Applied Mathematical Modelling: R. Touma

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

Applied Mathematical Modelling 000 (2015) 117

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

Well-balanced central schemes for systems of shallow water


equations with wet and dry states
R. Touma
Mathematics, Lebanese American University, Beirut, Lebanon

a r t i c l e

i n f o

Article history:
Received 23 February 2015
Revised 11 August 2015
Accepted 23 September 2015
Available online xxx
Keywords:
Unstaggered central schemes
Well-balanced schemes
Shallow water equations
Surface gradient method
Wetting and drying
Positivity preserving schemes

a b s t r a c t
In this paper we propose a new well-balanced unstaggered central nite volume scheme for
the shallow water equations on variable bottom topographies, with wet and dry states. Based
on a special piecewise linear reconstruction of the cell-centered numerical solution and a
careful discretization of the system of partial differential equations, the proposed numerical
scheme ensures both a well-balanced discretization and the positivity requirement of the water height component. More precisely, the well-balanced requirement is fullled by following
the surface gradient method, while the positivity requirement of the computed water height
component is ensured by following a new technique specially designed for the unstaggered
central schemes. The developed scheme is then validated and classical shallow water equation
problems on variable bottom topographies with wet and dry states are successfully solved. The
reported numerical results conrm the potential and eciency of the proposed method.
2015 Elsevier Inc. All rights reserved.

1. Introduction
Introduced by Nessyahu and Tadmor [1] in 1990, central schemes were meant to be simple and ecient tools for the numerical
solution of systems of hyperbolic conservation laws. Based on the staggered LaxFriedrichs scheme, the Nessyahu and Tadmor
(NT) central scheme evolves a piecewise linear numerical solution on two staggered grids and avoids the time consuming process
of solving Riemann problems arising at the cell interfaces. Furthermore, second-order quadrature rules along with a gradients
limiting process result in a second-order accurate and oscillations-free nite volume method for general hyperbolic systems of
conservation laws. Multidimensional extensions of the NT scheme on Cartesian grids [25] and unstructured grids [69] were
later developed and successfully applied to solve hyperbolic arising in aerodynamics and magnetohydrodynamics [5,10], as well
as balance law problems arising in hydrodynamics [1113].
One main disadvantage of using central schemes remains in the fact that a dual staggered grid is required to evolve the numerical solution, and this leads to synchronization issues whenever physical constraints are to be numerically forced, or if steady
states are to be handled in the case of balance laws. To overcome this problem, Jiang et al. [14] presented a rst adaptation of the
NT scheme that evolves the numerical solution on a single grid and another adaptation (known as unstaggered central scheme
UCS) followed in [13]. Based on a careful projection of the numerical solution obtained on the staggered cells, back onto the
original cells, the UCS method was successfully used to solve steady state problems arising in hydrodynamics, and ideal/shallow
water magnetohydrodynamics [15,16]. In this paper we develop a new well-balanced second-order accurate unstaggered central
scheme for the numerical solution of shallow water equation (SWE) problems with wet and dry states. The SWE system describes

Tel.: +9613834095.
E-mail address: rony.touma@lau.edu.lb

http://dx.doi.org/10.1016/j.apm.2015.09.073
S0307-904X(15)00607-1/ 2015 Elsevier Inc. All rights reserved.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
2

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

the motion of a free surface incompressible uid under gravity, over a variable bottom topography. The SWE system is widely
accepted as a mathematical model for the hydrodynamics of coastal oceans, simulation of ows in rivers, and can also be used to
simulate tsunami and inundation waves, dam breaches, and others. Ignoring the friction effects, the SWE system can be written
as a system of hyperbolic conservation laws with a source term describing the effects of the varying water bed or bathymetry;
this source term vanishes in the case of a at bathymetry. Recently, the research on numerical methods for the shallow water
equations became popular for two main reasons or challenges: The SWE system features equilibrium states/stationary solutions in which the nonzero ux divergence is exactly balanced by the source term. In their native form, usually most numerical
schemes fail to generate equilibrium state solutions and they generate nonphysical waves, oscillations, and instabilities because
the ux divergence and the source term lack well-balancing in their discretization. Well-balanced schemes were developed in
[12,1726] in a way to satisfy the steady state requirement such as the lake at rest problem. Another important feature in the
simulation of the SWE problems is the appearance of wet and dry areas that are due to the initial conditions or to the ow of
water in the computational domain. Here again most numerical schemes fail to handle the interaction between wet and dry
zones and generate negative water heights and other instabilities. The development of well-balanced schemes with wetting and
drying capabilities formed another challenge among the numerical community, and several numerical schemes were recently
developed [2730] to fulll both the lake at rest and the wetting and drying constraints.
In this paper we develop, analyze and implement a new unstaggered, well-balanced, non-oscillatory, and second-order accurate central scheme for the one-dimensional system of shallow water equations on irregular bathymetry with wet and dry zones.
The lake at rest constraint will be exactly satised at the discrete level by following the surface gradient method developed in
[12,26] which discretizes the water height according to the discretizations of the water level and the bottom topography functions. On the other hand wet and dry zones will be carefully treated by introducing a new technique that corrects the slopes of
the linearized water height function over the control cells in the forward and backward projections steps in a way to ensure both
water conservation and non-negative water height values. We show that the resulting scheme is a well-balanced scheme that
exactly maintains the steady state requirement at the discrete level when lake at rest problems are considered and also allows
proper wave run-ups and withdraws on coastal slopes and shorelines.
The rest of the paper is organized as follows: Section 2 is dedicated to reviewing the shallow water equations, their properties,
and their equilibrium states and constraints. In Section 3 we develop the well-balanced central scheme for the shallow water
equations with wet and dry states; we describe the well-balancing technique that ensures the lake at rest constraint and the
wet/dry treatment that allows a proper propagation of water waves on shores and islands. Section 4 is devoted to the validation
and application of the developed scheme; we perform several numerical experiments from the recent literature, and we conrm
the robustness and potential of the proposed scheme. We end the paper with some concluding remarks and perspectives for
future work in Section 5.
2. Shallow water equations
Shallow water equations are commonly used to mathematically model rapidly varying free surface ows such as dam
breaches, oods and inundation waves, tidal waves in oceans and lakes and many others. The system of shallow water equations is derived from the conservation principles of the mass and momentum, and under the main assumption of hydrostatic
pressure distribution, the SWE system is a time dependent two-dimensional system of hyperbolic balance laws. The conservative one-dimensional version of the SWE system reduces to


t u + x f (u) = S(u, x), t > 0, x  R,
u(x, 0) = u0 (x),

(1)

with x and t are the spatial and temporal independent variables, respectively. The computational domain  is an interval of
the real axis. Furthermore, the unknown vector solution u(x, t ), the ux function f (u), and the source term S(u, x) are given as
follows:

 

u(x, t ) =

h
, f (u) =
hv

hv
, and S(u, x) =
1
2
hv + gh2
2

0
gh

dz .
dx

(2)

Here h(x, t) denotes the water height, v(x, t) is the velocity in the x-direction, g is the gravitational constant, and z(x) denotes the
of system (1) vanishes
bottom topography function (Fig. 1). When the waterbed is at, i.e. z(x) = constant, the right hand side

and the resulting system reduces to a homogeneous hyperbolic system with real eigenvalues 1 = v gh and 2 = v + gh
and linearly independent eigenvectors.
The SWE system features equilibrium solutions that are solutions to the system


x (hv) = 0,

x (hv2 + gh2 /2) = dh

dz
.
dx

(3)

One particular equilibrium state that we would like to satisfy in this work is the lake at rest equilibrium state dened by


v = 0,

h + z = constant.

(4)

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

h(x,t)

z(x)

10

11

Fig. 1. Free surface ow over variable bottom topography, featuring a lake at rest state, wetting/drying, and a wave run up.

Numerical methods for the shallow water equations usually fail to satisfy the lake at rest equilibrium state unless a proper wellbalanced discretization of the source term is performed. Different schemes were recently [11,12,17,23,26,28,31] to satisfy the lake
at rest equilibrium state.
Another important feature that we would like to properly handle in this work is the ow of water on a wet and dry area
inside the computational domain; this usually arises while run-up of waves on shorelines as is illustrated in Fig. 1. Numerical
schemes, in their native forms, usually fail to properly handle the wet/dry states and they generate negative water heights before
they break out. In this work we aim to develop a new well-balanced unstaggered central scheme capable of wetting and drying
(WB-UCS-WD) that preserves the water height positivity constraint in the neighborhood of wet/dry states.
3. Well-balanced central schemes for the SWE systems with wet and dry states
In this section, we develop a new well-balanced central scheme for the shallow water equations on variable bottom topographies with wet and dry states. The proposed central scheme follows a classical nite volumes method construction. We start by
partitioning the computational domain  = [a, b] R using control cells Ci = [xi1/2 , xi+1/2 ] of equal width x = xi+1/2 xi1/2 .
The nodes xi , centers of the cells Ci , are obtained by setting xi = a + ix for i N. Similarly we dene the dual cells Di+1/2 =
[xi , xi+1 ] centered at the nodes xi+1/2 = xi + x/2. The time step will be denoted by t, and for n N we set and t n = nt.
Next, for a scalar function u(x, t), we will use the notation uni to denote the numerical estimate of u(xi , tn ), and for any scalar
function g(x) admitting point values we denote by gi the numerical approximate of g(xi ). Furthermore we will denote the discrete
spatial forward, backward and centered differences approximating gx (xi ) as follows:

gi+1 gi
,
x
gi gi1
D (gi ) =
,
x
gi+1 gi1
D0 (gi ) =
,
2 x
D+ (gi ) =

with gi 1 approximates g(xi x). Furthermore we introduce the average and jump notations, respectively as follows:

uni+1/2 =
[[u]]i+1/2

uni + uni+1

2
= ui+1 ui ,

uni =

uni1/2 + uni+1/2
2

[[u]]i = ui+1/2 ui1/2 .

In a similar way we will use the notations

[[u]]si = usi+1/2 usi1/2 ,

[[(us ) ]]i = usi+1/2

usi1/2 .

3.1. Well-balanced central nite volume method for the SWE system
Without any loss of generality, we assume that the numerical solution uni to system (1) is known at time tn at the centers
xi of the control cells Ci . In order to compute un+1
at time t n+1 = t n + t we will follow a classical nite volume approach that
i
evolves a piecewise linear numerical solution Li (x, t n ) dened at the nodes xi , and that approximates the solution u(x, t n ) on Ci .
Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
4

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

Li (x, t n ) is given by

Li (x, t n ) = uni + (x xi )(uni ) ,

x Ci .

(5)

(uni ) is a limited numerical gradient that approximates the spatial partial derivative x u(x, t n )|x=xi . In this work, we use van
Leers monotonized centered limiter (MC ), where the slope of the interpolant is calculated componentwise as follows: the kth
component (unk,i ) of (uni )

[[unk ]]i 1

(unk,i ) = minmod

x

unk,i+1 unk,i1

2 x

[[unk ]]i+ 1

x

here 1 2, and the minmod function is dened by

sign(a) min{|a|, |b|, |c|}, if a, b, c have the same sign,


0,
otherwise.

minmod(a, b, c) =

To construct the numerical solution un+1


at time t n+1 we start by integrating the hyperbolic balance law in system (1) on the
i
rectangle Rni+1/2 = [xi , xi+1 ] [t n , t n+1 ] and then we apply Greens theorem to the integral to the left; we obtain

( f (u)dt udx) =

Rn

t n+1

tn

i+ 1
2

xi+1

S(u)dxdt

xi

Expanding the integral to the left over the four sides of the rectangle Rn

i+ 12

xi+1
xi

u(x, t n ) dx +

t n+1

tn

f (u(xi , t ))dt =

tn

t n+1

f (u(xi+1 , t ))dt +

t n+1

tn

xi+1
xi

xi+1

xi

, we obtain

u(x, t n+1 ) dx

S(u) dxdt.

(6)

We assume that at time t n+1 and for x Di+1/2 the solution u(x, t n+1 ) Li+1/2 (x, t n+1 ) is a piecewise linear function dened at
the center xi+1/2 of the cells Di+1/2 ; the Mean-Value theorem gives

xi+1

u(x, t n+1 )dx = x Li+1/2 (xi+ 1 , t n+1 ) = x un+1


.
i+ 1
2

xi

In a similar way, since at time tn and for x Ci , the solution u(x, t n ) Li (x, t n ) is a piecewise linear function, the Mean-Value
theorem leads to

xi+1

xi

u(x, t n )dx =
=

xi+1/2
xi

x
2

u(x, t n )dx +

Li (xi+ 1 , t n ) +

xi+1/2

x

xi+1

u(x, t n )dx

Li+1 (xi+ 3 , t n ) := x uni+ 1 .


