_IOURNAL

OF GEOPHYSICALRESEARCH VOL. 70, NO. 18 SEPTEMBER15, 1965

Stream Function Representation

of Nonlinear Ocean Waves

R. G. D•

Chevron Research Company

La Habra, California

Abstract. An analytical stream function expressionrepresentinga nonlinear gravity water

wave is applied both to the representation of measured wave forms and also to nonlinear
theoretical waves. The stream function form is chosenso that it is a solution to the Laplace
equation and the bottom boundary condition; the parametersin the stream function expression
are chosenby a numerical perturbation procedure that provides a best fit to the kinematic
and dynamic free surfaceboundary conditions.The boundary condition errors associatedwith
the nonlinear stream function representation of four measured wave profiles are compared
with estimates of the correspondingerrors associatedwith a linear representation.The stream
function method is judged more accuratethan a linear method if the wave height is greater
than 50% of the breaking height. The stream function method is also applied to represent
theoretical waves for which only the wave height and p,eriodare available to characterizethe
wave profile. It is demonstratedthat the method representsan improvementover other avail-
able nonlinear procedures.The method can be employedto representwave conditionswhich
includea prescribeduniform steadycurrentand a specifiedpressuredistributionon the free
surface.

INTRODUCTION aspect of this processin both the analysisand

prediction stagesis an accurate descriptionof
The coastalengineerfrequently requiresin-
all necessarykinematical characteristicsof the
formation concerningextremelynonlinearocean
nonlinear waves.
waves, including the limiting breaking wave.
The type of information required by the engi- A method in which a stream function repre-
sentation of nonlinear ocean waves can be
neer depends on the particular application;
established from either a measured wave form
however,the investigatorin chargeof providing
or specified gross wave characteristics (wave
the informationis generallyconfrontedwith a
height, wave period, water depth) is demon-
two-stageprocessthat includes(1) analysisof
strated. All kinematical characteristics are de-
measurednonlinear ocean wave data and (2)
fined by and can be calculatedfrom the stream
predictionof the effectof nonlinearoceanwaves
function representation.The governinghydro-
on a particular processor design.Theseanalysis-
prediction stagescan be illustrated by discus- dynamicequationsare the basisfor establishing
the complete stream function representation
sionof the problemof wave forceson structures.
from either the measuredprofile (denotedcase
In this case,the analysisstageis the correlation
of measured wave forces with the wave kine- A) or the specificgrosswave parameters (de-
noted caseB).
matical flow field, which is usuallynot measured
but must be determined from the more easily
GOVERNINGHYDRODYNAMIC
EQUATIONS
measuredwave profile.The resultsof the analy-
sis are hydrodynamic(drag and mass) force The problem of determining an analytic
coefficientswhich relate the force on the piling function which representswater wave motion
to the water particle kinematics.In the predic- is expressedmathematically in the form of a
tion stage, for specifieddesignwave character- boundary value problem; that is, there is a
istics (height H, period T, and water depth h), differential equation that the function must
the kinematical flow field is calculated and satisfy on the interior of the wave boundaries
combinedwith the force coefficientsto predict and there are certain conditions on the bound-
the force on the structure. A very important aries that must be satisfied.The equationsde-
4561
4562 R.G. DEAN

fining the problem will be written for two- DIRECTION OF

dimensional
wave motion,'althoughthere are WAVE PROPAGATION

no complicationsin expressingthe corresponding ,,•"•--•,, p (x,i), IMPRESSED

PRESSURE
three-dimensional forms. The depth will be
considereduniform and the motion periodic.
,,"• ON
WATER
SURFACE
Conditions of a prescribednormal stressp(x) •' • , •.,...--.-
WATER
SURFACE •__• + x
on the free surface and a uniform steady cur-
rent U will be included in the formulation. The
terms usedin describingthe systemare defined w
in Figure 1.

