JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 88, NO. C10, PAGES 5925-5938, JULY 20, 1983

Transformationof Wave Height Distribution

EDWARD B. THORNTON

Naval Postgraduate School, Monterey,California 93940

R. T. Guz^

Shore Processes Laboratory, Scripps Institution of Oceanography, University of California

La Jolla, California92093

Thetransformation of random wave heights during shoaling, including waves breaking in thesurf zone, was measured withanextensive arrayofinstruments in thefield. TheinitiallyRayleigh height distributions in 10-mdepthwereobserved to be modified by shoaling and breaking into newdistributions whichare again nearly Rayleigh but withsome energy loss. Usinglocally measured H .... theRayleigh distribution

describes themeasured central moments of H/3 andH/owithaverage errors of -0.2% and -1.8%, respectively. The Rayleigh distribution is usedto describe the randomnatureof waveheights in a single-parameter transformation model based on energy fluxbalance. The energy losses associated with wave breaking areparameterized using observed breaking wave distributions coupled witha periodic bore dissipation model. Using incident waves measured in 10-mdepth asinputconditions, themodel predicts
Hrm sat shoreward locations withinarms errorof _+ 9%. Thesingle freeparameter of themodel, a constant

B representing thefraction offoam ontheface ofa wave, was chosen to best fit thedata. Theresulting large value ofB implies thatthesimple periodic boredissipation function substantially underestimates theactual
dissipation.

is calculated by usinga periodicbore formulation.The energy as a As wavesapproachthe breakpoint,wave-induced velocities flux equationis integratedto yield local valuesof Hrms function of offshore wave conditions. Both analytical and nuincrease and dissipation due to the bottom friction and/or In the last sectionthe modelsare percolation becomesincreasinglyimportant. But once the mericalmodelsare developed. in the waves startto break,turbulent dissipation of thewaveenergy is comparedwith resultsfrom random wave experiments laboratory and from an extensive set of field measurements. the dominantdissipative mechanism, and breakingprocesses
dominate the wave transformation. In contrast to monochro-

1.

INTRODUCTION

matic waves,thereis no well-defined breakpointfor random

waves.The largest wavestend to break farthest offshoreand smallwaves closer to shore. At eachspatialpoint thereare both

2.

EXISTING RANDOM WAVE TRANSFORMATION MODELS

broken and unbroken waves (sometimes having the same height),and the percentage of brokenwaves variesasa function of position. Most nearshore dynamical models forlongshore currents, rip currents,and flow over irregularbottom describe the wavesas

improvement in thesemodels wouldbe to more realistically water [Battjes and Janssen,1978] and will be further refined in this paper. includethe randomnatureof the waves. The objective of this and developed Wave heights aredescribed, in general, by a joint distribution paperis to characterize the transformation of the waveheight (or period or wave number,equivaprobabilitydensity function (pdf) from offshore to the shoreline of height and frequency all the above authors assume with a simple modelasa firststepin theevolution of dynamical lently). To simplifythe analysis, the waves are very narrow banded in frequency and coming modelshavinga probabilistic description for waves.Both analyticaland'numerical models are developed for describing the transformation of waveheights. The models arecompared with results from randomwaveexperiments in the laboratoryand
from the same direction,such that all wave heightsof the distribution are associated with an average frequency f and

monochromatic and ofconstant amplitude ateach location. An

There are two generic classes of randomwaveshoaling and breakingmodels.The earlier models[Collins,1970; Batties, 1972; Kuo and Kuo, 1974; Goda, 1975] describe shoalingas only dependenton the 'local' water depth and are described herefirst. A second type of modelis basedon integrating the energyflux balanceequationwith wave heightdependent on the shoaling processes alongthe integralpath startingin deep

meandirection 0. Therefore, starting in deepwater,the waves

are described by the single-parameter Rayleighpdf with the impliedassumptions of a narrowband Gaussian process. In the local models, the deep water wave heights are first in thefirstsection. Thenthetransformation of waves, including transformed into shallow water using shoaling theory in which dissipation dueto breaking andbottomfriction, isdescribed by all energy losses are neglected. The shoaled unbroken wave an energyflux balance model.The waveheightpdf of all waves locallyusing (broken and unbroken)is shownby the field data to be well heightdistributionis calculated described by the Rayleighdistributioneverywhere. The observed distributions of breakingand brokenwaveheights are fittedto simple analytical forms, andbreaking wavedissipation
from an extensive set of field measurements. Earlier models of random wave transformation are reviewed

p(H) - (KsHa) 2exp - KsHa

(1)

Copyright1983by the American Geophysical Union.

Paper number 3C0515. 0148-0227/83/003C-0515505.00

whereK sis a shoaling coefficient and Ha is the deepwater rms wave height. Eventually,the wavesreach such shallow water that they start to break,with the largestwavesbreakingfar5925

5926

THORNTON AND GUZA: TRANSFORMATION OF WAVE HEIGHTDISTRIBUTION

TABLE 1. Modified Probability Density Functions Du e toWave Breaking

AUTHOR DISTRIBUTION SHOALING BREAKER CRITERIA

COLLINS

,
, ,

LINEAR

Hd

(1970)
p(H)' '
(1972)

...
H b

..

(AFTER Le#EHAUT AND 0fl, 1957)

H

' x
H b H

u.E.

%= --

a,, (

p(H)

.....

K (1974) ond K /x x
Hb
(1975) Hbi
the modified distributions.

LINEAR H b = 0.63h
(UTO, 1974) (A"R A, ]95)

' Hb2 H

Thedotted lines represent theoriginal Rayleigh distributions andtheheavy lines represent

thestoffshore. Wave breaking is simulated by truncating the tail of the Rayleigh distribution based on various breaker criwaterdepthandtan/ isthebottomslope.

beseen later) in thatthedelta function at a particular breaking

waveheight[Collins,1970;Battjes,1972] is removed, but the

teriawhere, in general, H -- H(h,Ha,.I)where h isthelocal sharpcutoffat H isstillnonphysical.

Collins [1970]andBattjes [1972]used a sharp cutoff of the Rayleigh pdfwithall waves thatarebreaking or have already broken having heights equalto H. Requiring all the broken waves to havethesame heightH results in a deltafunction at H in thepdf(see Table1).Collins used linearshoaling theory and the breaking criterion afterLeMehautandKoh [1967]. Batties [1972]alsoused linear theory to shoal thewaves and applied thebreaking criterion based onMiche's [1954]formula forthemaximum height of periodic waves ofconstant form:
Goda[1975] assumed that wavebreakingoccurs with linearlyvarying probability overa range ofwave heights, resulting
in a modifieddistribution with a gradualcutoff of the distribution aroundH (Table 1). The spreading of breakers over a

range of heights crudely compensates for the simplification of using a single wave period in thebreaker height criterion andis
certainly morerealistic thanmodels whichassert that breakers are all the sameheight. Goda[1975] useda breaker criterion basedon laboratorydata which takes into accountbottom

slopeand wave.steepness in deepwaterHd/La [Goda,1970]

and calculatedthe wave shoalingusingthe monochromatic nonlineartheory of $huto [1974]. We note that the use of
monochromatic,nonlinear theories to shoal random waves

(whichhavebeencharacterized by a single frequency) is theoretically unjustifiable and introduces unnecessary numerical where 7 isanadjustable coefficient. In shallow water, (2)simply complications intoalready relatively crude models. reduces to
The commonidea of thesestudiesis to cut off the portion of

H= - tanh 0- kh

(2)

H, = 7h

(3)

the waveheight pdf beyond a breaker height, which is con-

trolledby thewaterdepth andotherfactors. Themethods differ in the techniques of cutoffand the formulaeusedto define statistics by using a sharply truncated conditional Rayleigh pdf breakerheights. A shortcoming of these earliermodels is that with the breaking waveheightsimplyproportional to local the calculated waveheights depend only on the localdepth.In waterdepth(3).Theyassumed the waves generally havesome application to cases wherethe depthis not monotonically heightsmaller than H afterbreaking and redistributed the decreasing, such asa barred coast, the predicted waveheights
Kuo andKuo [1974] modeled the effect of breaking on wave
brokenwavesacross the rangeof heights in proportionto the

probability of unbroken waves at each height. Stated in terms

decrease over the bar due to breakingand then increase in the

ofthe conditional wave height pdf,

t,(H) = t,(H/O_<H _<H,)= o(H) =0 H > H,,

deeper trough, simply following thelocal depth. Theincrease in

wave heightin the troughindicates a generation of energy whichis physically inappropriate. Modelsapplying the energy flux balance [Battjes andJanssen,1978]to calculate waveheights canproperly predict shoal-

