect to x and t, respectively, the following system is obtained:

∂2 h 1 ∂2 Q
+ =0 (1.26)
∂x ∂t B ∂x 2

∂2 h 2 Q ∂Q 2Q Q dK
+ 2 − =0 (1.27)
∂x ∂t K ∂t K 3 ∂t
Putting

∂K dK ∂h dK ⎛ 1 ∂Q ⎞
= = ⎜⎝ − B ∂x ⎟⎠ (1.28)
∂t dh ∂t dh

And eliminating ∂2h /(∂x∂t) from the two equations, we obtain

1 ∂2Q 2 Q ∂Q 2Q Q dK ∂Q
− + 2 + =0 (1.29)
B ∂x 2 K ∂t BK 3 dh ∂x

Or, by rewriting

∂Q ⎛ Q dK ⎞ ∂Q K 2 ∂Q 2
+⎜ − =0 (1.30)
∂t ⎝ BK dh ⎟⎠ ∂x 2 B Q ∂x 2

Equation (1.30) is a classical parabolic partial differential convection–diffusion

equation with one dependent variable Q(x, t). Thus, this quantity discharge is
convected with the velocity (Q / BK) (dK / dh) and diffused with a diffusion
coefficient K2/(2B|Q|). Therefore, if inertia terms are, in fact, negligible,
eqn (1.30) is a good model for flood propagation, and the associated flood routing
model is called the diffusion analogy model [2]. The successful application of
this model is subjected to fulfilling the hypotheses, what is verified by engineers
and scientists.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 11
1.3.4 Two-dimensional model for flood flow

Remembering that two-dimensional models only differ from one-dimensional ones

because the momentum conservation law was applied in two Cartesian directions,
there would not be any difficulty to formulate them.
According to Abbott [4], and neglecting Coriolis effects, the Saint-Venant equa-
tions in two-dimensions may be written as

∂h ∂ ∂
+ (uh ) + (vh ) = 0 (Continuity) (1.31)
∂t ∂x ∂y

( )
1/ 2
∂ ∂ 2 ∂ ∂ gu u2 + v 2
∂t
( )
uh +
∂x
u h + ( )
gh
∂x
(h + Z 0) +
∂y
( )
uvh +
C2
=0

(1.32)

( )
1/ 2
∂ ∂ 2 ∂ ∂ gv u2 + v 2
∂t
(vh ) +
∂y
( )
v h + gh
∂y
(h + Z0 ) + ∂x (uvh ) +
C2
=0

(1.33)
Equations (1.32) and (1.33) are the momentum (vector) equations in x and y
directions (plane), u and v are the velocity flow components in those directions
and C is the Chezy coefficient.
Besides the fact that it is not a simple matter to establish the correct boundary
conditions to the two-dimensional Saint-Venant equations set for a particular
problem (except to linearized), it remains an additional difficulty to associate
different and determined ‘bottom slopes’ as follows:

∂Z 0 ∂Z 0
S0 x = − S0 y = − (1.34)
∂x ∂y

Therefore, under such possible imprecision and difficulties, simpler models may
alternatively be employed, as will be seen in Section 1.6.
Two-dimensional models for varied unsteady flow represent flood movement on
a river basin, at the horizontal plan, taking into account equations written in two
orthogonal flow directions. Considering the occurrence of gradually varied flow,
two-dimensional flood routing equations can be obtained by analogy with shallow
waters equations, which result from vertical integration of the Navier–Stokes equa-
tions. In this representation, the term of resistance is deduced from Manning or
Chézy equations. The resultant system consists of three equations: one referring to
the mass conservation and two related to the momentum conservation in the
Cartesian plane directions. Using the same hypotheses adopted in the one-
dimensional flood flow modelling, for the Saint-Venant equations, and also

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
12 Flood Prevention and Remediation

disregarding Coriolis acceleration and the influence of the wind at surface waters,
which could appear in the two-dimensional movement, the following equations can
be written:
Continuity equation:

∂h ∂ (uh ) ∂ (vh )
+ + =0 (1.35)
∂t ∂x ∂y

Dynamic equation in the x-direction:

∂u ∂u ∂u ∂ (Z 0 + h ) n2
+u +v +g = − 4 3 u u2 + v 2 (1.36)
∂t ∂x ∂y ∂x R
Dynamic equation in the y-direction:

∂v ∂v ∂v
+u +v +g
∂ Zf + h n2 (
= − 4 3 v u2 + v 2
) (1.37)
∂t ∂x ∂y ∂y R

where h is the flow depth, u is the flow velocity on the x-direction, v is the flow
velocity on the y-direction, Z0 is the channel bottom level, n is the Manning friction
factor, and R is the hydraulic radius.
In terms of river basin flow, this model type allows the representation of large
inundated areas, where flows occur freely in any direction, with depths sufficient
to form a continuous sheet of water over flooded surfaces.

