Under consideration for publication in J. Fluid Mech.

1
Undular bore on a slope
G.A. El, R.H.J. Grimshaw and W.K. Tiong
Department of Mathematical Sciences, Loughborough University,
Loughborough LE11 3TU, UK
(Received )
We consider the propagation of a shallow-water undular bore over a gentle monotonic
bottom slope connecting two regions of constant depth, in the framework of the variable-
coecient Korteweg de Vries equation. We show that, when the undular bore advances
in the direction of decreasing depth, its interaction with the slowly varying topography
results, apart from an adiabatic deformation of the bore itself, in the generation of a
sequence of isolated solitons an expanding large-amplitude modulated solitary wave-
train propagating ahead of the bore. Using nonlinear modulation theory we construct an
asymptotic solution describing the formation and evolution of this solitary wavetrain. Our
analytical solution is supported by direct numerical simulations. The presented analysis
can be extended to other systems describing the propagation of undular bores (dispersive
shock waves) in weakly non-uniform environments.
1. Introduction
Description of the interaction of a shallow water wave with variable topography is a
classical and fundamental problem of uid mechanics. This problem has been thoroughly
studied for linear waves (see, e.g. Johnson (1997) and references therein) as well as for
isolated solitary waves (see e.g. Grimshaw (2007) and references therein). Much less is
known about the propagation of nonlinear wavetrains over obstacles. In spite of being
of considerable interest for applications, in particular, in coastal oceanography (see e.g.
2 G.A. El, R.H.J. Grimshaw and W.K. Tiong
Scotti et al. (2008), Tissier et al. (2011) and references therein) this problem remains
almost unexplored theoretically except for very special model cases where exact analytic
solutions are available (see Ostrovsky & Pelinovsky (1975), Miles (1983a), and Grimshaw
(2007)).
Nonlinear wavetrains are often generated in the form of unsteady undular bores con-
necting two dierent basic ow states and exhibiting solitary waves at the edges. Undular
bores are usually formed as a result of dispersive resolution of a shock or an initial dis-
continuity in uid depth/velocity (see e.g. Smyth & Holloway (1988), El et al. (2006),
Esler & Pearce (2011)) or due to the resonant interaction of a uid ow with variable
topography (see e.g. Grimshaw & Smyth (1986), Baines (1995), El et al. (2009)).
In the framework of a weakly nonlinear long wave paradigm, the propagation of a
shallow-water undular bore over a at bottom, in the absence of dissipation eects,
is asymptotically described by the slowly modulated periodic solution of the constant-
coecient KdV equation. The evolution of the modulation parameters (such as mean
value, amplitude, wavenumber etc.) in the undular bore is then described by an expansion
fan solution of the Whitham averaged equations (Gurevich & Pitaevskii (1974), Fornberg
& Whitham (1978)).
In the presence of variable topography, the relevant model for the weakly nonlinear
long-wave propagation is the variable-coecient KdV equation which, for slow topogra-
phy variations can be asymptotically reduced to the perturbed constant-coecient KdV
equation. The corresponding perturbed modulation equations were derived in El et al.
(2007). They take into account variations of the undular bore parameters due to varying
environment and bottom friction, in addition to the modulations resulting from spatial
non-uniformity of the initial or boundary conditions. Analysis of the behaviour of the
characteristics of the modulation system for the perturbed KdV equation in the vicinity
Undular bore on a slope 3
of the leading edge of the undular bore in El et al. (2007) has enabled one to identify
two possible scenarios of the undular bore evolution with respect to the behaviour of its
leading solitary wave:
(i) Weak interaction (local) scenario. In this scenario the undular bore propagates
so that its lead soliton evolves as an isolated, adiabatically varying solitary wave;
(ii) Strong interaction (nonlocal) scenario. The undular bore evolves so that its lead
soliton evolution is determined not only by the local variations of the topography or/and
the presence of bottom friction but also by the interaction with the entire nonlinear
wavetrain behind it.
The actual scenario of the propagation of an undular bore over a slope depends on the
slope polarity and on the relative values of the characteristic scales in the problem (the
typical slope value, the bottom friction coecient, the typical wavelength in the bore
and the typical spatial scale for modulations).
In this paper, we further explore the undular bore on a slope problem by considering
in detail a special case when the undular bore propagation occurs over a broad but
nite region of decreasing depth. Such a conguration is relevant to the modelling of
the near-shore tsunami propagation (see e.g. Madsen et al. (2008), Tissier et al. (2011)
and references therein) but is also of a broader signicance in the context of the general
description of the dispersive shock wave propagation in weakly inhomogeneous media.
The main result of the paper is that, if the undular bore advances in the direction
of decreasing depth, its interaction with the slowly varying topography results in the
formation of a sequence of isolated solitons an expanding modulated solitary wavetrain
attached to and propagating ahead of the bore. Importantly, this solitary wavetrain is not
part of the bore and its generation constitutes a non-adiabatic response of the undular
bore to the slow topography variations, which is not captured by the classical Gurevich-
4 G.A. El, R.H.J. Grimshaw and W.K. Tiong
Pitaevskii type solutions assumed in El et al. (2007). Here we construct an appropriate
extension of the traditional undular bore modulation theory to include an asymptotic
description of the generation and evolution of the advancing solitary wavetrains. Our
analytical solution is supported by direct numerical simulations.
2. Problem formulation
We start with the canonical model for a weakly nonlinear unidirectional shallow-water
wave propagation over uneven bottom, which is the variable-coecient KdV equation
A
t
+cA
x
+
c
x
2
A +
3c
2h
AA
x
+
ch
2
6
A
xxx
= 0 . (2.1)
This is written here in non-dimensional form, based on a length scale h
0
, a representative
depth, and a time scale
_
h
0
/g. Here A(x, t) is the free surface elevation above the
undisturbed non-dimensional depth h(x), while c(x) =
_
h(x) is the non-dimensional
linear long wave phase speed. Equation (2.1) was derived for surface gravity waves by
Johnson (1973b,a), and an analogous equation for internal waves by Grimshaw (1981).
The rst two terms in (2.1) are the dominant terms, and by themselves describe the
propagation of a linear long wave with speed c. The equation is derived using the usual
KdV balance in which the /x 1, A
2
, to which we add weak inhomogeneity
so that c
x
/c scales as
3
. For simplicity we ignore here the term describing eects of
bottom friction (see e.g. Miles (1983b) or El et al. (2007), this could be introduced later.
We can cast (2.1) into the asymptotically equivalent form
A

+
h

4h
A +
3
2h
AA
X
+
h
6
A
XXX
= 0 (2.2)
where =
_
x
0
dx

c(x

)
, X = t. (2.3)
Here h = h() explicitly depends on the variable which describes evolution along the
path of the wave. Formally we write A(x, t) =

