MATEMATIKA, 2008, Volume 24, Number 2, 211–230

c Department of Mathematics, UTM.

Modelling of Tsunami Waves

1
Nazeeruddin Yaacob, Norhafizah Md Sarif & 2 Zainal Abdul Aziz
Department of Mathematics, Faculty of Science, Universiti Teknologi Malaysia
81310 Skudai, Johor, Malaysia
e-mail: 1 [email protected], 2 [email protected]

Abstract The nonlinear dispersive model based on the forced Korteweg-de Vries
equation (fKdV) is developed from the approximation of Boussinesq shallow water
type model. This provides the possibility of observing, in particular, the process of
tsunami generation by atmospheric disturbances. The fKdV is then solved numerically
via an explicit finite difference method. From the simulations, the roles of nonlinearity,
dispersion and forcing terms in the process of tsunami generation are shown explicitly.
Keywords Tsunami waves; forced Korteweg-de Vries equation; finite difference method.

1 Introduction
The shallow water model basically describes the governing equations for the basic hydro-
dynamic model of tsunami generation by atmospheric disturbances (Akylas [1], Nosov &
Skachko [8]). This evolution equation is derived from the simplest theory of water waves
that reasonably approximates the behaviour of the real ocean and it comes from the system
of coupled partial differential equations (Pelinovsky et al. [9], Layton & van de Panne [7],
Guyenne & Grilli [4]).
An alternative approach to derive the appropriate evolution equation, which asymptoti-
cally approximates the Boussinesq equation, leads to the forced Korteweg-de Vries equation
(fKdV). The forcing term in the fKdV is assumed to be derivable from atmospheric distur-
bances. Various forms of this equation have been extensively studied (see Grimshaw et al.
[3], Pelinovsky et al [9]) and numerical results show that the solution contains the set of
solitary waves.
The behaviour of tsunamis on the open ocean is considerably more complicated than
the solitary wave model. Although some quarters doubt the relevance of soliton theory
to the modeling of tsunamis (e.g. Constantin & Johnson [2]), nonetheless, the approach
involving the solitary wave model gives an idea of the impressive scales involved in waves
in deep water (Sattinger & Li [10]). Therefore, a numerical study of this equation is done
in order to examine the phenomena of solitary wave in shallow water. In many cases, the
tsunami source moves with variable speed and direction. The fKdV equation has been
shown to describe the resonant mechanism of tsunami wave generation by atmospheric
disturbances moving with near-critical speed (Shen, [13]). In a series of papers, Shen [11]
& [12] summarized that there exist two supercritical solitary waves, and one subcritical
downstream cnoidal wave.
212 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

This paper aims to demonstrate the roles of nonlinearity, dispersion, and forcing terms
in the process of tsunami generation and also to develop a mathematical model in order
to study the characteristics of the tsunami mechanism. Predominantly this study aims
to portray the fKdV as a basic mathematical model of tsunami generation. The paper
is organized as follows. Section 1 is the introduction, section 2 describes the form of the
forced nonlinear dispersive equation, section 3 discusses the results and finally section 4 is
the concluding remark.

2 Nonlinear Dispersive Equation

Taking into account the resonant character of tsunami generation, the dynamical system can
be simplified. Here it is shown that by making further simplifying assumptions it is possible
to derive simpler nonlinear, dispersive wave models which can be used for our possible initial
numerical investigation. If it is assumed that waves travel in only the positive x-direction
then it can be shown that to a first approximation the waves have a steady form, i.e. the
nondimensional transformation,
X̃ = x̃ − t̃ (1)
can be introduced. Assuming that there is a slight variation with time or distance travelled,
the appropriate scaling is,
X̂ = σ X̃, (2)
T̂ = σ 3 T̃ . (3)
The free surface and horizontal velocity are scaled as before, according to the definitions
x̂ = σ x̃, t̂ = σ 3 t̃, η̂ = η̃ε (Ursell numbers) and û = ũε , respectively.
Substituting into a Boussinesq equation system (refer to Pelinovsky et al. [9]), leads to

∂ ūˆ ∂ ūˆ ∂ ūˆ ∂ η̂ σ 2 ∂ 3 ūˆ σ 4 ∂ 3 ūˆ 1 ∂patm

σ2 − + ε ūˆ + + − + = 0, (4)
∂ T̂ ∂ X̂ ∂ X̂ ∂ X̂ 3 ∂ X̂ 3 3 ∂ X̂ ∂ T̂
2 ερ ∂ X̂

∂ η̂ ∂ η̂ ∂
σ2 (1 + εη̂) ūˆ = 0,


− + (5)
∂ T̂ ∂ X̂ ∂ X̂
where patm is the pressure due to atmospheric disturbances.
To the lowest order of approximation both equations reduce to,

∂ η̂ ∂ ūˆ
+ O ε, σ 2


= (6)
∂ X̂ ∂ X̂
under the assumption that the initial conditions are consistent with this statement. Hence
by integrating, assuming that the order stays the same,

ˆ + O ε, σ 2 .


