Nonllinear transformation of waves in finite water depth

James M. Kaihatua)
Fluid Dynamics Branch, Remote Sensing Division (Code 7253), U.S. Naval ResearchLaboratory,
Washington, D. C. 20375
James T. Kirby
Center for Applied Coastal Research Department oj’ Civil Engineering, University oj’ Delaware,
Newark, Delaware 19716
(Received 27 December 1994; accepted 7 April 1995)
The formulation of a nonlinear frequency domain parabolic mild-slope model is detailed. The
resulting model describes two-dimensional wave transformation and nonlinear coupling between
frequency components. Linear dispersion and transformation characteristics are dictated by
fully-dispersive linear theory, an improvement over weakly-dispersive Boussinesq theory. Both the
present model and a weakly-dispersive nonlinear frequency domain model are compared to
laboratory data for both two-dimensional wave transformation and pure shoaling. It is found that, in
general, data-model comparisons are enhanced by the present model, particularly in instances where
the wave condition is outside the shallow water range. Q 1995 American Institute of Physics.

I. INTRODUCTION two free surface boundary conditions, and then integrating to

solve for the amplitude coefficients, which vary slowly in
The Boussinesq equations of Peregrine* are often used
time. The interaction coefficients which determine the degree
for nonlinear wave propagation in shallow water. These
of energy exchange among resonant triads were derived
equations have been treated in both the time domain (e.g.
without restrictions on the size of the dispersion parameter
Rygg”) and the frequency domain (Freilich and Guza3; Liu
kh. However, the time varying, spatially periodic formula-
et ~1.~; Kirby’). They provide reasonably accurate descrip-
tion is not suitable for most applications, since wave infor-
tions of the wavefield provided that the water is sufficiently
mation is usually taken in the form of time series from sta-
shallow (kh4 1, where k is the wavenumber and h the water
tionary gauges. These data are processed and stored by
depth). Unfortunately, errors in wave shoaling and celerity
means of a frequency-domain Fast Fourier Transform (FFT),
predictions become evident in intermediate or deep water,
since the linear dispersion relation and shoaling mechanisms which presupposes equal intervals of frequencies rather than
associated with these models become entirely inadequate wave numbers. Additionally, the waves over varying
outside the shallow water range. For example, the “consis- bathymetry can evolve rapidly in space, precluding the as-
tent” model of Freilich and Guza3 (which is a frequency sumption of spatial periodicity. Keller8 developed a set of
domain treatment of the Boussinesq equations of Peregrine’ equations describing the evolution of two interacting wave
in one dimension) has Green’s Law as its shoaling mecha- components. He demonstrated that, in the nondispersive
nism: limit, this same set of evolution equations can be derived
from the exact Euler equations, the nonlinear shallow water
-$ equations, and the Boussinesq equations. The model de-
a h
---ES . (1) scribed in this paper would match his model before the non-
i a0 i ( ho 1 dispersive limit is taken if the number of frequency compo-
This dictates a monotonic increase in wave amplitude a with nents is truncated to two and slow time dependence is
a decrease in h. Applying this shoaling relation to a typical reincorporated. Keller8 did not model his equations. Agnon
wind-wave spectrum would result in overshoaling in fre- et al.’ derived a one-dimensional nonlinear shoaling model
quency components which are in either intermediate or deep for time periodic, spatially varying waves from the boundary
water, where kh is not small. value problem.
In this study we wish to derive a nonlinear frequency We take a different approach in the present study to de-
domain parabolic mild-slope equation. This model would rive what is essentially a two-dimensional extension of the
have the correct linear dispersion and shoaling characteristics model of Agnon et a1.9 We begin with the boundary value
(i.e., that of fully-dispersive linear theory), while retaining problem for water waves and derive the basic model in Sec-
nonlinear coupling between frequency components. Several tion II. In Section III we assume periodicity in time and
investigators have previously worked on models similar to transform the problem into the frequency domain by factor-
the one described herein. Bryant6.7 formulated a model from ing out the time dependence, resulting in a nonlinear elliptic
the boundary value problem for water waves over a flat bot- mild-slope equation. In the process we explicitly formulate
tom by specifying spatially periodic Fourier expansions for the nonlinear coupling between frequency modes. Section IV
the variables C$and 7 which satisfy the Laplace equation and details the use of the parabolic approximation to develop
the bottom boundary condition, substituting these into the two-dimensional evolution equations governing the spectral
amplitudes. We show comparisons of the resulting model to
%-mail: kaihatu~shredder.ml.navy.mil both experimental data and weakly-dispersive spectral trans-

Phys. Fluids 7 (8), August 1995 1070~6631/95/7(8)/i 903/l 2/$6.00 Q 1995 American Institute of Physics 1903