[;o

o(H) aH

]-,
(4)

ing of waves overnonmonotonic bottomprofiles. For sim-

wherepo(H)is the pdf prior to truncation; p(H) is simplya truncated Rayleigh distribution renormalized to unity.Table 1 shows the originalRayleigh distribution with dottedlinesand
the truncated, renormalized Rayleighdistribution(4) in solid

plicity, Battjes andJanssen considered onlywaves normally

incident to a coastline with straight and parallelcontours, so
that

lines. Thedistribution byKuoandKuoismore realistic (as will

oEc,_ (e) 0x

(5)

THORNTON AND GUZA: TRANSFORMATION OF WAVE HIGHT DISTRIBUTION

$927

ments,evenfor a barredbeach.A deficiency of the Battjesand Janssen model is that the wave heightdistribution,although conceptually simple,is not a good representation of the measuredwave heightpdf because of the delta functionand truncation at Ho. The agreementbetweencalculatedand observed nrta s doesnot mean that the underlyingpdf's are similar, as Battjesand Janssen noted.
3. TRANSFORMATION OF WAVE HEIGHT
DISTRIBUTION MODEL

Fig. 1. Periodicbore usedto describe spillingbreakers.

where E is theenergy density, Cgis thegroup velocity, x is the

onshorecoordinate,and (e) is the averagedissipationper unit area.The energydensityand groupvelocityare givenby using linear theory relationships. Specification of the dissipationterm requires consideration of breakingwavedynamics. In general,wave breaking occursas a result of kinematic instability as the fluid velocity at the creastexceeds the wave speed, resultingin the crestcurlingover and injectingfluid at the surface.Dissipationrates and depth of turbulent penetration are dependent on the strengthand sizeof vorticesgenerated at the surface, which vary with the breaker type [Miller, 1976]. Turbulencecan penetrateto the bottom under plungingbreakers. For spilling breakers the turbulence is confined to a surface layer, primarily betweenthe crest-trough regionwhich at leastqualitativelyresembles the processes of a bore. For this reasonthe rate of energydissipation due to shallowwater wave breaking is usually modeledafter a bore, an approachoriginally suggested by LeMehaut6 [1962]. The detailsof the turbulencedynamicsin the bore (spillingbreaker) is avoided by applying conservationof mass and momentum at regionsof uniform flow upstreamand downstreamof the bore (Figure 1). The averagerate of energydissipationper unit area is calculated [Stoker, 1957]:

The model developedhere describes the transformation of the waveheightdistribution. The modelis similarin concept to that of BattjesandJanssen [1978], but it is extendedto describe realistically the transformation of wave heightpdf's as well as Hrm s. In addition, bottom friction is consideredin the dissipationfunction.For straightand parallelcontours the average energyflux balance,includingbore (e) and frictional dissi-

pation(er),isgiven by

OECg, - <e>+ <es> (x

where Cg,, isthex component of thegroup velocity.

(8)

The total energyflux in (8) is properlydescribed by usingan energydensityspectrumwith group velocities integratedover all frequencies and directions.Unfortunately,there is little theoreticalguidanceconcerning how to calculateenergyfluxesfor a broad-banded(in direction and frequency)nonlinear wave field with some wave breaking. Thornton and Guza [1982]
showed,for this same field data set, that the waves shoreward of

4 m depth can be frequencynon-dispersive acrossthe seaswell frequencyrange with phase speedsapproximatelygiven by

x/, which isthe same as the linear phase speed atthe spectral
peak. Thesemeasurements suggest usinga lowestorder model for the energy density and group velocity given by the linear wave theory relationships

(h2 - h,)3

(BH)3

where the wave height H is measured as the maximum to minimum of the bore, O is the volumedischarge per unit area across the bore, and B is a breaker coefficientof 0(1). The
coefficient B accounts for the differences in various breaker

E= pgHrms 2= ,o a H2p(H) dH

(9)

types and is considered as a function of the proportion of the foam regionon the faceof the breaker.The coefficient B will be the only unspecified parameter in the model and will be determined from the data.

c( Cv, = 1+sinh 2khJ cos 0

(10)

where 0 is the mean wave direction and k is the wave number

Various formulations have been suggested for the bore dischargeQ in (6). The simplest description of Q for wavesis for a linear periodicbore [Hwang andDivoky,1970]:
Ch

associated withaverage frequency f corresponding to thepeak

of the spectrum. Althoughthiscruderepresentation may not be theoretically justified,similar quasilinearapproximations have yieldedremarkablyaccuratemodelsfor set-up [Bowenet al., 1968; BattjesandJanssen, 1978]. At this point, we havethe governing energyflux relation(8),

Q-

(7)

with the dependence of EC,,on H,ms and h givenby (9) and

where C is the wave speedand L is the wave length.The bore dissipation functionis usedto describe only the breakingwaves
of the random wave distribution.

Battjes and Janssen[1978] use the dissipationfunction (6) with (7), but they reduce the dependence on the depth by assumingthat H/h = 0(1)= 1. They specify the percent of breaking waves at a particular location as simply the area under the delta functionat H of the truncatedRayleighwave height pdf. The dissipation function applied to the broken 3.1. Wave Height Distributions The Rayleigh wave height distribution was shown by wavesis substituted into the energyflux equation(5) which is numericallyintegrated.Good comparisons were obtained be- Longuet-Higgins [1952] to apply to deep water waves on the tween calculated rms wave heights and laboratory measure- assumptionthat the sea waves are a narrow-banded, linear

(10). The dissipationrate (eo)as a function of H and h, for a wave known to be breaking, is given by (6) with the bore discharge described by (7). In the following sections, we complete the determinationof % by specifying the pdf of wave heights and the probabilitythat a wave.ofa givenheightwill be breaking. Equation(8) canthenbe integrated and the resulting H,ms comparedwith observations.

5928

THOt4TO4 ^4D GUZ^: Tt^4SVOtMnXlO40V W^w

HEIGHT DISTRIBUTION

much a function of the tidal stage(mean range 1.5 m). Instrumentsinshoreof C22 wereuncovered at springlow tidesand all instruments wereimmersed at neaphigh tides. During the experiments, significantoffshorewave heights varied between60 and 160 cm. The averagepeak frequency of C42 .. :15.';.' '"'2:'':!4',;:'";:'?:r:. :,.?i':i'.....-': ,?'? the incident wave spectravaried little during the experiments and was about 0.07 Hz. Shadowing by offshoreislandsand Im-,'":C36':'C31 ':'C3'iff:'"",." ,,C22. ,_,...; -,. . offshorerefractionlimit the anglesof wave incidencein 10-m o- CURRT depth to lessthan 15 [Pawka et al., 1976]. The condition of 0 I0 20m METERS Fig. 2. Crosssection of surfzoneshowing instrument spacing and nearly normally incidentspilling(or mixed plunging-spilling) waves, breaking in a continuous way across the surf zone, elevationsrelative to measuredwaves on November 20, 1978 at Torrey Pines Beach, California. prevailedduring most of the experiments. Winds during the experiments weregenerallylight and variable in direction. Gaussianprocess. The Rayleighwave heightprobabilitydenWave heights were determined from the surfaceelevation sity functionis recordsusingthe 'zero-up-crossing method,'in which the wave height is defined as the difference of the maximum and minimum occurring between two consecutivezero-up crossings. The resultsusingthismethodare very sensitive to the definition of meanlevelabout whichthe zero-upcrossings are computed whichis entirelyspecified by Hrm s.The Rayleighdistributionis and to high-frequencynoise. In extreme cases,very lowappliedcorrectlyonly to linear waves[Longuet-Higgins, 1975]. frequencysignals, suchas the riseor fall of the tidesor surfbeat, The distributionderivedwith a linear assumption would not be can causethe troughsof the wavesto be abovethe calculated expected to hold for wavesapproaching maximum height,i.e., 'average'water level.It is important to definethe mean water closeto breaking,as in the surf region or for broader-banded level over a time scalewhich is of the order of only a few waves 'sea'conditions with whitecaps. if the interestis the heightstatistics of seaand swell.Therefore The theoretical Rayleigh distribution has been found by the data were first linearly detrendedto excludeeffects of the severalauthors[Chakrabarti and Cooley,1977;Forristal, 1978] risingand fallingof the tidesand then high-pass filteredwith a to overpredictthe number of large wavesin the tail compared cutoff frequency of 0.05 Hz (20-s period) to excludesurf beat. with observations. Various explanations includingnonlinearity High-frequency signals, eitherartificial(e.g.,electrical noise) or [Forristal, 1978; Tayfun, 1980; Longuet-Higgins, 1980], whitehigh-frequency capillarywaves,can resultin an increase in the capping [Tayfun, 1981], and finite bandwidth [Longuetzero-up crossings as the sea swell waves crossthrough zero, Higgins,1980] have beenexaminedas causes for the deviation with a concomitant increase in the number of waves counted. from a Rayleighdistribution. Tayfun's[1981] breakingeffect Therefore the data were also low pass filtered with a high and Longuet-Higgins'[1980] finite bandwidthmechanism both frequency cutoff(0.3 or 0.5 Hz) depending on sensor type and havesomesuccess in explainingfield data. Thesestudies seekto depth,as discussed below.Note that the choice of the rangeof explain deviationsfrom a Rayleigh distribution,but the relefrequencies examined will affectthe resultsand leadsto some vant point here is that wave heightsappear to be nearly Raysubjectivity. leigh under a much wider range of conditionsthan the strict The filteringwasaccomplished by Fourier transforming the assumptions of a narrow band Gaussian(linear)process would signals,zeroing out the Fourier amplitude coefficients in the imply. We now describe somefield experiments and show that filtered-outfrequencies, and inverse transforming the complex wave heightdata evenwithin the surfzone are reasonably well spectrum to obtainthe filteredtime series. The entire68-minute described by the Rayleighdistribution. record was transformed at one time to minimize the end effects

