J . Fluid Mech. (1987), vol. 185, p p . 523-545
Printed in &eat Britain
523
The runup of solitary waves
By COSTAS EMMANUEL SYNOLAKIS
School of Engineering, University of Southern California, Los Angeles,
California 90089-0242, USA
(Received 22 August 1986 and in revised form 9 May 1987)
This is a study of the runup of solitary waves on plane beaches. An approximate
theory is presented for non-breaking waves and an asymptotic result is derived for
the maximum runup of solitary waves. A series of laboratory experiments is
described to support the theory. It is shown that the linear theory predicts the
maximum runup satisfactorily, and that the nonlinear theory describes the climb of
solitary waves equally well. Different runup regimes are found to exist for the runup
of breaking and non-breaking waves. A breaking criterion is derived for determining
whether a solitary wave will break as it climbs up a sloping beach, and a different
criterion is shown to apply for determining whether a wave will break during
rundown. These results are used to explain some of the existing empirical runup
relationships.
1. Introduction
The problem of determining the runup of solitary waves on plane beaches usually
arises in the study of the coastal effects of tsunamis. Tsunamis are long water waves
of small steepness generated by impulsive geophysical events on the ocean floor or at
the coastline. Solitary waves are believed to model some important aspects of the
coastal effects of tsunamis well.
The process of long-wave generation and propagation is now well understood. The
process of long-wave runup and reflection is not. Although there is consensus that a
suitable physical model for this process is the formalism of a long wave propagating
over constant depth and encountering a sloping beach, there is little agreement on
the appropriate analytical formulation. The analytical studies of this problem can be
classified in two general groups depending on the approximation of the equations of
motion that was used in each study to determine the runup.
One group has studied variants of the Boussinesq equations. The state-of-the-art
numerical solutions of these equations for solitary-wave runup are those of Pedersen
& Gjevik (1983) and Zelt (1986) who solved the equations in Lagrangian coordinates,
and of Kim, Liu & Ligett (1983) who used boundary-integral methods.
The other group has used the shallow-water-wave equations. These two nonlinear
equations result directly from the Boussinesq equations, if the effects of dispersion
and vertical accelerations are neglected. A comprehensive review of various regular,
weak and apparent solutions of these equations may be found in Meyer (1986). The
classic solution of their linear form was given by Lewy (1946) for the problem of
periodic waves climbing up a sloping beach. Carrier & Greenspan (1958) derived a
nonlinear transformation to reduce the two equations to a single linear equation and
they solved several initial-value problems. Keller & Keller (1964) solved the linear
problem of a periodic wave propagating first over constant depth and then up a
524 C. E . Synolakis
sloping beach. Carrier (1966) used the Carrier & Greenspan transformation to
calculate in closed form the runup of a wave generated by a bottom displacement and
propagating over relatively arbitrary bathymetry. The state-of-the-art numerical
solution of the nonlinear equations was achieved with the work of Hibberd &
Peregrine (1979) ; they were able to calculate the runup of a uniform bore on a plane
beach.
Despite the quality of the analytical work, fundamental unresolved questions
about the runup of long waves still exist. The empirical relationship between the
normalized runup and the normalized wave height that has been established in the
series of laboratory investigations of Hall & Watts (1953), Camfield & Street (1969),
and Kishi & Saeki (1966) remains unexplained analytically. The results of the
available numerical solutions have not been compared with detailed amplitude-
evolution data from the laboratory, and, as a consequence, there is little conclusive
information about the relative importance of dispersion and nonlinearity during
runup. There is no realization of the differences in the runup behaviour of breaking
and non-breaking waves, and this has led to numerical results for non-breaking
waves to be compared with laboratory data for breaking waves. Compared with
recent advances in periodic-wave runup (Guza & Thornton 1982; Holman 1986), the
understanding of solitary-wave runup has been fragmented and incomplete.
In the present study an exact solution to an approximate theory will be presented
for determining the runup and the amplitude evolution of long waves on plane
beaches. A series of experiments will also be described and the resolution of some of
these questions will be attempted.
2. Basic equations and solutions
Consider a topography consisting of a plane sloping beach of angle p, as shown in
figure 1. The origin of the coordinate system is at the initial position of the shoreline
and 2 increases seaward. Dimensionless variables are introduced as follows :
2 = xd, i0 = h, d , r" = qd, 92 = u(gd)i, t" = t (d /g)$ .
q is the amplitude, u is the depth-averaged velocity, and h, is the undisturbed water
depth. The topography is described by
h,(x) = x t anp when x < cot@ ( 2 . 1 ~ )
and h,(x) = 1 when x > cotp. (2.1 b)
Consider a propagation problem described by the shallow-water-wave equations
h, + (hu), = 0 ( 2 . 2 ~ )
and ut+uu,+q, = 0, (2.2b)
where h(x , t ) = h,(x) +?I(%, t ) .