Differential Equation
,
The motion is considered irrotational and the Fig. 1. Wave system, definition sketch.
fluid incompressible.The wave system can be
representedby either the streamfunction½ (x,
(b) Dynamic free surface boundary condi-
z, t) or the velocitypotentialqb(x, z, t). These
tion. The dynamic boundary condition is the
functions are defined in terms of the velocity
unsteady form of the Bernoulli equation. For
components' the case in which there is a normal stress dis-
tribution p(x) (e.g. pressuredue to wind blow-
ing on the surface), the conditioncan be written
in termsof the velocitypotentialas
(1)

p(x).
Pg• •gg
1 [(u'-FU)• '-Fw•]
1
Each of these functions must satisfy Laplace's where p is the density=o
of waterz=v
and g is the
equation throughout the interior of the fluid;
toe.,
gravitational constant.
=o Summaryo• Boundary Value Problem
V• = 0
The problem is to find either a stream func-
Boundary Conditions tion or a velocity potential which satisfiesLap-
Bottom, z -- --h. If the bottom is imperme- lace'sequationand (3), (4), and (5).
able, the vertical velocity componentmust be The exact solution of this problem for the
zero at this surface: most generalconditionsis at presentimpossible;
however, for conditionsallowing certain ap-
• = --qb, = w = O, = proximations, it will be shown that for both
casesA and B a stream function representation
Free surface,z -- 7. Becauseof the uncon-
can be determinedwhich is considerablymore
strained nature of the free surface,two bound- accurate than other available methods.
ary conditionsmust be specified'
CASE A' ST•A• FUNCTION
(a) Kinematic free surface boundary condi- BASED ON MEASURED WAVE PROFILE
tion. The kinematic boundary condition re-
quires that the motion of the water surface be Suppose the water surface variation with
in accord with the velocity componentsof the time (wave profile) measured at a point is
water particles at the free surface. This con- known and that from the wave profile one
dition is wishesto infer other quantitative information
about the wave system.For example,the time
variation of the horizontal componentsof water
o--/
ov(u+ u)Ov= w z =. (4)
particle velocity and accelerationat a certain
 NONLINEAR OCEAN WAVES 4563

level could be the required information. SeverM 'wave' which is regarded as nonlinear and
differentapproacheshave been applied to effect composedof one fundamental componentand
this transformationfrom wave profile measure- the higher harmonicsof the fundamental. The
ments. characteristicsof the wave, as representedby
Early approachestreated one 'wave' at a a stream function expression,do not require
time by selectingthe period and height param- estimatesof wave height and period but are de-
eters which seemedto best fit the wave; linear terminedfrom the detailedmeasuredprofile.
wave theory was then used to calculate the
required information. For large waves,the same Stream Function Method: Problem Formulation
and Solution
procedurehas been usedexceptthat a nonlinear
wave theory (up to the fifth order) was sub- The nonlinear stream function method is
stituted for the linear wave theory. A dis- strictly applicableto a wave systemcomposed
advantageof this method is tha.t the analytical of a single fundamentaland the higher har-
representation based on the estimated wave monics.In effect, this impliesthat the wave
height and period corresponds to a symmetrical form travelswithoutchangeof shape;sucha
profile which is not generallyin uniformly good wave systemis renderedsteadyby selectinga
agreement with the asymmetrical measured coordinate systemtravelingwiththe wavespeed,
profile. It followsthat the calculatedvelocities C. The verticalvelocitycomponent is unaffected
and accelerationsare not necessarilyvalid. by this choiceof coordinate system;however,
An obvious extension of the linear method is the horizontalvelocitycomponentin the new
to Fourier analyze an entire profile record to system is u + U- C. The two free-surface
obtain the amplitude and phase spectra, then boundary
conditions
aresimplified
forthesteady
to modify each componentby the appropriate systemasshownin (6) and (7):
amplitude and phase factor to determine the
amplitudeand phasespectrafor the water parti- o•/o• = •/(u + u- c) (•)
cle velocities and accelerations. From these
p(x)
•g + •1 [(u+ •- c?+ w•]
spectrathe water particle motionsas a function
of time can be constructed. Reid [1957] and + • = const= Q (7)
Blumberg [1955] have developed a method
wherebythe equivalentresultscan be calculated Consider
the followMgexpression
for p(x, z)
directly without explicit calculationof spectra.
The approachinvolvesthe use of time domain
numerical 'filters' which are convolved with the
wave profile measurements.Each convolution
resultsin the value at one time of the quantity
+ •
n=4,6,8, ßßß sinh(n--
2)•(h+ z)
(velocity, etc.) for which the filter was de-
veloped. The filter method is based on linear ß X(n)•o•(n-- •) Z •
wave theory and is valid for a linear wave
systemwith a spectrumof arbitrary width.
An ideal solution to the case A problem
+X(n
+X)
sin(n--
2)w
Z ß• (s)
would be an extension of the filter method to The predictedwatersurfaceis definedimpliciW
account for the nonlinear features of the wave by settingz = W in (8).
system.It is doubtful, however,whether a feasi- •(x, .)
ble method can be devisedwhich will distinguish
between a given frequency componentbeing a
. = [(5/•)- v]- [(5/•)- v]
fundamental or, e.g., •he first harmonic of a
fundamental with half the frequency.The filter ß •
n=4,6,8, ßßß •inh(n-•)•(•+.)
characteristics would differ, depending on
whether the componentwere a fundamental or
a harmonic. ß X(n)•o•(n-- •)• •
The method to be described and demon-
strated in this paper is applied to a single +X(n+•)sin(n--•)
•Z ß• (0)
4564 R. G. DEAN

