FLOOD INUNDATION MODELLING

TO SUPPORT FLOOD RISK MANAGEMENT

(GARETH PENDER AND SYLVAIN NEELZ)
 The Problem: Hydro-Meteorological Risks
Hydrological
Hazards The Disaster Cycle
Hydrological extremes are
plainly crucial floods and
droughts;
Overall hydrological
characterisation may also be
important in terms of
background to other hazards
o Landslides/Mudflows/Debris
flows/Land subsidence
Risks are to human
wellbeing (health and
safety), water and food
supply, mobility/transport,
power supply, industrial
functioning, and
environmental quality
concerns
source: Swiss Federal Office for Civil Protection

Hydrological disasters have

INTRODUCTION
EUROPIAN FLOOD
MODELLING METHOD
Classification of flood inundation models (adapted from Pender 2006)

1D
Design-scale modelling, which can be of the
One dimensional St. Venant order of tens to hundreds of kilometres, depending on catchment
Equation size

Design-scale modelling, which can be of the

1D+ 1D plus a storage cell approach to

the simulation of floodplain flow
order of tens to hundreds of kilometres, depending on catchment
size; also has the potential for broad-scale application if used with
sparse cross-section data

2D minus the law of conservation of

2D- momentum for floodplain flow
Broad-scale modelling and applications where inertial effects
are not important

Design-scale modelling of the order of tens of kilometres. May

Solution for the two dimensional
2D
have the potential for use in broad-scale modelling if applied with
very coarse grids
shallow water equation

2D plus a solution for the vertical

2D+
Predominantly coastal modelling applications where 3D velocity
proles are important. Has also been applied to reach-scale river
velocities using continuity law modelling problems in research projects

3D
Solution for the three dimensional reynolds Local predictions of three-dimensional velocity elds in main
averaged-navier stokes equation channels and oodplains
1D and 2D SImulation
 Flood simulation

Computational hydraulics approaches:

Complexity Application scale
simple large
1D hydrostatic
1D hydrodynamic simplified (kinematic,
diffusion wave)
1D full hydrodynamic
1D/2D simplified hydrodynamic
1D/2D full hydrodynamic complex small
model setup
2D full hydrodynamic data requirements
computational demand

3D full hydrodynamic

Risk Analysis
Flood Hazard Assessment 9
 Flood simulation
Cross section over channel and floodplain
1D full hydrodynamic
Mulde_Test1 Plan: Plan 02 08/08/2008

.035 .11 . .11 .033 .035

140 0
3
5

Pros
135

130

Elevation (m)
Many software packages available, including free
125
software, e.g. HEC-RAS
Computationally efficient without consideration of 120

hydraulic structures 115

0 500 1000 1500 2000 2500 30
Station (m)

Interpolated cross sections

Cons
No representation of 2D floodplain flow
Derivation of cross sections time consuming
Interpolation to inundation areas

Application
River reaches with confined floodplains and parallel to
the river
Large scale
Source: HEC-RAS user manual
Risk Analysis
Flood Hazard Assessment 10
 Flood simulation

2D full hydrodynamic
Pros
Detailed process description
Precise calculation of h and v in areas with complex flow patterns
Realistic representation of floodplain processes, well suited for
urban environments
Mostly commercial software

Cons
Computationally demanding
Setup of computational mesh
Mostly commercial software

Application
Small scale, up to 500 km2
Risk Analysis
Flood Hazard Assessment 11
Source: Apel et al. 2009
 One dimensional 1D Flow Modelling

Saint venant equation

Mass conservation equation

The friction slope Sf is usually obtained from

a uniform flow formula such as Manning or
chezy.
 Assumptions for St. Venant Equations

Flow is one-dimensional
Hydrostatic pressure prevails and vertical
accelerations are negligible
Streamline curvature is small.
Bottom slope of the channel is small.
Mannings equation is used to describe
resistance effects
The fluid is incompressible
 Continuity Equation
Q = inflow to the control volume
q = lateral inflow

Q Rate of change of flow

x with distance
Q
Q dx Outflow from the C.V.
x

( Adx)
Change in mass
t

Elevation View
Reynolds transport theorem

d
0
dt c.v.
d V .dA
c.s.