W41

W38

W21

W29

p(H) =Hrms 2exp [--

(11)

3.2. Field Data Analysis

Experimentsmeasuringwave transformationwere conducted at Torrey PinesBeach,California,during November 1978. The details and various resultsof the experimentshave been described by Guza and Thornton [1980; 1981], Huntley et al. [1981], and Thornton and Guza [1982]. The beach is gently slopingand composed of moderatelysorted,fine-grainedsand. The beachprofile showsno well-developed bar structureand is remarkably free from longshoretopographicinhomogeneities. An extensive array of instruments wasdeployedto studynearshore wave dynamics.Measurementsdescribedhere are from sensors locatedalong an on-offshoretransectfrom 10-m depth to the inner surf zone. A crosssectionof a typical instrument transectinside3-m depth is shownin Figure 2. The waveshave been drawn to scale (vertical scale distorted 1:20), and they showthe horizontalwavelengths of dominant 14-second period swellas spatiallywell resolved. The sensors usedhereconsisted of 11, two-axis Marsh-McBirney electromagnetic current meters(denotedby C), four Stathem temperaturecompensated pressuretransducers (P), and four dual-resistance wire wave staffs(W). The shoreline and mean breakerpositions werevery

whichresultin spectral leakage and to obtainmaximum resolution givingverysharproll-off at the filter cutoffs. Since the averagewave period was about 14 s, the total
number of waves in the 68-minute record was about 300. Em-

piricalprobability density functions and heightstatistics of root

meansquare waveheightHrms, significant waveheight, H/3

(average oftheheights ofthe highest waves),/7/o (average of

the heights of the highest waves), and Hma x werecalculated
from the orderedsetof wave heights. To take advantageof the large number of current meters, currentdata wereusedwith linear theory to infer waveheights. The rationale for transformingthe velocitiesto infer surface
elevation is based on the earlier work of Guza and Thornton

[1980]. By intercomparing wave staffsand other sensors they showed, for this same data set, that linear theory spectral
transformations could be used to calculate surface elevation
location with less than a 20%

standard deviations either from pressure meters or current

meters at the same horizontal

error and, typically,lessthan 10%.

To obtain elevation time series from current measurements,

the complexFourier spectra of the horizontalvelocitycomponents U(f), V(f) were first calculatedand vectoriallyadded.

THORNTONAND GUZA: TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

5929

TABLE 2. WaveConditions Offshore andBreaker Type

rms

Wave

Height Ho,
Date
cm

Spectral

Hz
0.063 0.055 0.077 0.069 0.069 0.063

Depth,
cm
1069 1072 1088 1053 1050 1022

Width
C
0.66 0.72 0.57 0.63 0.51 0.76

Breaker
Type
spill spill spill spill/plunge spill/plunge plunge

Nov. 4 Nov. 10 Nov. 12 Nov. 17 Nov. 18 Nov. 20

35 56 88 38 49 50

The complexsurfaceelevationspectrum X(f) was calculated wave heights(Figure 3) is becausethe wave heightsinferred applying thelinearwavetheory transfer function H(f)' using current meters are low pass filtered with the highfrequency cutoff at 0.5 Hz, whichhasthe effectof roundingoff X(f) = H(f) V(f) (12) the peaksof the waves. Directly measured wave heightsusingwave staffscompare Only the seaswellband of frequencies is considered, so that the wave approach is almost normal to shore. Therefore it is as- slightlybetter with the Rayleighdistribution.For model coms described later, the sumed in the vectoradditionthat U(f) >>V(f), and the phase parisonsof observedand predictedHrm wave staffswere low-passfilteredwith a high frequency cutoff of the surface elevationis associated with the phaseof U(f) only. It is assumed that wavereflection is negligible. The com- at 0.5 Hz to be consistent with the other sensors. But for plex surfaceelevationspectrum was then inversetransformed comparisons here with Rayleigh statistics, the wave staff to obtain the surface elevation time series from which the wave measurementswere low-pass filtered with a higher highheight distribution was calculated.Surfaceelevationswere also frequencycutoff of 1 Hz in order to minimize filtering affects. inferredfrom pressure signalsby transforming the pressure The wave staffswere usually located around the mean breaker recordsusing linear theory. With increasing frequency and point and within the surf zone. The wave staff measurements to satisfythe strict depth, the signal to noise ratio of the surfacewave-induced give a most severetest of the necessity velocityand pressure signals decreases due to hydrodynamic theoreticalrequirementsof linearity and narrow bandedness for the Rayleighdistributionto be applicable. filtering.Simultaneously, the spectral transfer functions H(f) Table 2 gives offshore conditions of rmswaveheightHo,f, for both velocity and pressure exponentially increase. Therefore the surface elevationspectraobtainedfrom pressure and cur- spectralwidth at depth h, and the breaker type during each rent meterdata first decrease in energydensityfrom the peak experiment.The spectralwidth parameteris given by [Cart1956] frequency to higher frequencies, but 'turn up' at high fre- wrightand Longuet-Higgins, 2 quencies (e.g.,above 0.3 Hz for pressure signalsmeasured in 10-mdepth). The turn up is dueto noise beingamplified by the 2= mom4 m 2 (13) mom4 exponentially increasing transfer function H(f). Thusthehighfrequency filter cutoff was set at the frequency at which the wheremnare the variousspectralmoments. A value of near 1 inferred surface elevation spectrum turnsup, whichvariedwith is supposed to imply broad band waves,whereasa value 0 depth;the signals measured using deeper instruments had to be impliesnarrow-band waves. The Rayleigh distribution theoretmore severely filtered(,,0.3 Hz for 10-m-depth instruments) ically appliesonly for the case 0. Judging from the calcuthan shallowwaterinstruments. A highfrequency cutoffof 0.5 lated , the waves on all days were broad banded. To the Hz was applied to all instrumentsshallower than about 3-m contrary,visual observations indicatedthe waveson the 20th, depthfor Hrm scalculations. for instance, wereverynarrowbanded. The narrowbandedness was indicated by long-crested swell conditionsand by the 3.3. Comparison with RayleighDistribution classical 'groupiness,' or beating,of the narrow-banded waves. Six dayswereselected for analysis covering a widerangeof The groupiness wasexhibited by the wavesat breakinggoing conditions(see Table 2). Empirical pdf's of wave heights every severalminutesfrom 2-m heightsto essentially calm derivedfrom velocityand pressure measurements are com- conditions and back again to 2-m waves at the arrival of pared with the Rayleighpdf for selected depthson November another group of waves.The spectralwidth parameter, as 20(Figure 3).The Rayleigh pdfdepends onlyon thelocalHrm s- defined,doesnot appearto be a good indicator of band width SinceHrm sfirst increases towards the meanbreakerpoint and for shallowwaterwaves, whichare at leastweaklynonlinear,as then decreases, the modeand apparent width of the Rayleigh indicatedby the presence of spectralharmonics.In fact, the pdf first increases and then decreases when plotted against spectrafor November 20 [seeGuzaand Thornton, 1980] showa H/Ho (Figure 3). The width of 'bins'usedis the sameconstant narrow swellpeak and clear harmonic peaksat twice and three fraction of localHrm s for all empirical pdf's,so that as Hrm s times the swell frequency.The definition of does not disdecreases, the widthof thebinsof thepdf'sin Figure3 narrows. tinguish freehigh-frequency waves (say,chopfrom local winds), The Rayleighpdf appearsto qualitativelydescribe the mea- which indicate true broadbandedness, from harmonic peaks sured waveheights everywhere. The largest discrepancies of the which can be a consequence of the nonlinearity of a very measured waves with the Rayleigh pdf are deficits at the lowest narrow, energetic swellpeak. and highestwaves.But the bulk of the distribution is reasonThe cumulativeexceedance of wave heightdistributionsnorably well predicted,and thereforethe centralmomentssuchas malizedby Hmwere calculatedfor wave staff measurements H andHrm sshould be wellpredicted using the Rayleigh pdffor on November 4 (Figure 4). The waveson November 4 were modelcomparisons. Part of the reason for the deficitof higher relatively narrow band. The cumulative exceedance distri-