where•(x, •) is the value of the streamfunc- The problem is to selectthe unknown param-
tion on the free surfaceand is a constant(----•, eters in the stream function expressionso that
because the free surface is a streamline. For E• is minimized for some selected value of X.
notational convenience,the parameters L, T, In principle, the numerical perturbation pro-
and •,are defined,respectively,as X (1), X (2), cedureimproveson an availableapproximateso-
and X(3). lution through small changesin the unknown
The Laplace equation and the bottom and parametersof the solution.The initial approx-
kinematic free surfaceboundary condition (on imate solution was determined by estimating
V•) are satisfiedexactly by (8) for arbitrary the wave period from the recorda•d calculating
valuesof L, T, •, and the X(n) coefiqcients,the wavelength from small-amplitude wave
which at this stage are all regarded as un- theory. The initial X(n) for n > 2 were de-
knowns. The problem is determining these un- termined by minimizing E• on the measured
knownsso that the predictedV sequencefrom water surface by a direct least-squarespro-
(9) is in best agreementwith the measuredV cedure.With an approximatesolutionavailable,
sequenceand also so that the dynamic free at the jth iteration,small changesX' (n) in each
surfaceboundarycondition,equation7, is best of the parametersX(n) are chosenso that the
satisfied.This is accomplishedby a numerical resulting total error will be minimized by the
perturbation procedurewhich is best done on parameter changes.The resulting total error
a digital computer having reasonably high at the (j + 1)th iterationwill be
speeda•d large memory (e.g., the IBM 7094).
Numerical Perturbation Procedure i

The parametersin the streamfunctionexpres-

sion are determinedso as to provide a best fit OO
(i) )
to the measuredwave profile Vmand to a uniform
value Q of the Bernoulli constant. Consider a
total errorET composed
of two components--E•
which is that component due to the nonuni-
formity of the actual Bernoulli 'constant' and
E: which representsthe error componentdue to
deviationsbetweenmeasuredandpredictedwater
surfaces. These errors are defined as + ox(•) (•) - .m, (1•)
which involves the N unknownsX'(n). Exactly
ET = Ex + XE, (10)
N simultaneousequationsresult from the usual
where
least-squares
procedure'
•(•, •, •,, x(4), x(•), ..., x(•))
OX,-••-- 0
= II k (Q,- •)2 (11)
__

and

•(•, •, •, x(•), x(5), ..., X(N))

= I • (•m,
-- •,)"(1•) --0
OX'(•V)
The term X in (10) plays the role of a Lagrange
multiplier and providesa relative weight to the The partial derivativesindicatedin (14) mus•
two error components.The subscripti is a time be expandedby the chain rule; for example,
index for a sequenceconsistingof I values,and for oQ/oX(n), Q, is definedas
Q is definedby Q• =

- • • Q, (13)+ •1[(u,+ u- _LT)'

•1+ •-' (•)
+ w,
 NONLINEAR OCEAN WAVES 4565

But TABLE 1. ApproximateCharacteristicsof

Measured Wave Forms

u, =
Wave Wave Water
Wave Height Period Depth
No. H, ft T, sec h, ft
Therefore
I 7 7.2 33
OQi OQi OL OQ• OT 2 15 9.0 33