A(X, ) and h(x) =

h() but then omit
Undular bore on a slope 5
h
0
1
h
1
x
0
0 l x
A
0
Figure 1. Propagation of an undular bore over sloping bottom: problem setting
the tilde in (2.2). Equations (1) and (2) are asymptotically equivalent: they dier with
respect to terms of O(
7
), which is the same as the error term in both equations.
We shall suppose that
h(x) = 1 for x < 0 , h(x) = h
1
for x > x
1
, (2.4)
and varies monotonically in 0 x x
1
. We assume that x
1
1 Then for times t < 0
we shall suppose that we have an initial condition imposed in x < 0, that is
A(x, t = 0) = A
0
(x) for x < x
0
< 0 , A(x, t = 0) = 0 for x > x
0
, (2.5)
Thus initially we generate a solution of the constant coecient KdV equation, and the
aim is to see how this develops in x > 0. The special case when A
0
is a constant for x < x
0
generates an undular bore. In this paper we study the propagation of an undular bore
over a decreasing depth region, i.e. h
1
< 1. The setting under consideration is illustrated
in Figure 1. In terms of the signalling variables (2.9) the initial condition (2.5) becomes

A(X(t = 0) = , ) = A
0
(x) . (2.6)
However, because is a slow variable relative to X, we can infer that asymptotically the
6 G.A. El, R.H.J. Grimshaw and W.K. Tiong
initial condition for (2.2) is
A(X, = 0) = A
0
(X, 0) , (2.7)
where we have again omitted the tilde. Here we have used the fact that A
0
(x) is only
non-zero in x < 0 where = x, and so X = x at t = 0.
The governing equation (2.2) can be cast into several equivalent forms. That most
commonly used is the variable-coecient KdV equation, obtained here by putting
B = h
1/4
A, so that B

+
3
2h
5/4
BB
X
+
h
6
B
XXX
= 0 (2.8)
This form shows that equation (2.2) has two integrals of motion with the densities pro-
portional to B = h
1/4
A and B
2
= h
1/2
A
2
. These are often referred to as laws for the
conservation of mass and momentum.
It is often convenient to recast (2.2) into the standard KdV equation form with con-
stant coecients, modied by a certain perturbation terms. Thus we introduce the new
variables
u =
3
2h
2
A =
3
2h
9/4
B, S =
1
6
_

0
h(

)d

=
1
6
_
x
0
h(x

)
1/2
dx

. (2.9)
so that u
S
+ 6uu
X
+u
XXX
=
9h
S
4h
u . (2.10)
In this form, the governing equation (2.10) has the structure of the integrable KdV
equation on the left-hand side, while the eect of varying depth is on the right-hand side.
This structure enables one to use the modulation theory for perturbed integrable systems
developed in Kamchatnov (2004). Yet another convenient form for (2.2) is obtained by
putting
T =
1
6
_

0
d

h
5/4
(

)
=
1
6
_
x
0
dx

h
7/4
(x

)
(2.11)
U =
3B
2
so that U
T
+ 6UU
X
+(T)U
XXX
= 0 where (T) = h
9/4
(2.12)
Undular bore on a slope 7
In this formulation we assume that (T) = 1 for 0 < T < T
0
and (T) =
1
for T > T
1
with a monotonic variation in T
0
T T
1
. Note that the equations (2.2, 2.8, 2.10, 2.12)
are exactly equivalent. The initial condition for (2.10, 2.12) are
u(X, S = 0) = u
0
(X) = U(X, T = 0) = U
0
(X) =
3
2
A
0
(X) . (2.13)
Note that in all the equations (2.2, 2.8, 2.10, 2.12) we shall refer to , S, T respectively
as time, although they are in fact a variable describing evolution along the path of the
wave.
3. Undular bore transformation over the slope
When the bottom is at, so that = 1 in (2.12), we assume that the initial condition is
that U(X, T = 0) = H(X)U
0
, U
0
= 3A
0
/2. Then the decay of the initial discontinuity
at X = 0 leads to the development of an undular bore, an expanding slowly modulated
periodic wavetrain, asymptotically described by the Whitham modulation theory. The
local wave form of the undular bore is given by the periodic KdV solution
U = a{b(m) + cn
2
(
q

1/2
(X X
0
V T); m)} +d , (3.1)
where b =
1 m
m

E(m)
mK(m)
, a = 2mq
2
,
and V = 6d + 2a
_
2 m
m

3E(m)
mK(m)
_
. (3.2)
Here cn(x; m) is the Jacobi elliptic function of modulus m (0 < m < 1) and K(m), E(m)
are the elliptic integrals of the rst and second kind respectively, a is the wave amplitude,
d is the mean level, V is the wave speed, and X
0
is a constant dening the initial phase.
Note that we have retained in (3.1) and (3.3), in order to include the case when = 1
on the shelf. Also, if the characteristic scale for slow variations of with time is much
greater than that of the modulation parameters in the undular bore, expression (3.1)
remains asymptotically valid for
T
= 0. The value
1/2
/q can be identied as a measure
8 G.A. El, R.H.J. Grimshaw and W.K. Tiong
of the width of the wave humps while their spatial period (wavelength) is
L =
2K(m)
1/2
q
. (3.3)
Note that for 1m 1, we have L
1/2
/q. This family of solutions contains three free
parameters, which are chosen from the set {a, q, V, d, m}. As m 1, cn(x; m) sech(x)
and then the cnoidal wave (3.1) becomes a solitary wave, riding on a background level d.
On the other hand, as m 0, cn(x; m) cos x and so the cnoidal wave (3.1) collapses
to a linear sinusoidal wave (note that in this limit a 0).
The Whitham modulation theory assumes that the expression (3.1) describes a modu-
lated wave in which the amplitude a, the mean level d, the speed V and the modulus m are
all slowly varying functions of x and t. The outcome is a set of three nonlinear hyperbolic
equations for three of the available free parameters, chosen from the set (a, q, V, d, m),
or rather better, from an appropriate combination of them. These equations are often
called the Whitham equations.
The relevant asymptotic solution is then constructed in terms of the similarity variable
X/T, and is given by
X
T
= 2U
0
_
1 +m
2m(1 m)K(m)
E(m) (1 m)K(m)
_
for 6U
0
<
X
T
< 4U
0
, (3.4)
a = 2U
0
m, d = U
0
_
m1 +
2E(m)
K(m)
_
, q = U
0
1/2
. (3.5)
Note that this solution does not depend on the value of , the latter aects only the
wavelength of the underlying periodic wave (3.1). The wavenumber distribution in the
undular bore is then given by
k =
2
L
=
U
1/2
0