5930

THORNTON AND GUZA: TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

1.0

P4 = IO22 cm

P(H/Ho)

0.5

0.0
0 0 0.5 1.0 1.5 2.0

2.5

:3.0

1.0

C15

P(H/H0)

0.5

0.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

1.0

l
= 7 cm
0 0
-

P(H/H0)

0.5

0.0 0.5 1.0 1.5 2.0 2.5 3.0

2.0

1.5

0 0 0.5 1.0

C59
i

H = IO2 cm

1.0

P(H/H 0)
0.5

0.0

1.5

2.0

2.5

3.0

;2.5

;2.0

P(H/Ho)
1.5 1.0

C42 DEPTH

- 59 cm

0.5

0.0

0.0

0.5

1.0

, 1.5

2.0
o

I 2.5

3.0

H/H

Fig. 3. Empirical probability densityfunctionsusing pressure and current metersplotted againstRayleighpdf for November 20, 1978.H ois nrta sin ~ 10-mdepth= 50 cm.

bution emphasizes information in the high wave tail of the

distribution. The mean breaker line for these measurements

curveforcedthroughzero is drawn asa dashed line,but cannot

be differentiated from the 45 line. The difference between the

was betweenW29 and W21 so that W29 wasjust outsidethe

surf zone and the others were inside. Note that the differences

slopesof the 45 line and the regression line is the mean error over the rangeof measurements, whichis -i-0.3%.

with the Rayleigh(solidline) are exaggerated in the tail for this kind of plot. The largestwavesof the W29 distribution,measured just outside the surfzone,actuallyexceed the Rayleigh. For wavestaffson all 6 days,measured Hrm sand its approxi-

Measured H1/3 (average of the highest one third waves), Hl/o, and Hmax arecompared with theirrespective Rayleighderivedstatistics in Figures6-8, where for the Rayleighdistribution

mation Hrm s- 8V/o, where rno isthe surface elevation variance, are comparedin Figure 5. A 45 solid line is drawn to show a perfect correlation. The calculated correlation coefficientbetweenthe two variablesis 0.995.The linear regression

H/3 = 1.42Hrms H/o = 1.56Hrm

(14) (15)

THORNTON ANDGUzA' TRANSFORMATION OFWAVEHEIGHT DISTRIBUTION

5931

10 0

DEPTH
m (D '" +

IN

CM
200.0

82 185 177 225

150.0

Hm/$
100.0

'/

m.llm 10-2
50.0

lO

-3
o

1.0

.o

3,0
0,0
50.0

H/Hrm s
Fig. 4. Cumulative exceedence probability for wavestaffmeasurements plotted against Rayleigh pdfforNovember 4, 1978.

i
100.O

i
150.0

i
200.0

Hm/$ (RAYLEIGH )

Fig. 6. Measured/-/,/3 P lotted against Rayleigh statistic/-/,/3-

The maximumwaveheightfor the Rayleigh pdf wascalcu3.4. BreakingWave Height Distributions latedby using[CartwrightandLonguet-Higgins, 1956] The observations clearlysupportuseof the Rayleighdistri-

[(InN)1/2 + 0.2886 (In N)-1/2]Hrm s

where N is the total number of waves in the distribution.

in (16) butionas a goodmodelfor the waveheightpdf.However,

The

order to calculatethe breakingwave dissipation (eb), necessary for integration of the energy balance equation (8), we must correlation coefficients were 0.997, 0.988, and 0.924 forHu3, specifywhich wavesare breaking.Unfortunately,the imporHx/xo,and Hmax, respectively. The linearregression curves tance of measuringthe breaking wave distribution was not until after the Torrey Pinesexperiment. Thus adforced through zerogivean average percent errorof -0.2%, recognized The profile at SoldiersBeachgenerallyhas a moderateslope (1:30), oftenwith a,single crescentic bar and a steep (1:8) beacti

weremadeat Soldiers Beach, Monterey, -1,8%, and -6.8% for Hx/3,Hx/xo, and Hmax, respectively.ditionalmeasurements to obtain thenecessary breaking wave distributions. Theincreasing spread ofthepoints going from Hx/a to Hx/xo to California
Hma x may be because fewerpoints are usedto calculatethe statistic, Hma x beingthe extremewith onlyonepoint usedto calculateit. The heights may also be non-Rayleigh in the ex-

face (see Figure 9).Theoffshore surface elevation in 12-m depth using pressure sensor measurements. The surface treme tail.Theresults show 'hat thecentral moments ofHrm,, wasinferred elevations and velocities on a transect from offshore to the Hx/;andeven Hx/oarewellpredicted by using theRayleigh
distribution.Thereforeit is concluded that the Rayleighdistribution can be usedto give a reasonable description of waves evenin the surfzone,at leastfor the spillingbreakers measured at Torrey PinesBeach.

shoreline weremeasured usinga pressure sensor and current

meters mounted on a movable sled [see Sallenger, 1982]. The

base of the sled had dimensions of 4 x 5.5 m. A 10-m mast was

attachedto the baseand extended out of the water. The position and elevationof the sled relative to a baselineand MSL wereopticallysurveyed from shore.

200.0
200.0

150,0 150.0

Hrms
100.0

,.,1._1 NEAR

HI/io
100.O

,/

CURVE

50.0 50.0

0.0

i,

i ,

50.0

!00

i !50.0

200.0
0.0
i I i i ,i

JSmo
Fig. 5. MeasuredHrm s plotted againstthe approximation Hrm s--

50,0

tO0,O

150.0

200.0

H /o (RAYLE 16i-I)
Fig. 7. Measured H/toplotted against Rayleigh statistic Hx/o.

5932

THORNTON AND GUZA: TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

wherethe weightingfunctionW(H) _<1, to insurepb(H)_<p(H) in accordance with rule (2) above.The integral property((3) above)is
200.0

A =

p(H) dH

(18)

1S0.0

* ',
Hmax
100.0

CURVE

where A is the fraction of all waves which are breaking and that in deepwater,A-- 0 ash-- oo,and in surfzoneAb-- 1 as h-- 0, i.e., all wavesare breaking. The simplesthypothesis is that the wavesbreak in proportion to the distribution for all waves,so that
IV(H) = A (19)

whereA is independent of H and dimensionless.

50.0

Let

0.0

/W(H) -'gb --(Hrms n k, 7h}

Hma x (RYI_EIGH)

(20)

with n a variable to be determined from the observations. The

Fig. 8. Measured Hma xplottedagainst Rayleigh statistic Hma x.