OX(n)-OLOX(n)
-{-OTOX(n) 3
4
19
39
10.8
14.0
33
98

-Jr-
OQ• Ou•_}_
Ou, OX(n)
OQ• Ow• (18)
Ow, OX(n)
the boundaryconditionerrors are based; the
In (18), each of the first two terms is nonzero quantity X was thereforeset equal to 1.0 in
only if the X(n) representsthe particular the calculations.
variable of differentiation. It is of interestto comparethe errorsin the
Equations representing the counterparts of fits of the free surfaceboundaryconditionsob-
(18) are expressedin a similar manner for tained by this methodwith an estimateof the
OQ/OX(n)andOr/•/OX(n).The solution of (15) corresponding errors which would result from
for a particular measured profile rcsuks in linear wave theory representation.For the
quantitiesX'(n) which when addedto the X(n) smaller wa.ves,the nonlinearterms neglected
define the stream function from which any in the derivationof small-amplitudewave theory
feature of the wave system can be determined. will be of small relative importance,whereas
Changesin the parametersare usually small the assumptionof a single fundamentalcom-
after about three or four cyclesof the perturba- ponent will be of larger relative importance.
tion procedure. For large waves, the neglectednonlinearities
will be of larger relative importanceand the
Examples o/ Stream Function Representations assumptionof a singlefundamentalcomponent
o/Measured Waves will be of lesserrelative importance.It is there-
To illustrate the stream function method, fore expectedthat the lineartheorywill indicate
four measured storm and hurricane wave pro- a better boundary conditionfit for the smaller
files were selectedand the parametersin their waves and that the nonlinear method will pro-
fifth-order (N -- 13) stream function repre- vide a better fit for the larger waves.
sentations were calculated. The approximate
characteristicsof the measured wave profiles Definitiono/Free Sur/aceBoundaryCondition
Errors
are presentedin Table 1.
These data were measured in the Gulf of The free surface boundary condition errors
Mexico by Chevron Research Company. The for the linear and nonlinear methods are so
profile data were obtained by using a fixed defined that they provide a measure of the
step-type wave staff consistingof electrodes relative 'goodnessof fit' obtainedby use of the
spacedat 6-inch intervals along the staff; the two methods and therefore a measure of their
electrodesare effectively shorted out by the respective validities. The errors defined for the
rising water surface,thereby providingan indi- linear wave theory comprise those terms as-
cation of the instantaneous water surface to sumed negligible in derivation of the linear
within 0.5 ft. This type of wave staff has been theory; these errors therefore representa lower
describedin detail by Russell[1963]. bound on the linear theory errors becauseany
Becauseno data were available defining the calculationprocedurecan, at best, only approx-
mean current component or the pressure dis- imate the strict applicationof the linear theory
tribution on the wave surface, these quantities to the measuredprofile. All errors are calculated
in (7), (8), and (9) wereassumed to be zero. on the measuredfree surfaceW-
The references and dimensional units are the The error definitions for the linear and non-
samefor the two quantities(Q and V) on which linear methods are summarized in Table 2. The
4566 R. G. DEAN

TABLE 2. Errors in Free SurfaceBoundary Conditions

Dynamic Boundary Condition Kinematic Boundary Condition

Nonlinear Linear Nonlinear Linear

Local
error &.• ----Qi - Q •,i* --- 1/2g[u? •2,i • •]mi-- •]Pi •2,i* =
fo
xiuw/C• dx
Gross
error E• = 1/I Y: $•,i2 ----($•,i•) El* --• (($•,i*) "

8 terms are a measure of the various errors at profiles representwave conditionsfor which
particular local positions (denoted by the the energy spectraare believedto be narrow
subscript i) throughout the wave. The E in comparisonwith conditionswhich, for ex-
quantifiesare simply the mean-squarevaluesof ample, would prevail within a wave-generation
the local errors.With the exceptionof the linear area. The results, therefore, should not be
error in the kinematic boundary condition considereduniformly valid; testing of profiles
(8•,,* and E?), the bases for the errors are obtained under different conditions would be of
evident. valuein further assessingthis method.
In the derivation of the small-amplitudewave Referring to Table 3, we seethat the errors
theory, u is assumednegligible in comparison in the boundaryconditions indicate,as expected,
with C; i.e., in the kinematic boundary condi- that the linear wave theory providesa better fit
tion to the measured data for the smaller waves and

Ox-- --C(1-- u/C) (19)