1/2
K(m)
. (3.6)
Ahead of the wavetrain where X/T > 4U
0
, U = 0 and at this end, m 1, a 2U
0
Undular bore on a slope 9
and d 0; the leading wave is a solitary wave of amplitude 2U
0
relative to a mean level
of 0. Behind the wavetrain where X/T < 6U
0
, U = U
0
and at this end m 0, a 0,
and d U
0
; the wavetrain is now sinusoidal with the wavelength L = (/U
0
)
1/2
, which
holds throughout the wavetrain, so all waves behind the undular bore have the same
spatial wavelength (for = const, otherwise the wavelength slowly varies with T).
The solution (3.4), (3.5) is due to Gurevich & Pitaevskii (1974) (see also Fornberg
& Whitham (1978)), where it was obtained in terms of the Riemann invariants of the
Whitham modulation equations. It is important to note that the leading solitary wave in
the undular bore developing from an initial step represents asymptotically, as T , a
genuine isolated KdV soliton, not constrained by the interactions with the remainder of
the bore (see Khruslov (1976) or more recent investigation by Claeys & Grava (2010)).
If U
0
< 0 in the initial condition (3.1), then an undular bore solution analogous
to that described by (3.1, 3.4) does not exist. Instead, the asymptotic solution is a
rarefraction wave,
U = 0 for X > 0 ,
U =
X
6T
for 6U
0
<
X
T
< 0 ,
U = U
0
, for
X
T
< 6U
0
(< 0) . (3.7)
Small oscillatory wavetrains are needed to smooth out the discontinuities in U
X
at the
corners.
We concentrate here on the case U
0
> 0 producing an undular bore. If the slope is
suciently gentle, one could expect that the undular bore would undergo some adiabatic
change during the interaction with varying topography while retaining its structure as
a slowly modulated nonlinear periodic wavetrain with a soliton at the leading edge and
the linear vanishing amplitude wavepacket at the trailing edge. This initial assumption
10 G.A. El, R.H.J. Grimshaw and W.K. Tiong
that the interaction with the topography does not change the structure of the bore has
an immediate implication that the amplitude of the leading soliton for T > T
1
has the
same value a = 2U
0
as in the initial bore.
Indeed, one can readily infer from the KdV equation (2.12) that the jump [U] across the
whole propagating nonlinear wave structure is unaected by the depth variations. Let the
structure be conned to some interval X
a
(T) < X < X
b
(T) so that [U] = U(X
b
)U(X
a
).
Then it follows from (2.12) that [U]
T
=

T
_
X
b
Xa
U
X
dX = 0 provided U
X
= U
XXX
= 0
at X = X
a,b
(T). Thus, since the wavetrain advances into the undisturbed depth region,
U(X
b
) = 0, one has
[U] = U
0
for all T > 0 . (3.8)
Note that this result is not only unaected by the varying coecient (T) but is also
independent of the actual form of the structure. Now, if it is assumed that a single undular
bore emerges onto the shelf with =
1
then (3.8) implies that relevant modulation
solution for T > T
1
will have the same form (3.4) but with the constant X
0
generally
replaced by some function X
0
(m) so that it is no longer a centred fan but rather a more
general simple-wave solution of the Whitham equations. The function X
0
(m) then can be
viewed as a modulation phase shift acquired by the undular bore due to its interaction
with variable topography. We note that a similar phase shift arises in the interactions of
dispersive shock waves with rarefaction waves (Ablowitz et al. (2009); El et al. (2011) ),
which is to some degree analogous to the present problem of the transformation of the
undular over sloping bottom (see the end of this Section). Importantly, the relationships
(3.5) remain unchanged, which implies that amplitude of the leading solitary wave of
the transformed bore would then be 2U
0
, that is, unchanged from the value before the
variable depth region is encountered.
However, if one now accepts that the leading solitary wave in the undular bore evolves
Undular bore on a slope 11
over a slope as an isolated soliton (the weak interaction scenario, an assumption to
be conrmed in Section 6 by the analysis of the behaviour of the modulation Riemann
invariants near the leading edge of the undular bore), its amplitude must vary adiabat-
ically, a = 2U
0

1/3
to conserve the action ux
_
U
2
dX (see Section 4), so that for
T > T
1
the leading solitary wave amplitude is 2U
0

1/3
1
> 2U
0
for the decreasing depth
prole, which is clearly inconsistent with the described above evolution scenario based
on the assumption of a single undular bore emerging onto the shelf. To resolve the above
inconsistency an additional solitary wavetrain at the front of the undular bore is needed
to provide the gradual increase of the amplitude from 2U
0
at the undular bore leading
edge to the value 2U
0

1/3
1
implied by the action ux conservation for an isolated soliton.
Thus, the propagation of an undular bore over a broad region of slowly decreasing
depth leads to a non-adiabatic eect, the generation of a solitary wavetrain in front of
the bore. The adiabatic deformation of the bore itself is twofold: (i) the change of the
characteristic scale of the oscillations in the bore due to the change of the dispersion
coecient in (2.12); (ii) the occurrence of the additional slow modulation phase shift
X
0
(m) throughout the bore so that the relevant modulation solution generally represents
a non-centred simple wave of the Whitham equations.
The outlined undular bore transformation can be clearly seen in Figure 2 (right
panel) where we present the results of the numerical simulations of the original variable-
coecient KdV equation (2.1) with the following equilibrium depth prole
h(x) =
_

_
1 x < 0,
1 x 0 < x < 400,
0.64 x > 400.
= 0.0009 (3.9)
The initial condition for (2.1) was taken in the form A(x, 0) =
1
5
(1tanh(x+400)) so that
the wave breaking and formation of an undular bore occurs for x < 0. For comparison,
12 G.A. El, R.H.J. Grimshaw and W.K. Tiong
the left panel of Figure 2 presents the evolution of the undular bore generated by the
same initial condition but propagating over at bottom, h(x) = 1.
In Figure 3 the numerically found boundaries X

(T) of the undular bore along with the

leading edge X
s
(T) of the solitary wavetrain, are shown in the X-T plane of the evolution
according to equation (2.12) discussed in this Section. The variable coecient (T) was
taken in the form of a very smooth transition between = 1 and 0.367 over the
interval T 1300 and the initial condition was U(X, 0) = U
0
H(X), where U
0
= 0.5.
The non-monotonic behaviour of the trailing edge X

(T) in the interval 300 T 3000

is apparently due to the multiphase dynamics occurring at the rear part of the bore while
it propagates over the sloping region and continuing for some time after the bore emerges
onto the shelf (see plot 3 in the right panel of Figure 2). This transient alteration of the
dynamics does not aect the front, soliton, part of the bore which retains its regular
single-phase structure at all times. The single-phase slowly modulated wave behaviour
throughout the whole wavetrain is restored at suciently large times, T 3000.
One can clearly see the conrmation of the three main conclusions following from our
analysis so far: (i) the propagation of the shallow water undular bore into the region of
slowly decreasing depth leads to the generation of the large-amplitude solitary wavetrain
ahead of the bore; (ii) the undular bore edge speeds, X

(T), asymptotically restore their

original values X

(T) = cot(
1
) and X

+
= cot(
2
) upon the emergence of the bore onto
the shelf at T > T
1
; (iii) there are spatial shifts in the positions of the transformed
undular bore edges X