The sled was moved using a continuous 'clothesline' arrangement. A 1-inch(2.54cm)nylonropewasattached to both

ends of thesled. Thelineleading offshore wasfedthrough a blockaffixed to an anchor in approximately 10-mdepth; the linethenled backto shore where a winch could pull thesled
offshore.The line attached to the shorewardend of the sledwas

pulledon directlyto bringthe sledonshore. In thismannerthe sled could be positioned anywhere on the transect between
10-mdepth and the beach.

Measurements weretakenat a fixedlocation for 35 minutes,

and then the sled was moved to another location for the next

run.Waves visually observed to bebreaking or broken asthey passed the mastof the sledwereelectronically flagged and recorded simultaneously withpressure sensor data. Theflagged wave heights were measured fromthewave record (after having converted thepressure record to surface elevations) using the (21) 7h/L zero-up-crossing technique to obtaina breaking waveheight distribution. Wave height distributions forallwaves (including breaking and nonbreaking) and the corresponding breaking Thorntonand Guza [1982] showedthat for Torrey Pinesin wave height distributions (hatched area) forfourlocations (fur- the inner surfzone,an envelopecurverelatingrms wave height by thestoffshore at the top and proceeding shoreward downthe to depthwaswell approximated

form of (20) is motivated by two considerations. Firstly, the importance of the parameter Hrms/7h is expected because of the well-known depth limiting condition for shallow water monochromaticwaves,H = 7h. Secondly, (20) yieldsanalyticalsolutionsfor the transformation of Hrm son a planebeach. A problemwith (20) is that the likelihoodof a wavebreaking is independent of its height,sinceIV(H) = constant.Observations show that at a particular location the largestwavesare more likely to break. Hencea greaterproportion of the larger wavescontributeto the breakingwavedistribution,resulting in the breaking wave distributiongenerallybeing skewedto the higher waves relative to the Rayleigh. The skewing of the observedbreaking wave height distributions is obvious in Figure 10. A simplemodificationto (20) whichmore heavilyweights the largerwavesis givenby

IV(H--(--rmsX F1-exp <1 mn (--(m)2)l

figure)are shownin Figure 10. Sincethe data were not mea-

sured simultaneously but sequentially, thewaveheight values havebeen normalized by theoffshore rmswave height H0 to A value of 7 -- 0.42 was alsofound for the SoldiersBeachwaves
account for any variabilityin theincident waveconditions. The
and suggests similarity of the breaking wave processes in the Thereforea value of y -- 0.42 is Rayleigh pdf is superimposed on the measured waveheight inner surfzone at the two sites. distribution (solid line)andagain appears to represent thewave used.

Hrm s ' 0.42 h

(22)

heights well.The breaking waveheight distributions show that evenat the 5-mdepth, thewaves occasionally broke,andit was
not alwaysthe largest waves that broke.

Since thereis no theory for describing breaking wave distributions Pb(H), we simplyfit empirical expressions to the observed distributions. In describing pt(H) three ruleswill be

applied: (1)pt,(H) should resemble theobservations (Figure 10), (2) pt,(H)is a subset of the distribution p(H) for all waves, breaking and nonbreaking, and (3) the area underthe distributionis equalto thepercent of breaking waves; thisruleis a convenient definition to keep trackof whatpercent arebroken waves andmeans thatpb(H) isnot a pdf. The distribution of breaking waveheights canbeexpressed
asa weighting of the Rayleigh distribution for all waves:
Fig. 9.

-5

-lO
0

i
50

i
100

i
150

i
200

i
250

i
500

DISTANCE OFF-SHORE (M)

Beach profile at Soldiers Beach, California on August 24,
1981. Vertical lines indicate measurement locations.

Pb(H)= W(H)p(H)

(17)

THORNTON AND GUZA: TRANSFORMATION OF WAVE HEIGHTDISTRIBUTION

1.0

$933

,.2]
= 500 cm
0.0

b(H)dH

o i4 ,
0.5

'yh/Hrm s
Fig. 11. The percentobreakingwavesversus7h//'/r= s with 7 = 0.42 for pb(H)specified using(20) with . = 2 (solidline),. = 4 (dotted linc),using (21) (dashed linc)and measurements denoted by
crosscs.

described by (21)appears to givethe bestfit to the data and has

been drawn as a dashed line for all the distributions shown in
PTH = 284
0.4

cm

Figure 10. Equation (21) has no physical justificationand is simplya convenient empiricalexpression which fits the observations.Equation(20) doesnot fit the data as well but leadsto an analytic solutionderivedin the next section.

0.0

1.2

The p(H) givenby (17)-(21) are usedto calculate (%) to complete the modeldescription. Note that the Rayleigh pdf is completely specified by Hrms, and Hrm s is calculated by integrating the energyflux equation(8) from deep water to the
location of interest.

3.5. Energy Dissipation

0.8

223
0.4

cm

The energydissipation is considered primarily due to the conversionof potential wave energy to turbulent kinetic energy, whichis eventually lostto heatduringwavebreaking,

0.0

and secondarily due to bottomfrictionallosses. The energy dissipation in a breaking waveis modeled aftera periodic bore. The rateof energy dissipation perunit areafor eachboreusing thedescription for Q givenby (7) is

, = - pg

(BH)3

(23)

H = 207
0.4

cm

The average rate of energy dissipation is foundby addingup the dissipation for eachbrokenwavecalculated using (23)and dividingby the total numberof waves(including brokenand unbroken waves). In otherwords,the average rate of energy
dissipationis calculatedby multiplying the dissipationfor a single broken wave of height H by the probability of wave breakingat eachheight,asgivenby pgH). For the ensemble,

0.0

H/Ho
Fig. 10. Wave heightdistributions and breakingwaveheightdistributions (hatched area)normalized to offshore rmswaveheightH 0. The Rayleighdistribution is givenas a solidline and the empirical
breaking wavedistribution using (21)is givenasa dashed line.
1.2

B3f oH3p(H) dH (24)

0.8

Thepercen[age of breaking waves asa function of yh/Hrm s described using (20)withn = 2 and4 andusing (21)withn = 2
arecompared withthefieldmeasurements (Figure 11).Onlya
small percentof wavesare predictedto break until about

223

cm

7h/Hrm s 2, afterwhichthe waves veryquicklyreachsaturation, atwhich time allwaves arebreaking, i.e., 7h/Hrm s= 1. The comparisons in Figure11 suggest thatPb(H), described by either weighting functions (20)withn = 4, or (21),reasonablydescribes thepercent of all waves which arebreaking. The shapes of these breaking wave distributions arecompared with the measured breaking wavedistribution at 2.23-mdepth (Figure12); the pb(H)for this shape comparison havebeen adjusted sothat their areas arethesame asmeasured. Thep,(H)

0.0 0

H/Ho
Fig. 12. Wave heightand breakingwave height(hatched area) distributions compared with Rayleighdistribution (solid line) and breaking wavedistributions using (20)withn = 4 (dotted line)and(21) (dashed line).For shape comparisons the areas of empirical breaking waveheight distributions havebeen madeequalto measured area.

5934

THORNTON ANDGUZA: TRANSFORMATION OFWAVEHEIGHTDISTRIBUTION

d refers to deepwaterand K, accounts for Two.dissipation functions are considered. Substituting p(H), wherethe subscript described usingweightingfunction(20), with n = 4 into (24) refractiveeffects.The offshoresolution(35) and the shallow watersolution (34) are matched at h0.Transfor .ming backand and integratingyields expressing in terms of waveheight:

Althoughusing thisp(H) did not givethe bestfit to the data,its O<h<ho use leads to an analytical solution (described below) which allows us to easily explore the model behavior. The second The asymptotic case asthedepthgets veryshallow is dissipationfunction is obtained using(21) to describep(H),
giving

,ho3/ (36) Hrms _allSh9/lo[1__ h23/4( I y/ 82.)]-115

Hrm s' ailSh 9/1 as h0 (37)

3.6. Analytical Solution

An analyticalsolutionin shallowwater can be obtainedfor waves approachingnormally on a plane sloping beach. Although the solutionis only applicable in shallowwater, considerableinsightinto the workingsof the model can be obtained.Startingwith the energyflux equation(8) with E describedby (9) and describingthe bore dissipationusing (25) gives
dx 8 16

which says thatthewaveheight in theinnersurfzoneisrelated to the depthandindependent of theinitialconditions in deeper water.Thisresultis similarto theobservations at Torrey Pines Beach [Thornton and Guza, 1982] that wavesof all initial heights shoaling fromdeep waterbecome saturated in theinner
surfzonewith the heightsgivenby (22). 3:7. Frictional Dissipation