2.1. Linear theory
The system of equations (2.2) can be linearized by retaining the first-order terms
only. The following equation results :
Ttt - (Tzho), = 0. (2.3)
The solution for constant depth is
r when h,(x) = 1 , ( 2 . 4 ~ ) r ( x , t ) = e-ik(x+ct) + A eik(z-ct)
Runup of solitary waves
Maximum runup
525
FIQURE 1. A definition sketch for a sblitary wave climbing up a sloping beach. All variables
except B are dimensional.
and the solution for variable depth and finite at the shoreline is
~ ( x , t ) = B(k, /3) Jo(2k(x cot 8)’) e-ikct when h,(z) = x tan /3. (2.4b)
A, is the amplitude of the incident wave, A, that of the reflected wave, B is the
amplification factor, k is the wavenumber and c = 1. Keller & Keller (1964) presented
another steady-state solution for the combined topography defined by (2.1). They
matched the outer solution (2.4a) and its 2-slope to the inner solution (2.4b) and its
x-slope, at x = X, = cot p, the toe of the beach, and they derived the following values
for the reflected-wave amplitude and the amplification factor for an incident wave of
the form ~(z, t ) = A, e-i(z+et) :
and
. J0(2k cot/3)
A,( k, /3, A,) = A, exp [ - 2ik cot /3+ 2 arctan
2 exp [ - ik cot t!?l A,
= J0(2k cotl)-iJ1(2k cotp)’ B(k’B’
( 2 . 5 ~ )
(2.5b)
2.2. Exact solutions of the linear theory
This formalism will now be used to study the behaviour of more general waveforms
approaching the sloping beach. Since the governing equation (2.3) is linear and
homogeneous, then, as Stoker (1947) pointed out, the standing-wave solution (2.4)
can be used to obtain travelling-wave solutions by linear superposition. When a
boundary condition is specified, the solution follows directly from the Fourier
transform of the equation. For example, when the incident wave is of the form
q(X,, t) = @(k) e-Ikct dk ,
then the transmitted wave to the beach is given by
This solution is only valid when x 3 0; when x < 0, (2.3) does not reduce to Bessel’s
equation. To obtain details of the motion in this case, one must solve the nonlinear
set (2.2).
When initial values are available, it is necessary to use a different method. The
standard practice is to use Hankel transform techniques ; the substitution [ = xi
transforms (2.3) into the equation : va+ ( 1/C)qs = 4 cot Prtt, and the solution follows
526 C. E. Synolakis
directly as a Hankel transform integral (Carrier 1966; Tuck & Hwang 1972).
Alternatively, one may use the expansion
m
~ ( 6 , t ) = Z Cn(t)Jo(jn 5)
n-1
where jn is the nth zero of J,. The coefficients of the Fourier-Bessel series can be
found directly from the series expansion by using the orthogonality properties of the
Bessel functions : n f l
s is the integration variable. c,(t) can then be determined explicitly from
and the appropriate initial conditions. A similar solution is presented in Carrier,
Krook & Pearson (1966) for an equation that describes the motion of a hanging
chain.
2.3. Nonlinear theory
To solve the nonlinear set (2.2) for the sloping-beach case, h,(x) = x tan@, Carrier &
Greenspan (1958) proceeded to consider the independent variables x, t as functions of
the Riemann invariants of the hyperbolic system and, after some effort, they were
able to deduce the following hodograph transformation, (referred to as the C&G
transformation for short) :
$a u=- ,
U
( 2 . 7 ~ )
t = c o t @ -+A , (2 1 ( 2 . 7 ~ )
and 7 = i $ A - $ 2 (2.7d)
which reduces the set (2.2) to a single linear equation,
(a@.,), = @AA * (2.8)
The transformation is such that in the hodograph plane, i.e. the (u,A)-space, the
shoreline is always at Q = 0; this can be deduced easily by setting a = 0 in (2.7b) and
by observing that then x = -7 cot/3, which is an equality only valid at the shoreline
tip.
2.4. Exact solutions of the nonlinear theory
Equation (2.8) can be solved with standard methods. When a boundary condition is
specified, then the method of choice is the Fourier transform technique. Defining the
Fourier transform of $(a, A) as
Y(a, f ) = $(u, A ) e-iAE dA, s_9,
and, if Y(uo, &) = P ( f ) , then the bounded solution at u = 0 and u = 00 takes the
form
Runup of solitary waves 527
If an initial condition is available instead, then one may use Hankel transform
methods (Carrier 1966), or proceed as suggested in $2.2.