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
formation models in Section V. We present conclusions and
propose extensions of the work in Section VI. In Appendix (11)
4 1 we present an alternative derivation of the basic model
using the Lagrangian for water waves. In keeping with the
or the usual depth dependence dictated by linear theory. The
general nature of the discussion, we extend this formulation
frequency w, and the wave number k, are related by the
to include the effects of ambient currents in Appendix A 2.
linear dispersion relation:
II. BOUNDARY VALUE PROBLEM
m;=gk, tanh k,h. (12)
We consider a waveiield propagating over a spatially
varying bottom. The Cartesian coordinate system (x,y,z) is It is convenient to combine the two surface boundary condi-
located on the still-water level, with z considered positive
tions (8) and (9) into a single equation for 4 only. Eliminat-
upwards from the origin. The wavefield is denoted in terms
ing 17from the surface boundary conditions leads to:
of the free surface elevation g(x,y,t) (where t is time)
which is defined from the still water level z = 0. The water
depth is denoted h(x,y). The fluid is assumed to be inviscid #z= - i 4t,+$vh4jf+ fM,>f- &btj~r
and irrotational. In concert with this assumption, we use the [
water wave boundary value problem for the velocity poten-
tial 4: +v,*(q$v,cpj ; z=o. (13)
I
v2q5=v;++ cpzz=o; -la=ws?J, (2)
4Z= -V,h.V,& z= -h, In the manner of Smith and Sprink~,‘~ we will use Green’s
(3)
second identity on the variables f,, and 4, , as follows:
g.17+~i+4(vh~)2-t8(~2)2=0; z=‘17, (4)
77~-~z+v,77’vh~=o; z=q (5) Vh’(/y-3zvdn)
- I_:pZ&
where Vh denotes the gradient operation in the horizontal
coordinates (x,y) and g is the gravitational acceleration. = -fn&z[z=O+O(E,ti2) 04)
Subscripts denote differentiation.
We wish to retain leading order nonlinearity in the free
where (Y is a parameter characterizing the bottom slope. For
surface boundary conditions. Rather than scale and nondi-
our purposes it is assumed that
mensionalize the problem, we instead follow the approach of
Bryant6 and retain dimensional quantities, with the under-
standing that leading order nonlinearity is 0( e2), where &~O(E). (15)
e(=ka, where a is a characteristic wave amplitude) is the
nonlinearity parameter. We note here that the scaled, nondi- In this manner we can eliminate bottom boundary terms in
mensionalized problem had been addressed by Agnon et ~1.;~ comparison to O(2) terms from the free surface. Develop-
the approach here is entirely equivalent. ment of the linear part of the models is identical to Smith and
The free surface boundary conditions are both nonlinear sprinks; *e reference is made to their paper. We simply note
and applied at a position not known a priori; thus we expand that:
these in Taylor series about z= 0 and retain terms to
O(e2). The truncated boundary value problem is now: “f,(O) = 1, (16)

f,,(O)
=z,
V&p+ q&=0; -h=sz=zO, (6)
&= -V,h.V,& z=-h, (7) 07)
gv+ 9%+ 4(vh4)2+ s?q2+ 77&=0(& z=o, (8)
17t-&+Vh?pVh+ 17q&=o(E3); z=o. (9) 08)
Instead of using the approach of Bryant,6 who substituted an
appropriate form for c5 that would satisfy the Laplace equa-
tion and the bottom boundary condition, we instead use the
approach of Smith and Sprinks,” who used Green’s second (19)
identity to derive a linear mild-slope equation. We first as-
sume a superposition of solutions: where C, is the wave celerity and C,, the group velocity of
the nth component. The subscript z refers to partial differen-
4(x,y,z,i)=n~l fn(kn,h,z)~,(k,,o,,-~,y,t), (10) tiation with respect to z. Substitution of (lo), (13), (16), (17),
(18), and (19) into (14) and manipulation of the integrals
where w, is the radian frequency and k, is the wave number yields a time-dependent mild-slope equation with nonlinear
of the nth frequency component, and: coupling between modes:

1904 Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 that, in the linear limit, (20) approaches a representation of
~,,,-v,*.[(cc,),v,~,l+ 4 l- g 5, decoupled, independent waves. In the shallow water limit (as
i ni k,h approaches zero) we approach a frequency domain rep-

1
resentation of Boussinesq-type shallow water waves, as dis-
T s ~(Mdt cussed by Freilich and Guza.3. additionally, Bryant7 showed
.-; i [ “‘:“”2 G,&l,)t- 2 2
that a system’ like (20), with coefficients expressed as a
power series in E, will join smoothly to a solution for Stokes
waves in deep water for N= 3 components.

