Shallow Water

Finite Volume Method for Shallow Water
Equations
Oto Havle, Jir Felcman
Workshop Dresden-Prague on Numerical Analysis 2008
Contents
Shallow water equations
SWE without source term
SWE with topographic source term
Conclusion

A system of PDEs
h
+ div(hv) = 0,
t
1 2
z
(hvs ) + div(hvs v) + g
h = gh
,
t
2 xs
xs
z ... given function, g > 0 ... given constant
h
z
s = 1, 2.
Shallow water equations - a hyperbolic system

2
w X
+
f i (w ) = r(x, w )
t
xi
i=1
where

h
h
h
hv1
w
q
hv
hv2
q1
hv1
f 1 (w ) = h1 q12 + 21 gh2 = hv12 + 21 gh2
hv1 v2
h1 q1 q2
q2
hv2
f 2 (w ) = h1 q1 q2 = hv1 v2
hv22 + 12 gh2
h1 q22 + 21 gh2

0
r(x, w ) =
ghz(x)
Difficulties with Shallow Water Equations

I
existence / uniqueness theorems not available
dry areas - SWEs do not have sense for h = 0
source term

Flat bottom case z = const.
I
system of conservation laws
w X
+
f s (w ) = 0
t
xs
s=1

I
standard FVM discretization
Z
w (x, t
Di
k +1
Z
) dx
w (x, t k ) dx
Di
t k +1
+
tk
Di
2
X
ni f i (w )
i=1
{z
H(w |Di ,w |Dj ,nij )
dS = 0

I
standard FVM discretization
we need a numerical flux

wki w
on the cell Di
and time interval (t k , t k +1 )
wki +1 = wki
k X
|ij |H(wki , wkj , nij )
|Di |
jS(i)
Numerical flux
How to derive a numerical flux H(w L , w R , n)

I
Choose a new coordinate system such that n = (1, 0).
Linearize and solve a linear 1D Riemann problem.

w
w
+A
=0
t
x1
(
w L , x1 < 0
w (x1 , t = 0) =
w R , x1 > 0
Take the linear flux Aw at x1 = 0 and rewrite in the original

coordinate system.
How to linearize the Riemann problem

We adapt the Vijayasundaram flux,
g V (w L , w R ) = A+ (w ? )w L + A (w ? )w R
wL + wR
where w ? =
2
A1 (w ) = D f 1 (w )
(?)
As A1 (w )w 6= f 1 (w ), the 1D flux (?) is not consistent with the

Shallow Water Equations. We fix it with additional term

0
1
2
1
g V 1 (w L , w R ) = A+ (w ? )w L + A (w ? )w R gh??
2
0
hL + hR
where h?? =
2
SWE with no source term - Numerical results

Elementary Riemann problems in 1D
I
continuous solution (rarefaction wave) - EOC 0.75
discontinuous solution (contact discontinuity) - EOC 0.5
2D test problem - circular dam break
10
9
8
7
6
5
4
3
2
1
0
6
5
4
3
2
1
0
200
200
150
150
100
-200
50
-150
-100
0
-50
50
100
150
-50
-100
-150
-200
200
100
-200
50
-150
-100
0
-50
50
100
150
-50
-100
-150
-200
200
Source term - stationary solution

h
+ div(hv) = 0,
t
1 2
z
(hvs ) + div(hvs v) + g
h = gh
,
t
2 xs
xs
Stationary solution (lake at rest)
h(x, t) = H0 z(x),
Discrete stationary solution
H0 zi
w ki = 0 ,
0
v(x, t) = 0.

zi z D
s = 1, 2.
Discretization of source terms

Rewrite to a system of conservation laws with a nonzero RHS
2
w X
+
f s (w ) = r(x, w )
t
xs
s=1
Operator splitting
2
w X
+
f s (w ) = 0
t
xs
s=1
w
+
t
= r(x, w )
would not work. Stationary solutions are different.
Discretization of source terms

Difficulty with the source term
I
We use a piecewise-constant approximation of the

topography function z.
hi = H0 zi
The source term is now a distribution.

0
r(x, w ) =
ghz(x)
Source term must be taken into account when

approximating convective term.
Linearised Riemann problem with a source term

We construct a linear Riemann problem with nonzero RHS
w
w
+A
= (x)r,
t
x
(
w L, x < 0
w (x, 0) =
wR, x > 0
t >0
The matrix A is the same as before

A = A1 (w ? ) = Df 1 (w ? )
w? =
The right hand side has special form
0
r = gh? (zL zR )
0
1
(wL + wR )
2
Linear Riemann problem with a source term

To solve the Riemann problem
I
Use diagonal form A = UU1 to decouple equations.
u
u
+
= (x), t > 0
t
x
u(x, 0) = u0 (x)

I
Use characteristics to solve scalar problems.
= 0 = u(x, t) = u0 (x) + t(x),
K (x, t),
6= 0 = u(x, t) = u0 (x t) +
||
(
1, x(x t) < 0,
where K (x, t) =
0, x(x t) 0.