4

(7)

Eq. (7) denes the forward projection step that interpolates the numerical solution at time tn as follows:

 

uni+ 1
2

x n
1
=
L i xi +
,t
2
4
x

= ui+ 1

+ Li+1 xi+1

x
4



,t

[[(un ) ]]i+ 1 .

(8)

Eq. (6) can now be rewritten as

un+1
= uni+ 1
i+ 1
2

x

1
x



t n+1

tn

t n+1

tn

xi+1

xi

f (u(xi+1 , t ))dt

S(u)dxdt.

t n+1

tn


f (u(xi , t ))dt
(9)

As for the ux integrals in Eq. (9), they are approximated with second-order of accuracy using the midpoint quadrature rule as
follows:

t n+1

tn

n+ 12

f (u(xi , t ))dt t f (ui

),

(10)

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

where the predicted solution values at time t n+1/2 are obtained using a rst-order Taylor expansion in time as well as the balance
law (1) as follows:

u(xi , t n+ 2 ) u(xi , t n ) +
1

uni +

t 

t
2

ut (xi , t n )

f (u)x |(xi ,t n ) + S(u)|(xi ,t n )


(11)

2

t 
n+ 1
n
ui +
( fin ) + Sin := ui 2 ,
2

with ( fin ) approximates the ux divergence and Sin S(xi , uni ) approximates the source term at time tn on the cell Ci with
second-order of accuracy and is dened with the aid of a sensor function to ensure well-balancing as follows:
n
n
n
Sin = Si,L
+ Si,R
+ Si,C
,

with

Si,L

n
Si,R

n
Si,C

(12)

s2i (1 si ) (2 si )
0
,
ghni D (zi )
6

s2i (1 + si ) (2 si )
0
,
ghni D+ (zi )
2

(13)

si (si + 1) (si 1)
0
=
.
ghni D0 (zi )
6

Here 1 2 is the MC parameter used in the gradients limiting in the forward projection step as well as in the predictor
step.
The sensor function si , appearing in the discretization of the source term, forces the discretization of the source term (13) to
follow the same discretization of the term hx appearing in the ux divergence and is dened by

1 if h = D hi ,

1 if hi = D h ,
+ i
i
si =
0 if hi = 0,

(14)

if hi = D0 hi .

On the other hand, the integral of the source term in Eq. (9) is also approximated with second-order of accuracy using centered
differences and the midpoint quadrature rule as follows:

t n+1

tn

xi+1
xi

n+ 12

S(u(x, t )))dx dt t x S(ui

with


n+ 1
n+ 1
ui 2 , ui+12

S(

)=

n+ 1

, ui+12 )

ghn+1/2
D+ (zi )
i+ 1

(15)

Eq. (9) combined with Eqs. (10) and (15) reduces to


n+ 12

un+1
= uni+1/2 tD+ f (ui
i+ 1
2

) + tS un+1/2
, un+1/2
.
i
i+1

(16)

n+1 , is obtained at the center of the control cells


From Eq. (16), we see that the numerical solution un+1
1 , computed at time t
i+ 2

Di+1/2 . A backward projection step is therefore required to generate the numerical solution un+1
on the cells Ci as follows:
i
n+1

un+1
= ui
i

1
8

(un+1
) un+1
)
i+1/2
i1/2

n+1

= ui

x
8

[[(un+1 ) ]]i ,

(17)

n+1 
where (ui+1/2
) denotes a limited numerical gradient that approximates the spatial partial derivative x u(x, t n+1 )|x=x 1 . This
i+
2
completes the description of the central nite volume scheme. Next we show that the proposed nite volume scheme (16)
= uni+1/2 . In that context, we have the
preserves the lake at rest constraint of the SWE system (1); we rst show that un+1
i+1/2
following theorem:

Theorem 3.1. Assume that the numerical solution uni of the one-dimensional SWE system is generated using the nite volume method
(16), and assume further that uni satises the lake at rest constraint (3) at the discrete level at time tn , i.e.,

vni = 0, and hni + zi = constant, i,

(18)

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
6

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

then,
= uni .
(a) The predicted numerical solution obtained using Eq. (11) is such that un+1/2
i
n+1
on the dual cell Di+1/2 obtained using Eq. (16) and the forward projection solution obtained
(b) The computed solution at time t
using Eq. (15) are equal, i.e., un+1
= uni+1/2 .
i+1/2
Proof 3.1. First we establish (a). If for example in the prediction step (11) hx (xi , tn ) is discretized using the forward difference,
i.e., hx (xi , t n ) (hni ) = D+ hni , then the sensor function si takes on the value 1, and the discretized source term reduces to

n . In addition, since un satises the steady state requirement at time tn , then vn = 0 and f (un ) = f n = 0, 1 g(hn )2
Sin = Si,R
2
i
i
i
i
i

( fin )

T
0, ghni (hni ) .

=
Therefore the prediction step becomes
n+ 12

= uni +

ui

= uni +
uni

t
2

t






0
ghni D+ (zi )

ghni (hni )

so

0
ghni [(hni ) + D+ (zi )]



t
0
+
.
2x ghni [ (hni+1 + zi+1 ) (hni + zi )]

But since hni + zi = H = constant for all i, then (hni+1 + zi+1 ) (hni + zi ) = 0
leading to
n+ 12

= uni .

ui

A similar proof can be done for the other values of the sensor function sni .
n . Since the numerical solution un satises the lake at rest equilibrium
Next we establish (b), i.e., we show that un+1
1
1 = u
i
i+ 2

i+ 2

state (4) at time tn at the discrete level, i.e., Eq. (18), and taking into account part (a) of Theorem 3.1, then we get,

n+ 1
ui+12

f(

)=

0
1
g(hi+1
2

n+ 12

)2

and

Substituting in Eq. (16), we get

un+1
i+ 12

uni+ 1
2

t

x


+t

0
1
n+ 1
g(hi+12 )2
2

n+ 1

2
ghi+1/2
D+ (zi )

n+ 12

f (ui

)=

0
n+ 12

(hi

1
g
2

)2

0
1
n+ 1
g(hi 2 )2
2

Performing basic algebra operations, we obtain

un+1
i+ 12

uni+ 1
2

n+ 12

But since ui



0
t


1
1

.
n+ 1
2
2
x hn+
(hn+
+ zi+1 ) (hi 2 + zi )
i+1/2
i+1
n+ 12

= uni and hni + zi = H = constant, for all i, then (hni + zi ) (hni+1 + zi+1 ) = 0. Therefore, (hi

n
zi+1 ) = (hni + zi ) (hni+1 + zi+1 ) = 0, leading to un+1
1 = u
i+ 2

i+ 12

n+ 1

+ zi ) (hi+12 +

Remark 3.1. Theorem 3.1 ensures that if the numerical solution uni satises the steady state equilibrium state (4) at the discrete
at time t n+1 will satisfy the same equilibrium state, but only
level at time tn on the nodes xi , then the computed solution un+1
i+1/2
on the nodes xi+1/2 . This does not guarantee that the back-projected numerical solution uin+1 obtained using Eq. (11) will satisfy
the equilibrium state (4) at the discrete level. Therefore an additional treatment is required. In this work, we follow the surface
gradient method suggested by Zhou et al. [12,16,26], for SWE systems on completely ooded domains to the case of systems of
SWE systems with wet and dry regions.
Following the surface gradient method, we calculate the numerical derivative of the water height h(x, t) component in terms
of the water level function H (x, t ) = h(x, t ) + z(x). To do so, we assume that the bottom topography function values are known
at the cell-interfaces xi+1/2 , i.e., the values zi+ 1 are given at the nodes xi+ 1 . At the cell centers xi , we set zi = 12 (zi+ 1 + zi 1 ). The
2

waterbed function z(x) is reconstructed on the control cells Ci using linear interpolants as follows:

z(x) = zi +

x

(zi+ 12 zi 12 )(x xi ) x Ci .

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

Next, we linearize the water height function h(x) on the control cells Ci using the linearization of the water level function H(x).
For the linearization of the water level function H (x, t n ) Hin + (Hin ) (x xi ) on the cells Ci we limit the numerical derivatives of
Hin = hni + zi ; the numerical spatial derivative of the water height function is then obtained as follows:

(hni ) = (Hin )

x

(zi+1/2 zi1/2 ).

(19)

Eq. (19) is used in the forward projection step to calculate the rst component of uni+1/2 on the dual cells in Eq. (8).
Similarly for the back projection step of the water height component in un+1
on the original cells Ci (i.e., Eq. (17)), we rst
i+1/2
dene the water level values H n+1 at the centers of the dual cells as follows:
i+1/2

n+1
= hn+1
+ zi+1/2 ,
H i+1/2
i+1/2

where zi+1/2 = zi+1/2

(20)

z
zi+1/2 is the corrected bottom elevation due to the fact that the function z(x) is linear only on
2 i+1/2
the cells Ci and not on the staggered cells Di+1/2 .
) using the water level values H i+1/2
Next, we compute the limited numerical derivative of the water height function (hn+1
i+1/2
obtained over the corrected waterbed as follows:
n+1 
(hn+1
) = (H i+1/2
)
i+1/2

x

(zi+1 zi ).

(21)

This discretization of the spatial derivative of h(x, t) will ensure that the projected numerical solution un+1
back onto the original
i
cells Ci will satisfy equilibrium state requirement in the case of a the lake at rest problem as is stated in Theorem 3.2.
Theorem 3.2. Assume that the numerical solution un+1
of the system of shallow water Eq. (1) is computed using the nite volume
i
method (16) under the hypotheses of Theorem 3.1, and following the surface gradient method for the forward projection step (8), (19)
and the backward projection step (17), (20), (21) for the water height component, then the proposed nite volume scheme is a wellbalanced scheme that exactly preserves the lake at rest requirement at the discrete level (4) and therefore the equation uin+1 = uni holds
for all i.
Proof 3.2. To show that the lake at rest equilibrium state (18) remains maintained at time t n+1 provided it was as such at time
= uni , for all i. Below we present the proof for the h component. We
tn we will proceed component wise and show that un+1
i
note that in the lake at rest equilibrium state (4) the hv component does not change in time because v = 0 and because of the
well-balanced discretization.
We use the surface gradient method for both the forward and backward projection steps; for the forward projection we have

hni+1/2 = hni+1/2 +

x 
8


(hni ) (hni+1 ) ,

(22)

where the derivatives (hni ) are discretized using the water level function Hin = hni + zi as described in Eq. (19). Substituting in
Eq. (22) one obtains (while taking into account that Hin is constant for all i)

hni+1/2 = hni+1/2


1
zi+1/2 zi+1/2 .
2

(23)

Similarly, for the backward projection step we apply the surface gradient method and one obtains (while taking into account
n+1/2 = hn+1/2 + 
zi+1/2 is constant for all i)
that H
i+1/2
i+1/2

hn+1
= hn+1
+
i
i

1
[[z]]i1/2 + [[z]]i+1/2 .
8

(24)

= hni . Therefore we conclude that if the steady state (18) was satised at time tn , then
Substituting Eq. (23) in (24) leads to hn+1
i
n+1
.
it will remain as such at the next time t
3.2. Treatment of wet/dry states
When shallow water equation problems with wet and dry regions are considered, numerical instabilities usually arise unless
a proper care is taken while computing the numerical solution. In the case of unstaggered central schemes, the way the forward
and backward projection steps are performed is crucial for ensuring a physically admissible numerical solution. More precisely,
in the neighborhood of wet and dry regions, the appearance of spurious oscillations and other numerical instabilities is mainly
due to the gradient limiters that yield an interpolated numerical solution falling below the waterbed on the dual cells in the
forward projection step, or an interpolated water height lying below the water bed function on the original cells in the backward
projection step; in either case, the generated numerical solution becomes physically non-admissible, and therefore an additional
treatment is required. In this work we propose a new wetting/drying treatment procedure for the SWE systems that allows clean
forward and backward projection steps without yielding negative water heights, and conserves the total amount of water across
the computational domain. The proposed wetting/drying procedure is triggered when an interpolation had yield a negative water
Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
8

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

i+

z+

n
i

z + n

x
xi+3/2

xi+1

xi+1/2

xi+1/4

xi1/2

xi

z+

z+

i+

z+

i+
1

Fig. 2. The corrected forward projection step avoids negative water height values hni+1/2 and conserves the cells water average value hni in Ci with hni+1 = 0 in Ci+1 .

xi+3/2

xi+1

xi+3/4

xi+1/2

xi1/2

xi

Fig. 3. Correcting the forward projection step leading to hni+1/2 = 0 while conserving the cells water average value hn1 in the cell Ci with hni+1 = 0 in cell Ci+1 .

height value, then we redene the slope of the water height interpolant appropriately on the neighboring control cells in a way to
ensure a non-negative water height interpolated value, which guarantees a physically admissible numerical solution. To better
describe the effects of wet and dry states on the interpolated water height values, we consider the following case where the
numerical solution uni = (hni , hni uni ) is such that hnk > 0 for k {i 1, i} and hni+1 = 0.
Performing a forward projection step such as Eq. (8) may lead to a negative water height value (hni+1/2 ) in uni+1/2 if the slope of
the interpolant in (x) that approximates h(x, tn ) for x Ci is too steep as is illustrated in Fig. 2. The water height interpolant in (x)
on the cell Ci at time tn is dened by

in (x) = hni + (x xi )(hni ) , x Ci .


in (x)

h(x, t n )

in (xi )

(25)
hni

hni

Note that

for x Ci and
=
> 0 on a wet cell. For a dry cell Ci (i.e., if
negative water height hni+1/2 in the interpolated solution arises for example when in (x) is too

= 0), we set in (x) = 0 x Ci . A


steep so that in (xi+1/4 ) < 0 and

n (x
n
n
n
i+1
i+3/4 ) < |i (xi+1/4 )|. This results in hi+1/2 = 0.5 i (xi+1/4 ) + i+1 (xi+3/4 ) < 0. A wet cell Ci neighbor to a dry or nearly dry
cell Ci+1 , may result in such a situation. This is illustrated in Fig. 2 where we show the water level function z + in over the wet cell
n over the dry cell C
n
Ci and the water level function z + i+1
i+1 . Fig. 2 shows also the corrected water level function z + i (dotted
line) over the wet cell Ci that ensures hni+1/2 = 0 on the dual cell Di+1/2 . Fig. 3 shows a similar case where the forward projection
Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

n (x)
step leads to a negative water height value hni+1/2 whenever hni = 0 and hni+1 > 0; the corrected water height interpolant i+1
n
for x Ci+1 will guarantee non-negative values of hi+1/2 .
The treatment we propose in this work is as follows. Without any loss of generality, we can assume that the water height
component hni in the numerical solution uni at time tn is positive or zero. Our challenge is to generate the numerical solution uin+1

in a way the water height hn+1


remains non-negative. Explicitly, we need to make sure that the forward projection step leading to
i

uni+1/2 and the backward projection step leading to uin+1 fulll the non-negativity requirement of the water height. We propose
the following modications to the forward and backward projection steps. If in the forward projection step, the water height
hni+1/2 in the cell Di+1/2 becomes negative, then we set hni+1/2 to be zero and we correct the slopes of the water height interpolants
in the cells Ci and Ci+1 to ensure water conservation across the computational domain. Two cases might arise: If hni  (for some
hn

i
preassigned tolerance  = 109 ) we re-linearize the water height function on cell Ci by taking hx (xi , t n ) x/4
, i.e., we set

hn

i
(hni ) = x/4
, otherwise if hni <  , we set hni = 0 and hx (xi , t n ) = (hni ) = 0. Similarly, if hni+1  we re-linearize the water height

hn

, otherwise we set hni+1 = 0 and hx (xi+1 , t n ) = 0. Furthermore, to


function h(x, tn ) on the cell Ci+1 and we take hx (xi+1 , t n ) i+1
x/4
maintain well balanced property of the developed scheme, the discretization of zx in the source term in Eqs. (1) and (2) should
mimic the discretization of hx . We therefore extend denition of the sensor si in Eq. (14) as well as the discretization of the
source term Sin in Eq. (13) to include the following two cases, whenever a negative water height value hni+1/2 had been detected,
as follows.
hni
z
zi
If hni >  , then in addition to setting (hni ) = x/4
, we set zx (xi ) i+1/2
(x/2) for the discretization of zx (xi ) in the approximation
hn

i+1/2
, we set zx (xi ) i+1
of the source term. Similarly if hni+1 >  , then in addition to setting (hni+1 ) = i+1
x/4
(x/2) . A similar treatment
n+1
is to be applied for the backward projection step that generates hi . This will ensure both a non-negative forward projected

, while maintaining the well-balanced discretization


water height value hni+1/2 and a backward projected water height value hn+1
i
of the source term according to the discretization of the ux divergence. We nally note that the proposed treatment of wet dry
states ensures a physically admissible numerical solution with non-negative water height while maintaining the well-balanced
property of the numerical base scheme, but it does not necessarily ensure the lake at rest constraint along with wet/dry interaction constraint, simultaneously. In fact, and as is discussed in [32] it is not easy to make the scheme still satisfy the lake at rest
constraint after a wetting/drying treatment.

t =10
f

6
100 grid points
Exact solution

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fig. 4. Lake at rest problem: prole of the water level at time t = 10.

Table 1
Lake at rest problem: experimental order of
convergence measured in the L1 -norm.
N

L1 error

Order

50
100
200

1.96E02
3.73E03
8.28E04

2.39
2.17

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
10

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

Water level and velocity

Front position

15
Waterbed
Water level
Velocity

Exact solution
Numerical solution

4
10

=0

=0

2
5
1

1
15

10

0
x

10

15

0.5

1
time t

1.5

1.5

1.5

(a) = 0
Water level and velocity

Front position

15
Waterbed
Water level
Velocity

Exact solution
Numerical solution

4
10

=/60

=/60

2
5
1

1
15

10

0
x

10

15

0.5

(b) = /60

Water level and velocity

1
time t

Front position

15
Waterbed
Water level
Velocity

Exact solution
Numerical solution

=/60

=/60

10
3

5
1

1
15

10

0
x

10

15

0.5

(c) = /60

1
time t

Fig. 5. Dam break over an inclined plane.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

11

4. Numerical experiments
In this section we apply the developed well-balanced numerical scheme and solve classical one-dimensional shallow water
equation problems with at or variable bottom topography functions.
4.1. Lake at rest problem with smooth bottom topography
The rst experiment features a lake at rest problem with a smooth bottom topography. This experiment is meant to validate
the well-balanced property of the proposed scheme and its second-order of accuracy. The bottom topography is dened by the
function z = sin2 ( x) over then interval [0, 1]. The initial water level is set to be H (x, 0) = h(x, 0) + z(x) = 5 and the initial
velocity is v(x, 0) = 0. The numerical solution is computed at the nal time t f = 10 on 100 grid points and is compared to the
exact solution (lake at rest solution). The obtained numerical results are reported in Fig. 4, where we compare the water level
obtained on 100 grid points ( curve) to the exact solution of the problem (solid curve). Both curves are in perfect match thus
conrming the well-balanced property of the proposed scheme.
The L1 error and the order of convergence of the numerical solution on an increasing mesh was computed; the obtained
results, reported in Table 1, validate the order of accuracy of the numerical scheme.
4.1.1. Dam break over a plane
This experiment features a dam break over inclined planes with various angles of inclination as suggested in [28,31,33]. The
computational domain is the interval [15, 15] which we discretize using 200 grid points. The bottom topography function is
dened by the function z(x) = x tan , where denotes the inclination angle. In this example we consider three inclinations
= /60 (downhill inclination), = 0 (at), and = /60 (uphill inclination). The initial conditions for this test case are as
follows:

u(x, 0) = 0,

h(x, 0) =

1 z(x),
0,

x < 0,
otherwise.

We apply the proposed scheme and compute the numerical solution at time t = 2. The obtained results are reported in Fig. 5(a)
(c), for = /60, = 0, and = /60, respectively. The left column shows the water elevation and velocity along the waterbed at the nal time, while the right column shows the front location at each time t 
[0, 2] obtained using the proposed
numerical scheme and compared to the exact analytical front position given by x f (t ) = 2t g cos () 0.5gt 2 tan (); the numerical front location is dened to be the rst cell-center (from left to right) where the water height exceeds the value  = 109 .
The obtained numerical results are in perfect match with their corresponding analytical solution and in perfect agreement with
the results reported in [28,33], thus conrming the potential of the proposed scheme to handle shallow water equation problems
on variable topographies with wet and dry states. The water mass m(t) at each time step t on the computational domain [15,15]
is computed and compared to the initial water mass m0 . The obtained results are reported in Fig. 6 where we show the graph of
the curve m(t)/m0 on 50, 100 and 200 grid points. The obtained results conrm the conservation of the water mass as x tends
to zero.
4.1.2. Parabolic bowl
Now we consider the parabolic bowl problem as considered in [33]. This problem was previously considered in the context of
shallow water equations with friction term in [33]. The computational domain is the interval [5000, 5000] and the water bed

1.0006

1
50 points
100 points
200 points

0.9998

1.0005

50 points
100 points
200 points

0.9996
0.9994

m(t)/m0

m(t)/m0

1.0004

0.9992

1.0003

1.0002
0.999
1.0001

0.9988
0.9986
0

0.5

1.5

2.5

1
0

0.5

1.5

2.5

Fig. 6. Dam break over an inclined plane: graph of the curve m(t)/m0 on the interval 0 t 2 obtained on 50, 100, and 200 spatial grid points.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
12

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

is the parabolic bowl dened by z(x) = h0 (x/a)2 , where h0 = 10 and a = 3000 are two constants. The initial water height h(x) is
given by the equation

h(x, t = 0) = h0

Bx
B2

2g 2a

8h0
z(x),
g

where B = 5 is a constant. The initial velocity v(x, t = 0) is set to zero. The exact solution of this problem at any time t is given by
the equation

Bx
B2
B2
cos (2t )

h(x, t ) = h0
4g
4g 2a

8h0
cos (2t ) z(x),
g

30

30
Waterbed
Numerical Solution
Exact solution

25

20
Water elevation

Water elevation

20

15

15

10

10

0
5000

4000

3000

2000

1000

0
x

1000

2000

3000

4000

0
5000

5000

(a) Water height at t = 1000

3000

2000

1000

0
x

1000

4000

5000

Water elevation

20

15

15

10

10

4000

3000

2000

1000

0
x

1000

2000

3000

4000

0
5000

5000

(c) Water height at t = 3000

4000

3000

2000

1000

0
x

1000

2000

3000

4000

5000

(d) Water height at t = 4000

30

30
Waterbed
Numerical Solution
Exact solution

25

25

20

20
Water elevation

Water elevation

3000

Waterbed
Numerical Solution
Exact solution

25

20

15

10

4000

3000

2000

1000

0
x

1000

2000

3000

(e) Water height at t = 5000

4000

5000

Waterbed
Numerical Solution
Exact solution

15

10

0
5000

2000

30
Waterbed
Numerical Solution
Exact solution

25

Water elevation

4000

(b) Water height at t = 2000

30

0
5000

Waterbed
Numerical Solution
Exact solution

25

0
5000

4000

3000

2000

1000

0
x

1000

2000

3000

4000

5000

(f) Water height at t = 6000


Fig. 7. Parabolic bowl problem.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

13

a2 cos (t ) and x = a Ba2 cos (t ), where


and the exact location of the water is located between the points x1 = a B2gh
2
2gh

2gh0 /a. The computational domain is partitioned using 200 control cells and the numerical solution is calculated using
the proposed WB-UCS-WD scheme at the nal time t = 6000. The obtained numerical results are reported in Fig. 7 where we
show the water height at different times. The exact solution at each time is also shown in Fig. 7, where we see a perfect match
between our generated results and the exact reference solution.

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

0
2

1.5

0.5

0
t=0

0.5

1.5

0
2

(a) Water height at t = 0


4

3.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

0.5

0
t=0.148

0.5

1.5

0
2

(c) Water height at t = 0.148


4

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

0.5

0
t=0.49

0.5

1.5

0
2

(e) Water height at t = 0.49


4

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

0.5

0
t=0.148

0.5

0
t=0.07

0.5

1.5

1.5

(g) Water height at t = 1.25

1.5

0.5

0
t=0.278

0.5

1.5

1.5

0.5

0
t=0.8

0.5

1.5

(f) Water height at t = 0.8

0
2

0.5

(d) Water height at t = 0.278

0
2

(b) Water height at t = 0.07

3.5

0
2

1.5

0
2

1.5

0.5

0
t=1.5

0.5

1.5

(h) Water height at t = 1.5

Fig. 8. Symmetric double dam-break problem.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
14

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

1.02

0.98

m(t)/m0

0.96

0.94

0.92

0.9
50 gridpoints
100 gridpoints
200 gridpoints

0.88

0.86
0

3
t

Fig. 9. Symmetric double dam-break problem: graph of the curve m(t)/m0 on the interval [0,6] obtained on 50, 100, and 200 grid points.