η̂ = ū (7)

Adding equations (4) and (5) and using the relation (7), we have

∂ η̂ ∂ η̂ σ 2 ∂ 3 η̂ 1 ∂patm
2σ 2 = O σ4 ,


+ 3εη̂ + + (8)
∂ T̂ ∂ X̂ 3 ∂ X̂ 3 ερ ∂ X̂
Modelling of Tsunami Waves 213

where it is assumed from the definition of the Ursell number that ε = O σ 2 . In dimen-


sionless variables this reads,

∂η ∂η ∂η ∂3η ∂f
+c + αη +β 3 = , (9)
∂t ∂x ∂x ∂x ∂x
where
β = ch2 6,
 
α = 3c 2h, f = −hpatm /2cρ. (10)
Equation (9) is the forced version of the Korteweg-de Vries equation and currently it is
considered as one of the basic nonlinear equations of mathematical physics. As mentioned
before, η (x, t) describes the free surface profile, f 0 (x) is the forcing function, x is the spatial
coordinate along the channel, and t is time. Its solution contains the set of solitary waves
and dispersive tails (variable wave packets, refer Shen [13]).

3 Results and Discussions

In this section, a numerical scheme is developed for the simulation of the fKdV equation. By
using an explicit finite difference method, this equation is solved by using a numerical scheme
in Matlab 6.1 (for example Hanselman & Littlefield [5]). The results of the computations
are presented for various initial conditions and time.

3. 1 Finite Difference Method

The finite difference method is developed by first partitioning the spatial domain into a set
of non-coincident points or nodes here, and for simplicity of presentation, the discussion is
limited to a constant spacing between adjacent nodes,

xi = x1 + (i − 1) ∆x, i = 1, 2, . . . , N, (11)

xupper − xlower
∆x = , (12)
N
and ∆x is termed the grid size or grid spacing.
The fKdV is solved numerically by this explicit finite difference method, where the basic
form being suggested by Vliegenthart [16]. The following notations are used.

∂η 1 1 d3 η
(ηi,j+1 − ηi,j ) − ∆x2 3 + o ∆x4 ,


= (13)
∂t ∆t 6 dt

∂η 1 1 d3 η
(ηi+1,j − ηi−1,j ) − ∆x2 3 + o ∆x4 ,


= (14)
∂x 2∆x 6 dx
∂ 3η 1
∂x3
= 2(∆x)3
(ηi+2,j − 2ηi+1,j + 2ηi−1,j − ηi−2,j )
5 (15)
7
∆x2 d η5 + o ∆x4 ,


+ 60
dx
ηi+1,j + ηi,j + ηi−1,j
η̄ = . (16)
3
214 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

Discretisation of the differential equation turns it to a difference equation. Substituting

(13)-(16) into fKdV equation (9) and only taking the first order term, we obtain
ηi,j+1 − ηi,j ηi+1,j − ηi−1,j
+c
∆t 2∆x
 
3c ηi+1,j − ηi−1,j ηi+1,j + ηi,j + ηi−1,j
+
2h 2∆x 3 (17)
!
2
ch ηi+2,j − 2ηi+1,j + 2ηi−1,j − ηi−2,j
+ 3 = f0.
6 2 (∆x)
Rearranging equation (17), we have
c ∆t c ∆t
ηi,j+1 = ηi,j − (ηi+1,j − ηi−1,j ) − (ηi+1,j − ηi−1,j ) (ηi+1,j + ηi,j + ηi−1,j )
2 ∆x 4h ∆x
ch2 ∆t
− (ηi+2,j − 2ηi+1,j + 2ηi−1,j − ηi−2,j ) + f 0 .
12 (∆x)3
(18)
This scheme is consistent with equation (9) and the truncation error is
h i h i
3 2
o (∆t) + o ∆t (∆x) .