I
The solution is a function, unless v1 = c.
|v | < c
... subcritical
|v | = c
... critical
|v | > c ... supercritical
p
gh
h
w = hv1
hv2
c=

I
The solution is a function, unless v1 = c.
For the purposes of deriving a numerical flux, ignore the

distributional case.
The linear flux Aw at x = 0 is

1
1
A
lim w (x, t) +
lim w (x, t)
2 x0+
2 x0
1
= A+ w L + A w R + (sgn A)r ? .
2
New numerical flux

The numerical flux (discretization of convective term)
g conv (w L , w R , zL , zR )
1 2
= A+
1 (w ? )w L + A1 (w ? )w R 2 gh?? e 2
1
gh? (zR zL )(sgn A1 )e 2
2
In fact, the numerical flux does not depend on sgn 2 .
p
1 = v1 c, 2 = v1 , 3 = v1 + c,
c = gh
Therefore, it is continuous in the subcritical case |v | < c,
1/c
sgn A1 (w ? )e 2 = v1 /c .
v2 /c
Discretizing the source term

We discretize the source term proper
t k +1
t k +1
Z Z
Z Z
r(x, w (x, t)) dx dt ghi
Di
tk
Di
k

0
dx dt
z(x)
tk
= ghi
0
z(x)
nij

dS
Di
k
ghi
|ij |zij? nij
jS(i)
where zij? =
zi + zj
is an approximation to z at the interface ij .
2
The numerical scheme

Both convective and source have a quasi-flux form
X
k
|ij |(. . . )
jS(i)
We can write
w ki +1 = w ki

k X
|ij |H total w ki , w kj , zi , zj , nij
|Di |
jS(i)
Theoretical properties
I
The scheme is conservative with respect to h.
The discrete "lake at rest" is a stationary solution.
Numerical results
The scheme seems to work in the subcritical case |v | < c, but
not in the supercritical case. It gets better if we replace
h?? =
hL + hR
2
with
h?? =
h
qR
1+ 2
2 hL
< 1,
+
h ,
L
1 2
2 hR ,
1 < < 1,
v1,?
c?
> 1,
g conv (w L , w R , zL , zR )
1 2
= A+
1 (w ? )w L + A1 (w ? )w R 2 gh?? e 2
1
gh? (zR zL )(sgn A1 )e 2
2
Summary
Numerical scheme for Shallow Water Equations.
I
Simple first order Finite Volume scheme.
Handles topographic source term.
Future work
I
Enhance with mesh adaptivity.
Investigate the case h = 0.
Summary
Numerical scheme for Shallow Water Equations.
I
Simple first order Finite Volume scheme.
Handles topographic source term.
Future work
I
Enhance with mesh adaptivity.
Investigate the case h = 0.
Thank you for your attention.

Shallow Water

Uploaded by

Copyright:

Available Formats

Shallow Water

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Shallow Water

Uploaded by

Copyright:

Available Formats

Finite Volume Method for Shallow Water

Workshop Dresden-Prague on Numerical Analysis 2008

Shallow water equations

SWE without source term

SWE with topographic source term

Shallow water equations

Shallow water equations - a hyperbolic system

Shallow water equations - a hyperbolic system

Difficulties with Shallow Water Equations

existence / uniqueness theorems not available

dry areas - SWEs do not have sense for h = 0

Shallow water equations - a hyperbolic system

system of conservation laws

Shallow water equations - a hyperbolic system

system of conservation laws

standard FVM discretization

H(w |Di ,w |Dj ,nij )

Shallow water equations - a hyperbolic system

system of conservation laws

standard FVM discretization

we need a numerical flux

How to derive a numerical flux H(w L , w R , n)

Choose a new coordinate system such that n = (1, 0).

Linearize and solve a linear 1D Riemann problem.

Take the linear flux Aw at x1 = 0 and rewrite in the original

How to linearize the Riemann problem

As A1 (w )w 6= f 1 (w ), the 1D flux (?) is not consistent with the

SWE with no source term - Numerical results

continuous solution (rarefaction wave) - EOC 0.75

discontinuous solution (contact discontinuity) - EOC 0.5

2D test problem - circular dam break

Source term - stationary solution

Discretization of source terms

would not work. Stationary solutions are different.

Discretization of source terms

We use a piecewise-constant approximation of the

Source term must be taken into account when

Linearised Riemann problem with a source term

The matrix A is the same as before

The right hand side has special form

Linear Riemann problem with a source term

Use diagonal form A = UU1 to decouple equations.

Linear Riemann problem with a source term

Use diagonal form A = UU1 to decouple equations.

Use characteristics to solve scalar problems.

= 0 = u(x, t) = u0 (x) + t(x),

Linear Riemann problem with a source term

Use diagonal form A = UU1 to decouple equations.

Use characteristics to solve scalar problems.

The solution is a function, unless v1 = c.

|v | > c ... supercritical

Linear Riemann problem with a source term

Use diagonal form A = UU1 to decouple equations.

Use characteristics to solve scalar problems.

The solution is a function, unless v1 = c.

For the purposes of deriving a numerical flux, ignore the

The linear flux Aw at x = 0 is

New numerical flux