(T) relative to those that would have taken place in the orginal
bore in the absence of the variable topography.
It is interesting to note that the generation of an advancing soliton train in front of
the undular bore was recently observed in Ablowitz et al. (2009) for one of the cases
of the overtaking interaction of a dispersive shock wave with a rarefaction wave studied
Undular bore on a slope 13
in the framework of the constant-coecient KdV equation. Another interesting feature
seen in the corresponding numerical plots in the Ablowitz et al. (2009) paper is the
presence of a transient multi-phase behaviour for the undular bore interacting with the
rarefaction wave. This behaviour can also be clearly seen in our Figure 2 (plots 4 and 5
in the right panel). These similarities between two apparently dierent problems are not
surprising when one observes that, in our present formulation the undular bore essentially
propagates through the rarefaction region of decreasing depth. The essential dierence
is, of course, that in our problem the prole h(x) of the rarefaction region is xed and
described by variable coecients in the KdV equation while the evolving prole of the
hydrodynamic rarefaction wave in Ablowitz et al. (2009) is given by the relevant solution
(3.7) of the constant-coecient KdV equation.
1
4
G
.
A
.
E
l
,
R
.
H
.
J
.
G
r
i
m
s
h
a
w
a
n
d
W
.
K
.
T
i
o
n
g
0 500 1000 1500 2000 2500 3000 3500 4000 4500
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 260
0 1000 2000 3000 4000 5000 6000
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 260
0 500 1000 1500 2000 2500 3000 3500 4000 4500
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 600
0 1000 2000 3000 4000 5000 6000
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 600
0 500 1000 1500 2000 2500 3000 3500 4000 4500
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 1600
0 1000 2000 3000 4000 5000 6000
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 1600
0 500 1000 1500 2000 2500 3000 3500 4000 4500
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 3600
0 1000 2000 3000 4000 5000 6000
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
x
A
t = 3600
Solitary wavetrain Undular bore
Figure 2. Left panel: an undular bore propagating over a at bottom; Right panel: propagation of the same (initially) undular bore over an uneven
bottom with the prole h(x) given by (3.9). The sloping region is located in the interval x [0, 400]
Undular bore on a slope 15
6000 4000 2000 0 2000 4000 6000 8000 10000 12000
0
500
1000
1500
2000
2500
3000
3500
4000
X
T
Before slope
On slope
After slope
X
s
X
+
X

1

2

1

2
Figure 3. X-T plane of the evolution of an undular bore according to the KdV equation (2.12)
with variable dispersion coecient (T) = h
9/4
(T) where h(T) =
9
50

1 tanh

T450
150

+ 0.64
and the initial condition U(X, 0) = 0.5H(X): numerical simulation data. The undular bore
and the solitary wavetrain are conned to [X(T), X+(T)] and [X+(T), Xs(T)] respectively. The
dashed line shows an extrapolation of the leading edge X+(T) of the initial undular bore so one
can see the spatial shift of X+(T) due to the interaction with the slope.
4. Formation and evolution of the solitary wavetrain
In this section we construct an asymptotic description of the solitary wavetrain forming
in front of the undular bore that advances into the decreasing depth region. We start
with the standard results for the adiabatic evolution of a single soliton in a slowly varying
environment and then proceed with the analysis of the sequence of solitary waves attached
to the leading edge of the undular bore.
It is well-known that a solitary wave propagating over a slowly varying bottom will
deform adiabatically so that its amplitude varies as h
1
. To demonstrate this it is best
to use (2.2), (2.8) or (2.12) as these have the conservation law for wave action ux. Thus,
16 G.A. El, R.H.J. Grimshaw and W.K. Tiong
for instance,
d
dT
_

U
2
dX = 0 . (4.1)
The slowly varying solitary wave for (2.12) is
U a sech
2
{(X (T))} , V =
T
= 2a = 4
2
, (4.2)
where the amplitude a etc. are slowly varying function of T. Substitution into (4.1) gives
a
2

= 2
2

3
= constant , (4.3)
and so
2/3
and a
1/3
. Thus a h
3/4
and noting that U = h
1/4
A, the
result follows. The leading wave in the undular bore is a solitary wave whose amplitude
for 0 < T < T
0
is 2U
0
. We will assume as the undular propagates and deforms over
the slope, this leading solitary wave will behaves as if detached from the undular bore
(the weak interaction scenario, El et al. (2007) and Section 5 below. Then its amplitude
a = 2U
0

1/3
and it will emerge onto the shelf region T > T
1
with an amplitude
2U
0

1/3
1
. Its trajectory is
X = (T) =
_
T
0
4U
0
(T

)
1/3
dT

, (4.4)
and it reaches T = T
1
at X = X
1
= (T
1
).
More generally, a slowly-varying solitary wavetrain propagating over a zero background
can be constructed as a modulated periodic lattice of solitary waves (4.2) with a large
spatial period so that the solitons overlap only in the regions of their tails, where the
functions are exponentially small, and thus can be regarded as non-interacting. We note
that in fact, a periodic lattice of solitons

nZ
sech
2
((x +nL)) converges pointwise to
the cnoidal wave solution (3.1) for any period L > 0 (see e.g. Whitham (1984)). The
condition L
1
for such a soliton lattice is equivalent to the condition 1 m 1 in
Undular bore on a slope 17
the more conventional representation of the periodic solution (3.1) in terms of the Jacobi
elliptic function).
Thus, locally, the solitary wavetrain can be described, up to a constant phase shift, by
an asymptotic expression
0 < X < 1/ : U a sech
2
() ,
T
= V ,
X
= , (4.5)
where V = 2a = 4
2

2
; (4.6)
U( + 1) = U() .
The term locally here implies distances comparable with a single spatial period L =
1/. In a modulated solitary wavetrain, a and are slowly varying functions of X and
T (i.e. they vary on the scale much larger than 1/) so that (4.5) can be viewed as a
natural extension of (4.2) for a single slowly varying soliton, and it reduces (within each
period) to that expression when one sets = 1. The fact that the initial phase is not
xed in (4.5) is not essential for our consideration as we shall be interested only in the
behaviour of slow modulations a(X, T), (X, T). Also note that expression (4.5) agrees
with (3.1) in the limit when m 1 if one assumes q =
1/2
.
The modulation equations for the amplitude a and the soliton train wavenumber
then follow:
_
a
2

_
T
+V
_
a
2

_
X
= 0 , (4.7)

T
+ (V )
X
= 0 . (4.8)
The system (4.7), (4.8) was obtained by Grimshaw (1979) (see the zero mean, A = 0,
reduction of equation (2.19)) using a multiple-scale expansion of the variable-coecient
KdV solution with the leading term in the form (4.5). The amplitude equation (4.7)
can also be obtained directly, using averaging of the KdV mass conservation law over
18 G.A. El, R.H.J. Grimshaw and W.K. Tiong
the periodic soliton lattice (4.5) or via the averaged Lagrangian approach (Whitham
(1974)). It can also be shown to be consistent with the soliton (m 1) limit of the full
Whitham modulation system derived in El et al. (2007) for the perturbed KdV equation
(2.10). It should be stressed that, while the amplitude equation (4.7) represents an exact
reduction of the full modulaton system where one sets m = 1, equation (4.8) is valid only
asymptotically for 1 m 1 i.e within the range of validity of the approximation (4.5).
Using the relations (4.5), equations (4.7), (4.8) can be written in the form
A

+ 2AA
X
= 0 , A =
_
a
2

2
_
2/3
= a
1/3
, (4.9)

+ (2A)
X
= 0 , (4.10)
where =
_
T
0
(T

)
1/3
dT

. (4.11)
Remarkably, the system (4.9, 4.10) for A(X, ) and (X, ) has the same form as system
(4.7, 4.8) for a(X, T) and (X, T) in the case when =
0
= constant (and so a =
4
0