2O1•
CERRORS
IN
KINEMATIC
I [ I'*'-*-
NONLINE
This equationcanbe approximatedby

OV w wu
•.•E
•omFREE
SURFACE
BOUNDARY
CONDITIO
I LJe-•
•L
NEAR
S2'
(t)
Ox C C" (20) :z=I I0 -

DYNAMICI
I0F BERRORS
IN I"-'-"
NONLINEAR,
81(t)
__•l
'"LINEA
81'
I
FREESURFACE
BOUNDARY
CONDITI•
(t)
Only the first term in (20) is includedin the o • 5o1---•
I

derivation of the linear theory. To obtain an

error of the same form as that defined for the
nonlinearmethod,the term uw/C' is integrated
as shown in Table 2 to result in the error 3•.••
in •/due to neglectingthis term in the formula- 80
tion.
The wave profiles and the associatednon- 60
•AMEASURED
AND/] Ii"--"•mCTEO
ø NONLINEA
PROFILE
?7
I
linear and linear errors for waves 1, 2, 3, and 4
are presentedin Figures 2, 3, 4, and 5. The
squareroots of the grosserrors associatedwith õ20
each of the waves are summarized in Table 3.

Discussiono,• Stream Function Representation -20

o• Measured Waves
-4O
In evaluating the significanceof the errors I 2 3 4 5 6 7 8
TI•E, (SEC.)
associatedwith the linear and nonlinear repre-
sentations,it should be noted that the wave Fig. 2. Stream function representationof wave 1.
 NONLINEAR OCEAN WAVES 4567

2.01I•cERRORS
INKIIIEMATIC
FREEI I I basis of consideration
I•NONLINEAR,))2(t)] of the relative errors in
m• SURFACE
BOUNDARY
CONDITION
I I • ,A,
LINEAR both the dynamic and kinematic free surface
boundary conditions, the nonlinear stream
function method appears more accurate for
waves with a nonlinearity index exceeding0.5.
It should be noted that a different value of k
.01 I /
I BERRORS FREE I• I I I•NONLINEAR, would changethe relative emphasison kinematic
INDYNAMIC
20m.-•.
SURFACE
BOUNDARY
CONDITION
--5"- I• •LINEAR and dynamic boundary condition errors in •he
calculations.For example,it shouldbe possible
z
.•z
to adjust .• for a given wave so that the gross
kinematic and dynamic boundary condition
errors are the same. It is doubtful, however,
.PREDICTED
PROFILES
i I • MEASURED
PROFILE,?7
m whether any such adjustment of .• would ap-
I0 ß
preciably affect the conclusionthat the non-
linear method is better for a nonlinearity index
exceeding0.5.
CASE B: STREAM FUNCTION
REPRESENTATION OF THEORETICAL WAVES

When the detailed description of the water

surface is not available, for example, when
designing a structure, an estimate of various
wave characteristics•na.y be necessaryto de-
termine wave forceson pilings,wave cresteleva-
0 • 2 3 4 5 6 ? 8 9 tions,etc. In suchcases,it is customaryto select
TIME, (SEC)

Fig. 3. Stream œunctionrepresentation oœwaYe 2.

F CERRORS
INKINEMATIC I i 11
• NONLINEAR,
20 FREE
SURFACE
•,lBOUNDARY
CONDITION
I II• •LINEAR
, •2 t)
the nonlinear strealm function results in smaller 0 L -- ....
errors for the larger waves.As previouslymen-
tioned, the error definitionsprobably tend to •B ERRORS
INIDYNAMIC NONLINEAR,
•q (t)

favor the linear wave theory to somedegree. • FREE

SUR-
• FACE
BOUNDARY
CONDITION
I1•
•LINEAR
•
It is of interest to attempt to determinethe 40
m ! •-•l m' , i (t)
wave conditions above which the stream func-
tion method results in smaller errors and below
14.0 --A: MEASURE[
which the linear method results in smaller AND
I PREDICTED
NONLINEAR
PROFILE,
?•p
PREDICTE(
errors. An index of the nonlinearity of the 120 • PROFILES I I .... MEASURED
PROFILE,
?•m
various waveswas formed by dividing the wave ioo

heightsin Table i by the breakingwave heights ._. 80

correspondingto the individual wave periods • 60
and water depths.These indicesare plotted in
.-J 4o
Figure 6 versusthe ratios of nonlinearto linear o
a:: 2o
errors shown in Table 3.
ß .

Figure 6 indicatesthat the dynamicboundary <> o

condition is best satisfied by the nonlinear •20

streamfunctionrepresentationfor wave heights -40

whichare greaterthan about 30% of the break-

ing wave height.The error in kinematicbound-
ary conditionis best satisfiedby the stream 4

TIME,
5 6

(SEC.)
-60

-:80
• 8 9 10
I
function method for wave heightsgreater than
about 70% of the breakingwave height.On the Fig. 4. Stream function representation of wave 3.
4568 R.G. DEAN

• Ioom•
cb m--: /
c:
ERRORS
IN
KINE#ATIC
I I I
FREE
SURFACE
BOUNDARY
CONDITION
I i ,
I I"•"NøNLINEAR,
, i • • ........