To complete the solution of (2.8) an appropriate initial or boundary condition has
to be specified explicitly. Carrier & Greenspan (1958) presented a general solution to
an initial-value problem where the initial velocity u(x, to) is zero. Spielvogel (1974)
used this solution to derive the evolution of a wave during rundown, assuming an
initially exponentially shaped runup profile. Carrier & Greenspan also presented
certain solutions with u(x, to) =I= 0, but for very specific initial conditions.. Tuck &
Hwang (1972) also solved the initial-value problem.
In general, it is difficult to specify initial or boundary data on the sloping beach
without making restrictive assumptions about the solution ; a boundary condition
requires specification of the solution at (zo, Vt), and an initial-condition specification
at (to,Vx), but in practice the solution is only known at (zo 2 cotp, t < to), where to
is the time when the wave reaches the z-location xo. Even when boundary or initial
conditions are available in the (x,t)-space, the process of deriving the equivalent
conditions in the (u, A)-space is not trivial.
These difficulties have restricted the use of the Carrier & Greenspan formalism to
problems that can be reduced directly to those solved by Carrier & Greenspan. This
is unfortunate because some of the problems described can be circumvented. Carrier
(1966) demonstrated how to specify a boundary cohdition when reflection from the
beach is negligible. Another method will be presented here to specify a boundary
condition including reflection.
2.5. Approximate solution of the nonlinear theory
Carrier (1966) pointed out that far from the shoreline nonlinear effects are small. The
transformation equations can then be simplified by neglecting O(uz) terms. To the
same order, $,, + kz and $Ju 4 +A. Using these approximations, the set of
equations (2.7) becomes
U '
u=- $u 'I = a$,,, x = &uz cot/?, t = -$A cotp. (2.10)
These equations are uncoupled and allow direct transition from the (a,A) to the
(z, t)-space.
One method for specifying a boundary condition in the physical space is to use the
solution of the equivalent linear problem, as given by (2.8). This is formally correct
to the same order of approximation as (2.10). The obvious choice for the specification
is the seaward boundary ; it is desirable to use the linearized form of the equation of
motion at the furthest possible location from the initial position of the shoreline
where the Carrier & Greenspan formalism is valid. This is the point z = X, = cot /?
and it corresponds to the point cr = uo = 4 in the (c,h)-space. Then (2.10) implies
that v(Xo, t ) = *$,,(4, A). The boundary condition F(&) in the (a, /\)-space is then
determined from (2.29) by repeated application of the Fourier integral theorem.
Assuming that $(uo, A) + 0 as A+ f co, then the solution of (2.8) follows:
- . * .
the substitution K = (2/X0)& was used to simplify the expression.
528 C. E. Synolakis
2.6. Runup invariance in the linear and nonlinear theory
It is interesting to compare the predictions of the linear and nonlinear theory for the
maximum runup and minimum rundown. It will be proved that they are identical.
The maximum runup according to the linear theory is the maximum value attained
by the wave amplitude at the initial position of the shoreline x = 0, or
@(k) exp [ - ik(X, + ct)] dk
J0(2kX0) - iJ1(2kX0) 'I(0,t) = 2 (2.12)
In the nonlinear theory the maximum runup is given by the maximum value of the
amplitude at the shoreline 7(xs,A). (xs is the x-coordinate of the shoreline and it
corresponds to t~ = 0.) To determine q(xs,A) one can use (2.7d) to obtain
(2.13)
@(K) exp [ - iKX,( 1 -+A)]
dK-$:;
J,(2KX0) - iJ,(2KX0)
us = dxs/dt is the velocity of the shoreline tip.
Setting us = 0 and t~ = 0 in the transformation equations (2.7) reduces them to
At the point of maximum runup the velocity of the shoreline tip becomes zero.
u = 0, 'I = : $ A , x = -'I cotp, t = -y cotp. (2.14)
Substitution of these values in (2.13) reduces it to (2.12), proving that the maximum
runup predicted by the linear theory is identical with the maximum runup predicted
by the nonlinear theory. The same argument applies at the point of minimum
rundown where the shoreline tip also attains zero velocity. This invariance was first
noted by Carrier (1971) who observed that the maximum runup is given correctly by
a linear theory (presumably (2.8)) as the maximum value of g = a$*. However,
Carrier did not present an actual comparison between linear and nonlinear theory for
polychromatic waveforms, and this result has been largely unrecognized. It is
unexpected, because, as will be shown later, the amplitude-evolution data derived
using the linear and the nonlinear theory differ most at the initial shoreline (see, for
example, figure 5 ) .
