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

Eective Coastal Boundary Conditions for Water Waves

W. Kristina
1,
, O. Bokhove
1
, E. van Groesen
1,2
1
Department of Applied Mathematics, University of Twente, The Netherlands,
2
Lab-Math Indonesia

w.kristina@math.utwente.nl
Introduction
Fig. 1: Japan tsunami on 11 March 2011 (left) and harbor wave (right).
Numerical models of tsunami propagation from the location of
generation to the arrival at coastlines must deal with the ne
scale details in the generation region, the proper large-scale
long-wave propagation across the oceans to the coast, and the
subsequent ne-scale run-up and run-down ooding nearshore.
It is a daunting task for numerical models to capture such a
variation in spatial and temporal scales. Some shallow water
tsunami models therefore approximate the coastline, or large
tracts thereof, with an inpenetrable vertical wall at a certain
depth contour. Obviously, the reection properties of such a
boundary condition can be very unrealistic. We wish to alleviate
this shortcoming by an investigation of partially transparent,
so-called eective boundary conditions, instead of these
solid-wall boundary conditions.
Fig. 2: Illustration of the domain decomposition for our eective boundary
treatment.
In one horizontal spatial dimension, an outline of the desired
mathematical modelling is sketched in Fig. 2. In the deep ocean
for x > 0 and x < B with horizontal coordinate x, denoted as
the simulation area, we model the wave propagation numerically
and provide the information of incoming wave at x = B by
using observation operator O. In the coastal zone for
x [B, L], denoted as the model area, we model (M) the wave
propagation (semi) analytically and use the information that
account for the reected waves to be inuxed back into the
simulation area using inux operator I.
Linear Variational Boussinesq Model:
Continuum Case
Nonlinear, potential ow water waves are described by the
variational principles in [1, 2]. Following [3] the variational
principle for the linearized Boussinesq equations
0 =

T
0
L[, , ]dt =

T
0

L
0

t

1
2
g
2

1
2
h|
x
|
2

x

1
2
|
x
|
2

1
2

2
dxdt
(1)
with the functions (x)
2
3
h(x), (x) =
8
15
h(x), and
(x) =
4
3h(x)
.
When we ignore the underlined terms in (1), the linearized
shallow water equations are obtained as limiting system. Taking
variations with respect to , , , we nd the equations for
Linear Variational Boussinesq Model (LVBM) as follows:

t
+
x
(h
x
) +
x
(
x
) =0 (2a)

t
+ g =0 (2b)

x
(
x
) +
x
(
x
) =0 (2c)
with the dispersion relation is given by

2
= ghk
2

1

2
k
2
h(k
2
+ )

. (3)
Linear Variational Boussinesq Model:
Semi-Discrete Case
We a priori divide the domain into two intervals, from x [0, B]
and from x [B, L], such that we are forced to consider the
boundary conditions coupling these two intervals. The region
x [0, B] will be approximated using a classical Galerkin nite
element expansions

h
(x, t) =
j
(t)
j
(x)

h
(x, t) =
j
(t)
j
(x)

h
(x, t) =
j
(t)
j
(x)
(4)
concerning N
k
elements.
To ensure continuity and a unique determination, we substitute
(x, t) =

(x, t) +
N+1
(t)
N+1
(x)
(x, t) =

(x, t) +
N+1
(t)
N+1
(x)
(x, t) = (x, t) +
N+1
(t)
N+1
(x)
(5)
with
N+1
in an element N
k
+ 1 for x [B, L] and with

(B, t) = (B, t) =

(B, t) = 0.
For linear polynomials, use of (4) and (5) into (1) yields
0 =

T
0
M
kl

1
2
gM
kl

1
2
S
kl

1
2
A
kl

l
B
kl

1
2
G
kl

l
+

L
B

t

1
2
g
2

1
2
h(
x
)
2

1
2
(
x
)
2
(
x
)(
x
)
1
2

2
dxdt
(6)
where we introduced mass and stiness matrices M
kl
, S
kl
, A
kl
,
B
kl
, G
kl
, and used endpoint conditions
k
(0) = (T) = 0,
connection conditions (B, t) =