•z• 0
•z•
-50

• •
•z
• IO0•B:
ERR•S
IN
DYNANIC
'
m FREE•RFACE
BOUN•RY
•NDITION• ) m•NONLmNEAR
m
• • LINEAR
zO O,

Dz•-50 300 --A' #EASUREDAND

250
PREDICTED
PROFILES
!"•"-•
ml: :PREDICTED
•LINEAR
•7•FILE,•7p(t)
)4EASURED
PROFILE,
m (t)

2O0

•_.•5o
• ioo
,,
o

m50

-150
o i 2 5 4 5 G

TIME, (SEC.)

Fig. 5. Stream function representationof wave 4.

a wave whoseprofile is characterizedby some travel without change of shape in water of

grossdimension,usuallythe vertical distance uniform depth; a moving coordinatesystemis
from the trough to the crest (wave height). chosenwhich reducesthe problemto one of
The wave profile and other characteristics are steady motion. The differentialequationand
then determinedso that goodfits are provided boundaryconditions are the sameas described
to the specifiedboundaryconditions. This wave for caseA, equations2, 3, 6, and 7. In general,
is usuallyregardedas being reasonablyrepre- the wave height (or equivalentlythe Bernoulli
sentative of all 'natural' wave profiles of the constant),wave period,water depth, uniform
chosenwave height and period. Waves which currentmagnitude,and pressuredistributionon
are specifiedby a singlegrossdimension of the the free surfaceare regardedas specifiedquan-
wave profilewill be denoted'theoretical'waves tities. The stream function solution for case B
to indicate that the detailed profile is deter-
mined from theory rather than from measure-
ments.
TABLE 3. GrossBoundary Condition Errors
Associated with Stream Function and
In this sectionthe problem of stream func- Linear Wave Theories
tion representationof nonlinear theoretical
waves is described and the method is demon- Wave
strated for the caseof near-breakingwaves. A No. (E•)
discussion is also presentedof stream function
representation
of nonlinearwavesfor whicha I 0.345 0.250 0.407 0.096
2 0.317 0.866 0.578 0.373
pressure
distribution
on the freesurfaceand/or 3 0.473 1.817 0.594 0.682
a uniform current are specified(see Figure 1). 4 1.348 3.592 1.501 1.225
As in caseA, the wave form is considered to
 NONLINEAR OCEAN WAVES 4569

is essentiallythe same as for caseA; however, to determine two sets of coefficients,one for
becausethe period is not an unknownin the the flow field and one for the wave profile.
presentcase,the solutionwill be written in the Also, the velocity potential approach doesnot
followingslightlydifferentform' implicity satisfy the kinematic free surface
boundarycondition.The goodness of fit to this
boundary condition was one of the considera-
tions entering into Chappelear'schoiceof co-
N-1 efficients.
•rn
The problemis to find L, •, and the X's so
-+- •'. sinh
-• (hJFz)
n•2,4,6, ßßß
that the dynamic free surfaceboundary condi-
tion is best satisfied and also so that the amounts

I
ß X(n)cos x of water above and below the mean water level
are equal.The procedure,as before,is to assume
that an approximate solution is available and
+ X(n+ 1)sin x 1 (21) that small changes(denoted by primes) are to
be made so that an improved solution results.
and
The problem is more direct if Q is specified
X(1) ----L rather than the wave height. The error, E•
(L, •, X(1), X(2), ß ß ß , X(N q- 1) ), in the
dynamicfree surfaceboundaryconditionis
The stream function solution satisfies the
differentialequationand the bottom and free
surface kinematic boundary conditions for
arbitrary valuesof L, •., and the X's. Two, of
the advantagesof usinga streamfunctionversus
a velocity potential representationare evident
(23)
.• OOQ,
-Jr- X(n) Xt(n))-
F•]
2
defines The first two terms inside the error expression
at this point.First, oneset of coefficients
all characteristicsof the wave system. Second, representthe value of the local Bernoulli term,
the stream function expressionexactly satisfies includingthe effectof the changesX'(n). The
the kinematic free surfaceboundary condition. X'(n) are determined accordh•gto the least-
Chappelear [1961] investigateda similar ve- squares criteria, that is, so, that E• will be a
locity potential method of representingnon- minimum.
linear theoretical waves and found it necessary This operation results in N simultaneous
equations with N unknowns.The revised X's
after the (j + 1)th iteration are typically
I
X (i+'}(n)= X (i}(n)-+-Xt(n) (24)
q•./ WAVE
NO,
The wave profileat the (j + 1)th iteration, but
basedon the value of ½,(J•,is found by setting
z -- •/in (21), i.e.,
, \
\ KINEMATIC
BOUNDARY
X\..,•..../-
\
CONDITION 3](i+1)
\