III. TIME-HARMONIC WAVE PROPAGATION IN TWO

+w&,tvh~dl 9 cm DIMENSIONS
1?I
where the notation On on the right-hand side of (20) indi- We wish to develop a series of evolution equations for
cates that the nonlinear coupling occurs between the mode the propagation of time-harmonic waves in two spatial di-
under consideration (n) and two other modes in the spectrum mensions. Hence, we can completely factor out the time de-
(I and m). .Explicit relationships between I, m, and y1follow pendence by assuming:
the des of resonant triad interaction theory (Phillips”).
These restrictions will be imposed in the next ‘section. $n(XYfjk&e-iut n + +“* pnt
, ,
2
Despite the fact that we have decomposed the wavefield
into individual components, we have not made any assump- and using resonant triad interactions to formulate the nonlin-
tions concerning. the behavior of these components; they ear coupling between frequency components (e.g. Phillips*‘).
could represent propagating or standing ‘waves with charac- This yields a time-harmonic wave equation with modifica-
teristics that vary slowly in time,and space. It is noted here tion by nonlinearity:

A
- ~l&WhZ+E- “ly’“” (6+ op,+~+ w;,lj@&+~
1. (22)

We note here that the solution of Bryant6 can be recov-

ered by assuming spatial (rather than temporal) periodicity in $X(x,yj=z A~(x,yje-jlk,(x,~)dx, (24)
(20). While we have incorporated an assumption of time pe-
riodicity into our problem, we have still not explicitly speci-
fied the spatial variation of the wavefield. Additionally, (22) where the complex amplitude A, is assumed to be a slowly
is elliptic, with the linear terms comprising the elliptic mild- varying function of the spatial coordinates (x,y), and the
slope equation model of Berkhoff12 and Smith and Sprinks.” wave is assumed to be traveling primarily in the + x direc-
Numerical solution of the equation in this form would re- tion. Substituting (23) and (24) into (22) gives:
quire fine resolution (many gridpoints per wavelength) and
complete specification of the boundary. This is not well
suited for coastal propagation problems in which the shore-
line location is not known a priori and the modeled region
+ [ (cc,),(A,e’lk,(X.y)dX)y]ye -ifk,(X,y)dX
can be on the order of kilometers in size.

IV. PARABOLIC APPROXIMATION

We will restrict the problem to that of a progressive

wave field with the following form:

&,*(x,y)
= - E An(X,y)eiJk~(x,y)d*, (2%
where

Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby 1905

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 (32)

= ~[o~k~k,_l+(kr+k,-l)(o,-rk,+olk,-,)o.l which gives:

A,(x,y) = a IIix 9yjej(fk(x).dx-fk,(x.y)dx) (33)
w%-lf&j, (26) and:
A;(x,y) =a;(x,y)e -i(lk(x),dx-fk,(x.y)dx). (34)
Substituting this into (25) yields: ’

2~ikCC,L~,, -2ikCC,),ik,-k,)a,-ti(kCCg!nxan
+ [wcgM4,1y
(27) n-l
=$ 1z 1Rala,_le’f(kl+~-,-,)dX
Because (25) is still elliptic, we need to explicitly invoke N-n
the parabolic approximation (Radder;13 Lozano an4 Liu14). +2C ~a~a,+~a’f(k,+r-~-~)~~ .
We have assumed that the wave propagates primarily in the (35)
I=1
+x-direction; thus we retain fast wave-like variations in the
x-direction but not in the y-direction. The fast variations in The two-dimensional parabolic equation (35) was modeled
the &irection are accounted for by me complex exponential with the Crank-Nicholson numerical scheme. This scheme
in (23) and (24). We use the scaling approach of Yue and is unconditionally stable fop linear ‘problems, and is second-
Mei15 to order the derivatives of A, as follows: order accurate in x and y.’ The paraboljc model (35) was
converted to finite differences much like the Kadomtsev-
Petviashvili (KP} model ‘of J&r et al. ,4 where the numerical
scheme was detailed; thus, it will not be shown here. For the
case of unidirectional wave shoaling we reduce the model to
one dimension:
WC,L
Anx+ 2(kCC,), An
since the fast variations in x have already been factored out.
The ordering of the y-derivative in (29) allows us to keep
d2A, / G’y2, thereby allowing us to model the slow phase-like
variations of the wave in the y-direction that occur when it is N-n
turned at a small angle to the’x-direction. Additionally, since +2‘2. SA:Anflejf(k,+j-kl-k.)dx (36)
we assume the wave to propagate primarily in the 171

x-direction, changes in the amplitude A, would be due where we revert back to the A, notation since the reference
mostly to x-derivatives of the depth h. Thu$ the order of wave number k, is identical to k,. This one-dimensional
bottom slope in the x-direction (h,j, as well as x-derivatives model is identical to that of‘Agnon et a1.i9which was derived
of depth-dependent properties (e.g,, C, C,), should also be using continuous Pourier integrals rather than discrete Fou-
O( e2). This allows us to order the relative amplitudes of the rier transforms, as we have done here. We model (36) with a
first two terms of (25) as: fourth-order. Runge-Kutta scheme with error checking and
adjustable stepsize control.
E4C(CC,),A,,I,~E2C2i(kCCg)nAnxlr (30)

where the subscript x refers to differentiation with respect to V. COMPARISONS TO DATA