Frictionaldissipation at the bottomboundary layer can easily beincluded andis described herefor completeness but is
shownto be a minor dissipation mechanism. Once wavesstart

d1 PgHrms2C" __ /4.h5 Hrms ? (27)

h

to break,the average rate of frictional energy dissipation for a single waveis calculated by assuming the usual.quadratic formulation for bottom shear stress,

The independentspatial variable can be transformedfor a planesloping beachusing

pcv,'lvl

(38)

where theh sqbscript refers to bedandc$is thebedfriction

coefficient..Again assuming a verynarrow-band wavespectrum

x=ta n/
v/oh and defining '
(27) is rewritten

(28)sothatall waves havethesame average period andapplying

lineartheoryto describe the wave-induced velocityat the bed, the frictionalenergydissipation for a single waveis

Bypptying the shallow water linear approximation Cg - C =

y '- Hrms2h TM (29)

e= pc k, sinh khJ

1( 2nf 3H3

(39)

dh-2 ' tan/ 27/4 (30) dy 3() '1f'B 3h y71_

(30) is easilyintegratedto give

The expected (average) frictional dissipation for the ensemble (all waves) is calculated usingthe Rayleigh probability distribution{17),sinceall wavescontribute to the velocities at the
bed:

__y-5/2 ._ 1h_23/4 +const.

a

(31)

(es) --PCs '( ksinh kh,l Hap(H) dH

where

=pcs 16x/ L sinh kh J

a =- B3 f fi 23() 'a ?' tan

I [2nf,m,1 3

(40)

Dissipation dueto breaking andfriction canbecompared for (32) theshallow watercase (sinh kh 2n(hg-)/'f) withinthesurf

Sincethe shallowwater approximationwasmade,the offshore zone using thebreaking wave dissipation function (25)and(40) boundary condition is defined where the shallow water ap(es) 740 /' h7/, proximationbecomes valid (withinonepercent):

Hrms 4

Y= Yo = Ho2ho TM at h0 _< -
,

to (33) In the innersurfzone,(37) appliesand(41)reduces

After applying the outer boundary condtion to specifythe integrationconstant, the complete solutionfor y is

-- L J

mtJ-0

y = 8215[(h - 23/4 __ho - 23/4 + Y012/5

is sen to be relativelymore impor(34) The frictionaldissipation tant on a mildersloping beachand for lowerfre,quency waves

The solution can be stated in terms of the deep water wave

conditions by assuming conservation of energy flux nd using rl [Si, finitedepthlear theoryseaward of h0. Torrr i

aswould b pctd. r mi = 0.01, cceptd oil t L, 1977), rl rrtti ch or= 0.07H, . = 0.02,
B = 1.0(tlly drlopedbores), th tctiol iitio i 1 th 3% ot th diiptio d to brti for &pth ter
20 ithi th rr zo. Similar comrio

y0 =

TM =

gm

(35) th

THORNTON AND GUZA' TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

5935

1.0

'

0.8

respectively.The results,along with the curve for no dissipation, i.e., energyis conserved, are comparedwith laboratory data from Battjes and Janssen[1978] (Figure 13). Only the portion of the curvenear the beachis shown.The lab data were measuredon a 1' 20 beach slope; the generatedwaves were random, with mean frequencyof 0.407 Hz and a deepwater rms wave height Ha of 12.6 cm..The waves broke as plunging breakers. The waves were one-dimensionalin the laboratory

channel sothatC = C in (10).A value of 0.4wasmeasured

0.0 0 2 4. 6 8 0

h/Hd (crn)
Fig. 13. Model generated Hrm sversus distance offshore normalized by deep water wave height(H a = 12.6cm) plotted againstlaboratory data from Battjesand Janssen [1978]. Solid line is M2, upperdashed line is linear wave theory with no dissipation, and dot-dashedline is
MI.

obtained for laboratory beaches.The solution indicates that frictionaldissipation is negligible comparedwith the dominant wave breakingdissipation, exceptin the very shallowest water as h 0, say, in the run-up region, where the boundary layer effects dominate.Our analysisis not concerned with the run-up regionso that frictionaldissipation will be neglected.
3.8. Numerical Model

from Figure 13. Model fitting suggests that B = 0.8 is reasonable for both models.For the same B value, model M2 is more dissipative. This is expectedbecause the breaking wave distribution usingweightingfunction(21) in M2 is skewedto higher wavesrelativeto that for MI. Thus whentaking high moments to calculatethe breakingwave dissipationas in (25), M2 dissipation will be greater. The departure of the model resultsfrom the measuredlab valuesat the shallowest depthsis partly because the model does not accountfor waveset-up.It is clear that increasing the mean

depthwith set-upwouldincrease the theoretical Hrms, bringing the inner surf zone data into closeragreementwith the model
results.

For the completesolution starting in arbitrary depth, for general bottom profiles,numerical integration must be used. For the numericalsolution the more accuratedescriptionof pb(H), describedusing (21) resulting in breaking wave dissipation (26),will be employed. The energyflux balanceequation (8) is solvedby substituting the bore dissipation function(and bottom friction dissipation function) and numerically integrating from offshore to the shoreline. Several numerical schemes were investigated. A fourth-order Runge-Kutta coupled with a fourth-order Adams-Moulton extrapolation scheme[Gerald, 1978] was used as a standard. In the end, it was found that the simplestforward steppingschemeis sufficientlyaccurate, where

The analytical solution is not applied to the plane sloping laboratory beach becausethe shallow water solution is not applicableuntil the depthis 15 cm (h/Ha = 1.2 in Figure 13) for the short, 2.5-s-periodwaves.However, the analytical model does have applicationto the field. For example,for the relatively long-periodwavesincident at Torrey Pines Beach with mean period rangingbetween13 and 18 s, the corresponding shallowwater depthlimits are 4 and 8 m.
4.2. Model Comparisons with Field Data

For comparisonpurposesthe bottom contours at Torrey Pines can be considered straight and parallel and the waves normally incident. Guza and Thornton [1980] performed refraction and shoaling sensitivitymodel testing using linear refraction.Waves of 0.067 Hz and varying angle of incidence from 0to + 15wererefracted from 10- to 3-m depthusingthe ECgxl2 = Cgxl + () I Am + %> I Am (43) measuredbathymetry at Torrey Pines Beach.The testsshowed the percent differencebetweenlinearly shoaledwave heights Starting from the deepest measurement, say, location 1 where (within the 15 angularspread)on the measured topography Hrms, and fare given, theintegration andpredicted quantities and normally incident waveson plane parallel contourswas areobtained asfollows. CxJ.2 andE1 arecalculated using the lessthan 5% for any directionalband. Thereforethe wavescan lineartheoryrelationships (9) and(10)since Hrms, 1 and hi and be approximated as normally incident and shoaled over the h2 are known. The dissipation(e)J is calculatedusing(26) measured bathymetry on the instrument transect. This sim-

whereHrm s is calculated at 1. E2 (and henceHrms,2) is then

predicted,sincethis is the only remainingvariable.The data have shownthe Rayleigh distributiondoesa goodjob of describingwave heightseverywhere, so the pdf at location 2 is

plifies theanalysis, since C = Cg in (8).

The model M2 is comparedwith the 6 days of data representinga relativelybroad range of wave conditions(seeTable 2). Optimal model coefficients B determinedby iteration were soughtto represent individualdaysand also a singlevalue was soughtto represent all days.The sumsof the squareerror of the model Hrm scomparedwith the measured Hrm swerecalculated
TABLE 3. Optimal B valuesfor Model M2
Number of Instruments 16 12 8 14 12 12

completely specified using thepredicted Hrms, 2.The modelalso

predicts the fractionof waveswhichhavebrokenand their pdf. Further spacesteppings yield similar predictionsat all shoreward locations.

In the applicationof the model, the coefficient ? = 0.42 has been applied as determinedfrom the data. The only underdeterminedcoefficientthen is B, which is found by model fitting. B and ? could have beencombinedinto one coefficient to be determinedfrom the data, but it was felt that greater insightis retainedby treatingthem separately. It is expected that B = 0(1)if the modelis performing properly.
4. MODEL COMPARISONS

Date

Percent

Error

4.1. Laboratory Data

Nov. 4 Nov. 10 Nov. 12 Nov. 17 Nov. 18 Nov. 20

1.5 1.7 1.3 1.5 1.5 1.6

7.3 5.3 7.0 7.9 9.3 6.3

The numericalenergy flux balance model was run using dissipation functions (25) and (26),denotedmodelsM 1 and M2

All days
Lab

1.5
0.8

8.6
6.1

74
9

5936

THORNTON AND GUZA' TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

80[
4020-

shown in Figure16.The modelresults do not appearto depend on the magnitude of Hrms, i.e.,asthe waveheightincreases, the differences from the 45 line do not increase. The percentage error betweenpredictedand measured Hrm s plotted against distanceoffshore(Figure 17) showsthat the model predicted Hrm s values are within + 20% of the measuredvalues and usually much less. The error standard deviation (standard
error) is 0.086. Hrm smeasurement errors.aredue to sensorerrors and errors
I

100

200

i 500

400

500

i 600

inherent in transformingthe velocity and pressuremeasurementsto surface elevations with linear wave theory. The current meter and pressuresensorcalibration errors were of the
order of + 5%. The combined error associated with linear

X(m)

Fig. 14. Model generated Hrm s versusdistancex plotted against

measured values for November 10, 1978. Solid line is M2 without

friction,uppersolidline is linearwavetheorywith no dissipation. The measuredHrm s values were determinedfrom sea surfaceelevations measured directly with a wave staff (circle)and by applyinglinear theory to the velocity(crosses) and pressure (squares) measurements to
infer sea surface elevations.

theory and sensorerror is less than + 20% [see Guza and Thornton,1981,Figure 5], whichis not substantially different
from the error between model and measurements. Therefore much of the difference between model and measurements could be due to measurement errors.
4.3. Discussion

for all sensors for eachof 6 days and for all 6 days usingvarious values of B. The optimal values of B, the number of instruments,and the standarderror comparedwith the field data are given in Table 3. The value of B obtainedfor the laboratory data of Battjes and Janssen[1978] is also given. Sincemost of the instrumentswere located from just outsidethe surf zone to the beach,B is weightedto fit the data betterin this region;this also coincideswith the zone in which most of the dissipation and changeoccur. Model M1 was also run and an optimal value was found for B = 1.72 with a standard error of 8.3% for all days. Even though model M 1 givesa slightlylesserror than M2, the latter is the preferredmodel becausethe breaking wave distribution used appearsto be more appropriate (Figure 12). Therefore model M2 will be usedto demonstrate the data comparisons. Examplesof model M2 comparisons with measurements for the 10th and 20th of Novemberare shownin Figures14 and 15. A value of B equal to 1.5, the averagefor all days, has been used. Using optimal B values for the particular days given in Table 3 would give slightly better fits, particularly in the nearshore region. The waves on November 10 were relatively broadbandedsea while the waveson November 20 were very narrow-banded swell. A comparison with conservation of energy for linear waves outside the surf zone is shown for November 10; the difference between the model and conservation of energyis slightuntil verynear the surfzone.
For November 20 the model was run with and without

Why doesthe modelseemingly work so well while incorporating a numberof grossly simplifying assumptions? The basic ingredients making up the model are the energyflux equation and a bore dissipation function. The energy flux balanceequation (8) is a correctstatementof the physics, but the linearized theoryis only valid to the first order in nonlinearity, whichis certainlyof questionable accuracy near breaking.Even so, the conservationof linearized energy flux (no dissipation)was shown to do a goodjob of predicting the shoaling of Hrm suntil very near the surf zone, both in the lab (Figure 13) and in the field at Torrey PinesBeach(Figure 14 and Guzaand Thornton [1980]). The model predictswave saturationconditionsin the inner surf zone wherethe wave heightis stronglya functionof the

depth.It is difficultto discriminate between the h9/dependence on H,nsin the innersurfzonepredicted by the analytical model and the linear dependence suggested by Thorntonand Guza [1982] using the data. The differencebetween the two
curves is not significant over the interval of measured wave

heights in the inner surfzone(100- to 15-cmdepth),giventhe

scatter of data.

dissipation due to friction.Adding bed friction dissipation by

Sincethe predictions of H,ns usingthe modelsdescribing p(H) in two differentwaysgaveessentially the sameresults, it is concluded the model resultsare not stronglysensitive to the formsof p(H).The differences in p(H) are compensated for by usingdifferentvaluesof B. The accuracy of the modelis dependent on the selection of B.
80-

usinga uniform bed shearstress coefficient cs equalto 0.01

resulted in a maximum additional decreasein wave height
about the mean breaker line of less than 3% which is consistent

with our earlieranalysis. Again,because frictionaldissipation is of secondary importance and complicates the analysis by introducinganother unspecified coefficient, it is not includedin the
final results.

60

40

Figures14 and 15 showthat the modelis capableof predicting the wave height increasedue to shoalingand subsequent decrease due to wave breaking.The wavespeak up to a maximum, a point definedhere as the mean breaker line, and then decreases. In the inner surf zone, all the wavesbecomelocally depth controlled, and changesin the depth are reflectedin changes in wave height.The 'wiggles' in the resultsin the inner surf zone for November 20 (Figure 15) are due to variationsin the bottom profile. All the measuredHrm s versusmodel predictedvalues are

20

0
o 100 200 300 400 500

I
600

x (rn)

Fig. 15. Model M2 generated Hrm s versus x plottedagainst measuredvaluesfor November20, 1978.The measured Hrm s were determinedfrom seasurface elevations measured usingwave staffs(circles) and by applyinglinear theory to the velocity(crosses) and pressure
(squares) measurements to infer seasurface elevations.

THORNTON AND GUZA' TRANSFORMATION OF WAVE HEIGHT DISTRIBUTION

5937

100 80 10 -

NOV

0.20

0
o

17 18 20

no

0.15

0.10

20 0

o9/
20 40 60 80 100

0.05

0/ i i i i i
Hr= (model,} Icm)
Fig. 16. Hrms measured versus Hrm smodel.

0.00
1.0

1.2

1.4

1.6

i.8

2.0

Fig. 18. ModelM2 errorprediction compared withTorreyPines

data for all daysfor variousvaluesof B.

Several of these other For the two very differentapplicationsof the model M2 to the pation function have been suggested. formswere investigated, but it wasdecidedthat the field and lab data, B valuesof 1.54(averagefor all 6 days)and dissipation 0.8 were obtained.SinceparameterB represents the percentage physicalrationale for using theseother bore dissipationfunctions was not justified on the basisof the data presentlyavailof foam on the face of the wave, which is a measure of the physicalmodel has beenpresented intensity of breaking, it is expectedB would depend on the able.Thereforethe simplest breaking wave characteristics, but no such correlation was and comparedwith data. found, which is apparent by comparingB with various wave 5. SUMMARY AND CONCLUSIONS parameters in Tables2 and 3. The sensitivityof the model to the selectionof B usedfor all The Rayleighdistributionis shownto givesurprisingly good days is shown in Figure 18, where various values of B are estimates of wave height statistics, even Hmax, for the spilling plotted against standard error. The curve shows that a varibreakersmeasuredat Torrey Pines Beach. The percent mean ation of + 25% about the optimal value B results in an inerrors over the measuredrangescomparedwith the Rayleigh
creased model error of less than 10%. Therefore the model fit to

data is not overly sensitive to the sleetionof B. The conditionB g 1.0 is expectedon physicalgrounds,with B = 1 correspondingto fully developed bores. Since (e0) is

statistics of H1/3, Hl/o, and Hma x were -0.2, -1.8, and

-6.8%, respectively. The results show the Rayleigh distribution to slightlyoverpredict the numberof waves in the tail of the distribution,but it is nevertheless able to predictthe central