2 ..
•'•A

DYNAMIC
BOUNDARY
'**'•""•:•,•4
CONDITION
••
o
o .2 .4 .6 .8 I.O

NONLINEARITY
INDEX, H/H b

Fig. 6. Free surface boundary condition errors (25)

versusnonlineartry index. -Jr-
X(i+z)(n-Jr-
1)sin-•-x
4570 R.G. DEAN

After the •/(J*•) sequence

is generated,½u.• is oA
F c ERRORS
INKINEMATIC I I I
determinedsothat the •/u.•) sequenceresultsin
no net change in the mean water level. The j ! FREE
SURFACE
•-• 0'21BOUNDARY
[ CONDITION
•- I,,':"'•,
I I [IFUNCTION
STREAM [
ERRORS't--
I IDENTICALLY
ZEROALL
PHASE
ANGLES
expression
for ½•u+•)is therefore
2l _o.1 I ............
. ....,
- L fon=2,4,6,
Y',... sinhn7r
-• (hq- •.
•
40[• BERRORS
!
INDYNAMIC
FREE
SURFACE
BOUNDARY
I I
CONDITION
I I
I
I
I
I
I
I

ßIX(n) n•r
cos-•
x•- X(n•- 1)sin
nw ax v .... ß
........
i.... ....... -"-'
.............
(26) 1 I
where the superscriptson the right-hand side
have been droppedfor convenience. This inte-
zo.. 90'0
FA•AVE
P • 80.OL
PROFILE
CHAPPELEAR'S
SEVENTH
ORDER
SOLUTION
u.J'-' SEVENTH
ORDER
SOLUTION
grationis •pproxim•ted by Simpson'srule.
At this stage, one cycle of improving the
stream function solution has been completed.
• M ANWAT LEVEL
The solution is not the best lea•-squares solu-
tion, partly because• was determinedat a
different stage from that •t which the other
unknownswere detersned. In general,however,
"'Q
•.c
0'00 20 40 •0 8Q IOOmE 140 160
PHASE ANOLE (DEGREES)
additional cycleswill result in successivelyim-
Fig. 7. Comparison of seventh-orderwave rep-
proved solutionsuntil some'near-best'solution resentations by two different methods.
is obtained. For near-breaking w•ves, it was
found necessaryto d•mp the specifiedch•ges
(X'(n)) by some f•ctor between zero and 1 solutionare comparedwith thoseof the velocity
so that the solution would not become unstable. potential solutionas determinedby Chappelear.
This instability is not unco•on in perturba- Note that the errors in the kinematic boundary
tion solution •pplications, particularly if the conditionsare identically zero for all phase
representedphysical process can become un- angles for the stream function method. The
stablefor certainlimiting conditions. errors in the dynamic boundary condition for
the stream function method are significantly
lessthan for the velocitypotentialmethod.
Examples o.f Stream Function Representation The largest stream function errors occur near
of Theoretical Waves the crest, where greatest variations in the wave
profile and water particle velocity occur. This
Example 1. Zero uniform current, uniform
is a consequence of a direct application of the
surface pressure,nonlinear seventh-o.rderwave.
least-squares procedureto evenlyspacedpoints.
It can be shownthat if the pressurespecified The errors could be distributedmore evenly
on the free surface is symmetric about the z throughoutthe wave either by employinga
axis, then the boundary conditions,equations6 weightedleast-squares procedure,in whichthose
and 7, are such that they will be satisfiedby a pointsnear the crestwouldbe assigned a greaier
symmetric stream function solution. For this relative value, or equivalently by selecting
situation all the coefficients of the sine terms points with a relatively finer spacingnear the
are zero. crest.

Chappelear[1961] appliedhis velocitypoten- Example 2. Zero uniform current, uniform

tial method to severalwave conditions,the most surface pressure,near-breaking eleventh-order
nonlinear of which was H = 34.4 ft, T = 15.3 wave. To illustrate the applicability of the
sec,and h = 48.8 ft. To comparethe velocitypo- method to near-breakingwaves, an eleventh-
tential and streamfunctionmethods,thesesame order stream function solution was calculated
wave conditionswereselectedfor example1. for the followingwaveconditions:
H = 38.8ft,
In Figure 7, the resultsof the stream function h -- 48.6 ft, and T -- 18 sec.
 NONLINEAR OCEAN WAVES 4571

34.648 (BERNOULLICONSTAN : WAVE PROFILE

.MEAN WATER LINE

--iI-•6 -14-2
t •,2
/4 6 8 IO
eee
ee
eee
eeee
ee
eeee
eeeeeee
eeJ J eeeeeeeee
ee
eeeee
ßeeeeeeeeeeee
-I0

I WAVE
CHARACTERISTICS,
h : 48.6 FT. -20 NUMERICAL
SOLUTION

-30
I H: 38.80 FT.
T: 18.0 SEC.

-40
I L:844.3õ
FT.
I /..f OCEAN
BOTTOM
Fig. 8. Eleventh-order stream function representation of near-breaking wave.

The wave profile and variation in calculated ary conditionswith varying degreesof accuracy.
dynamic boundary condition (Bernoulli 'con- Chappelearfound that for h/T • = 0.199, the
stant') are shown in Figure 8. The difference range in the Bernoulli 'constant' was 26% of
between the wave erest and the specifiedBer- the wave height; the correspondingrangesfor
noulli constant is 0.78 ft. The Stokes' breaking larger values of h/T • were much less. Corre-
criterion,u = C at the erest,wouldimply that spondingboundaryconditionerrorsfor Lairone's
r/: Q at the crest.It is interestingthat there investigationare not available.
are no secondarycrestsin the strea•mfunction
Uniform Current and PrescribedFree
wave profileas foundin other investigations of
SurfacePressure
the form of steep waves.The secondarycrests
are probably a result of attempting to represent The stream function can readily be applied
nonlinear waves by analytic forms which are when a uniform steady current and/or a pres-
inadequatelynonlinear. suredistributionon the free surfaceis prescribed
The ratios h/T • and H/h for example 2 are as part of the specifiedwave conditions.An
0.15 and 0.798, respectively.The ratio H/h is asymmetrical pressure distribution about the
above the usually accepted shallow water origin would require that sineterms be included
breakingratio of 0.78. Chappelear[1959], in a in the stream function expression.
theoretical investigationof breaking wave con- It is problematicalwhethera prescribedpres-
ditions,founda ratio H/h -- 0.87 to apply for sure distribution would be the same as the
h/T • -- 0.199. Lairone [1963], however, in a expected pressure distribution for the wave
theoreticalinvestigationof solitary wave theory, profile associatedwith the solution.If not, the
found a breakingratio H/h -- 0.727.The actual implication is that the wave and pressuredis-
ratio is still an open question;the differentre- tribution could not coexist under equilibriu'm
ported ratios are undoubtedlya result of the conditions; that is, the pressure distribution
various methodssatisfying the required bound- would causechangesin wave height and period.
4572 R.G. DEAN

REFERENCES cnoidal, solitary and Stokes' waves, Beach Ero-

sion Board, Tech. Mere. 133, February 1963.
Blumberg, R., Analysis of ocean waves and wave
Reid, R. O., Correlation of water level variations
forc.esby a filtering technique, Ph.D. disserta- with wave for•es on a vertical pile for non-
tion, Department of Oceanography,A and M periodic waves, Proc., Sixth Con/. on Coastal
College of Texas, May 1955. Eng., chapter 46, pp. 749-786, 1957.
Chappelear, J. E., On the theory of the highest Russell, T. L., A step-type recordingwave gage,
waves, Beach Erosion Board, Tech. Mere. 116,
Proc., Con/. on Ocean Wave Spectra, pp. 251-
July 1959.
258, Prentice-Hall, EnglewoodCliffs,N.J., 1963.
Chappelear, J. E., Direct numerical calculation of
wave properties, J. Geophys. Res., 66(2), 501-
508, 1961.
Laitone, E. V., Higher approximationsto non- (Manuscript received November 28, 1964;
linear water waves and the limiting heights of revised May 18, 1965.)