Journal of the Faculty of Environmental Science and Technology,

Okayama University
VoUl, No,l, pp,1518, March 2006
Numerical Techniques for Simulation of Tsunami Based
on Finite Elements
Masaji Watanabe
l
, Ying Liu
2
and Ming Jun Wang
3
(Received December 27, 2005)
Numerical techniques to simulate tsunamis are described. Partial differential equations are re-
duced to a system of ordinary differential equations to which appropriate numerical solvers can be
applied. The techniques are illustrated with an example in which tsunami due to an earthquake
is simulated.
Key words: tsunami, finite elements, numerical simulation
1 INTRODUCTION 2 REDUCTION OF TSUNAMI MODEL
where Xj and Yj are the coordinates of the jth node.
(Pi(Xj,Yj)=bij, i=1,2, ... ,n, j=1,2, ... ,n,
A mathematical model for simulation of tsunami is the
system of partial differential equations:
M= t;udz, N= [(vdz.
J-h Lh
(1)
aM o(
at +9 (h +() 8x = 0
8N 8(
tit +9 (h +() 8y = 0
8( + 8M + aN = 0
at ax 8y
Suppose that a region in xy plane is divided into m
triangular elements with n nodes. Let (Pi (x,y) be the
continuous piecewise linear basis function that satisfies
Those equations are based on the continuity equation
and the equations of motion to analyze long waves
(Dean and Dalrymple, 1991). Here, z = (and z = -h
correspond to the water surface and the ocean floor,
respectively. Let u and v denote x-component and Y-
component of the velocity, respectively. M and N are
defined by
The earthquake of magnitude 8.9 occurred in an off-
shore area of the north end of Sumatra around 1 a.m.
GMT on December 26,2004. The tsunami caused by
the earthquake reached the maximum height of 10 m,
and devastated the coastal areas of Indian Ocean (De-
cember 27, 2004, YOMIURI ONLINE, in Japanese).
The number of the deaths due to the disaster exceeded
230,000 (January 25, 2005, YOMIURI ONLINE, in
Japanese). The establishment of warning systems to
prevent damages from such disasters is essential.
Techniques based on shallow water equations have
been proposed to analyze tsunami waves (Kawahara
et al., 1988, Kodama et al., 1993, Lye, 2005) In this
paper, a technique based on finite elements is intro-
duced to simulate tsunami waves. In the following
sections, techniques to analyze flows generated in Ko-
jima Lake as well as numerical results are introduced.
In Section 2, a technique to reduce partial differential
equations to a systemof ordinary differential equations
are described. In Section 3, those techniques are illus-
trated with an example. A tsunami in Nankai Trough
area is simulated. In Section 4, the difference between
the technique introduced in this paper and some other
methods is discussed.
lGraduate School of Environmental Science, Okayama Uni-
versity.
2Department of Environmental and Mathematical Sciences,
Faculty of Environmental Science and Technology, Okayama
University.
3Graduate School of Natural Science and Technology (Mas-
ter's Course), Okayama University.
15
16 J. Fac. Environ. Sci. and Tech .. Okayama Univ.11 (1) 2006
Then M, N, (, and h can be given by
n
M (x,y, t) :::::!
I:Mj (t)j (x,y)
j=l
n
N(x,y,t) :::::! I:Nj (t)j (x,y)
j=l
n
((x,y,t) :::::!
I:(j (t) j (x, y)
j=l
n
h (x, y) :::::!
I:hjj (x, y)
j=l
Substitution of these expressions into the system (1)
leads to
The value of the partial derivatives on the right-
hand sides at (x, y) = (Xi, Yi) can be approximated
by the average of the partial derivatives over the ele-
ments which have the node in common as one of their
vertices. Then the system (2) can be regarded as a
system of ordinary differential equations.
For (x, y) = (Xi, Yi), these equations become
dM
i
a(
dt = -g (hi + (i) ax
dN
i
a(
----;It = -g (hi + (i) ay
d(i aM aN
- -----
dt ax 8y
(2)
Figure 1: Topography of the ocean floor.
on this rectangular grid was set. It consists of 160000
elements and 80601 nodes.
Figure 1 shows the topography of the ocean floor
generated by the values of h introduced into the com-
putational analysis. The depth of the ocean in the area
ranges approximately from 0 m to 4500 m. Figure 2
shows the surface of the sea initially set. It is based
on Gaussian hump (Lye, 2005).