proportional to B3 in (23),the resultthat the optimalvalueof

B = 1.54 implies the simple periodicbore dissipationfunction A model describing the transformation of random wave underestimates the dissipationby almosta factor of four. Howheights was developedbased on energy flux balance. Dissiever,from Figure 18, usinga B value of 1.2 would only slightly pation is considered due to wave breaking and bed friction. degradethe model fit and result in only a 70% underestima- Wave breaking is characterized after periodic bores. The tion. Stive [1983] made detailed measurementsof breaking random nature of the wave heightsis described usingthe Raywavedisipationfor monochromatic wavesbreakingon a gently leigh distribution everywhere, as suggested by the data. An sloping(1'40) plane laboratory beach.He concludedfrom an empiricalbreakingwave heightdistributionbasedon the field evaluation of laser doppler flow field measurements that there data is usedto definewhich wavesthe bore dissipation function was a strong resemblance betweenthe internal mean and turis applied. The model is capableof predictingthe increasein bulent flow fields of quasisteadybreakers,bores, and weak averaged waveheightdue to shoalingand subsequent decrease hydraulicjumps. But he found that the classicalperiodic bore due to wavebreaking. Bottomfrictiondissipation usingcs formulation,as usedin this analysis,underestimated the mea- 0.01 results in a maximum wave decreaseof 3% occurring sureddissipationrates by 30% to 50%. This laboratory result about the mean breakerline comparedwith the inviscidshoalis qualitativelyconsistent with the field results. ing. Because it is of secondaryimportance and introducesa As mentioned earlier, various other forms for the bore dissisecondunspecified coefficient, bed friction is not included.The model has only one adjustableparameter,B, which is a mea0.3 sureof the intensityof wavebreaking. NOV 4 The model is comparedboth with laboratory data and the 10 0.2 extensiveset of field measurements collectedat Torrey Pines 12 17 Beach, California. The model is able to predict rms wave 18 O.l +x 20 heights to within a standard error of 8.6% throughout the regionfrom offshoreto the beach.Although good comparisons o + o are obtained,the resultssuggest simplebore theory underestix A o matesthe dissipation. The underestimates are compensated by adjustingthe B coefficient.

statistics ofH/3 andeven H/o well.

-O.l +oax x
-, 2
-0.3
0 200

o
Acknowledgments. This work wassupported by the Officeof Naval
I I

Research, Coastal SciencesBranch, under contract numbers NR 388-

00

800

114 (for E. B. Thornton) and N00014-75-C-0300(for R. T. Guza), the

National Science Foundation under contract number DAR-8023425

DEPTH (cm)

Fig. 17. Percent error between modeland measured Hrm s as a function of distance offshore.

(for E. B. Thornton),and the SeaGrant NearshoreSediment Transport Study (project number R/CA-N-4D). R. J. Seymour obtained the bathymetric data.The staffof the ShoreProcesses Laboratory(Scripps

5938

THORNTON AND GUZA' TRANSFORMATIONOF WAVE HEIGHT DISTRIBUTION

Institution of Oceanography)installed and maintained the offshore sensors and data acquisitionsystem.R. L. Lowe was the principal engineer. A. Sallengerand his crew at the U.S. GeologicalSurveyare thanked for their cooperationin obtaining the breaking wave distributions. D. O. Burych aided substantiallyin the data processing. A critical review by J. A. Battjes,which lead to an improvementin the model,is gratefullyacknowledged.
REFERENCES

distributionof wave heights,Proc. Civ. Eng. Oceans, 3, 1211-1231,

1974.

LeMehaut6, B., On non-saturated breakers and the waverun-up,in Proceedings of the8th International Conference Coastal Engineering, pp.77-92, American Society of Civil Engineering, New York, 1962. LeMehaut6, B.,andR. C. Y. Koh, On thebreaking of waves arriving at
an angleto the shore, J. Hydraul.Res.,5(1),541-549, 1967.

Longuet-Higgins, M. S.,On the statistical distribution of theheights of

seawaves,J. Mar. Res.,11(3),245-266, 1952.

Battjes,J. A., Set-updue to irregularwaves, in Proceedings of the 13th InternationalConference CoastalEngineering, pp. 1993-2004, American Societyof Civil Engineers, New York, 1972. Battjes,J. A., and J.P. F. M. Janssen, Energy lossand set-updue to breakingof random waves,in Proceedings of the 16th International Conference Coastal Engineering, p. 569, American Society of Civil Engineers, New York, 1978.
Bowen, A. J., D. L. Inman, and V. P. Simmons, Wave 'set-down' and waveset-up,d. Geophys. Res.,73,2569-2577, 1968. Cartwright, D. E., and M. S. Longuet-Higgins,The statisticaldistri-

Longuet-Higgins, M. S., On the joint distribution of the periods and amplitudes of seawaves, J. Geophys. Res., 80, 2688-2694, 1975. Longuet-Higgins, M. S., On the distributionof the heightsof sea waves: Some effects of nonlinearity andfinitebandwidth, J. Geophys.
Res.,85, 1519-1523, 1980.

Miche, R., Mouvements ondulatoires desmersen profoundcur constanteou decroissante, Ser.3, Issue363, Wave Res.Lab., Univ. Calif.
at Berkeley,1954.

butionof the maximaoa randomprocess, Proc.R. Soc.LOndon Set.

A, 237, 212-232, 1956. Collins, J. I., Probabilities of breaking wave characteristics, in Pro-

Miller, R. L., Role of vortices in surfzone prediction: Sedimentation and waveforces, beach and nearshore sedimentation, Spec. Publ.24, Soc. of Econ.Paleontol. andMinerals, Washington, D.C., 1976.
Pawka, S.S., D. L. Inman, R. L. Lowe, and L. Holmes,Wave climateat

TorreyPinesBeach, Calif.,Tech.Pap. 76-5,Coast.Eng.Res.Center,

Fort Belvoir, Va., 1976.

ceedings of the 13thInternational Conference Coastal Engineering, pp. 399-412, AmericanSocietyof Civil Engineers, New York, 1970. Chakrabarti, S. K., and R. P. Cooley, Statistical distributions of periodsand heightsof oceanwaves,$. Geophys. Res.,82, 1363-1368,
1977.

Forristall, G. Z., On the statisticaldistribution of wave heights in a storm,d. Geophys. Res.,83, 2353-2358, 1978. Gerald, C. F., AppliedNumericalAnalysis, Addison-Wesley, 1978. Goda, Y., A synthesis of breaker indices(in Japanese), Trans. dpn. Soc. Civ. Eng., 13, 227-230, 1970. Goda, Y., Irregular wave deformation in the surf zone, Coast. Eng. dpn.,18, 13-26, 1975. Guza, R. T., and E. B. Thornton, Local and shoaled comparisons of sea surfaceelevations,pressures, and velocities,d. Geophys.Res., 85,
1524-1530, 1980.

Sallenger, A. H., Jr., P. C. Howard,C. H. Fletcher, III, and P. A. Howd, Profile,waveand currentmeasurement system for the high energy nearshore, J. Mar. Geol., in press, 1983. Shemdin, O. H., K. Hasselman, S. V. Hsiao, and K. Herterich,Nonlinear and linear bottom interactioneffects in shallowwater,in Turbulence Fluxes through theSeaSurface, WaveDynamics andPrediction, editedby A. Favre and K. Hasselmann, pp. 347-372, Plenum,New
York, 1978.

Shuto, N., Nonlinear long wavesin a channelof variable section,

Coast.Eng. Jpn., 17, 1-12, 1974.

Stive,M., Energydissipation in wavesbreaking on gentleslopes, J.

Coast.Eng.,in press,1983. Stoker,J. J. Water Waves, Interscience, New York, 1957.

Tayfun, M. A., Narrow-band nonlinear seawaves, J. Geophys. _Res., 85,

1548-1552, 1980.

Guza, R. T., and E. B. Thornton, Wave set-up on a natural beach,d. Geophys. Res.,86, 4133-4137, 1981. Huntley, D., R. T. Guza, and E. B. Thornton, Field observations of surf beat, 1, Progressive edge waves,d. Geophys. Res., 86, 6451-6466,
1981.

Tayfun,M. A., Breaking-limited waveheights, J. Waterw.,Port, Coast OceanDiv. Am. Soc.Civ. Eng.,107(WW2), 59-70, 1981. Thornton, E. B., and R. T. Guza,Energy saturation andphase speeds measured on a natural beach,J. Geophys. _Res., 87, 9499-9508, 1982.

Hwang, L.-S., and D. Divoky, Breaking wave set-up and decay on gentle slopes,in Proceedings of the 12th InternationalConference Coastal Engineering,pp. 377-389, American Society of Civil Engineers, New York, 1970. Kuo, C. T., and S. T. Kuo, Effect of wave breaking on statistical

(ReceivedOctober 9, 1981' revisedMarch 25, 1983' accepted March 28, 1983.)