We then simulate the difference equation (18) with an initial condition of form
ηi,0 = 12 sec h2 (i∆x − x0 ) . (19)
Another initial condition, for the development of undular bore in shallow water, can also
be set for equation (18) as
ηi,0 = 1/2 [1 − tanh(xi − 25)/5] . (20)
3.2 Stability of the Numerical Scheme
Explicit finite difference method is simpler to formulate, however it is less stable com-
pared to the other schemes of finite difference method such as the implicit method. A
numerical scheme is said to be stable if the graph of the solution does not change dramat-
ically in a short time span, otherwise it is said to be unstable (Tiong & Zainal, [15]). As
shown by Vliegenthart [16], the numerical scheme is only conditionally stable and expected
to be stable for ∆t ≤ ∆x3 .
The accuracy of the numerical calculations was checked in various ways; first, the known
exact linear solution was compared with the corresponding numerical solution. Secondly,
the numerical scheme was successfully tested against known soliton solutions of the KdV
equation. Finally, the use of refined integration steps ∆t, ∆x and a larger computational
domain for x did not alter the results within the accuracy reported here, especially in the
domain x.
In order to investigate influences of mesh size on the numerical simulation for the KdV
equation, two different meshes ( ∆t = 0.0005, ∆t = 0.001) are considered. However, using
these values to our numerical scheme, we found that the numerical scheme is unstable (see
Figure 1 (a) and Figure 1(b)). From these figures, these show the influences of different
mesh sizes on the magnitude and distribution of simulated wave crests and trough. It is
observed that the smaller mesh size gives better results.
Modelling of Tsunami Waves 215

(a)

(b)

Figure 1: Numerical Demonstration of the Instability of a Solitary Wave of the KdV Equa-
tion (a) ∆t = 0.0005, (b) ∆t = 0.001
216 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

3.3 Numerical Results

The numerical procedure below is carried out with ∆x = 0.1. For the present problem,
∆x = 0.1, ∆t = 0.25 × 10−3 are sufficient for stability, in accordance with Vliegenhart [16].

3.3.1 Simulation on the Forced KdV

We now proceed to simulate the solitary wave evolution for KdV equation with the forcing
in motion. To demonstrate the roles of solitons in the forced dynamics, the assumption
in the scheme has been made that the forcing term being introduced is sufficiently weak
(Pelinovsky et al. [9]), so that the soliton retains its form in the process of interaction.
We consider for the cases h = 1.0 m and h = 0.5 m. Until up to t = 300, it is found
that the solitary wave does not change its shape. However, at (c) t = 1000, the soliton
interaction begins and from Figure 2(c) the non-linear steepening dominates. Now, it is the
dispersion term that is small and we can see the effect of the nonlinear ‘convection’ term.
Where η is large the ‘convection’ is rapid, but out in front where η is small the ‘convection’
is slower. This allows the fast peak to catch up with the slow front end, causing wave
steepening. (An effect like this causes ocean waves to steepen and then break at the beach).
In the case of three solitons, the wave packet has very small amplitude. It is readily seen
that the peaks of the soliton lie approximately on a straight line. This is due to the fact
that the velocity of the soliton is proportional to its amplitude, and so that the distances
covered by them will be proportional to their amplitude.
The phenomena of a solitary wave moving into a region of decreasing depth and produc-
ing solitons are shown in Figures 2(a)-2(c), 3(d)-3(f), 4(g) and 5(a)-5(f). From the above
result, it shows many similarities with the standard theory of the KdV equations. Standard
KdV theory then predicts that exactly n solitons will eventually appear on the shelf if
 −4/9
1
h= n (n + 1) (21)
2

where the soliton amplitude is 2Am2 /n (n + 1) , m = 1, 2, . . . , n, where A is the amplitude

of the initial solitary wave. This prediction is made by Johnson [6].
In our simulation of the fKdV equation, we have found that for Case 1 at h = 1.0 m,
3 solitons were produced with an oscillatory tail (Figure 3(f)). In this case, according to
equation (21), this corresponds to 1 < n < 2, and 2 solitons would be produced. For Case
2 at h = 0.5 m, the case where the depth is a bit shallower than Case 1, 6 solitons were
produced with an oscillatory tail behind them (Figure 5(e)). According to equation (21),
this corresponds to 2 < n < 3, i.e. 3 solitons would be produced. It is seen that in the
simulations, equation (21), which is derived for the variable coefficient KdV equation is
not compatible for the above cases (also see Tiong & Zainal, [15]). This certainly requires
further examination.
Now, we focus on the effect of forcing in soliton. Due to the effect of forcing, the solitary
wave amplitude and position will change. Numerical simulations of this equation show
that the imposition of forcing generates soliton, as well as nonlinear wave packet. In our
simulation of the forced Korteweg-de Vries equation, it seems that the soliton amplitude
grows at the moment of interaction and recovers after interaction. Figures 2, Figures 3 and
Figures 5 (a)-(f) show the evolution of the wave disturbance at different values of h. Solitons
Modelling of Tsunami Waves 217

are successively generated at N < 150 and propagate in front of the pressure distribution:
after each soliton reaches certain equilibrium amplitude, a new soliton of slightly smaller
equilibrium amplitude is released. Immediately, behind the pressure excitation, a trough
appears (to balance more or less the amount of water used to form a soliton), and at larger
distances from pressure, the wave disturbance is highly oscillatory with a larger amplitude.
As shown by the numerical solution of the fKdV equation, the simplified model given
here yields a physical representation of the interaction between the soliton and forcing in
motion. In reality, the number of waves interacting with the atmospheric disturbances can
be high and they may interact between them, sometimes forming large amplitude waves.