2
), i.e. for the constant-coecient KdV equation see Whitham (1974), Ch. 16.
When there is no X-variation, A, are constants, and the result (4.3) is recovered.
The general solution of (4.9, 4.10) is found by using characteristics,
A = constant , on
dX
d
= 2A, (4.12)
and
d
d
= 2A
X
=
A

A
. (4.13)
Note that the system (4.12), (4.13) has only one multiple characteristic family and all
the characteristics are straight lines in the X- plane.
We dene the position of the trailing edge of the solitary wavetrain as the line X =
X
+
(T) where a = 2U
0
. This denition is consistent with the location of the leading edge
of the undular bore for the at bottom propagation case, and initially X

+
(T
0
) = 4U
0
.
However, for the varying bottom case the line X = X
+
(T) is not associated with the
trajectory of a particular solitary wave as the solitary waves must be allowed to cross
Undular bore on a slope 19
this boundary to enable the formation of the advancing modulated solitary wavetrain;
therefore
0 < X

+
(T) < 4U
0
for T
0
< T < T
1
. (4.14)
Thus, the following boundary condition must be satised for the solitary wavetrain am-
plitude equation (4.9),
A = 2U
0

1/3
on X =

X() , (4.15)
where

X() = X
+
(T()); T() being the inverse relation of (4.11), so that (T()) = .
The latter relationship requires that (T) varies monotonically from 1 at T = T
0
to
1
at T = T
1
.
The solution for A is
A = A
0
(
0
) = 2U
0

1/3
(T(
0
)) , X

X(
0
) = 2A
0
(
0
) , (4.16)

0
[0,
1
] being a parameter on the curve X =

X(). Formally, elimination of the
parameter
0
from (4.16) yields A as a function of X, . The solution (4.16) is dened
for

X() < X < X
s
, where X
s
= 4U
0
is the trajectory of the leading soliton in the
solitary wavetrain (see (4.4)), having the amplitude a = 2U
0

1/3
, i.e. A = 2U
0
. For
X > 4U
0
we have A = 0. Calculating the derivative A
X
we obtain:
A
X
=
A

0
(
0
)
2A

0
(
0
)(
0
) + [

X

(
0
) 2A
0
(
0
)]
. (4.17)
Owing to (4.14) [

X

(
0
) 2A
0
(
0
)] < 0, therefore to guarantee the existence of the
obtained solution for all X, ome must have A

0
< 0. This, by (4.16), (4.11) implies

(T) < 0. Then our solution represents a rarefaction fan emanating from the curve
X =

X(
0
). The condition

(T) < 0 (decreasing depth) thus can be viewed as the

condition of the formation of an expanding solitary wavetrain in front of the bore. Using
A(X, ) dened by (4.16) the corresponding general solution for is then found from
20 G.A. El, R.H.J. Grimshaw and W.K. Tiong
(4.13), that is
=
0
_
1 +
2A

0
(
0
)(
0
)

(
0
) 2A
0
(
0
)
_
1
. (4.18)
where
0
is the value of on the curve X = X
+
(T(
0
)) =

X(
0
) and
0
(X, ) is dened
by (4.16).
Generally, to nd the curve X = X
+
(T) for T
0
< T < T
1
one needs to solve the full
perturbed modulation system from El et al. (2007). However, it is instructive to assume
that
X

+
(T) 4U
0
for T
0
< T < T
1
, (4.19)
and thus

X

(
0
) 2A
0
(
0
). This behaviour can be formally justied for functions (T)
varying suciently fast on a typical time scale of the solitary wavetrain modulations (but
still being slow functions on the time scale of a single soliton). Indeed, since X

+
(T
0
) =
4U
0
, in order to satisfy (4.14) one must have X

+
(T) < 0 ( it is clear that signX

+
(T) =
sign

(T)), then the result follows as an asymptotic behaviour of X

+
(T) for suciently
large T. Thus we have
X
+
(T) X
+
(T
0
) = 4U
0
T
0
for T
0
< T < T
1
. (4.20)
Our numerical simulations show that behaviour (4.20) establishes itself quite quickly even
for rather slow functions (T). One can see this in the numerical plot corresponding to

T
5 10
4
shown in Figure 3. For larger values of
T
, say
T
10
3
10
2
, the
boundary X
+
(T) becomes stationary almost immediately as variations of begin at
T = T
0
; X
+
resumes its motion at T = T
1
and gradually restores its initial velocity
4U
0
. Thus (4.19) and (4.20) can be safely used in the solution (4.18), (4.16) for a broad
range of the slope values specied in terms of (T). The schematic behaviour of the
boundaries X
+
(T) and X
s
(T) illustrating the asymptotic formulation of the problem of
the generation of the solitary wavetrain on the given boundary X = X
+
(T) = X
+
(T
0
)
Undular bore on a slope 21
Figure 4. Schematic behaviour of the boundaries X+ and Xs of the solitary wavetrain
generated at the leading edge of the undular bore on a slope.
for T
0
< T < T
1
is shown on Figure 4. Thus, using (4.19) we have to leading order

0
_
1
A

0
(
0
)(
0
)
A
0
(
0
)
_
1
=
0
_
1
2
3

(
0
)
(
0
)
(
0
)
_
1
, (4.21)
where (
0
) (T(
0
)) so that

(
0
) =
T

1/3
(
0
) < 0 and therefore solution (4.21)
exists for all X, . Then the leading edge of the undular bore, that is also the trailing
edge of the solitary wavetrain, emerging on the shelf is X
+
(T) 4U
0
T
0
+ 4U
0
(T T
1
)
for T > T
1
and the phase shift
+
= X
0
(1) (see Section 3) can be estimated as
+

4U
0
(T
1
T
0
). Of course, one can guarantee the linear behaviour of X
+
(T) 4U
0
T only
for T T
1
when the slowly modulated structure of the undular bore fully adjusts itself
to the shelf region with =
1
.
Let
1
= (T
1
). Then on the shelf where T > T
1
, >
1
we have =
1
, =

1
+ (T T
1
)
1/3
1
. The leading edge of the solitary wavetrain on the shelf is X
s
=
4U
0
= 4U
0
(
1
+(T T
1
)
1/3
1
). We note that for T > T
1
both both boundaries X
+
(T)
and X
s
(T) conning the expansion fan are characteristics and the total number of solitary
waves in the train for T > T
1
does not change with time. The value

0
(
0
)
U
1/2
0
4
1/2
(
0
)
I , where I 0.6569, (4.22)
22 G.A. El, R.H.J. Grimshaw and W.K. Tiong
is found by examining the undular bore structure near its leading edge (see Appendix).
Now, from (4.6), (4.9) we have (X, T) =
1

1/3
(A/2)
1/2
and so the slowly varying
solitary wavetrain (4.5) is fully dened. As T , T
1/3
1
, X
+
4U
0
T, X
s

4U
0

1/3
1
T and the asymptotic solution is,
4U
0
T < X < 4U
0

1/3
1
T : A
X
2
, or a
X
2T
, (4.23)

g(X/(2T))

g(A)

, or
G(a)
T
. (4.24)
Here g(A) = 3
0
(
0
)/(2

(
0
)), where
0
(A) is found from the solution A = 2U
0

1/3
(
0
)
(see (4.16)). As a matter of fact, g(A) is only dened for the variable coecient region
0 <
0
<
1
, where

() = 0 and where the generation of the solitary wavetrain occurs.

The function G(a) = g(a
1/3
1
) has the meaning of the distribution function over ampli-
tude in the solitary wavetrain so that G(a)da is the number of solitons with amplitudes
in the interval [a, a + da] (see Whitham 1974). Since the total number of solitons N in
the train remains constant for T > T
1
it can be estimated by the formula
N
_
2U0
1/3
1
2U0
G(a) da . (4.25)
In concluding this section we note that if

(T) > 0 (increasing depth) our solution

(4.21) yields that the interaction of the solitary wavetrain with the increasing depth
bottom prole topography would result in an increase of the density of solitary waves
enhancing thus their interaction and eventually rendering invalid the basic assumption
about the isolated character of solitary waves in the train. Also, the amplitude prole
specied by the solution (4.16) with

(T) > 0 will develop a breaking singularity at some

=
c
also making this solution physically invalid. All this suggests that one should not
expect the generation of the chain of individual non-interacting solitons in front of the
bore if

(T) > 0. Instead, this case appears to be relevant to the realisation of the
second (nonlocal) strong interaction scenario of the undular bore evolution discussed
Undular bore on a slope 23
U
0

3

1
X

X
U
0

3

2

3
X

X
s
Figure 5. Schematic behaviour of the Riemann invariants in the modulation solution. Left:
regular undular bore (before the slope, T < T0); Right: undular bore with an advancing soliton
train conned to [X+, Xs] (after the slope, T T1)
in El et al. (2007) and outlined in the Introduction. We plan to study this case in our
future publications.
The obtained description of the transformation of an undular bore over a slope is most
readily illustrated using a diagram showing the behaviour of the Riemann invariants

j
(X, T), j = 1, 2, 3 of the Whitham equations in the combined modulation solution
obtained in Sections 3 and 4. These Riemann invariants
1

2

1
can be ex-
pressed as certain functions of any three independent modulation parameters from the
set {a, q, V, d, m} introduced in Section 3 but they are most conveniently represented in
terms of the roots b
1
b
2
b
3
of the polynomial, dening the Riemann surface associ-
ated with the cnoidal wave periodic solution (3.1) (Whitham 1965), see formulae (5.4) in
the next section. The periodic solution parameters a, q, V and d are expressed in terms
of these Riemann invariants as
a = 2(
2

1
) , q =
_

1
, V = 2(
1
+
2
+
3
) , d =
1
+
2

3
+4(
3

1
)
E(m)
K(m)
.
(4.26)
Then the Gurevich-Pitaevskii modulation solution for the undular bore is given by
1
= 0,

3
= U
0
, while
2
= U
0
m is dened by the same expression (3.4). In the solitary wavetrain
24 G.A. El, R.H.J. Grimshaw and W.K. Tiong
we have
2
=
3
to leading order so the asymptotic solution (4.23) as T assumes
the form
3
X/(4T) in the interval 4U
0
T X 4U
0
T
1/3
1
.
The outlined behaviour of the Riemann invariants in the modulated solutions before
and after the slope, is schematically shown in Fig 5. This behaviour will be conrmed by
direct numerical simulations in the next Section.
5. Numerical simulations
In this section we shall describe the numerical conrmation of the three main assump-
tions used in the modulation analysis of Sections 3 and 4: (i) the assumption that the
undular bore on a slope can be described by a slowly modulated periodic solution of
the KdV equation; (ii) the weak interaction scenario ensuring the behaviour of the
leading solitary wave in the undular bore on a slope as an isolated KdV soliton; (iii)
the assumption that the wave structure forming in front of the undular bore is indeed
a solitary wavetrain (rather than part of a traditional cnoidal bore described by the
standard Gurevich-Pitaevskii solution).
For the numerical simulations we shall be using the variable-coecient KdV equation
in the form (2.8),
B

+
3
2h
5/4
BB
X
+
h
6
B
XXX
= 0 (5.1)
with the dependence h() corresponding to the depth prole (3.9),
h() =
_

_
1.0 : < 400
_
1
(400)
2
_
2
: 400 < < 844.44
0.64 : > 844.44
, = 0.0009. (5.2)
As was already mentioned, equation (5.1) is exactly equivalent to the equation (2.12)
used for our asymptotic analysis. The present form (5.1) is slightly more convenient for
numerical simulations as it produces less waves for the same running time interval. The
Undular bore on a slope 25
initial condition is taken in the form of a smooth step,
B(X, 0) =
1
4
(1 tanh(X/10)) (5.3)
Equation (5.1) was solved using the method of lines where the spatial derivatives are
discretised using second order accurate nite dierence approximation to reduce the
governing partial dierential equation to a system of ordinary dierential equations.
This system is then solved using the fourth order Runge-Kutta method (see e.g. Schiesser
(1991)).
Assuming that locally, undular bore can be described by the periodic solution (3.1)
we shall use the numerical solutions to extract the parameters corresponding to the
modulation Riemann invariants
j
introduced in (4.26), which are expressed in terms of
the basic wave parameters b
1
, b
2
and b
3
as (see Whitham (1965))

3
=
b
2
+b
3
2

2
=
b
1
+b
3
2

1
=
b
1
+b
2
2
(5.4)
Here b
2
B
min
and b
3
B
max
are easily found from the numerical data. The third
parameter b
1
can be obtained from the numerical values of the local spatial period
(wavelength) L, which for the variable-coecient KdV equation (5.1) is given by the
formula
L =
4h
9/8
K(m)
_
3(b
3
b
1
)
where m =
b
3
b
2
b
3
b
1
. (5.5)
We expect that the variables
j
will demonstrate the qualitative behaviour shown in
Figure 5 which will be a conrmation of the validity of the modulation analysis presented
in Sections 3 and 4.
The results of our numerical simulations are presented in Figures 6 10. Each of the
Figures 6 9 corresponds to a particular value of time, =
i
and consists of three
plots: the top plot shows the numerical solution for B(X,
i
), the middle plot shows
26 G.A. El, R.H.J. Grimshaw and W.K. Tiong
spatial behaviour of the numerical Riemann invariants
j
and the bottom plot shows
the behaviour of the numerically determined value of the modulus m.
Figure 6 corresponds to the initial undular bore, = 400, with the numerical values
of the Riemann invariants
j
(the middle plot) distributed according to the Gurevich-
Pitaevskii solution (3.4) with = 1 and the modulus running from 0 to 1 across the
bore (see the bottom plot). The Riemann invariant distribution agrees with the diagram
shown on Fig. 5a.
Figure 7 corresponds to the bore propagating over the slope, = 800, and one can
observe the presence of the region with
2
=
3
in front of the bore. Since we have m = 1
in this region (see the bottom plot), the corresponding wave structure should be identied
with the solitary wavetrain. The amplitude a =
3

1
in this solitary wavetrain increases
towards the leading edge as predicted by our analysis (the comparison for the amplitude
of the leading solitary wave is presented in Figure 10).
In Figure 8 we present the distributions corresponding to = 1400, i.e. after the
slope. While the front part of the bore and the advancing solitary wavetrain retain their
structure presented in Figure 7 one can observe the occurrence of the new multi-phase
(presumably, two-phase) region in the rear part of the bore. As a matter of fact the
modulus m is not dened for this region (see the bottom plot).
Figure 9 shows the plots at = 3400 corresponding to the long-time behaviour of the
transformed wavetrain. The distribution of the Riemann invariants agrees with diagram
in Figure 5b, conrming our theoretical predictions. One can see that the behaviour
of the upper envelope in the solitary wavetrain is markedly dierent from that in the
undular bore behind it and agrees with the asymptotic formula (4.23).
Note the behaviour of the Riemann variables
2
and
3
near the leading edge of the
undular bore in all plots. One can see that
d
dX

2
< and
d
dX

3
< , which corresponds
Undular bore on a slope 27
to the weak interaction scenario identied in El et al. (2007). This conrms our main
assumption that the leading soliton in the undular bore behaves as an isolated solitary
wave in the propagation over the decreasing depth region. The direct quantitative con-
rmation of this assumption can be seen in Figure 10 where we present the comparison
for the amplitude of the adiabatically varying isolated solitary wave on a slope with the
numerical values of the leading solitary wave amplitude in the modulated wavetrain (ini-
tially a single undular bore) propagating over the same slope. In both cases the initial
value of the amplitude was the same, a = 1. For the variable-coecient KdV equation
(5.1) the adiabatic variations of the solitary wave amplitude are given by formula (see
Section 4).
a = a
0
_
h
0
h()
_
3/4
(5.6)
where h
0
and a
0
are the initial depth and the solitary wave amplitude respectively.
Relationship (5.6) is an exact counterpart of the relationship (4.3) for the slowly varying
solitary wave of the variable-coecient KdV equation in the form (2.12). One can see
an excellent agreement which provides direct quantitative conrmation our assumption
that the leading solitary wave of the undular bore evolves as an isolated solitary wave
over the variable depth region when h

(x) < 0.
28 G.A. El, R.H.J. Grimshaw and W.K. Tiong
300 250 200 150 100 50 0 50 100 150 200
0
0.2
0.4
0.6
0.8
1
1.2
X
B
= 400
300 250 200 150 100 50 0 50 100 150 200
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
X

3
300 250 200 150 100 50 0 50 100 150 200
0
0.2
0.4
0.6
0.8
1
X
m
Figure 6. Initial undular bore (before slope), = 400. Upper plot: B(X); Middle plot: the
modulation Riemann variables 1 2 3 obtained from the plot for B(X) assuming a local
representation of the wave in the form of a cnoidal periodic solution of the constant-coecient
KdV equation; Bottom plot: the modulus m = (2 1)/(3 2) as function of X.
Undular bore on a slope 29
400 300 200 100 0 100 200 300 400 500
0
0.2
0.4
0.6
0.8
1
1.2
X
B
= 800
400 300 200 100 0 100 200 300 400 500
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
X

3
400 300 200 100 0 100 200 300 400 500
0
0.2
0.4
0.6
0.8
1
X
m
Figure 7. Same as in Figure 7 but for = 800 (undular bore on the slope). One can see the
solitary wavetrain (m 1) forming between X 350 and X 500.
30 G.A. El, R.H.J. Grimshaw and W.K. Tiong
200 0 200 400 600 800 1000 1200
0
0.5
1
1.5
X
B
= 1400
200 0 200 400 600 800 1000 1200
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
X

3
200 0 200 400 600 800 1000 1200
0
0.2
0.4
0.6
0.8
1
X
m
Figure 8. Same as in Figures 6,7 but for = 1400 (after the slope). The solitary wave train is
located between X 700 and X 1200. A multiphase behaviour can be seen around X = 100.
Undular bore on a slope 31
2000 1000 0 1000 2000 3000
0
0.5
1
1.5
X
B
= 3400
2000 1000 0 1000 2000 3000
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
X

3
2000 1000 0 1000 2000 3000
0
0.2
0.4
0.6
0.8
1
X
m
Figure 9. Same as in Figures 6,7,8 but for = 3400 (long-time behaviour after the slope). The
linear behaviour of the envelope in the solitary wavetrain agrees with the asymptotic prediction
(4.23).
32 G.A. El, R.H.J. Grimshaw and W.K. Tiong
300 400 500 600 700 800 900 1000 1100 1200
0
0.5
1
1.5

a
Figure 10. Comparison for the amplitude of the adiabatically varying solitary wave on a slope
(formula (5.6) with h0 = 1, a0 = 1 ) solid line; and the numerical data (circles) for the leading
solitary wave in the modulated wavetrain (initially, a single undular bore on the same slope).
6. Discussion
In this paper, we have considered a shallow water undular bore propagation over a
sloping bottom in the framework of the appropriate variable-coecient KdV equation.
We have shown that, when the undular bore advances into the decreasing depth region
its interaction with the varying topography results in the formation of a sequence of
isolated solitary waves, an expanding modulated solitary wavetrain propagating ahead
of the bore and having the amplitude greater than that of the leading solitary wave
in the undular bore. This important non-adiabatic eect was apparently overlooked in
the previous studies including our paper El et al. (2007), where a similar problem was
considered in a general theoretical framework with the account of eects of turbulent
bottom friction.
Using several assumptions (conrmed by detailed numerical simulations) about the
structure of the undular bore propagating on a slope we have constructed an asymptotic
modulation solution describing the generation and propagation of the solitary wavetrain
Undular bore on a slope 33
ahead of the undular bore. Our analysis can be extended to include the eects of small
dissipation (e.g. turbulent bottom friction).
Importantly, our general approach is not conned to the KdV dynamics and can be
applied to other systems describing the propagation of undular bores (or, more generally,
dispersive shock waves) through weakly non-uniform environments. We also stress that
the availability of the full modulation solution (an analog of the Gurevich-Pitaevskii
solution (3.4)) for the initial at-bottom undular bore is not a pre-requisite in our
analysis, and a similar study can be undertaken for the systems where the initial evolution
of the undular bore is described by a non-integrable dispersive equation (see El (2005)
for the relevant generalisation of the Gurevich-Pitaevskii problem). In particular, one
can consider the propagation of a fully nonlinear shallow water undular bore over a slope
in the framework of the appropriate variable-coecient Su-Gardner (Green-Naghdi or
Serre) equations (see El et al. (2006) for the corresponding at-bottom undular bore
theory).
Appendix: Determination of
0
We assume here that function (T) is suciently slow so that for each T the undular
bore can be locally described by a single-phase solution (3.1). As our numerical simula-
tions show, this is not entirely true for the whole wavetrain but denitely holds for the
front part of the bore. The wavenumber in the bore is given by the formula (3.6),
k =
U
1/2
0

1/2
K(m)
. (6.1)
If we consider the solitary wavetrain as a small wavenumber asymptotic of the modulated
cnoidal wave we must have (see Whitham (1974))
1 m 1 :
k
2
, (6.2)
34 G.A. El, R.H.J. Grimshaw and W.K. Tiong
which suggests a continuous matching between k/(2) in the undular bore and =
0
in
the solitary wavetrain at the leading edge X
+
(T) of the undular bore. This continuous
matching, however, is not possible for the following reason.
The asymptotic behaviour of k in the modulation solution (3.4) near the leading edge
X = X
+
(T) is (Gurevich & Pitaevskii (1974))
k
2

1/2
U
0
1/2
ln(1/(s
+
s))
, (6.3)
where s = X/T, s
+
= X
+
/T = 4U
0
. This implies that |k
X
| while k 0 when
X X
+
. Thus the wavenumber varies rapidly near X = X
+
. On the other hand, the
typical spatial scale of the variations of modulations in the advancing solitary wavetrain
is much greater than that in the undular bore. Therefore it is natural to require matching
of with the mean value of k/(2) across the undular bore front, which one can naturally
dene as the (soliton) part of the bore propagating to the right, i.e. 0 < X < X
+
(note
that X = 0 is the characteristic of the modulation system separating the right- and left-
propagating parts of the characteristic fan (3.4)). The mean value of k across the bore
front is then
k =
1
X
+
X+
_
0
kdX =
1
4U
0
1
_
m

k(m)
_
ds
dm
_
dm, (6.4)
where s(m) is given by the modulation solution (3.4) and m

0.6415 the value of

m on the characteristic X = 0 is the root of the equation s(m) = 0. Then, using (6.1)
we obtain
k =
U
1/2
0
2
1/2
I, (6.5)
where
I =
1
_
m

(m)
K(m)
dm 0.6569 , (6.6)
W(m) = 1 +m
2m(1 m)K(m)
E(m) (1 m)K(m)
. (6.7)
Undular bore on a slope 35
Thus we have

0
=
k
2
=
U
1/2
0
4
1/2
I . (6.8)
REFERENCES
Ablowitz, M.J., Baldwin, D.E. & Hoefer, M.A. 2009 Soliton generation and multiple
phases in dispersive shock and rarefaction wave interaction. Phys. Rev. E 80, 016603.
Baines, P.G. 1995 Topographic eects in stratied ows. Cambridge University Press.
Claeys, T. & Grava, T. 2010 Solitonic asymptotics for the Korteweg-de Vries equation in the
small dispersion limit. SIAM J. Math. Analysis 42, 21322154.
El, G.A., R.H.J., Grimshaw & Kamchatnov, A.M. 2007 Evolution of solitary waves and
undular bores in shallow-water ows over a gradual slope with bottom friction. Journ.
Fluid Mech. 585, 213244.
El, G. A. 2005 Resolution of a shock in hyperbolic systems modied by weak dispersion. Chaos
15, 037103.
El, G. A., Grimshaw, R. H. J. & Smyth, N. F. 2006 Unsteady undular bores in fully nonlinear
shallow-water theory. Phys. Fluids 18, 027104.
El, G. A., Grimshaw, R. H. J. & Smyth, N. F. 2009 Transcritical shallow-water ow past
topography: nite-amplitude theory. Journ. Fluid Mech. 640, 187 215.
El, G. A., Khodorovskii, V. V. & Leszczyszyn, A. M. 2011 Refraction of dispersive shock
waves. submitted .
Esler, J. G. & Pearce, J. D . 2011 Dispersive dam-break and lock-exchange ows in a
two-layer uid. Journ Fluid Mech 667, 555585.
Fornberg, B. & Whitham, G. B. 1978 A numerical and theoretical study of certain nonlinear
wave phenomena. Philosophical Transactions of the Royal Society A 289, 373404.
Grimshaw, R. 1979 Slowly varying solitary waves. I Korteweg-de Vries equation. Proc. Roy.
Soc. 368A, 359375.
Grimshaw, R. 1981 Evolution equations for long nonlinear internal waves in stratied shear
ows. Stud. Appl. Math. 65, 159188.
36 G.A. El, R.H.J. Grimshaw and W.K. Tiong
Grimshaw, R. 2007 Internal solitary waves in a variable medium. Gesellschaft f ur Angewandte
Mathematik 30, 96109.
Grimshaw, R. H. J. & Smyth, N. F. 1986 Resonant ow of a stratied uid over topography.
J. Fluid Mech. 169, 429464.
Gurevich, A. V. & Pitaevskii, L. P. 1974 Nonstationary structure of a collisionless shock
wave. Sov. Phys. JETP 38, 291297.
Johnson, R.S. 1997 A modern introduction to the mathematical theory of water waves. Cam-
bridge University Press.
Johnson, R. S. 1973a On an asymptotic solution of the Korteweg - de Vries equation with
slowly varying coecients. J. Fluid Mech. 60, 813824.
Johnson, R. S. 1973b On the development of a solitary wave moving over an uneven bottom.
Proc. Camb. Phil. Soc. 73, 183203.
Kamchatnov, A.M. 2004 On Whitham theory for perturbed integrable equations. Physica
D188, 247261.
Khruslov, E. A. 1976 Asymptotics of the solution of the Cauchy problem for the Korteweg-de
Vries equation with step-like initial data. Math. USSR-Sb. 28, 229248.
Madsen, P. A., Fuhrman, D. R. & Sch affer, H. A. 2008 On the solitary wave paradigm
for tsunamis. J. Geophys. Res. 113, C12012.
Miles, J.W. 1983a Solitary wave evolution over a gradual slope with turbulent friction. J. Phys.
Oceanog. 13, 551553.
Miles, J.W. 1983b Wave evolution over a gradual slope with turbulent friction. J. Fluid Mech.
133, 207216.
Ostrovsky, L.A. & Pelinovsky, E.N. 1975 Refraction of nonlinear sea waves in a coastal
zone. Akad. Nauk SSSR, Izv. Atmos. Ocean Phys. 11, 3741.
Schiesser, W. E. 1991 The numerical method of lines: integration of partial dierential equa-
tions. Academic Press.
Scotti, A., Beardsley, R.C., Butman, B. & Pineda, J. 2008 Shoaling of nonlinear internal
waves in Massachusetts bay. J. Geophys. Res. 113, C08031.
Undular bore on a slope 37
Smyth, N.F. & Holloway, P.E. 1988 Hydraulic jump and undular bore formation on a shelf
break. J. Phys. Oceanog. 18, 947962.
Tissier, M., Bonneton, P., Marche, F., Chazel, F., & Lannes, D. 2011 Nearshore dy-
namics of tsunami-like undular bores using a fully nonlinear Boussinesq model. Journal of
Coastal Research Special Issue 64, 603607.
Whitham, G.B. 1965 Non-linear dispersive waves. Proc. Roy. Soc. London A 283, 238261.
Whitham, G.B. 1984 Comments on periodic waves and solitons. IMA Journ. Appl. Math. 32,
353366.
Whitham, G. B. 1974 Linear and Nonlinear Waves. J. Wiley and Sons.