3. The solitary-wave solution
The results of the previous section will now be applied to derive a result for the
maximum runup of a solitary wave climbing up a sloping beach. A solitary wave
centred at x = X , at t = 0 has the following surface profile:
(3.1)
H
d
~ ( x , 0) = - sech2 y(z -X, ) ,
where y = (3H/4d)4. The function @(k) associated with this profile is derived in
Synolakis (1986) and it is given by
@(k) = Qk cosech (ak) e'"1, (3-2)
where a = n/2y. Substituting this form into (2.12) and defining as &(t) the
dimensional surface elevation at the initial position of the shoreline, then the
following relationship results :
dk
exp [ik(X, -X,-ct)]
J0(2kX0) - iJ1(2kX0)
!!@ = "r k cosech (ak)
d 3 -m (3.3)
Runup of solitary waves 529
This integral can be calculated with standard methods of applied mathematics ; its
convergence and evaluation is discussed in the Appendix. The integration result is
The series can be simplified further by using the asymptotic form for large arguments
of the modified Bessel functions. When 4X,y B 1, then
(-l)n+1n~exp[-2y(X,+X,-ct)n]. (3.5)
This form of the solution is particularly helpful for calculating the maximum runup.
The series in (3.5) is of the form
z (-l)n+lnixn;
its maximum value occurs at x = 0.481 = e-0.732. This value defines the time t,,
when the wave reaches its maximum runup as
m
n-1
t,, = - X1+X,--
C Y 0-366). Y
The value of the series (3.5) at t,, is s, and s,, = 0.15173. Defining as W the
maximum value of ' ( t ) and evaluating the term 8(n~3)bm, then the following
expression results for the maximum runup :
- = 2.831 (cot/?)t
W
d (3.7)
This equation will be henceforth referred to as the runup law. It is formally correct
when (H/d)f >> 0.288 tan/? - the assumption implied when using the asymptotic form
of the Bessel functions - and when the series converges as discussed in the Appendix.
The same result can also be derived by calculating $A from the nonlinear-theory
solution (2.11) and then using the appropriate equations (2.14) for the shoreline
motion.
To derive surface profiles in the entire flow domain i t is necessary to use the
nonlinear theory and solve the transformation equations (2.7). The solution is given
by
u = - $n
g
J,(&nvX,) exp [iK(XI -X, +-
K cosech ( a ~ ) J0(2Kx,) - u1(2Kx,) 'no)1dK-~2, (3.8b)
x = cot/?($-+ ( 3 . 8 ~ )
and t = cot/3($-:n). (3 .8d)
530 C . E . Synolakis
- 13
' 0
FIQURE 2. The function f l A ( c , A ) defined by (3.9) for a solitary wave with H/d = 0.019 up a
beach with co tp = 19.85.
The integrals in these equations can be evaluated directly for given (a, A) using the
formalism of the Appendix. The function $A is shown in figure 2 as a function of
(a,A) for the case when H / d = 0.019, X, = 19.85 and X, = 37.35. The amplitude
evolution predicted by (3.8b) is discussed in 85.2, where it is compared with
laboratory data obtained with the methods discussed next.
4. Experimental equipment
A series of laboratory experiments was conducted to investigate the validity of the
results derived in the previous sections. The experiments were performed in the
40 m wave-tank facility of the W. M. Keck Laboratories of the California Institute
of Technology. The facility consists of a wave tank, a wave generation system and
a wave measuring system. The experimental equipment is detailed in Hammack
(1972), Goring (1978) and Synolakis (1986).
The wave tank has glass sidewalls and dimensions 37.73 m x 0.61 m x 0.39 m. At
one end of the tank and at a distance of 14.68 m from the wave generator a sloping
beach was constructed; it consisted of a ramp made out of anodized aluminium
panels with a hydrodynamically smooth surface. The ramp was supported with a
wooden truss which was surveyed periodically to ensure that it conformed to the
Runup of solitary wuvcs 53 1
initial 1 : 19.85 slope. For brevity, the ramp will be referred to as the laboratory
beach.
The wave generation system consists of a piston attached to vertical wave plate
which moves horizontally and displaces the adjacent fluid thereby generating waves.
The piston is driven by a doubly actuated hydraulic cylinder controlled by a servo-
valve which can be programmed to produce any desired piston motion. The system
was interfaced with a PDP11/23 microprocessor which provided the required piston
trajectory. The generation algorithm developed by Goring ( 1978) and generalized by
Synolakis (1986) was used to compute the trajectory. The laboratory equipment
produced near-perfect solitary waves ; near-perfect refers to waves that conform to
the Boussinesq profile (3.1) and have a tail that does not exceed 2% of the wave
height H.
To measure wave heights in the constant-depth region parallel-wire resistance-
type wave gauges were used. The gauges are constructed of steel wire of diameter
0.025 cm and can be calibra