X. Since we are keeping terms to O(i2) we drop the first
To verify the performance of the model, we will com-
term in (25). Additionally, we need to factor out any y ,de-
pare it to available experimental.data. The first comparison
pendence from the phase function. This must be done since
will be to the data of Whalin,? who conducted a laboratory
we are only integrating the phase in x, but the wave number
experiment to investigate the &nits of linear refraction
k, is a function of both x and y . There are several ways ‘to
theory. He generated sinusoidal waves of 1,2, and 3 second
address this; we choose the method of Lozario and Liu,13
periods and ran them over bathymetry that resembled a tilted
where they defined’ a y-averaged wave number En(x) as a
cylinder. The bottom contours and tank dimensions are
reference phase function. Thus, we, rewrite (23) and (24) asf
shown in Figure 1. We only compare the 2 and 3 second
cases; the 1 second case will be discussed later in this sed-
&(x,y) = - $ a,(x,y)e’Jk‘Jx)d~, tion. The wave parameters for the cases used in the compari-
n son are shown in Table I, where the’tank depth used in cal-

1906 Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 -‘~ 0
0 5 IO 15 20 4
0 5 10 15 20 2s
x 04 xW

-0.1

-0.51 I
0 5 10 15 20 25
x Cm)

FIG. 1. Wavetank layout of experiment of Whalin.16 (Top) bottom contours;

(bottom) centerline depth.

culating the nonlinearity parameter ka and the dispersion

parameter‘kh were chosen to maximize the values of these
parameters Lk was calculated by (12) using the primary bar-
manic]. Wl~alin’~ placed gages at certain locations along the
0
centerline, and recorded the amplitudes of the first three har- 0 6 10 15 20 5
*. x (m)
monics. This experiment demonstrated the inadequacy of
non-diffractive linear refraction theory as a general modeling
methodology; ‘wave ray crossing downwave of the top of the FIG. 2. Comparisons between models and Whalin’s experiment, T=3 s,
ao=0.68 cm. Nonlinear mild-slope model (-), I@ model of Liu et ~1.~
tilted cylinder would indicate an infinite waveheight accord- (- -). data of Whalin16 (0). (Top) first harmonic; (middle) second har-
ing to linear refraction theory, .which would be impossible. monic; (bottom) third’haimonic.
Diffraction is very prevalent there. Additionally, using a lin-
ear sinusoidal wave as .input generates higher harmonics
since.the nonlinear surface boundary conditions cannot be
satisfied by a single harmonic. ”
We’ ran the parabolic nonlinear mild-slope model (35) Figures 2 through 4 show the comparisons between the
and the Rl? parabolic frequency domain model of Liu et al.” models and Whalin’s data for the case of T= 3 seconds. It is
against the data ‘of Whalin using the parameters shown in apparent that neither model predicts the first harmonic am-
Table I. This KP model is essentially a frequency, domain plitudes particularly well. This seems to’ be endemic of most
treatment of a reduction. of the Boussinesq equations of data-model comparisons done in the literature where Wha-
Peregrine,’ and so ‘has lowest-order dispersion and Green’s lin’s T= 3 second data were used ie.g., .Rygg;’ Madsen and
Law shoaling. We used N= 5 harmoni& for the T= 3 second SBrensen;i7 Nwogu’*). Liu et aL4 maintain that the relatively
case and N= 3 harmonics for T= 2 seconds. A grid spacing short evolution distance for a 3 second wave period (roughly
of 4x= 12 cm and Ay=8 cm was used for both models. No two wavelengths) may at least partially ‘violate the si0Wiy
substantial improvements were noted when the stepsizes varying ‘assumption used to derive their frequency domain
were halved. KP model. In any case, both models show reasonable com-
parison with the T=3 second data for the second and third
harmonics.
The T=2 second case is more demonstrative of the ad-
TABLE I. Wave parameters of experiment of Whalinr6 used in data-model
comparisons (hi=457 cm; h,=15.2 cm). vantages of the dispersive model. The value of the dispersion
parameter kh in the deep portion of the tank is 0.733, which
afl may violate the shallow water assumption used in the KP
(cm) ku kh model of Liu et al.4 Figures 5 through 7 show’these results.
3 0.68 0.0068 0.464 It is apparent that (35) models the first harmonic amplitudes
3 0.98 0.0098 0.464 better than the KP model. The overprediction of these ampli-
3 1.46 0.0146 0.464 tudes by the KP model is most likely due to the inherent
2 0.75 0.0120 0.733
?a 1.06 0.0169 0.733
Green’s Law shoaling.’ The prediction of the-second and third
2 1.49 0.0238 0.733 harmonic amplitudes are equivalent between the two’mbdels,
despite the relatively high values’of kh in the deep portion of

Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihafu and J. T. Kirby 1907

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 o,o@&----.-L- -_-- ._.L_ _---$-.-A ooL.-.- -.- ; -A.-L--_ - --..-I
0 5 10 20 25 10 15 20 25
x (4 x Cm)

---- I

-- 15 20 26 0.005ooL ; “yy6 ---- ,-1,

x ON x(m)