3.3.2 Simulation of Undular Bore

If we change slightly the initial condition in the fKdV equation i.e. using (20) instead, then
another simulation will be formed, which is known as the undular bore. The term bore is
used for a moving discontinuity. The bore or hydraulic jump has exactly the same dynamic
function as shock wave that is it converts the kinetic energy in the incident flow into some
other forms.
In shock wave, the energy goes into random motion of the molecules, and generates heat.
In the bore, the energy then becomes transformed into a random motion of the water on
a small scale, and produces turbulence. Nonetheless, in some circumstances part or all of
the energy goes into generating surface waves, giving an undular bore situation. Figures 6
(a-f) show these possibilities.
For some initial conditions it is true that an initial wave of elevation will steepen and
eventually form a hydraulic jump or bore, as mentioned before, and this will happen when
the nonlinearity is quite strong, as is shown in Figure 6 (c). But if the initial wave of
elevation is not very marked, then other terms in the full equation have an effect and a
moving equilibrium state may be reached which propagates unperturbed shape. Another
reason why a bore may not form in certain circumstances is that friction (due for example,
to turbulence or obstructions) may be relatively important, and this can remove energy at
such a rate that no discontinuity needs to be formed.
Tsunami can thus be considered as a big solitary wave, i.e. soliton. Hence, tsunami
may possesses similar properties of soliton such as representing a wave of permanent form,
localized, and interact strongly with other solitons. In our simulations, we observe that
for large t, the graph consists of a finite number of solitons depending on the depth and
initial condition. When a soliton propagates into a different depth, as a tsunami leaves the
deep water and moves to shallower water, it will deform. This situation is certainly due to
the nonlinear and dispersive effects, which depend on the water depth, and will no longer
balance each other anymore as in the case of constant depth. The larger the nonlinearity,
as observable from the simulations, the bigger is the amplitude produced. This case also
explains the freak wave formation on shallow water waves. Freak wave is always a wave
with small ratio of nonlinearity to dispersion, so it is almost a linear dispersive wave (e.g.
Tan [14])
As obtained in section 3.3.1, the influence of nonlinear effects on the large amplitude
appearance is demonstrated and its probability of occurrences increases as nonlinearity
becomes significant. Consequently, the results obtained thus far would allow the study of
unidirectional waves in shallow water within the precepts of KdV model, and which can be
related to tsunami waves. Such numerical experiments demonstrated dependence of wave
218 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

statistics and ratio of nonlinear effect to dispersion, and what leads to the growing of the
asymmetry in wave field with time increasing of the contribution of the large waves in total
distribution.
Since the velocities of the solitons are proportional to their amplitude, solitons are
arranged in order of increasing amplitude, i.e. the largest are in front. The rapidly wave
packet (or tail) is seen to remain behind in accordance to the dispersion equation. It can be
shown that from our simulation, if we begin from an initial profile, such as of Figure 2(a)
and Figure 5(a), but with the taller wave somewhat to the left of the shorter, then the
development observed is as depicted in Figure 5(c). In this case, the taller wave catches up
and becomes intact and undistorted.

4 Conclusions
We have discussed the basic modelling of tsunami waves. The characteristics of tsunami
have been compared and the problem of tsunami is formulated into a mathematical model,
with a few physical considerations taken into account. Shallow water theory is seen as a
viable theoretical model to describe the properties of tsunami waves. Nonetheless from the
results above, we can conclude that
(a) Tsunami wave can be shown to act as a soliton and solitary wave and to possess
similar properties of solitons.
(b) As the depth of sea decreases, the number of solitons increases as the nonlinear
effects become larger than the dispersive effects.
(c) The larger the nonlinearity, the bigger amplitudes of the waves are produced.
(d) Due to the effect of forcing, the solitary wave and its position will change. A trough
will appear (to balance more or less the amount of water used to form a soliton) when the
soliton interact with the weak forcing.
(e) The velocities of the solitons are proportional to their amplitudes, and solitons are
arranged in order of increasing amplitude, i.e. the largest are in front.