(x - xO)2 + (y - YO)2 < 4(72
(x - XO)2 + (y - YO)2 4(72
Here (xo, Yo) corresponds to the east longitude 134
and the north latitude 32, a = 10, (7 = 10000. The
initial values of (i were generated according to this
expression of (. The initial values of M
i
and N
i
3 NUMERICAL RESULTS FOR TSUNAMI
SIMULATION
A coordinate system whose origin corresponds to the
east longitude 130 and the north latitude 30 and unit
length corresponds to 1 meter was set. A tsunami was
simulated over a rectangular region
Xmin X Xmax , Ymin Y Ymax,
where Xmin = 200000, X
max
= 1000000, Ymin = 10000,
and Ymax = 410000. The region lies between east lon-
gitude 131 and 140 and north latitude 30 and 36,
and contains a part the north edge of Nankai 'Though.
The interval [Xmin, x
max
] was divided into 400 equally
spaced intervals. Similarly, the interval [Ymin, Ymax]
was divided into 200 equally spaced intervals, and a
rectangular grid was set. A finite element mesh based
were set.
The fourth-order Adams-Bashforth-Moulton predictor-
corrector (PECE mode) was implemented to solve the
system (2) numerically in conjunction with the fourth-
order Runge-Kutta method to generate approximate
solutions in the first three steps (Lambert, 1973). Fig-
ures 3, 4, 5, and 6 show the transition of the surface
of the sea.
4 CONCLUSION
It has been shown that partial differential equations of
fluid dynamics can be reduced to system of ordinary
differential equations to simulate tsunami implement-
ing a finite element mesh. It has also been shown that
ODE solvers can be applied to solve those ordinary
t.lasaji WATANABE et al. I :\umerical Techniques for Simulation of Tsunami 17
Figure 2: Initial surface of the sea.
~
Figure 3: Surface of the sea at t = 60.
.... .-
_. .-.--.
Figure 4: Surface of the sea at t = 300.
Figure 5: Surface of the sea at t = 600.
18 J. Fac. Emiron. Sci. and Tech .. Okayama Univ.ll (1) 2006
Figure 6: Surface of the sea at t = 900.
differential equations effectively. The values of spatial
derivatives at each node was replaced with average val-
ues of the derivatives over the elements that have the
node in common as one of vertices (Okamoto et al.,
1992) and the method of collocation (Mitchell et al.,
1977) was implemented. The technique has an advan-
tage over a method based on weak formulation that
leads to a system of linear equations with a stiffness
matrix. It also has an advantage over a finite differ-
ence method for it can be adapted to the structure of
a coastline.
ACKNOWLEDGMENTS
Data concerning the depth of the ocean are based on
the Ocean Sounding Chart, G1406, produced by the
Hydrographic Department, M. S. A., Japan, 1985, and
Bathymetric Chart, 6602, Nippon, Offing of Tokai-Kii,
compiled from the various sources of the Hydrographic
Department, M. S. A., Japan, to 1992, 1993, Maritime
Safety Agency, Tokyo. The articles of YOMIURl ON-
LINE cited in this paper were featured in the web site,
http://www.yomiuri.co.jp/features/india/ .
REFERENCES
Dean, R. G. and Dalrymple R. A. (1991): Water Wave
Mechanics for Engineers and Scientists, Advanced
Series on Ocean Engineering - Volume 2, World
Scientific, Singapore.
Kawahara, M., Kodama, T., and Kinoshita, M. (2005):
Finite element method for tsunami wave prop-
agation analysis considering the open boundary
conditions, Computers & Mathematics with Ap-
plications, 16 (1/2) (1998), 139-152.
Kodama, T. and Kawahara, M. (1993): A finite el-
ement analysis of shallow water flow with open
boundary condition, Computational Mechanics,
Valliappan, Pulmano & Tin-Loi (eds) 1993
Balkema, Rotterdam, ISBN 90 5410 333 7 (1993),
11-18.
Lambert, J. D. (1973): Computational Methods in
Ordinary Differential Equations, John Wiley &
Sons, Chichester.
Lye, K. H. (2005): Modeling propagation of 2004
Asian Tsunami: A theoretical Analysis, Inter-
national Conference of Reservoir Operation and
River Management, Guangzhou & Three Gorges,
China, September 17-23, 2005.
Mitchell, A. R. and Wait, R. (1977): The Finite El-
ement Method in Partial Differential Equations,
John Wiley & Sons, Chichester.
Okamoto, T, Kawahara, M., loki, N., and Nagaoka,
H. (1992): Two-dimensional wave run-up anal-
ysis by selective lumping finite element method,
International Journal for Numerical Methods in
Fluids, 14 (1992), 1219-1243.