0
0 5 10 15 20 25
x Cm) x O-4

FIG. 3. Comparisonsbetweenmodelsandwhalin’sexperiment, T=3 s, FIG. 4. Comparison betweenmodelsandWhalin’sexperiment, T=3 s,

ao=0.98cm. Nonlinearmild-slopemodel(-), KP modelof Liu et aL4 ao=1.46cm. Nonlinearmild-slopemodel(-), KF’ modelof Liu et aL4
(- -). dataof Whalin14
(0). (Topj first harmonic;(middle)secondhar- (- -), dataof whalin16(0). (Top!Fist harmonic;(middle)Secondhar-
monic;(bottom)thirdharmonic. manic;(bottom)thirdharmonic.

the tpk (kh = 1.92 for the second harmonic and kh=4.94 results demonstrated by the T=2 second case is sufficient
for ,the third). This equivalence may be due to the fact that attestation to the importance of incorporating dispersion and
these harmonics are input with zero amplitude and allowed
shoaling characteristics in nonlinear wave transformation
to grow initially due to nonlinear energy exchange between
which are valid for arbitrary water depth.
components. For such a relatively short slope, ‘it is possible
that nonlinearity, rather than first-order quantities lik& shoal- We also compare our one-dimensional model (36) to
ing and refraction, control the evolution of these harmonic Case 2 of the wave shoaling experiment of Mase and
amplitudes, and as such are not affected by the disparity Kirby.lg This experiment was conducted to study the shoal-
betwZen a dispersive shoaling mechanism and Green’s Law. ing and breaking characteristics of irregular waves. Case 2 of
‘I& would not be the case for a broad spectim of irregular this experiment was notable in that the peak frequency of the
waves shoaling over a long slope, as we shall se&in the next simulated wave spectrum was actually in fairly deep water
set of compesons. (kh = 1.9), well outside the probable range of validity for
It can be argued that the T= 1 second case of Whalin16 Boussinesq theory. The experimental setup is depicted in
would be a more severe test case for (35). The value of kh Figure 8, which was taken from Wei and Kirby.“’ Mase and
for this case is 1.92 in the deep portion of the tank. Both Kirby I9 generated Pierson-Moskowitz spectra at the wave-
Madsen and SBrensen17and Nwogu” compared their time maker in two different experiments; Case 2 had a peak fre-
domain dispersive Boussinesq equations to this case; they
quency of 1 Hz. This choice of peak frequency, combined
both demonstrate reasonable agreement with the data. How-
with the broad-banded spectral shape, allowed significant en-
ever, the effectiveness of the frequency domain formulation
used here is contingent upon the relative size of the argu- ergy in frequencies that were well into the deep water range,
ments of the complex exponentials in the nonlinear right- thus providing a demanding test for dispersive shoaIing mod-
hand side of (35). These arguments can become large in deep els.
water, invalidating the assumption of slow spatial variation We will be taking the model through the surf zone, re-
used in deriving the model. This problem will likely arise in quiring implementation of a dissipation mechanism in our
the simulation of the T= 1 second case; thus we do not com- model. We use the methodology of Mase and KirbyI to
pare our model to it. We feel, however, that the favorable formulate this dissipation, which revises (36) to:

1908 Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 0.014
‘,,-~~z=7y--j _--.

i$ ..:;>pF-y=+ ;
, x

o.ooeL- I 1
0 5 10 15 20 25 0 5 10 15 20 25
x ON x (m)

0.015
0.01
I-- t ,/$q&
z< poo - \ I
0.005

0
0 5 10 15 20 25
x ON

5
4
E
z3
2
1

0 0
0 5 10 15 20 25 0 5 10 20
x Cm) x Cm)

FIG. 5. Comparisons between models and Whalin’s experiment, T=2 s, FIG. 6. Comparisons between models and Whalin’s experiment, T=2 s,
no=0.75 cm. Nonlinear mild-slope model (-), KF’ model of Liu et ak4 ao=1.06 cm. Nonlinear mild-slope model (-), KP model of Liu et aL4
(- -1, data of Whali# (0). (Top) first harmonic: (middle) second har- (- -), data of Whalin16 (0). (Top) first harmonic; (middle) second har-
monk; (bottom) thiid harmonic. - monic; (bottom) third harmonic.

data of Mase and Kirby” and are close to values found from
field data by Thornton and Guza.” The dissipation function
(41) is the so-called “simple” probabilistic dissipation model
of Thornton and Guza.21The frequency distribution of the
RAIA,_lz’s(kI+k’-l-k.)dx dissipation mechanism is divided into CY,~,which drains an
equal amount of energy across all frequencies, and LY,~,
which weights the dissipation higher toward higher frequen-
+2:x: SA;EAn+leiS(k,+r-kl-kn)dx (37) cies; this is shown in (38). The distribution (~,t represents
the f2 dissipation trend found from the data. A detailed deri-
where: vation of this dissipation model is found in Mase and
Kirby.” We will compare (37) to the “consistent” shoaling
(38) model of Freilich and Guza,3 which was equipped with the
same dissipation mechanism as described above.
(39) Figures 9 through 12 show comparisons between the
data, the nonlinear mild-slope model (37), and the “consis-
f;eak%~4~2 tent” shoaling model. Most of the waves began to break at
%I = (P(x) - %o) (40) h = 17.5 cm, so Figs. 11 and 12 detail the effectiveness of the
%f%Li" '
dissipation function as well as that of the model equations. It
36 B3fpea&n, is apparent that the nonlinear mild-slope model compares
P(x) = ~ , (41) very well to the data for much of the frequency range
4va
throughout most of the domain. The “consistent” model, on
the other hand, exhibits a strong deviation from the spectral
Km,=2 45i (42) peak even at the first slope gage (d= 35 cm). This deviation
is due to the overshoaling of frequency components in
and where fpeakis the peak frequency of the spectrum. The deeper water than that for which Green’s Law (1) is valid. As
free parameters B, F and y are assigned values of 1.0, 0.5 the waves shoal and break, deviation of both models from
and 0.6, respectively. This was based on comparisons to the the data becomes more pronounced. This may be due to pos-

Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby 1909
Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 o,olL.-_----l--.. $---.--d
0 5 10 2n 25
x (4

0.02

0.015

5 0.01

0.005

0 ---(I. -_
0 5 10 15 20 25 f IHd
x (ml
FIG. 9. Comparison of shoaling models to Case 2 of Mase and Kirby.*’
Experimental data (-), nonlinear mild-slope model (- -), consistent model
0.01
of Freilich and Guza3 (-w-). (Top) input spectra at d=47 cm; (bottom)
7-- d=35 cm.

thus insuring the validity of the linear terms of the model in

0.002 .‘b\ ;‘b arbitrary water depth. This represents an advance over
0
frequency-domain Boussinesq models, which are weakly dis-
0 5 10 15 20 25
x cm)
persive and thus limited to shallow water. The resulting

FIG. 7. Comparisons between models and Whalin’s experiment, T=2 s,

ao=1.49 cm. Nonlinear mild-slope model (-), KP model of Liu et al’
(- -), data of Whalin16 (0). (Top) first harmonic; (middle) second har-
monic; (bottom) third harmonic.
‘~7 ’ ------I

sible inadequacies of the dissipation mechanism. Overall,

however, one may conclude from both data-model compari-
sons that the parabolic nonlinear mild-slope model performs
f (Hz)
better than equivalent models with lowest-order dispersion
and Green’s Law shoaling. This is primarily due to the mild-
slope formulation, which retains fully-dispersive linear id!
theory for the dispersive and shoaling characteristics. 1