Plan View
 Continuity Equation (2)
Q A
q 0 Conservation form
x t

(Vy ) y
0
x t
Non-conservation form (velocity is dependent
variable)
y V y
V y 0
x x t
 Momentum Equation
From Newtons 2nd Law:
Net force = time rate of change of momentum
d
F dt Vd VV .dA
c .v . c.s.

Sum of forces on
the C.V.

Momentum stored Momentum flow

within the C.V across the C. S.
Forces acting on the C.V.
Fg = Gravity force due to weight
of water in the C.V.
Ff = friction force due to shear
stress along the bottom and
sides of the C.V.
Fe = contraction/expansion
force due to abrupt changes in
the channel cross-section
Fw = wind shear force due to
frictional resistance of wind at
the water surface
Elevation View Fp = unbalanced pressure forces
due to hydrostatic forces on the
left and right hand side of the
C.V. and pressure force exerted
by banks

Plan View
 Momentum Equation
d
F dt Vd VV .dA
c .v . c.s.

Sum of forces on
the C.V.

Momentum stored Momentum flow

within the C.V across the C. S.

1 Q 1 Q 2 y


g g (So S f ) 0
A t A x A x
 Momentum Equation(2)
1 Q 1 Q 2 y
g g ( S o S f ) 0
A t A x A x
Local Convective Pressure Gravity Friction
acceleration acceleration force force force
term term term term term

V V y
V g g (So S f ) 0
t x x
Kinematic Wave

Diffusion Wave

Dynamic Wave
 Momentum Equation (3)
1 V V V y
So S f
g t g x x
Steady, uniform flow

Steady, non-uniform flow

Unsteady, non-uniform flow

 Applications of different forms of momentum equation

V V y
V g g ( So S f ) 0
t x x

Kinematic wave: when gravity forces and friction forces

balance each other (steep slope channels with no back water
effects)
Diffusion wave: when pressure forces are important in
addition to gravity and frictional forces
Dynamic wave: when both inertial and pressure forces are
important and backwater effects are not negligible (mild
slope channels with downstream control)
 Dynamic Wave Routing

Flow in natural channels is unsteady, nonuniform

with junctions, tributaries, variable cross-sections,
variable resistances, variable depths, etc etc.
 Solving St. Venant equations
Analytical
Solved by integrating partial differential equations
Applicable to only a few special simple cases of kinematic waves

Numerical
Finite difference
approximation
Calculations are performed
on a grid placed over the (x,t)
plane
Flow and water surface
elevation are obtained for
incremental time and
distances along the channel

x-t plane for finite differences calculations

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

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

x x
Cross-sectional view in x-t plane

x-t plane
h0, Q0, t1 h1, Q1, t1 h2, Q2, t2

t h0, Q0, t0 h1, Q1, t0 h2, Q2, t0

x x
 Finite Difference Approximations

Explicit
Implici
uij 1 uij 1 uij
u u t u u
j 1 j 1 j
uij1
t t i i 1 i
t 2t
Temporal derivative
Temporal derivative
uij uij1 uij1
u uij11 uij 1 uij1 uij
x 2x (1 )
x x x
Spatial derivative
Spatial derivative

Spatial derivative is written using terms on Spatial and temporal derivatives use
known time line unknown time lines for computation

Solution scheme
Equations are transformed to a set of implicit finite difference
equations over a computational grid
alternating Q - and H points, where Q and H are computed at each
time step
numerical scheme - 6 point Abbott-Ionescu scheme
i-1 i i+1
Time
Time step n+1
Q
h7
6
Time step n+1/2
Q
Q h5
4
h1 h3
2
Time step n

Center point Space

Example
 2D Saint Venant Equations

Obtained from Reynolds Navier-Stokes equations by depth-

averaging.
Suitable for flow over a dyke, through the breach, over the
floodplain.