(B, t) =

(B, t) = 0, and
no-normal through ow conditions at x = 0, L.
Taking variation of
k
,
k
, and
k
for the (N + 1)st
elements, the equations arising from (6) for x [0, B] are
M
kl

l
S
kl

l
B
kl

l
+
k(N+1)
(h
x
+
x
)|
B+
=0 (7a)
M
kl

k
+ gM
kl

k
=0 (7b)
A
kl

l
+ B
kl

l
+ G
kl

l
+
k(N+1)
(
x
+
x
)|
B+
=0 (7c)
with Kronecker delta symbol
kl
, one when k = l and zero
otherwise, and the equations (2) for x [B, L].
Eective Boundary Condition over Flat
Bathymetry
Considering harmonic analysis for the LVBM in the case of , ,
, and h are constants and focusing on long waves, we get an
approximation for the linear VBM observation and inux
operator as follows
O(, ) (
t
c
0

x
c
0

2h

x
) 2g
inc
(t) (8a)
I (, ) (
t
+ c
0

x
+ c
0

2h

x
) 2g
refl
(t) (8b)
We use as a simple model M the uniform translation as in
linear shallow water theory such that

refl
(t) = M(
inc
)(t) =
inc
(t ). (9)
with reection coecient [0, 1] and delay = 2(L B)/c.
But since the LVBM is a dispersive model, instead of
c = c
0
=

gh, we use as propagation speed c = c
p
, the speed
of waves at peak frequency, which is determined via the LVBM
dispersion relation.
Fig. 3: We compare numerical results between a simulation in the whole
domain (dashed line) and one with the EBC (solid line) with B = 1km. Wave
proles at times t = 0, 20, 40, . . . , 120s are shown on the left both simulations,
and on the right a snapshot at t = 110s. The deviations are caused by the
approximation to the dispersive wave propagation in the model area
[B, L] = [1, 1.2]km.
Eective Boundary Condition over Slowly
Varying Bathymetry - WKB approximation
Using higher order WKB approximation, the reection
model M over slowly varying bathymetry is given by

refl
(t, B) =
t/2
0

inc
(t 2)

B() d (10)
with

B() =

0

b()d and

b() = b ((x)) depends on
the bathymetry according to
b (x) =
1
4
c

(x)
2
+
1
2
c (x) c

(x) . (11)
Simplied Aceh bathymetry with
transparent coastline
Fig. 4: Approximation (dashed line) of Aceh bathymetry (solid line) in
the entire domain (left) and an inset (right).
Fig. 5: Comparisons between simulation in the whole domain (dashed
line) and using the EBC (solid line) for several time steps (left) and
snapshot at t = 4800s (right).
Wind Waves and Harbour Reections
Fig. 6: Comparisons between simulation in the whole domain (dashed
line) and using the EBC (solid line) for wind waves in a harbor for
several time steps (left) and snapshot at t = 195s (right).
Conclusions
The EBC for linear VBM has been formulated based
on well-known shallow water theory and the WKB
approximation.
Using variational principles, the EBC has been
integrated with a nite element treatment in the
simulation area, and analytical, asymptotic methods
in the model area.
Some numerical test cases of increasing complexity
has been done to verify and validate our approach.
Acknowledgement
This work is under the Nearshore Tsunami Modelling and Simulations
project funded by the Dutch Science Foundation (NWO).
References
[1] J.C. Luke: A variational principle for uids with a free surface,
J. Fluid Mech., 27, 395, 1967.
[2] J.W. Miles: On Hamiltons principle for surface waves, J. Fluid
Mech. 83:153158, 1997.
[3] G. Klopman, E. van Groesen, M. Dingemans: A variational
approach to Boussinesq modelling of fully non-linear water
waves. J. Fluid Mech., 657, 3663, 2010.

You might also like