4.5

4.5

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

0
2

1.5

0.5

0.5

1.5

2.5

0
2

1.5

0.5

t=0

4.5

4.5

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

0.5

0.5

1.5

2.5

0
2

1.5

0.5

t=0.19

2.5

0.5

1.5

2.5

2.5

(d) Water height at t = 0.3

4.5

4.5

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

0
2

1.5

t=0.3

(c) Water height at t = 0.19

(b) Water height at t = 0.1

(a) Water height at t = 0


5

0
2

0.5
t=0.1

1.5

0.5

0.5

1.5

2.5

0
2

1.5

0.5

t=0.42

0.5

1.5

t=0.49

(e) Water height at t = 0.42

(f) Water height at t = 0.49

Fig. 10. Dam break problem with irregular bumps.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117

15

4.1.3. Two symmetric dam break problems


We consider now a symmetric double dam-break problem facing a steep triangular bump. The computational domain is the
interval [2, 2] and reective boundary conditions are set at the endpoints of the domain. The variable waterbed function z(x)
and the initial water height function h(x, 0) are given by

if x < 1,
2 + x if 1 x 0,
z(x) =
2 x if 0 x 1,

1
if x 1


h(x, 0) =

and

2 if x < 1,
0 if 1 x 1,
3 if x > 1.

The initial water velocity v(x, 0) is set to zero. We discretize the computational domain using 100 grid points and compute the
numerical solution at the nal time t = 1.5 using the MC( = 1.5) limiter. The obtained results for the water height are reported
in Fig. 8(a)(h). Since the early computations, we see two symmetric rarefaction water waves propagating toward the line x = 0.
At the time t = 0.148 the two waves meet in the center of the computational domain and then they reect to form two shock
waves propagating toward the left and right boundaries of the computational domain. By the time t = 1.25 these two shock
waves have reected at the boundaries of the domain and have formed two new rarefaction waves propagating again toward the
center of the domain. This process is repeated in perfect symmetry about the line x = 0 as is shown in Fig. 8(g) and (h). We have
also computed the water mass m(t) across the computational domain at each time step and we compared the obtained results
to the initial water mass m0 . Fig. 9 shows the curve m(t)/m0 over the interval [0, tf ] on 50, 100 and 200 grid points. The obtained
results conrm the conservation of the water mass as x tends to zero.
4.1.4. Dam break problem facing irregular bumps
We consider for our last experiment a dam break problem facing two irregular bumps with wet and dry states. The computational domain is the interval [2, 2.5], and the dam is located at the point x = 1.5; the initial water height is h(x, 0) = 4 if
x < 1.5 and h(x, 0) = 0 elsewhere. The waterbed features a triangular bump facing the dam and a trapezoidal bump follows
5

4.5

4.5

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

0
2

1.5

0.5

0.5

1.5

2.5

0
2

1.5

0.5

t=0.67

(a) Water height at t = 0.67


5

4.5

4.5

3.5

3.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

0.5

1.5

2.5

1.5

2.5

(b) Water height at t = 1.95

0
2

0.5
t=1.95

0.5

1.5

2.5

0
2

1.5

0.5

t=3.55

0.5

t=10

(c) Water height at t = 3.55

(d) Water height at t = 10

Fig. 11. Dam break problem with irregular bumps (continued).

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

ARTICLE IN PRESS

JID: APM
16

[m3Gsc;October 22, 2015;13:26]

R. Touma / Applied Mathematical Modelling 000 (2015) 117


1.05
50 gridpoints
100 gridpoints
200 gridpoints

0.95

0.9

0.85

0.8

0.75

0.5

1.5

2.5

3.5

Fig. 12. Dam break problem with irregular bumps: graph of the curve m(t)/m0 on the interval [0,3.5] obtained on 50, 100, and 200 spatial grid points.

toward the right boundary of the domain as is shown in Fig. 10(a). Reecting boundary conditions are set on both boundaries of
the domain. The proposed scheme is applied and the numerical solution is computed at the nal time t = 10. The obtained results are reported in Figs. 10(a)11(d) where we show the water height at different times. As is shown in Fig. 10(a)(e) the water
front is moving from left to right crossing the both bumps, then reects at the right boundary of the computational domain, and
goes backward to cross again both bumps before reaching the left boundary (Figs. 10(f)11(d)). At the time t = 10, we see that the
water is spread in the regions separated by the two bumps to form three isolated lakes where water waves keep on propagating
back and forth in each of them.
The water mass m(t) across the computational domain was computed at each time step, and compared to the initial water
mass m0 . The obtained results are reported in Fig. 12 where we show the graph of the curve m(t)/m0 over the interval [0, tf ] on
50, 100 and 200 spatial grid points. The obtained results conrm the conservation of the water mass as x tends to zero.
5. Conclusion
In this work, we developed a new one-dimensional central nite volume method for the numerical solution of systems of
shallow water equations on variable waterbed with wet and dry states. The proposed scheme is a well-balanced unstaggered
central nite volume scheme especially designed to satisfy the lake at rest equilibrium state of the shallow water equations and
is capable of handling wet and dry states over irregular topographies. The well-balanced component of the scheme is ensured
thanks to the surface gradient method through a proper discretization of the water height in terms of the water level function. As
for the wetting/drying capability of the scheme, it is ensured through a proper interpolation of the water height components in
the forward and backward projection steps, and negative water heights at the wet/dry fronts are avoided by redening the slopes
of the interpolants in terms of the water bed elevations at the front locations. The resulting scheme is a second-order scheme
that evolves a piecewise linear numerical solution on a single grid, avoids the resolution of the Riemann problems at the cell
interfaces, is well-balanced at the discrete level due to a proper linearization of the source term similar to the discretization of the
ux terms, preserves the positivity of the water height. The proposed numerical scheme is then validated and classical shallow
water equation problems on variable waterbeds with wet and dry states are successfully solved. The obtained results show
very good agreement with corresponding ones appearing in the recent literature thus conrming the potential and eciency
of the proposed methods to handle general shallow water equation problems. We are currently working on a two-dimensional
extension of the proposed scheme that allows a physically admissible interaction between wet and dry states in two space
dimensions while preserving the well-balanced property of the numerical base scheme.
References
[1] H. Nessyahu, E. Tadmor, Non-oscillatory central differencing for hyperbolic conservation laws, J. Comput. Phys. 87 (2) (1990) 408463.
[2] P. Arminjon, M.-C. Viallon, Gnralization du schma de Nessyahu-Tadmor pour une quation hyperbolique deux dimensions despace, C. R. Acad. Sci.
Paris 320 (1995) 8588.
[3] A.N. Guarendi, A.J. Chandy, Nonoscillatory central schemes for hyperbolic systems of conservation laws in three-space dimensions, Sci. World J. 2013 (2013).
[4] G.-S. Jiang, E. Tadmor, Nonoscillatory central schemes for multidimensional hyperbolic conservation laws, SIAM J. Sci. Comput. 19 (6) (1998) 18921917.
[5] R. Touma, P. Arminjon, Central nite volume schemes with constrained transport divergence treatment for three-dimensional ideal MHD, J. Comput. Phys.
212 (2) (2006) 617636.
[6] P. Arminjon, M.-C. Viallon, Convergence of a nite volume extension of the Nessyahu-Tadmor scheme on unstructured grid for a two-dimensional linear
hyperbolic equation, SIAM J. Numer. Anal. 36 (1999) 738771.
[7] P. Arminjon, M.-C. Viallon, A. Madrane, A nite volume extension of the Lax-Friedrichs and Nessyahu-Tadmor schemes for conservation laws on unstructured
grids, Int. J. Comput. Fluid Dyn. 9 (1) (1998) 122.
[8] G. Jannoun, R. Touma, F. Brock, Convergence of two-dimensional staggered central schemes on unstructured triangular grids, Appl. Numer. Math. 92 (0)
(2015) 120.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

JID: APM

ARTICLE IN PRESS
R. Touma / Applied Mathematical Modelling 000 (2015) 117

[m3Gsc;October 22, 2015;13:26]


17

[9] R. Touma, G. Jannoun, Non-oscillatory central schemes on unstructured grids for two-dimensional hyperbolic conservation laws, Appl. Numer. Math. 62 (8)
(2012) 941955.
[10] J. Balbs, E. Tadmor, C.-C. Wu, Non-oscillatory central schemes for one-and two-dimensional MHD equations: I, J. Comput. Phys. 201 (1) (2004) 261285.
[11] A. Canestrelli, A. Siviglia, M. Dumbser, E.F. Toro, Well-balanced high-order centred schemes for non-conservative hyperbolic systems. Applications to shallow
water equations with xed and mobile bed, Adv. Water Resour. 32 (6) (2009) 834844.

L. Sopta, Balanced central NT schemes for the shallow water equations, in: Proceedings of the Conference on Applied Mathematics
[12] N. Crnjari
c-ic,
S. Vukovic,
and Scientic Computing, Springer, 2005, pp. 171185.
[13] R. Touma, Central unstaggered nite volume schemes for hyperbolic systems: Applications to unsteady shallow water equations, Appl. Math. Comput. 213
(1) (2009) 4759.
[14] G.-S. Jiang, D. Levy, C.-T. Lin, S. Osher, E. Tadmor, High-resolution nonoscillatory central schemes with nonstaggered grids for hyperbolic conservation laws,
SIAM J. Numer. Anal. 35 (6) (1998) 21472168.
[15] R. Touma, Unstaggered central schemes with constrained transport treatment for ideal and shallow water magnetohydrodynamics, Appl. Numer. Math. 60
(7) (2010) 752766.
[16] R. Touma, S. Khankan, Well-balanced unstaggered central schemes for one and two-dimensional shallow water equation systems, Appl. Math. Comput. 218
(10) (2012) 59485960.
[17] F. Alcrudo, P. Garcia-Navarro, A high-resolution Godunov-type scheme in nite volumes for the 2D shallow-water equations, Int. J. Numer. Methods Fluids
16 (6) (1993) 489505.
[18] F. Bouchut, Nonlinear Stability of Finite Volume Methods for Hyperbolic Conservation Laws: And Well-Balanced Schemes for Sources, Springer Science &
Business Media, 2004.
[19] A.I. Delis, T. Katsaounis, Numerical solution of the two-dimensional shallow water equations by the application of relaxation methods, Appl. Math. Model.
29 (8) (2005) 754783.
[20] P. Glaister, Approximate Riemann solutions of the shallow water equations, J. Hydraulic Res. 26 (3) (1988) 293306.
[21] L. Gosse, A well-balanced ux-vector splitting scheme designed for hyperbolic systems of conservation laws with source terms, Comput. Math. Appl. 39 (9)
(2000) 135159.
[22] R.J. LeVeque, H.C. Yee, A study of numerical methods for hyperbolic conservation laws with stiff source terms, J. Comput. Phys. 86 (1) (1990) 187210.
[23] S. Noelle, N. Pankratz, G. Puppo, J.R. Natvig, Well-balanced nite volume schemes of arbitrary order of accuracy for shallow water ows, J. Comput. Phys.
213 (2) (2006) 474499.
[24] S. Noelle, Y. Xing, C.-W. Shu, High-order well-balanced nite volume WENO schemes for shallow water equation with moving water, J. Comput. Phys. 226
(1) (2007) 2958.
[25] M.E. Vzquez-Cendn, Improved treatment of source terms in upwind schemes for the shallow water equations in channels with irregular geometry, J.
Comput. Phys. 148 (2) (1999) 497526.
[26] J.G. Zhou, D.M. Causon, C.G. Mingham, D.M. Ingram, The surface gradient method for the treatment of source terms in the shallow-water equations, J.
Comput. Phys. 168 (1) (2001) 125.
[27] A. Balzano, Evaluation of methods for numerical simulation of wetting and drying in shallow water ow models, Coastal Eng. 34 (1) (1998) 83107.
[28] J.M. Gallardo, C. Pars, M. Castro, On a well-balanced high-order nite volume scheme for shallow water equations with topography and dry areas, J. Comput.
Phys. 227 (1) (2007) 574601.
[29] M.J. Castro, A.F. Ferreiro, J.A. Garca-Rodrguez, J.M. Gonzlez-Vida, J. Macas, C. Pars, M.E. Vzquez-Cendn, The numerical treatment of wet/dry fronts in
shallow ows: application to one-layer and two-layer systems, Math. Comput. Model. 42 (3) (2005) 419439.
[30] J.M. Greenberg, A.-Y. Leroux, A well-balanced scheme for the numerical processing of source terms in hyperbolic equations, SIAM J. Numer. Anal. 33 (1)
(1996) 116.
[31] A. Bermudez, M.E. Vazquez, Upwind methods for hyperbolic conservation laws with source terms, Comput. Fluids 23 (8) (1994) 10491071.
[32] G. Bollermann, G. Chen, A. Kurganov, S. Noelle, A well-balanced reconstruction of wet/dry fronts for the shallow water equations, J. Sci. Comput. 56 (2)
(2013) 267290.
[33] Y. Xing, X. Zhang, C.-W. Shu, Positivity-preserving high order well-balanced discontinuous Galerkin methods for the shallow water equations, Adv. Water
Resour. 33 (12) (2010) 14761493.

Please cite this article as: R. Touma, Well-balanced central schemes for systems of shallow water equations with wet and dry
states, Applied Mathematical Modelling (2015), http://dx.doi.org/10.1016/j.apm.2015.09.073

You might also like