Assumptions: hydrostatic pressure distribution, small

channel slope
t x y
(hu)(hu)
h (hu(hv)
) (huv)
2
h z b u 2
2

0 continuityeq.
gh gh gn 2 u x momentumeq.
t x y x x v h 13
(hv) (huv) (hv 2 ) h z u2 2

gh gh b gn 2 y momentumeq.
t x y y y v v
1

h 3
Module 6
 Solutions to St. Venant equations

Solutions to
St. Venant equations

Method of Finite Difference Finite Element

characteristic methods Methods
s
Explicit Implicit

Further reading - Note: NPTEL (Computational Hydraulics),

http://www.efm.leeds.ac.uk/CIVE/CIVE3400/stvenant.pdf.
Module 6
 Solutions to St. Contd
Venant equations
Analytical
Solved by integrating partial differential equations

Applicable to only a few special simple cases of kinematic waves

Numerical
Finite difference approximation

Calculations are performed on a grid placed over the (x, t) plane

Flow and water surface elevation are obtained for incremental time
and distances along the channel

Module 6
Finite Difference Scheme (FDS)

Time, t

Downstream
Upstream
boundary

boundary
Distance, x
x-t plane for finite differences calculations
Module 6
 Finite Difference Scheme (FDS) Contd
x-t plane

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

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

x x

h0, Q0, t1 h1, Q1, t1 h2, Q2, t2

Cross-sectional view
in x-t plane t h0, Q 0, t 0 h1, Q 1, t 0 h2, Q 2, t 0

x x

Module 6
 Explicit Implicit
Temporal derivative Temporal derivative

ui j u
j j j
u u j 1 u u ui 1 1 i1

t i1
2t
t j
i i
u
t
Spatial derivative Spatial derivative

u u j u u j 1
ui1 u i j 1 j u
ui1
j

x
i1
x x (1 ) x
i

i1

2x

Module 6
 Explicit FDS

Reach
Time
Time stepstepj
Time j+1

Time step j-1

Space

Spatial derivative is written using terms on known time line j

Module 6
 Implicit FDS

i-1 i i+1
Time
Q Time step j+1
h7
6
Q Time step j+1/2
Q h5
4

h1 2 h3
Time step j

Center
point Space

Spatial and temporal derivatives use unknown time lines for computation j+1

Module 6
 Momentum Analysis in an Open Channel

For a constant mass and a per unit width consideration in a rectangular channel,

L
P1
mvs qv2 v1 Wsin

v1
P2
v2
PRF P
2 f s 1 Rf
W

Module 6
Momentum Analysis in an Open Channel Contd

Here,

Rf is the frictional resistance.

P1 and P2 are pressure forces per unit width given by:

2
y
P 2 2
2

Combining terms we get:

y1 2 W sin q(v
f
Considering a short section

2
so
2
y that R is negligible
R v ) slope is
and the channel
f
2
small so that sin 0, the equation can be written as:
1

y 12 y 22
qv1 qv 2
2 2

Module 6
 Momentum Analysis in an Open
Contd
Channel
2 2
i.e, y qv1 y qv 2 M
1 2

2 g 2 g
y
M is the specific force plus Specific force plus momentum curve
momentum and is constant for
q1
both y1 & y2.
There are two possible depths q2
y2
for a given M called sequent
depths. The depth associated y = yc
with the minimum M is yc. yc
y1

Mc M M
Module 6
 Hybrid 1D/2D model
Based on raster DEM
1D Kinematic or diffusion wave routing in channel
Once bankfull depth is exceeded calculates a flux to floodplain cells
using Mannings equation or 2D diffusive wave to route water over
complex floodplain topography

Model discretization of floodplain and

channel topography

In-channel flow routed using a 1D wave

equation
Once bankful depth is exceeded water
can flow laterally over adjacent low
lying floodplains according to
topography and free surface gradient
 Inundation is based on a simple continuity
equation:
dV
Qup Qdown Qleft Qright
dt
where:
V = cell volume
t = time
Qup, Qdown, Qlef and Qright = flow rates in each direction
into (positive Q) and out of (negative Q) the cell
i j
Flux between cells
calculated using the
Manning equation
Ai , j Ri2,j/ 3Si1,/j 2
Qi , j
n
where:
Qi,j = flux between two cells i and j,
Ai,j = cross sectional area at the cell
interface,
Ri,j = hydraulic radius at cell interface,
Si,j = water surface slope between cells,
n = Manning friction coefficient.
 HYBRID 1D/2D MODEL