Acknowledgment
This research was partially supported by MOHE Fundamental Research Grant, Vot. No.78082.

References
[1] T.R. Akylas, On The Excitation of Long Nonlinear Water Waves By A Moving Pres-
sure Distribution, Journal Fluid Mechanics, 141(1984), 455-466

[2] A. Constantin & R.S. Johnson, Modelling Tsunamis, J. Phys. A: Math. Gen, 39(2006),
L215-L217.

[3] R. Grimshaw, E. Pelinovsky & X. Tian, Interaction of a Solitary Wave with an External
Force, Physica D, 77(1994) 405-433.

[4] P. Guyenne & S.T. Grilli, Computation of Three-Dimensional Overturning Waves in

Shallow Water, Dynamics and Kinematics, in Proc. 13th Int. Offshore and Polar Eng.
Conf. Honolulu, Hawaii, 2003.
Modelling of Tsunami Waves 219

[5] D. Hanselman & B. Littlefield, Mastering Matlab R 6, A Comprehensive Tutorial and

Reference, Prentice Hall, Upper Saddle River, N.J., 2001
[6] R.S. Johnson, The Korteweg-de Vries Equation with Variable Coefficients, Lecture in
Applied Mathematics, American Mathematical Society, 15, 1974.
[7] A.T. Layton & van de Panne, A Numerically Efficient and Stable Algorithm for Ani-
mating Water Waves, Visual Computer, 18(2002), 41-53
[8] M.A. Nosov & S.N. Skachko, Nonlinear Tsunami Generation Mechanism, Natural Haz-
ard and Earth Sciences, 1(2001), 251-253.
[9] E. Pelinovsky, T. Talipova, A. Kurkin & C. Kharif, Nonlinear Mechanism of Tsunami
Wave Generation by Atmospheric Disturbances, Natural Hazard and Earth Sciences,
1(2001), 243-250.

[10] D.H. Sattinger & Li, Yi, Matlab Codes for Nonlinear Dispersive Wave Equations, Tech-
nical Report, at http:// www.math.usu.edu/∼dhs/, 1998.
[11] S.S. Shen, Locally Forced Critical Surface Waves in Channels of Arbitrary Cross Sec-
tion, Journal of Fluid Mechanics, 42(1991), 147-188.
[12] S.S. Shen, Forced Solitary Waves and Hydraulic Falls in Two Layer Flows, Journal of
Fluid Mechanics, 234(1992), 583-612.
[13] S.S Shen, A Course on Nonlinear Waves, Kluwer Academic Publisher, (1993) 53-70.
[14] W.N. Tan, Theory of Bichromatic Wave Groups Amplitude Amplification using Implicit
Variational Method, PhD Thesis, Universiti Teknologi Malaysia, Skudai, 2006.
[15] K.M. Tiong & A.A. Zainal, On The Splitting of A Solitary Wave Propagating Over a
Slowly Varying Topography, in Prosiding Simposium Kebangsaan Sains Matematik XI,
Amran Ahmed et al (eds.), (2003) 149-157, UMS, Kota Kinabalu.

[16] A.C. Vliegenthart, On Finite Difference Methods for the Korteweg de Vries Equation,
Journal of Engineering Mathematics, 5(1970), 137-155.
220 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

Case 1: h = 1.0 m

(a)

(b)
Modelling of Tsunami Waves 221

(c)

Figure 2: (a-c) Show an Evolution of Solitary Wave at Shallow Water

222 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

(d)

(e)
Modelling of Tsunami Waves 223

(f)

Figure 3: (d)-(f) Show Collision of Three KdV Soliton; the Evolution of the Initial Profile
u (x, 0) = 12 sech2 x, at (d) t = 3000, (e) t = 5000, and (f) t = 10000 (continued)
224 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

(g)

Figure 4: The Evolution of Forced KdV at (g) t = 15000

Modelling of Tsunami Waves 225

Case 2: h = 0.5 m

(a)

(b)
226 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

(c)

(d)

The wave starts to steepen earlier when the depth is decreasing.

Modelling of Tsunami Waves 227

(e)

(f)

Figure 5: (a)-(f) Show the Evolution of Forced KdV at (a) t = 0, (b) t = 500, (c) t =
1000, (d) t = 30000, (e) t = 5000, (f) t = 10000
228 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

(a)

(b)
Modelling of Tsunami Waves 229

(c)

(d)
230 Nazeeruddin Yaacob, Norhafizah Md Sarif & Zainal Abdul Aziz

(e)

(f)

Figure 6: (a)-(f) Show the Evolution of Undular Bore at Shallow Water