1.4 Some gaps between one-dimensional dynamic models

and reality
In order to support some ideas, we begin this section with few words based on
somewhat philosophical thoughts on real physical processes and one-dimensional
models approach.
Generally, models see what the modeller wants them to see. More than this,
models are always blind to some aspects of reality because of simplifications
assumed on their formulation.
This blindness condition, however, should not only be seen as a limitation. On
the other hand and in most cases, it can be an advantage or a convenience for
engineers and scientists. Frequently, we can distinguish several types of blindness
aspects in fluvial hydraulics, mainly in one-dimensional mathematical models for
flood flow simulations, as will be discussed next in this chapter. We can call all
these situations as gaps between one-dimensional dynamic models and the reality.
One should always remember a holiday experience; you took a picture of a beautiful
river, for instance. The picture of the river could never be like the real one, once it
is just a plain sheet of paper (R2), a two-dimensional approximation of a three-
dimensional (R3) reality (or four-dimensional, R4, considering the time of that
picture shot).

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 13
A blind man cannot see, but he has a more sensible audition, a finer touch and
smelling capacity, e.g. he is more sensitive when considering his other capabili-
ties, excluding vision. In this way, when one uses one-dimensional mathematical
models (blind for the other two dimensions), he is able to become, in a short time,
very sensible to the model response, advantages and drawbacks, as well as the
model behaviour in the presence of changing any of its parameters. He becomes
more strongly linked to the model’s reality.
The first apparent gap may be related to the fact, not always reported by math-
ematical modellers, that discrete cross-river sections do not match with computa-
tional points in a river reach. In fact, when one applies a numerical method, often
finite-differences, to solve the one-dimensional Saint-Venant equations, the flow
direction x is discretized by finite increments Δx. The distances between the
increments are associated to the discrete cross sections, where flow-dependent
variables are calculated by the model. However, each discrete sub-stretch or reach
has a representative computational point, generally located at the middle of the
reach, and actually at this point the model would associate the calculated flow-
dependent variables. The nodal point is supposed to represent the sub-stretch
while the discrete cross section is actually the location where the numerical
model evaluates the flow-dependent variables. In other words, when the model
attributes a static variable, such as depth or water surface elevation, to a given
cross section it seems to be right; however a moving or dynamic variable such as
velocity or discharge may be more adequate to be attributed to a sub-stretch
instead of a cross section. Therefore discharge and velocity should be evaluated
at different points than depths and water surface elevations. This situation is
presented in Fig. 1.6.
Another conflict between one-dimensional mathematical dynamical models for
flood flow and reality may arise when one considers an average bottom slope in
the dynamic equation as representative of that slope on an entire river stretch.
There are cases where the bottom slopes vary considerably along the river course,
and perhaps it might be better to adopt variable bottom slope. However, this eval-
uation may be difficult, by lack of surveys for example. Anyway, this can consti-
tute another gap between model and physical reality, as shown in Fig. 1.7.
Taking flow velocity as a dependent variable is another situation in which reality
is often ignored, or in which approximation can be seen much more than a mere
simplification. Even considering that, as mentioned before, cross-section velocity is
more adequate than cross-section flow depth or water surface elevation, it is verified
that flow velocity in a section is not uniform. Therefore, any model using u instead of
flow discharge at a given flow cross section will be assuming a strong simplification,
as flow velocity varies with depth and width at the section, as can be seen in Figs 1.8
and 1.9. In many real flood flow cases, where the flood movement is relatively slow,
as occurs in most rivers not having steep bottoms, the gaps mentioned above do not
cause significant errors if it would be possible to compare real flow-dependent variables’
values with those obtained by one-dimensional dynamic mathematical models.
Besides, one-dimensional models are restricted in their essence to ignore variations
in any dependent flow variable with other dimension orthogonal to the main flow

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
14 Flood Prevention and Remediation

Δx
Δx
Δx Δx

calculation
sections

Figure 1.6: Discrete river cross sections and computational point.

Figure 1.7: Variable bottom slope in a river.

direction. Therefore, some gaps are inherent to the model formulation and one must
be aware of the associated limitations for each application case.
In this section, we will not enumerate or describe all gaps or incompatibilities
between mathematical models and physical reality, but only give the most relevant
examples where these gaps appear in a more clear way. Therefore, we will follow
below with few additional particularities that show how some equations may

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 15

Figure 1.8: Flow velocity variation with depth in a river cross section.

Figure 1.9: Flow velocity variation with cross section width.

ignore what really happens in river nature in the presence of singularities and flood
wave movement.
Take, for example, the flow width B, which appears in some flow equations, as
representative of river sub-stretch or a river reach. This property may not be constant
along the reach; there can be either sudden changes in river cross section as shown

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
16 Flood Prevention and Remediation

in Fig. 1.10, or small river islands at the reach, making it very difficult to attribute a
unique flow width in this case, as one can see in Fig. 1.11. If the model’s discrete
sections are spaced in a way in which the flow width is constant at each section,
there is not much to do in order to represent this variation, and one must be aware
that the mathematical model might be blind to this feature of physical reality.
Another difficulty, although this one can be corrected in the continuity equation,
is the flow width variation inside the same river cross section, when the section is
compound or irregular, as shown in Figs 1.12 and 1.13. In such a case, one must
define a dynamic or live width where the flow really occurs, and a storage or dead
width where dynamic effects of the flow can be neglected. This situation is not
exactly a gap between the model and reality; however it needs special attention.
For variables or parameter variations in a same flow cross section, there is still
a special case where taking average flow depth or roughness coefficient at a river
cross section may not represent what actually happens in physical reality. First,

Figure 1.10: Sudden river cross-section variation.

Figure 1.11: Small river islands as obstacles to the flow.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 17

Main Channel

Figure 1.12: River spilling water out from the main channel to the overbanks.

Figure 1.13: Dynamic river cross section.

consider what can happen to flow depth in a compound cross section, like the ones
shown in Figs 1.14 and 1.15. Different flow depths can be distinguished and trans-
versal water surface slopes may vary (Fig. 1.15).
In a same river cross section, the water surface slope presents different behav-
iours depending if the flood is in its rising or falling stages. This fact may not affect
directly the one-dimensional mathematical models; however, it is partially respon-
sible for the loop in the rating curve at the section, which can be used in the com-
putational or numerical model as downstream boundary condition in sub-critical
flow situations. This error type and a looped rating curve are shown in Figs 1.16
and 1.17.
Another conflict observed, as mentioned before, is associated to the fact that a
unique roughness value is considered for a given river cross section. Some studies
[5, 6] have shown that there may be a variation in the coefficient with flow depth,
even for almost a regular section. Moreover, if the river cross section increases its
width until reaching the floodplain zone, there will be a discontinuity in the rough-
ness coefficient, as presented in Fig. 1.18.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
18 Flood Prevention and Remediation

h2
h1

Figure 1.14: Different flow depths in a compound cross section.

(a)

(b)
Figure 1.15: Transverse water profiles slopes in (a) rising stage and in (b) falling
stage.

observed
computed

Figure 1.16: Errors caused by looped rating curve.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 19
h,Z Steady flow rating curve
Looped or dynamic
rating curve

Falling

Rising

Figure 1.17: Looped rating curve.

Figure 1.18: Roughness coefficient variation in a river cross section.

To conclude this section, consider a finite difference solution for the one-
dimensional dynamic equations. We have two independent variables – space and
time – represented graphically in a horizontal plane. Therefore if one constructs
the graphic with the vertical axis representing any flow-dependent variable, the
unknown solution will be a surface in R3, e.g. in space. Figure 1.19 is a good illus-
tration of what goes in essence when finite differences are applied in terms of
method’s concepts. We can see that the surface of solutions is approximated by
secant planes, or in other words R3 is approximated by several R2 elements. This
is not exactly a gap between the model and the physical reality, but rather a dimen-
sionality reduction embedded into the numerical method, therefore it can be
looked more as a simplification inherent to the solution’s method. Models based
on this method will be briefly discussed in the next section.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
20 Flood Prevention and Remediation

Surface of solutions

points of
solution
Secant
planes

Discretizing mesh nodal point

x
Figure 1.19: Spatial sketch for the finite difference method.

1.5 Models based on the finite-difference method

The simplest way to approach the concepts involved with finite differences is
based on the mathematical definition of derivatives. To illustrate the main idea, we
may recall the definition of the derivative of a function f(x) at a point x0 as

⎛ df ⎞ f ( x0 + Δ x ) − f ( x0 )
⎜⎝ dx ⎟⎠ = lim (1.38)
x= x
Δ x →0 Δx
0

Although to mathematicians this could seem rather correct, there is an apparent

loss of real sense in the fact that a quantity in the right hand is tending to zero,
or, in other words, is an infinitesimal quantity. This is an abstraction such as defin-
ing something at a point or speaking about infinity. Engineers understand this
meaning; however, they cannot treat these concepts in a practical way. Numeri-
cal methods came to rescue, reality and finite differences (and finite elements as
well) consider small but finite (non-infinitesimal) increments or intervals Δx. The
order of magnitude of the errors and the finite difference expressions may also be
obtained by truncated Taylor series [7]. Indeed, real problems involve functions of
more than one independent variable, but the main idea remains, resulting generally
in a set of partial differential equations discretized by finite differences.
Assuming a one-dimensional time-dependent problem and recalling what is
shown in Fig. 1.19, the problem domain is discretized by the rectangular mesh as
demonstrated in Fig. 1.20. For this type of flood flow situation and sub-critical
flow, in order to assure the solution uniqueness, it is necessary to prescribe the
appropriate initial and boundary conditions as presented in the same figure. To

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 21
summarize time-dependent finite-differences methods, there are two kinds of
approximations: explicit and implicit. Explicit approximations are simpler, as
shown in Fig. 1.21; however, they are subjected to stability conditions. On the
other hand, implicit finite difference approximations lead to a set of equations
involving several unknowns (e.g. functions at a following time step ‘i + 1’), but
most implicit schemes are generally stable in numerical terms, mainly if adequate
weighting factors are used.
Stability analysis is performed by Fourier components and amplifications factors
of the numerical scheme, and then the Neumann condition is imposed [8]. It is
cited just in this chapter because mathematical aspects will not be approached in
detail due to the fact that it is easy to find them in literature [9, 10].
To conclude this section, it must be stressed that not only derivatives are approxi-
mated by finite differences. In fact, when applying the method, both derivatives

t t
Line of upstream Line of downstream
boundary conditions boundary conditions

i+1,j-1 i+1,j i+1,j+1

i
i,j-1 i,j i,j+1
i-1
...
i=1 x
j = 1... j-1 j j+1 Line of initial conditions N

Figure 1.20: Domain discretization for the finite-difference method.

Δx

i+1
Δt

x
j-1 j j+1

Figure 1.21: Graphical sketch for the explicit diffusive method.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
22 Flood Prevention and Remediation

and the function itself are written as depending on the discrete function values at
the nodal domain discrete points. For example, the Preissmann’s generalized
implicit scheme [11] applied to a one-dimensional time-dependent problem results
in the expressions:

fqy ≅ f ( x, t ) = q ⎡y fi+j +11 + (1 −y ) fi+j 1 ⎤ + (1 −q ) ⎡y fi j +1 + (1 −y ) fi j ⎤ (1.39)

⎣ ⎦ ⎣ ⎦

∂f fy −fy
∂t
= i+1 i =
Δt
1 ⎡
Δt ⎣
( ) (
y fi+j +11 − fi j +1 + (1 −y ) fi+j 1 − fi j ⎤
⎦ ) (1.40)

∂f f j +1 − fqj
∂x
= q
Δx
=
1 ⎡
Δx ⎣
( ) (
q fi+j +11 − fi+j 1 + (1 −q ) fi j +1 − fi j ⎤
⎦ ) (1.41)

where Δt, Δx are the finite increments in time and space, respectively; i, j are the
discrete notations for time and space, respectively; and θ, ψ are the weighting fac-
tors in directions of time and space, respectively.
The most used version of Preissmann’s scheme considers 0.5 for the weighting
factor in space, and is unconditionally stable for θ greater than 0.6 [11]. It must not
be confused with stability for explicit schemes. The stability condition for the
explicit version of discrete Saint-Venant equations was posed by Courant,
Frederichs and Lewy (CFL) [12] and resulted in the well-known CFL condition. It
comes from the first study by Courant and Friedrichs [13], stated for pure diffusion
problems. These problems do not present any difficulty in applying Neumann’s
analysis, which is the general name of the stability analysis. In fact, for a one-
dimensional diffusion equation, it is very easy to introduce a particular solution
based on Fourier component with separate independent variables. Then, this solu-
tion is introduced in the discrete partially implicit general approximation for the
diffusion equation and the amplification factor is evaluated. It is imposed that its
modulus has to be smaller or equal to one and the Courant condition for explicit
schemes is obtained [14].
An advantage of Preissmann’s scheme (Fig. 1.22) when applied to Saint-Venant
modelling is that the resulting discrete equations solved at a given time step do not
involve simultaneously all same unknowns. Therefore, applying the Newton–
Raphson method, to solve the set of discrete equations, results in band matrices of
coefficients, with no more than four non-zero elements in a same row. This feature
implies that efficient algorithms related to time and storage computer savings
become quite important to be developed.

1.6 Pseudo-two-dimensional flood control model

This type of model deals with flood simulation in a particular way. Although
representing a two-dimensional area, flow equations are written in a one-dimensional
form. Pseudo-two-dimensional models consider the main rivers of a basin, their

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 23
f (x,t)

Surface of solutions
j
f i+1
t
f θj
ψ
f i+1
f ij i+1 f θψ j+1
f i+1

f θj+1

i f iψ
f i j+1

ψ Δx θΔt
Δt
j
Δx
j+1
x
Figure 1.22: Sketch for the generalized Preissmann’s implicit scheme.

flood plains and secondary drainage paths, as well as upper reaches of the basin,
altogether in an integrated representation. This feature allows the establishment of
a flood flow net, covering the basin surface as a system, where flow patterns are
governed by physical characteristics of the basin itself. In this context, the basin
is modelled by cells, with free shapes, capable to represent homogeneous areas,
carrying physical information about basin landscape. Rivers, plains, roads, natural
depressions, among other natural or constructed features, are divided into cells,
which are used to model flow and storage conditions.
The mesh of cells composes a complete representation of the basin. Each cell
communicates with its neighbouring cells by hydraulic laws, respecting modelled
local flow characteristics. In this type of model, topography and land use play a
fundamental role in flood flow simulation. These are the starting points for the
solution of the proposed problem and the main conceptual difference between
two-dimensional modelling, which departs from formal two-dimensional equa-
tions, and pseudo-two-dimensional modelling, which begins with surface interpre-
tation. This latter case makes it very important to have a sound knowledge of
topographical and hydraulic aspects related to the simulated basin behaviour.
In this case, cell modelling does not necessarily consider the formation of a
continuous free surface flow. This feature implies different possibilities of flow,
following patterns that may be independent from the channels net and governed by
natural or constructed elements over the landscape.
All cells are able to store a certain amount of water. The superficial water area
is defined by the water level in the cell and for its natural contours, such as levees,

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
24 Flood Prevention and Remediation

roads, natural land elevations, etc. The volume of water contained in each cell is
directly related to its own water level. The continuity equation for a cell i, in a
certain time interval, can be written as follows [2]:
t+ Δt t+ Δt
Δ Vi = ∑ ∫ Qi,k dt + ∫ Pi dt (1.42)
k t t

where ΔVi is the water volume variation in cell i, Qi.k is the discharge between cells
i and k, and Pi is the discharge obtained from effective rainfall over cell i.
The volume variation in a cell i, in a time interval t, is given by the balance
between inflow, from contributing neighbouring cells and rainfall runoff transfor-
mation, and outflow, to other neighbouring cells. Equation (1.43) is obtained by
expressing the volume stored inside the cell i, in the considered time interval, as a
function of the superficial area ASi of this cell

Zi (t + Δ t )
Δ Vi = ∫ ASi dZ i (1.43)
Zi (t )

Considering only the first-order terms and assuming that the relation (1.44) is
valid, it is possible to combine eqns (1.42), (1.43), and (1.44) to rewrite the continuity
equation, leading to eqn (1.45) in a discrete form

∂Asi
Δ Z i < < Asi (1.44)
∂Z i

dZ i
Asi = Pi + ∑ Qi,k (1.45)
dt k

In a general way, considering that flooding of plains occurs slowly, it can be

assumed that the discharge through cells is not influenced by the inertia terms.
This hypothesis allows the representation of discharge as a function of the water
levels inside two connected cells, as shown in eqn (1.46)

Qi,k = Q (Z i , Z k ) (1.46)

Figure 1.23 shows a schematic view of a region modelled by cells.

All modelled equation terms depend on the water level in each cell. In this situ-
ation, it is possible to have a cell initially dry. The cells may be flooded during the
calculation, depending on rainfall and exchanges among other neighbouring cells.
Equation (1.45) can be written for each one of the N cells of a basin, which
forms a system of N non-linear ordinary equations. The dependent variables of the
system are the N water levels. Equation (1.46) can represent any suitable hydraulic
discharge law for the particular modelled case. The resultant model is essentially

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 25

Figure 1.23: Schematic view of a region modelled by cells.

a looped one, with various possible flow directions, determining a pre-defined

pattern. Boundary conditions have been defined at the limits of the modelled
region and can assume records of measured water levels or of known discharges,
as well as a mathematical relation between discharge and water levels. It is also
necessary to define the initial condition of the system at the beginning of the
simulation.
In the 1960s, the first relevant pseudo-two-dimensional model was proposed
and implemented. This simulation was held for Mekong delta river by Societé

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
26 Flood Prevention and Remediation

Grenobloise d’Études et Applications Hydrauliques (SOGREAH), in a work for

UNESCO. Zanobetti and Lorgeré presented the resultant model at La Houille
Blanche [15]. This application divided the basin in storage cells, which were able
to represent river reaches and plain areas, in order to simulate flow patterns in
inundated Mekong delta areas, considering both natural floods and floods modi-
fied by a regulation dam, which would act in flood damping, flow regularization
for navigation purposes and irrigation. The modelled area had about 50,000 km2
and was inhabited by approximately 10 million people.
Other applications of this model can be found in several study cases: Mopipi
marsh [16]; Mono river basin [17]; Senegal river basin [2, 17]; Mfolozi river and
Santa Lucia estuary [18]; upper Rhône river basin [2]; and Parana river basin in
Yacyreta/Apipe [19].
In 1990, a different type of cell model was developed for heavy floods simula-
tion, including dam break [20]. In this case, nearby dam zones were treated par-
ticularly, considering the shock wave and privileging inertia effects. Distant dam
zones were considered to be characterized by great flood plains, terrain topogra-
phy, land use, presence of dykes, among other factors; hence shock waves can be
considered to have suffered significant dissipation. In this way, contiguous rectan-
gular cells, where water levels were computed for its centre and considered valid
for the entire cell region, could represent distant flood plain.
A first version of the cell model type was developed in Brazil to represent Pan-
tanal Matogrossense, a large marsh in west-centre region of Brazil [21]. Later, this
model was adapted to an urban environment, gaining a series of new facilities,
being presented by Mascarenhas and Miguez in 2002 [22].

1.7 Cell model for urban basins

A city and its drainage net can be divided into cells, constructing a two-dimensional
flow net where the flow occurs freely in different directions inside the modelled area,
departing from one-dimensional relationships for water exchanges. This superficial
two-dimensional flow net is linked to a second two-dimensional flow net, represent-
ing the under-surface drainage conducts. In this context, it is possible to consider that
pseudo-two-dimensional mathematical modelling achieves a status of a pseudo-three-
dimensional spatial modelling of the basin once these two nets are vertically linked.
The cells are able to represent the watershed scenery, composing complex struc-
tures. The definition of a varied set of flow type links, which represent different
hydraulic laws, allows the simulation of several flow patterns that can occur in an
urban landscape. Therefore, the task related to the topographic and hydraulic
modelling depends on a pre-defined set of cell types and the set of possible links
between cells.
The pre-defined set of cell types considered in MODCEL, presented by
Mascarenhas and Miguez [22], is listed below:

• River or channel cells – type of cell used to model the main free open channel
drainage flow, in which the cross section is taken as rectangular and may be
simple or composite.
WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 27
• Underground gallery cells act as complements to the drainage net.
• Urbanized superficial cells used to represent free surface flow over urban flood-
plains, as well as to storage areas linked to each other by streets. Alternatively,
these cells can also represent slope areas, although considering a little storage
capacity, associated to a possible flooded area, different from the total drainage
area. In this case, they are designated to receive and to transport the rainfall
water to the model’s core. Urbanized plain cells can also simulate a broad crest-
ed weir to the spilled waters from a river to its neighbour streets. This cell type
presents a gradation level for the compound rectangular cross section, assum-
ing a certain pre-defined urbanisation pattern, as shown in Fig. 1.24.
• Natural superficial cells are similar to the preceding ones; however, they have
a prismatic shape, without considering any kind of urbanisation.
• Reservoir cells used to simulate water storage in a temporary pond or reservoir,
which presents a curve for the elevation versus surface area. From this curve, it
is possible to evaluate the stored volume variation, departing from the water
depth variation. The reservoir cell type plays the role of damping an inflow
discharge.

A schematic vertical plane cut in an urban basin showing a cell model represen-
tation can be seen in the next chapter (Fig. 2.24).
Typical hydraulic links between cells can be summarized as shown below [14]:

• River/channel link is related to river and channel flows. It may eventually also
be applied to flow over the streets. More specifically, it corresponds to the free
surface flow represented by the complete Saint-Venant dynamic equation.
• Surface flow link corresponds to the free surface flow without inertia terms, as
presented in Zanobetti et al. [23].
• Gallery link represents the free surface flow in storm sewers, as well as the
surcharged flow conditions. Free surface flow is modelled using Saint-Venant

Urbanisation pre-defined pattern

by the model user, depending on
the case study:
% of cell area at bottom level
% of cell area at h1 level
% of cell area at h2 level

Urbanised
Cell Bottom
h2
h1

Figure 1.24: Gradation levels in the surface of an urbanized cell – storage pattern
representation.
WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
28 Flood Prevention and Remediation

dynamic equation. On the other hand, when galleries become drowned, pres-
sure flow conditions are given by energy conservation law, using Bernoulli
equation.
• Inlet gallery link/outlet gallery link – computed flow conditions define if the
inlet/outlet is drowned or not, also considering the possible occurrence of local
head losses.
• Broad crested weir link represents the flow over broad-crested weirs. It is used,
mainly, to represent the flow between a river and its margins.
• Orifice link represents the classic formula for flow through orifices.
• Street inlet link promotes the interface between surface and storm drain gallery
cells. When not drowned, this link acts as a weir conveying flow from streets to
galleries. When drowned, it considers flow occurring through a certain number
of orifices associated to the street inlets.
• Reservoir link combines an orifice, as the outlet discharge of a reservoir, with a
weir, that can enter or not in use, depending on reservoir operation.
• Stage-discharge curve link corresponds to special structures calibrated at a labora-
tory and basically relates a discharge with a water level, in a particular equation.
• Pumping link allows pumping discharges from a cell to another, departing from
a starting pre-defined operation level.
• Flap gate link simulates flows occurring in the direction allowed by the flap
gate opening, and can often be used when modelling regions protected by
polders.

1.8 Concluding remarks

Flood flows, as any natural phenomenon, actually are time-dependent and three-
dimensional in space. However, mathematical modelling of flood flows may be
simplified in order to represent the relevant flow patterns regarding the dimen-
sionality concept. Mathematical models act as built artificial systems, generally
representing real physical systems, which are able to give a response to a set
of input data. These input data are modified or transformed on the built system
through the model parameters that have to be calibrated in order to produce reliable
outputs for the physical system or, in this case discussed here, for the watershed
(Fig. 1.25).
The representation shown in Fig. 1.25 applies to any mathematical model,
despite they are one-, two or pseudo-two-dimensional, and represents a crucial
phase for the model application, named parameters calibration.
Regarding dimensionality, in this chapter, some ideas related to simplifications
over the three-dimensional approach were presented. We have seen that actually a
rigorous two-dimensional mathematical model for flood flow may be cumber-
some. In fact, one-dimensional mathematical models can be sometimes more
adequate to represent complex flows that have locally one-dimensional behaviour
(Fig. 1.26).
On the other hand, when flood flow presents non-negligible two-dimensional
behaviour, the pseudo-two-dimensional model may be applied, mainly in cases

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 29

Figure 1.25: Representation of parameters calibration system.

-Locally 1-D Flow Model

Figure 1.26: Apparent complex flow simplified to one-dimensional approach.

where one can observe logical one-dimensional links that constitute cell connec-
tions or when does not occur a real unique surface of flow, that is, when local
features act as partial barriers to flow and different patterns of independent flow
are simultaneously observed. These cells, as presented, are consequence of the
domain discretization process.
Therefore, we can summarize the simplified mathematical flood flow models as
presented in Fig. 1.27, where we have disregarded rigorous three- and two-
dimensional mathematical models. It must be stressed, however, that few situa-
tions may require at least two-dimensional mathematical models, while other
cases, like flood flow in an urban area, may require locally pseudo-three-
dimensional approaches (pressure flow in galleries), as it will be seen in Chapter 2.
Anyway, the correct identification of the flood flow patters will lead to the
development of the one-dimensional and pseudo-two-dimensional mathematical
flood flow models.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
30 Flood Prevention and Remediation

River Basin
Mass
Conservation Hydrological Models
(continuity)
Physical
Laws 1-D Network Model
Momentum
Conservation
(dynamic)
Pseudo 2-D Looped Model

Figure 1.27: Summary of simplified flood flow mathematical models.

It was also shown that flow equations can be simplified in most real cases, by
neglecting less important terms, leading to mathematical forms that can be solved
through a well-suited numerical method.
Concluding this chapter, special attention must be given to the simplest flood
flow model, based only on the mass conservation law, leading to the well-known
hydrological models. These models, however, neglect the flow dynamic pattern,
and must be applied only to situations where storage effects are relevant, rather
than dynamic flood flow properties. The engineer must be aware regarding the
prevalent flow directions, in order to correctly choose the most appropriate mathe-
matical flood flow model, taking into account available data and macroscopic flow
behaviour. Obviously both physical laws, concerning mass and momentum conser-
vation, are true.

References

[1] Cunge, J.A., On the subject of a flood propagation computation method

(Muskingum Method). Journal of Hydraulic Research, 7(2), pp. 205–230,
1969.
[2] Cunge, J.A., Holly Jr, F.M. & Verwey, A., Practical Aspects of Computational
River Hydraulics, Pitman Advanced Publishing Program: London, 1980.
[3] Saint Venant, B., Théorie du Movement Non Permanent des Eaux, Avec
Application aux Crues des Riviéres et à La Introduction des Marées dans
Leur Lits, Comptes Rendus des Séances de L´Académie des Sciences, 73,
pp. 147–154 & 237–240, Paris, 1871.
[4] Abbott, M.B., Computational Hydraulics. Elements of the Theory of Free
Surface Flows, Pitman Publishing Limited: London, 1979.
[5] Chow, V.T., Open Channel Hydraulics, McGraw-Hill Book Company:
New York, 1959.
[6] Cowan, W.L., Estimating hydraulic roughness coefficients. Agricultural
Engineering, 37(7), pp. 473–475, 1956.
[7] Carnaham, B., Luther, H.A. & Wilkes, J.O., Applied Numerical Methods,
John Wiley & Sons: New York, 1965.
[8] Von Neumann, J. & Richtmyer, R.D., A method for the numerical calculation
of hydrodynamic shocks. Journal of Applied Physics, 21, pp. 232–237, 1950.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)
 Simplified Models for Flood Flow Simulation 31

[9] Gunaratnam, D.J. & Perkins, F.E., Numerical Solution of Unsteady Flow
in Open Channels, Hydrodynamics Laboratory, Massachusetts Institute of
Technology Report n. 127, Massachusetts, 1970.
[10] Richtmyer, R.D. & Morton, K.W., Difference Methods for Initial Value
Problems, Interscience Publishers: New York, 1957.
[11] Lyn, D.A. & Goodwin, P., Stability of a general Preissmann scheme. Journal
of Hydraulic Engineering, 113(1), pp. 16–28, 1987.
[12] Courant, R., Friedrichs, K.O. & Lewy, H., On partial differential equations
of Mathematical Physics. IBM Journal of Research Development (translated
1967), 11, pp. 219–234, 1948.
[13] Courant, R. & Friedrichs, K.O., Supersonic Flow and Shock Waves, Inter-
science Publishers: New York, 1948.
[14] Mascarenhas, F.C.B., Toda, K.; Miguez, M.G. & Inoue, K., Flood Risk
Simulation, series: Progress in Water Resources, Vol. 10, WIT Press: South-
ampton, UK, 2005.
[15] Zanobetti, D. & Lorgeré, H., Le Modele Mathématique du Delta du Mékong,
La Houille Blanche, n. 1, 4 and 5, 1968.
[16] Hutchison, I.P.G. & Midgley, D.C., Mathematical model to aid management
of outflow from the Okavango Swamp, Botswana. Journal of Hydrology,
19, pp. 93–113, 1973.
[17] Cunge, J.A., Two-dimensional modeling of flood plains (Chapter 17).
Unsteady Flow in Open Channels, eds. K. Mahmood, V. Yevjevich., Water
Resources Publications: Fort Collins, USA, 1975.
[18] Weiss, H.W. & Midgley, D.C., Suite of the mathematical flood plain models,
Journal of the Hydraulics Division, ASCE, 104(HY3), pp. 361–376, 1978.
[19] Major, T.F., Lara, A. & Cunge, J.A., Mathematical modeling of Yacyreta-
Apipe Scheme of the Rio Parana, La Houille Blanche, n. 6 and 7, 1985.
[20] Gallatti, M., Braschi, G., Di Fillipo, A. & Rossi, U., Simulation of the
inundation of large areas of complex topography caused by heavy floods.
Hydraulic engineering software applications, Proc. Third Int. Conf. Hydraulic
Eng. Software, Massachusetts, Computational Mechanics Publications,
Southampton, UK, 1990.
[21] Mascarenhas, F.C.B. & Miguez, M.G., Large flood plains modeling by a
cell scheme: application to the Pantanal of Mato Grosso-Brazil. Second Int.
Symp. Eng. Hydrol., ASCE, San Francisco, USA, 1993.
[22] Mascarenhas, F.C.B. & Miguez, M.G., Urban flood control through a
mathematical cell. Water International, 27(2), pp. 208–218, 2002.
[23] Zanobetti, D., Lorgeré, H., Preissmann, A. & Cunge, J.A., Mekong Delta
mathematical program construction. Journal of the Waterways and Harbours
Division, ASCE, 96 (WW2), pp. 181–199, 1970.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press
www.witpress.com, ISSN 1755-8336 (on-line)