Interaction between 1D Overland Network and 2D

Overland Network
 HYBRID 1D/2D MODEL
1D-1D Hybrid 1D-2D
 MODEL PARAMETERIZATION

Channel routing parameters

Channel Variable
Measured data at outlet Geomorpho- Fitting curve
routing channel
(discharge, top width, logical parameter
parameters routing
cross-section) relations adjustment
at outlets parameters

Spatially variable basin properties Observed

(slope, area, drainage density) outlet
hydrographs

46/59
Model discretization of floodplain and
channel topography

In-channel flow routed using a 1D wave

equation
Once bankful depth is exceeded water
can flow laterally over adjacent low
lying floodplains according to
topography and free surface gradient
 TERRAIN GEOMETRY

Good terrain representation is important for any type of

floodplain delineation.
Good is a function of the type of modeling being made
and type of terrain morphology (flatter terrains need higher
accuracy).

Mapping the Zone, National Research Council, 2009,

National Academies Press, Washington, D.C., 122 pp.
Elevation Data for Floodplain Mapping, National
Research Council, 2007, National Academies Press,
Washington, D.C., 152 pp.
 TERRAIN GEOMETRY

1. Blending of Different Data

Sources
2. Use of TINs vs DEMs
3. Methods for creating
hydrologically correct DEMs
 TERRAIN GEOMETRY

Typically many terrain data sets are used in

the calculations of the flood boundaries
Floodplain boundaries require special
attention at the intersection of different
Insert Graphic
topographic data
showing
Data
sets
Shelving of

49
 TERRAIN GEOMETRY
LIDAR for measuring terrain
LIDAR is a powerful tool in the
professional mappers
toolbox.
LIDAR can be used to produce
a wide variety of products
Good project design ensures
product suitability for end
user application

50
 TERRAIN GEOMETRY

DEMs are Derived from TINs and is a generalization of the data within
Defined Cell Size
In general, DEM data requires more smoothing routines than does TIN
data
TINs can be used to reduce generalization of data

Insert Graphic
Insert Graphic
showing TIN Data
showing DEM Data

50 ft

51
 TERRAIN GEOMETRY

How to compute it? Statistical inference (Sampling theory)

Differences between z DEM

Check points selection
and z from an independent
(finite sample N)
source of higher accuracy

DEM quality evaluation

RMSE and ME Check Points Error

calculation (ZDEMi -ZCPi = ei)

Newcastle, September
2007
 TERRAIN GEOMETRY

Types of errors

1. Blunders or Outliers
2. Systematic (bias) errors (constant
offset)
3. Random errors (random fluctuations
in the measurements)

Newcastle, September
2007
 CONCLUSION

1. 1D techniques remain appropriate for situations where a clearly defined one-

dimensional pathway exists, such as rivers and pipe systems.

2. In situations where no such clearly defined single pathwayexists, recourse

mustbe made to 2D methods. Here a choice is required between those
techniques employing the full shallow water equations and those based on
simplified equations, such as diffusive or kinematic wave equations. The
con- sequences of this choice, when models are applied to floodplains with
complex geometries, are pres- ently unclear; however, ongoing research
(Environ- ment Agency, 2009) will soon clarify such choices.
 CONCLUSION

3. Research is continuing to deliver improve- ments in 2D modelling methods.

In particular in the areas of trans-critical flow simulation, hybrid methods
(linking 1D and 2D models) and discre- tization techniques. Additionally,
the systems approach required by flood risk management and the
uncertainties associated with predicting pres-sures and flood sources dictate
a need for multiple simulations to facilitate a probabilistic approach to
uncertainty analysis.

4. There is therefore a press- ing need for faster model predictions through
either the development of accurate model emula- tion techniques or the use
of parallel processing to speed up 2D modelling methods. Research is on-
going in each of these areas