VI. SUMMARY
~~~:
This study detailed the development of a nonlinear para-
bolic frequency-domain mild slope equation. The linear char- _---.
1 1.5 2 2.5 3
acteristics of the model utilize fully-dispersive linear theory, f (Hz)

IO2

Wave Paddle I
10’
. .
z,,o
B
B Id
-
8
IO'*
0 0.5 1 1.5 2 2.5 3
f (Hz)

FIG. IO. Comparison of shoaling models to Case 2 of Mase and Kirby.lg

Experimentaldata (-), nonlinear mild-slope model (- -), consistent model
FIG. 8. Experimental setup of Mase and Kirbylg (taken from Wei and of Freilich and Guza3 (-.-). (Topj d=30 cm; (middle) d=25 cm; (bottomj
Kirby20). d=20 cm.

1910 Phys. Fluids; Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
 f (Hz) f (W

$10* FE loo

glo.l glo.l
$7
104 lo.=
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
f (Hz) f 6-W

f (Hz1 f (Hz)

FIG. 11. Comparison of shoaling models to Case 2 of Mase and Kirby.” FIG. 12. Comparison of shoaling models to Case 2 of Mase and Kirby.lg
Experimental data (-), nonlinear mild-slope model (- -), consistent model Experimental data (-), nonlinear mild-slope model (- -), consistent model
of Freilich and Guza3 (-.-). (Top) d= 17.5 cm; (middle) d= 15 cm; (bottom) of Freilich and Guza3 (-.-). (Top) d=lO.O cm; (middle) d=7.5 cm; (bot-
d=12..5 cm. tom) d=5 cm.

model was compared to experimental data and a frequency- APPENDIX: LAGRANGIAN DERlVATlON OF THE
MODEL IN A NONSTATIONARY DOMAIN
domain Boussinesq model. In general the present model
demonstrated improved comparisons with the data over the 1. Lagrangian derivation of Equation (20)
Boussinesq model, especially in instances where the data
A variational principle governing the motion of fluid de-
were outside the shallow water range. This was especially
scribed by (2) is given by Luke:‘”
apparent in the comparisons to the data of Mase and Kirby,”
where most of the frequencies were in intermediate or deep
water. We also added a surf zone dissipation mechanism to
s L(qqh;x,t)dxdt=O, W)
the model to enhance the comparisons to the experiment of
Mase and Kirby.” where
One drawback of the parabolic formulation used in this
study is the limitation to small angles of incidence. Future L= w
work will likely involve overcoming this limitation and in-
vestigating other possible dissipation mechanisms, as well as is the Lagrangian density of the motion, and {x, t} is the
generalizing this dissipation to two horizontal dimensions. propagation space of the waves. Since variations of L with
respect to arbitrary S# and Sg will reduce L by one order of
E, we need to retain terms to O(e3) in L, after which linear
terms will arise due to variations of O(ez) quantities of L..
ACKNOWLEDGMENTS Taking 77 and (b to be O(E) and expanding L about z = 0
then gives
The first author was supported by an ASEE National
Science and Engineering Graduate Fellowship from the Of-
fice of Naval Research while a Ph.D. candidate at the Uni-
versity of Delaware. The second author was supported by
University Research Initiative Grant No. DAAL-03-92- +t+ ; &+;(vh+)2+;b$‘z)2 - 643)
GO116 from the Army Research Office. 2-O

Phys. Fluids, Vol. 7, No. 8, August 199.5 J. M. Kaihatu and J. T. Kirby 1911
Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
Variation of L with respect to 77gives the dynamic free sur- for progressive waves. The absolute frequency w, may be
face boundary condition (8) directly. Varying L with respect written in terms of the intrinsic frequency CT,,according to
to C$then yields
w,-cr,+k,*u, (AlO)
4z(W,dz+ v(Wh
cr,= (gk,tanhk,h)1’2. (All)
I, /i $( Sqbj,, ‘=’ We proceed to 0( e2) by assuming forms for Cpand 17
according to

+TlG(
4%II z=odxdt. 644)

After one partial integration, the integrand of the second in- =40+W+1T (A12)
tegral becomes
1?= 170f(4’171* (A13)

vi- 17t+#%l
z=o
w),=o(~4) (-45) Here, v. represents the 0( 1) setdown due to the strong cur-
rent U. The O( 1) water depth is thus given by
due to (5). The remainder of the equation can be manipulated h=h,+ vo, where now h, is the still water depth. We use
by recognizing that only terms proportional to & in the the Lagrangian approach of this appendix. Accordingly,
0( E’) portion of (A4) will contribute to the linear portion of (A12) and (A13) are substituted into (Al) to obtain
equations for 4, after variations with respect to +, . We can LIp=L&+(E)L; f(2)Li
then write (A4) as: 6414)
where
ftjJ(-V,-Il)j=:dZbz&+
j_oj-$-hb7n,i(si) L, =g hi-h3
0
2 + 31U12(v0+~o>~ W5)

-vh’(~vh+)(G6)
11 z=o
dxdt=O, W)
G=g170771+
I R”(p~,+U.Vh~,jdz+~~U12~~,
ho i
6416)
where we must set SC$/,=,= SC& to obtain nonlinear terms
contributing to the 5, components. Substituting for vn, from
(8) then gives

+(~l,+uavh+lj 71

‘10

= -~h+#$Vh~k;(Vh~):- ;(4,):
I

+$.4tj;t 9 WY
1 n I z=O

where we have used (16) through (19). Substituting for (b +~(vh~,)‘+(~,321,11, - 6417)
1
from (10) then gives (20).
Variations of the integral of LA with respect to p. and ~$0
2. A nonstationary domain: Effect of ambient currents
yield the relations
The results of the preceding section may be conveniently
extended to the case of a moving domain containing an am-
c418)
bient current U(x,y)={U,V}=VhCpO. The current is as-
sumed to be constant over depth, consistent with irrotation-
ality and a slowly varying depth h(x,y). It is further Vh*(@O+ oO)u}=o, NW

assumed that U=vh&=O(l); the solution to O(E) for a as required for a steady current. L; is altered from (A3) by
single wave component from (10) is then known to be the advective term. Variation with respect to q1 yields the
dynamic boundary condition
4= 4o+f& h>Zj&z(kl ,%I AYJ) L48)
with k%+{ %+“I( ~)~~vh#d2+i(r’l,~2/ Iv0

‘39
=O(e3j. 6420)

1912 Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
Varying the integral (A17) by c,& and eliminating the kine- linearized form of the wave-current equation of Turpin
matic boundary condition yields et al.% after neglecting time dependence.-In this derivation,
we have assumed that waves propagate primarily in the
,x-direction, and we have further restricted the large current
ff[f t x
~“‘bV’O,W~+
-ho
assumption by specifying that

-
0771 (A24)
~+Pw-hl

as in KirbyZ3 and Booij.= Consistent with the first assump-

tion, it would be appropriate to calculate (T, according to
dxdt=O, (A21)
IJ
‘lo
vn=ian-knU 6425)

where the last term is O(E~,CY)-O(F) due to (A19). The since propagation has been assumed to be nearly colinear
operator with the x-direction.

D d
-G ,+u.v,.
Dt

Eliminating v1 and substituting -the expansion of (pr from ‘D.H. Peregrine, ‘Long waves on a beach,” J. Fluid Mech. 27,815 (1967).
(A12) gives the desired form of the wave equation for the ‘O.B. Rygg, “Nonlinear refraction-diffraction of surface waves in interme-
n’th component at w, : diate and shallow water,” Coast. Eng. 12, 191 (1988).
3M.H. Freilich and R.T. Guza, “Nonlinear effects on shoaling surface grav-
2- ity waves,” Proc. R. Sot. London Ser. A 311, 1 (1984).
04 “P. L.-E Liu, S.B. Yoon, and J.T. Kirby, “Nonlinear refraction-diffraction of
g+wh.u)* -vhwc,>nMt~
waves in shallow water,” J. Fluid Mech. 153, 185 (1985).
‘J.T. Kirby, “Modelling shoaling directional wave spectra in shallow wa-
ter,” Proceedings of the 22nd International Conjketzce on Coastal Engi-
neering, Delft, the Nemerlands, 1990 (American Society of Civil Engi-
neers, New York, 1990). pp. 109-121.
“PJ. Bryant, “Periodic waves in shallow water,” J. Fluid Mech. 59, 625
(1973).
7P.J.Bryant, “Stability of periodic waves in shallow water,” J. Fluid Mech.
66, 81 (1974).
“J.B. Keller, “Resonantly interacting water waves,” J. Fluid Mech. 191,
529 (1988).
“Y. Agnon, A. Sheremet, J. Gonsalves, and M. Stiassnie, “Nonlinear evo-
lution of a unidirectional shoaling wave field,” Coast. Eng. 20, 29 (1993).
‘OR. Smith and T. Sprinks, “Scattering of surface waves by a conical is-
+v,-( 2 V,dm) ..,.( % V/j+)]]; WW land,” J. Fluid Mech. 72, 373 (1975).
“O.M. Phillips, The Dynamics of the Upper Ocean, 2nd edition (Cambridge
University Press, Cambridge, 1980).
This extends (20) to include an ambient current U where “J ..:C W Berkhoff, “Computation of combined refraction-diffraction,” Pro-
ceedings bf the 13th International Conference on Coastal Engineering,
kU/w=O(l). The linearized form of (A22) is the wave- Vancouver, British Columbia, Canada (American Society of Civil Engi-
current equation of Kirby.“3 Proceeding directly to the neers, New York, 1972). pp. 471-490.
coupled parabolic equations following the method of section 13A.C. Radder, “On the parabolic equation method for water-wave propa-
II, we obtain gation,” J. Fluid Mech. 95, 159 (1979).
14C. J. Lozano and P L.-F. Liu, “Refraction-diffraction model for linear
surface water waves,” J. Fluid Mech. 101, 705 (1980).
2i&,,+ U)AnX--2r,JC,,+ U)&-k,)A, “D.K.P. Yue and C.C. Mei,.“Forward diffraction of Stokes waves by a thin
wedge,” J. Fluid Mech. 99, 33 (1980).
16R.W. Whalin, “The limit of application of linear wave refraction theory in
a convergence zone,” Research Report No. H-71-3, U.S. Army Engineer

+i-;i(%q+(g-)jV%[
(C%$),) Waterways Experiment Station, Vicksburg, MS? 1971.
r7P.A.-Madsen and 0. R. SQrensen, “A new form of the Boussinesq equa-
tions with improved linear dispersion characteristics. Part 2. A slowly-
Y varying bathymetry,” Coast. Eng. 18, 183 (1992).
‘*O. Nwogu, “Nonlinear transformation of multi-directional waves in water
1 of variable depth,” unpublished manuscript (1994).
=- R’AIAn-lei(kIo+kn-l,-k,o)x 19H. Mase and J.T. Kirby, “Hybrid frequency-domain KdV equation for
4 random wave transformation,” Proceedings of the 23rd International Con-
ference on Coastal Engineering, Venice, Italy (American Society of Civil
1 “/ Engineers, New York, 1992), pp. 474-487.
+-Es rAIAn+lei(k~+~o-k~O-k,o)X, (A23) sOG:Wei and J.T. Kirby, “A time-dependent numerical code for extended
21
Boussinesq equations,” J. Waterways, Port, Coast. Ocean. Eng. (ASCE),
in press (1994).
where R’=R(a,,~~-l,crl,k,_l,kl) and so on. The linearized, **E.B. Thornton and R.T. Guza, “Transformation of wave height distribu-
one-dimensional (in x) form of (A23) is equivalent to the tion,” J. Geophys. Res. 88, 5925 (1983).

Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T Kirby 1913

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
=J.C. Luke, “A variational principle for a fluid with a free surface,” J. Fluid depth and current on the evolution of a Stokes wavepacket,” J. Fluid
Mech. 27, 395 (1967). Mech. 132, 1 (1983).
=J.T. Kirby, “A note on linear surface wave-current interaction over slowly ?%. Booij, “Gravity waves on water with non-uniform depth and current,”
varying bathymetry,” J. Geophys. Res. 89, 745 (1984). Report No. &l-l, Department of Civil Engineering, Delft University ‘of
%F.-M. Turpin, C. Benmoussa, and CC. Mei, “Effects of slowly varying Technology, Delft, the Netherlands, 1981.

1914 Phys. Fluids, Vol. 7, No. 8, August 1995 J. M. Kaihatu and J. T. Kirby

Downloaded 22 Oct 2002 to 128.175.90.36. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp