JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 96, NO.

Cll, PAGES 20,561-20,569, NOVEMBER 15, 1991

BreakingWavesAffecting Microwave Backscatter

2. Dependenceon Wind andWave Conditions
A. T JESSUPt AND W. K. MELVILLE

R. M. ParsonsLaboratory,Massachusetts
Instituteof Technology,Cambridge

W. C. KELLER

U.S.Naval ResearchLaboratory,Washington,D.C.

Thispaperis the secondof a two-partserieson usingmicrowavetechniques for detectionandcharacteriza-

tion of wave breaking. The statisticsof seaspikesdetectedusingthe methoddescribedin Part 1 are investi-
gatedasfunctionsof frictionvelocityu. andof a ReynoldsnumberRe. basedon u. andthe dominantsurface
wavelength.For vertical (VV) andhorizontal(HH) polarization,the frequencyof seaspikesandtheir contri-
butionto the meanradarcrosssectionshowa roughlycubicdependence on u.. The percentageof wave crests
producingseaspikes,Pss,is consistent
with theRe. exponentof 1.5 reportedby otherinvestigators.
At high
frictionvelocities
(u. between
40 and50cms-t), seaspikes
contribute
between
10%and15%to themean
radarcrosssectionfor VV polarizationandbetween15%and25% for HH polarization.Thedatasuggestthat
the averageradarcrosssectionof an individualseaspikedoesnot dependon u.. Measurements
of Doppler
frequencyandbandwidthareusedto investigatethekinematicsof thebreakingprocess.

1. INTRODUCTION whereu. is the frictionvelocity,L is a lengthscalecharacteristic

of the sea surface, and v is the kinematic viscosity of air. The
In Part 1 [Jessupet al., this issue]we showedthat a detection
theoreticalformulation [Toba, 1972] is supportedby field mea-
schemebased on thresholdsin radar crosssection and Doppler
surementsof whitecapcoverage[Toba and Chaen, 1973] and
bandwidth consistently identified individual breaking events.
wind tunnelmeasurements of the percentageof breakingcrests
Here, in Part2, statisticsof the breakingeventsdetectedarecom-
[Tobaand Kunishi,1970]. For thefieldmeasurements, the white-
paredwith modelsand measurements of wave breakingby other
investigators.Our analysisis basedon measurements from the
cap coverage wasproportional to Re•. '5, whereaswe have
estimatedanRe. exponentof 1.4 for theirlaboratorydata.
SAXON-CLT experiment(SyntheticApertureRadar and X-Band
OceanNonlinearities
experiment
at theChesapeake
LightTower) A cubic
dependence
onu. isequivalent
toaRe•.
'5dependence
which occurred in the fall of 1988. Most field measurements of if we assumethe fetch-dependent
scalingrelationsuggested
by
wave breakingare basedon the fraction of affectedsurfacearea,
Kitaigorodskii
[1970]andsummarized
byPhillips[1977]:
referred to as the whitecap coverage [Monahah, 1969, 1971; 1

Monahan and O'Muircheartaigh, 1986; Toba and Chaen, 1973] O•U.

- 2.2 (2)
(see also Wu [1988], Monahah and Woolf [1989], and Wu g
[1989]), which is expectedto vary cubically with friction velo-
city [Wu, 1979]. An alternativemeasurefor the degreeof wave
wheretOou./gis the dimensionless
frequencyof the waveheight
breaking is the percentageof breaking crestspassinga fixed
spectral
peak and xg/u.
2 is the dimensionless
fetch. The
point.Holthuijsen andHerbers [1986] andToba etal.[1971]wavelength
usedsimilar techniquesto count the numberof breakingcrests
Xocorresponding
tothepeak
frequency
ofthesur-
facedisplacement
spectrum
is proportional
to the frictionvelo-
identifiedvisually by an observer.Longuet-Higginsand Smith
city u.:
[1983] andThorpeand Humphries[1980] usedwire wavegauges
1
to detectabruptchangesin sur.
face elevation attributedto break-
5 x 2 u.
ing waves. Weissmanet al. [1984] correlatedincreasesin the (3)
g
energy of high-frequency waves with visual observationsof
breaking.
Toba and colleagueshave proposedthat the dependenceof whentooin (2) is given by the deepwaterdispersionrelation. If
wave breaking on the combinedeffectsof wind stressand wave Xois usedfor thelengthscaleL, thenRe. isproportional
to u.2.
conditionscanbe parameterizedby a Reynoldsnumber: Thus the Re. exponentof 1.5 estimatedby Toba and Chaen
[1973] for the fraction of breakingcrestsis consistentwith the
u.L cubic dependenceon friction velocity found for other measures
Re. - (1)
v of wave breaking. Note that the Re. basedon Xo in (3) incor-
poratesthe friction velocity and fetch dependenceinto a single
parameter.Thusa formulationin termsof Re. effectivelyrelaxes
•Presently
atthe
Applied
Physics
Laboratory,
Universit),
ton, Seattle, Washington. ofWashing-
the
constraint
ofalinear
dependence
ofLonu..
In Part 1, we addressedthe validity of using the frequencyof
Copyright
1991bytheAmerican
Geophysical
Union. seaspikes
caused
by individual
breaking
wavesasa measure
of
Paper
number
91JC01994. thedegree
of wavebreaking.
Phillips
[1988]predicted
a cubic
0148-0227/91/91JC-01994505.00 dependenceon friction velocity for both the frequencyof sea
20,561
20,562 JESSUP
ETAL.:BREAKING
WAVES
A•'ECT•OMICROWAVE
BACKSCATI'ER,
2

spikesand their contributionto the meanradar crosssection. TABLE1. LeastSquares,

Orthogonal
Regression
Coefficients
Jessupet al. [1990]presentedthefirstquantitative
analysis
of sea forPowerLawsof FormY= CiXai ListedWith95%
ConfidenceLimits andCorrelationCoefficientsPi
spikesassociated with breakingwavesover a rangeof environ-
mental conditions. Sea spikeslikely to be associatedwith large-
Relation Wave Conditions ai log Ci Pi
scalebreakingeventswereidentifiedby a crosssectionthreshold
basedon Dopplerfrequencyinformationandpartiallyverifiedby Nvs.u. unimodal 2.9+0.6 1.1 0.91
videorecordings.The dependenceof the seaspikefrequencyon
Pssvs.Re. all 1.4+0.2 -7.0 0.94
friction velocity and the contributionof sea spikesto the mean unimodal 1.2+0.3 -5.9 0.84
radarcrosssectionweregivenfor datacollectedin theNorth Sea.
•J•,vvs.u. all 2.0+0.1 -2.1 0.98
The resultsfor that limited data setwere consistentwith Phillips'
[1988] model.
•a vs.u. all 1.9+ 0.1 -2.2 0.96
Melville et al. [1988] and Loewen and Melville [1991] meas- •søs
(VV)vs.u. unimodal 3.2+0.7 -3.3 0.89
ured microwave scatteringand soundgeneratedby breaking •jo
ss(HH)vs.u. unimodal 3.2+ 0.8 -3.3 0.87
wavescontrolledin a laboratory.The dissipationdue to breaking
correlatedalmostlinearly with boththebackscattered microwave
powerandthe radiatedacousticpower. Theselaboratoryresults
and Phillips' [1988] analyticalmodel suggestthat information tion, thepower-lawrelationsare expressed
in termsof his dimen-
about the breaking process itself may be inferred from sionless
frictionvelocity•. = [u.2rdg]
v', where•: is the elec-
tromagneticwave number and g is the accelerationof gravity.
microwave measurementsof breaking waves. If their results
apply to field measurements, The plotsof the computedquantitiesversusfriction velocity have
then microwaveobservationsof
breakingwavesmayyield importantdynamicandacousticinfor- two x-axis scales: the top scale is the dimensionalfriction velo-
mation on the wave field and the upperocean. cityu. (cms-1), andthelowerscaleis thedimensionless friction
velocity•..
2. RESULTS AND DISCUSSION
2.1. Dependenceon Wind and WaveConditions
The results discussed in Part 1 showed that a microwave
scatterometer cannot detect all waves that break near the
The log-logplot in Figure la showsthe dependence of the sea
spike frequencyN on friction velocity for the entire data set
illuminated area. However, the number of breaking waves
describedin Part 1. Although N tends to increasewith u., the
detectedusingbandwidthinformationwasapproximatelypropor-
amountof scattersuggeststhat a straight-linefit would be inap-
tional to the total numberof seaspikescausedby breakingwaves
propriate. Note, however, that the scatter is characterizedby
(see Table 2, Part 1). Therefore if the number of sea spikes
clusteringaccordingto run number. The scatteris significantly
causedby breakingwavesis a valid indicatorof the degreeof
reducedby excludingdata with bimodal wave height spectra,as
wave breaking, our methodprovidesa relative measureof this
is shownin Figure1b. In this case,a linear fit to the data seems
phenomenon.
reasonable; for the relation
In Part 1 we tested four detection schemes based on radar cross
sectionor bandwidthmaxima (or both) corresponding to indivi-
dual waves. The resultsin this paper are basedon the scheme N = C• i•., (4)
usingboth the radar crosssectionandbandwidthmaxima,which
detected
thehighest
percentage
of seaspikes
caused
bybreakingtheexponent
a• (with95%confidence
limits)is2.9+0.6,andthe
waves.
Theediteddatasetconsisted
of thirty-eight
1-hourcorrelation
coefficient
p• is 0.91.Although
these
datacover
a
records covering various sea conditions(see section 2.3 and limited range of friction velocities and have relatively large
Table 1 in Part 1). The data were divided into seven subsets, confidencelimits, they supportPhillips' [1988] predictionof a
called runs, made up of 1-hour recordsoccurringclose in time. cubic friction velocity dependencefor the frequency of sea
We described the wave conditions for each run in terms of the spikes.The resultsare comparableto our measurements from the
numberandlocationof major peaksin the wave heightspectrum. North Sea [Jessup et al., 1990], which had friction velocity
Someof the spectrawere unimodaland somebimodal. Unimo- exponents between 2.7 and 3.6 and correlation coefficients
dal spectraindicateseasdominatedby eitherswell (runs1 and6) between0.84 and 0.91. The decreasein scatterfrom Figure 1a to
or wind waves (runs 11 and 12), whereasbimodal spectraindi- Figurelb suggeststhat friction velocity alone does not ade-
cate mixed seas(runs4, 5, and 9). Bimodal spectrasuggestthe quatelyparameterizewave breaking.
possibilityof breakingdue to the interactionof swell and wind The roughnessReynoldsnumberRe. incorporatesboth wind
waves. and wave conditionsinto a single parameter. The Re. depen-
To help identify variations that might be caused by the denceof the percentageof wave crestsproducingseaspikes,Psi,
differing wave conditions,each of the sevenrunsis identifiedby is shownfor the entire data setin Figure2a andfor runswith uni-
a differentsymbolin the plotsin thispaper. Someof theseresults modalspectrain Figure2b. The characteristic wavelengthL used
are presentedon log-log plots fitted with a straightline illustrat- in computingRe. correspondedto the peak frequencyof the sur-
ing a power law relation. These lines are the result of a least face displacementspectrumwhen computedusingthe dispersion
squares,orthogonalregressionanalysis,following the method relation. When the surfacedisplacementspectrumwas bimodal,
describedby Casella and Berger [1990]. Orthogonalregression the higher-frequencypeak was used. In contrastto Figure1, the
is more appropriatethan linear regressionwhen both variables scatterin Figure2 is not significantlyaffectedby excludingruns
involved contain random errors. The analysis provides an with bimodalspectra.The expression
exponenta which is a measureof the slope of the line, 95%
confidenceintervalson that slope,and a correlationcoefficientp. P• = C2Re*•2 (5)
The regressionparametersfor the detection schemeused are
listedin Table 1. For comparisonwith Phillips' [1988] formula- yields an exponenta 2 of 1.4+0.2 (with 95% confidencelimits)
 JESSUP
ETAL.' BREAKING
WAVESAFFECTIN6
MICROWAVE
BACKSCATI'ER,
2 20,563

u,/cm S-1 correlation for the entire data set is not obvious and simply may
be a result of the larger rangeof Re. in Figure2a. Although the
20 30 40 50
correlation coefficients for the regressionson u. and Re. are
103 I
i
comparable, the greater dynamic range and improvement in
•:1 (a) - confidence limits for Re. suggest Re. is better suited for
X:q parameterizingwave breakingthan is u..
5-
_
<>:5 Previousauthorshave reportedthe wind dependenceof Pssin
z•:6 terms of U•0, the wind speedreferencedto a height of 10 m.
Y:9 They basedthe total number of wave crestsused in computing
_ Ps, either on zero crossingsin surfacedisplacementmeasure-
•:11
ments [Holthuijsen and Herbers, 1986] or on an average wave
•:12
period [Toba et al., 1971]. For our data, we define the total
• 102- number of wave crests as the record length (in seconds)times
z -
_

fp•k, thefrequency
(in hertz)of thepeakof thesurface
displace-
ment spectrum. When the surface displacementspectrumwas

102
_ _

_
(D:i (a)
5-
_

<3>:5 -
101
z•:6 -
10o
Y:9
2-
)•:11 X:• mm -
u, / cm S-1
20 30 40 50 • 101-
X:12 •• •(!) •
rJ• - xy
103 --
I
i 13_ -

:_
_

(b) - 5-
z•:6
_

5-

10o I I I I I I • I
• 102 - 102 I I
Z - _

5- z•:6

_ _

101 I I
½ 101-
10o 2 3 13_ -
_

5-
Fig.1. Frequency
of seaspikes,N (numberperhour),versusfriction _

velocity u. for (a) all wave conditionsand (b) unimodal wave condi-
tions. The slopeof the orthogonalregression line (with 95% confidence
intervals)for Figure lb is 2.9_+0.6with a correlationcoefficientof 0.91
(seeequation4 and Table1). For this and similar plots, each symbol _
corresponds to 1 hour of data, with different symbolsfor each run (see
Table 1, Part 1).
10o
106
for the datain Figure2a and 1.2+ 0.3 for thosein Figure2b, with Re.
corresponding correlationcoefficients
P2 of 0.94 and0.84. These
exponentscomparereasonablywell with the exponentsof 1.4 and Fig.2. Percentage
of crests
producing seaspikes,Pss,versustherough-
1.5 found by Toba and Kunishi [1970] and Toba and Chaen nessReynoldsnumberRe. for (a)all waveconditions and(b)unimodal
waveconditions.The correspondingslopesof the orthogonalregression
[1973]. Note thatalthoughtheReynoldsnumberincludesdepen- lines (with 95% confidenceintervals) and correlation coefficientsare
denceon waveconditions,it doesnot reflectthenumberof peaks 1.4___
0.2 and 0.94 in Figure2a and 1.2-+0.3 and 0.84 in Figure2b (see
in the wave height spectrum. Thus the reason for the better equation(5) andTable 1).
20,564 JESSUPET
AL.'BREAKING
WAVES
AFF•INO MICROWAVE
BACKSCATTER,
2

4O u, / cm S'1
O Holthuijsen& Herbers[1986]
[] Tobaet al. [1971] 5
I I I I I
101
I
2
I I I
5
ß SAXON-CLT -10
3O

ß ø
-15 - VV
POLARIZATIO
•x ' _

x
-20 - x x -
•EL• •0' OO&

)4
)4• _
-25 -
I I
0 5 10 15

Ulo/ rns'1
-30 -
Fig. 3. Comparison of resultsfor SAXON-CLT measurement (solidtri-
angles)with thoseof Holthuijsenand Herbers[1986] (opencircles)and
Toba et al. [1971] (open squares)as a percentageof breakingcrests
versusU •0, thewindspeedreferenced to 10m. -35 I i i
'1 i

2
, , • ' 100 ;•
bimodal,thehigher-frequency
peakwasusedto determine
fp•k. u. / cm S-1
In Figure3, we compare our measurementswith the results
reported by Holthuijsen and Herbers [1986] and Toba et al. 101
I
2
i i i
5
i
-10 '
[1971]. The previouslypublisheddata (open symbols)and the
HH POLARIZATION
SAXON-CLT data (solid triangles)are plotted on linear axesas
the percentageof breakingcrestsP,, versusU10. Consideringthe
-15 - _

inherent differences and uncertainties in the three measurements,

we concludethat the agreementis reasonable.

2.2. Mean and Sea SpikeRadar CrossSections -20 -

xX -

The meanradarcrosssectionc5
ø versusfrictionvelocityis
shown in Figure4 for VV and HH polarization. The angle ½
between the radar look direction and the wind direction ranged
from 0 ø to 25 ø for the SAXON-CLT measurements. The
-25 -

x
_
expected
variation
of c5
ø with½for ½in thisrangeis lessthan •x _
-30 -
1 dB, according to the Seasat-A satellite scatterometermodel x

SASSII [Wentzet al., 1984]. The calibrationaccuracyof the

scatterometer
usedin the SAXON-CLT experimentis estimated / (b)
to be+l dB. -35 i i i I i

For
2
, , .• ' 100
~a 3
C7øvv
= C3u. (6)
Fig. 4. Meannormalizedradarcrosssectionversusfrictionvelocityfor
and (a) VV and (b) HH polarization.Eachsymbolrepresents a l-hour aver-
age. For VV polarizationthe slopeof the orthogonalregression line
~a 4
oø. = C• u. (7) (with 95% confidenceintervals)is 2.0_+0.1andcorrelationcoefficient0
= 0.98;for HH polarization
theslopeis 1.9+ 0.1 andp = 0.96 (seeequa-
a 3 (with 95% confidencelimits) and P3 are 2.0+0.1 and 0.98, tions (6)and (7)).
respectively,andan andP4 are 1.9+0.1 and0.96.
We previouslyoutlinedtwo methodsof computingthe totalsea 0 ~a6
(Jss= C6 u. (9)
spikecontribution[Jessupet al., 1990] which associated
the indi-
vidual seaspikecontributionswith the areaunderthespikeabove the exponents(with 95% confidencelimits) are a5 = 3.2-+0.7 and
a specificvalue. For method1, the lower boundwasc•ø, the
a6 = 3.2+ 0.8; thecorrelation
coefficients
areP5= 0.89andP6=
long-term mean radar cross section; for method2 it was the
0.87. TheseresultssupportPhillips' [1988] predictionand are
smallerof the local minimaon eithersideof the seaspikepeak. comparable to results reported by Jessup et al. [1990] of
Ingeneral,
thecontribution
givenbymethod2 isslightly
greater
exponents
between
3.3and3.8 andcorrelation
coefficients
thanthatgivenbymethod1. Becausethefrictionvelocity
between
0.88and
0.93.Note
that
thevalues
forc5,ø•
areroughly
dependence
iscomparable
using either
method, onlytheresults
equal
forVVand HHpolarization,
whichisconsistent
withasea
for method2 arepresentedhere. spikepolarizationratio of unity.
Figure5 showstheseaspikecontribution
C5sø•
versus
friction The roughlycubicdependence
0
on frictionvelocityof botheL,,
velocityfor unimodalwaveconditions.For VV polarization, the seaspikecontributionto the meanradarcrosssection,andN,
~a 5
Osøs
= Csu. (8) theseaspikefrequency,impliesthatthesetwo quantifiesmaybe
linearlyrelated. Sucha linearrelationshipsuggests thattheaver-
and for HH polarization, ageradarcross section of anindividual seaspike,cs,,/N,
0 might be
 JESSUP
ETAL.'BREAKNO
WAVESAFFECTNO
MICROWAVE
BACKSCATI'ER,
2 20,565

u,/cm S-1 the centroidsof the distributions,corresponding

to the average
20 30 40 50 individualcrosssections,do not vary significantlywith friction
-20 I
i
I I velocity.
• ' 0 is of interest because current
The relative importanceof c5,
(a)
(!):1 models of microwave backscatter from the ocean surface
• :6 generallydo not includereturnsfrom breakingevents[Donelan
•:11 and Pierson, 1987]. Figure8 showsthe dependenceof the frac-
5• :12
tionalseaspikeradarcross section0 0 on friction velocity for
c5,,/c5
VV andHH polarizationfor all waveconditions.For HH polari-
zation the detectedsea spikescontributealmost 25% of the
received power, whereasfor VV polarizationthey contribute
o
only about15%. (The maximumcontributions calculatedusing
o
method1 are approximately5% lessthanthesecalculatedusing
method2.) A greaterfractionalsea spikecontributionfor HH

u. / cm S-1
20 30 40 50
-4o I I I I
VV POLARIZATION (a)
VV POLARIZATION

-30 I

-42 -
10o

u,/cm s-1
20 30 40 50
-20
I I
i
-44 -

(b) 0'1
0:1
X :4
• :6
<3> :5
•:11 -46 -
• :6
5• :12 Y :9
•:11
X :12
-48 i i
1 2

u, / cm S-1
20 3O 4O 5O

I I I
(b)
HH POLARIZATION
HH POLARIZATION

-30 -42 -

10o
o
Fig. 5. Sea-spikecontributionOs, versusfrictionvelocityfor unimodal
wavedatafor (a) VV and(b) HH polarization.The slopesof the orthog- -44 -
onal regression
lines (with 95% confidenceintervals)andthe correlation
coefficients are 3.2 +_0.7 and 0.89 and 3.2 +_0.8 and 0.87 for VV and HH 0:1
polarization,respectively(seeequations(8)and (9)and Table1). X :4
4:> :5
-46 - • :6
Y :9
independentof friction velocity [Jessupet al., 1990]. The quan-
•:11
tity cs•ø,/N
for the SAXON-CLTmeasurements
is plottedas a
functionof friction velocity in Figure6 for VV and HH polariza- • :12
tions. The scatterof approximately+1 dB indicatesthat the aver- -48 I I I
agecontributionof an individualseaspikeis independent of u.. 1 2 3
Thissuggests
thattheprobability
density
function
p (c5•ø,)
may d,
alsobeindependent
of u.. Figure7 shows
linearplotsof p(cs•ø,)
Fig.6. Average
seaspikecontribution
O•ø•/N
versus
friction
velocity
for VV and HH polarization. The distributionsinclude all runs for (a) VV and (b) HH polarizationfor all datafrom Part1. The scatter
and cover friction velocities of 25 to 45 cm s-1 in 5 cm s-1 inter- of approximately_+1dB indicatesthat the averagecontributionof an
vals. The changesin the distributionsare relatively small, and individualseaspikeis independentof frictionvelocity.
20,566 JESSUP
ETAL.:BREAKING
WAVESAFFECTINO
MICROWAVE
BACKSCA'ITER,
2

0.04 As is shownin Part 1, maxima in the mean Doppler frequency

•25<u*<:•0:__ tend to be correlatedwith large spikesin the radar crosssection
(a) attributedto breakingwaves. The averageof the maximum mean
30<u*<35:
Dopplervelocity(i.e., the maximumfirstmome_nt of theDoppler
35<u*<40: spectrum)correspondingto detectedseaspikesUm•,,is given by

40<u*<45: - 1 N

VV POLARIZATION
0.02
where Um•,, is the line-of-sight velocity correspondingto the
maximummean Doppler frequencyassociatedwith a sea spike.
The scalingof this averagevelocity associated with the detected
breakingeventsmay be relevantto the kinematicsof the break-

u, / cm S-1
2O 3O 4O 5O
I I

O'1 VV POLARIZATION (a)

•25<u*<310'__ X :4
(b)
30<u*<35:,• <3> :5
• :6
35<u*<40: Y :9
40<u*<45: •:11
X :12

0.02
HH POLARIZATION 0.2-

0
I I I
0.04 •! 1
0 1 2
u, / cm S-1
Osøs
x 108 2O 30 40 5O
o
0.4
Fig.7. Probability densityfunctions forOssfor(a)VVand(b)]Hit
polarization covering the friction velocity range 25-45cm s- in
O'1 HH POLARIZATION (b)
5 cms-1 intervals
forthedataused
in Figure
6. Theinvariance
of the
distributionsis consistentwith Figure6. X :4
<3> :5
• :6
Y :9
polarizationthanfor VV polarizationis consistent
with a polari-
•:11
zationratio thatis greaterthanunityin themeanbut nearunity
for individualseaspikesassociated with breaking.Theseresults
agree with previous maximum contributionsof 10% and 20% 0.2- X '12
ø•Iøt::> ••,•
calculatedusingmethod1 for VV and HH polarization,respec-
tively [Jessupet al., 1990].
0 }X yYYy
2.3. Kinematicsof the BreakingProcess
When the illuminated area is small comparedwith the
wavelengthof thedominantsurfacewave,theDopplerfrequency
providesa measureof the seasurfacevelocity. For nonbreaking
0 I I I
waves,the measuredvelocity is dominatedby the line-of-sight 1 2 3
orbitalvelocityof the long surfacewaves. For breakingwaves
we expectthat the fluid-particlevelocityat the crestis compar-
ableto the phasespeed[LoewenandMelville, 1991]. This sug- o o
Fig.8. FractionalradarcrosssectionOsslO versusfrictionvelocityfor
geststhat the phasespeedof the dominantsurfacewave may be (a) VV and (b) HH polarization.The detectedseaspikescontribute
an appropriateparmeter for scalingDopplervelocitymeasure- nearly15%and25% of thecrosssection for VV andHH polarization,
mentsof breakingwaves. respectively.
 JESSUP
ETAL.:BREAKING
WAVES
AFFECTING
MICROWAVE
BACKSCATI•R,
2 20,567

ing process.If the averagevelocity is dominatedby the large I I

velocitieswithin the whitecapregion,it may indicatethe scaleof

the breaking. Sincewe expectthat the velocitynear the crestof a
breakingwave is nearly horizontal,we have assumedthat Uma,,is
90
• , , ,
due to a completely horizontal velocity. Furthermore,since the m

crestvelocityof a breakingwaveisexpectedto be comparable

to
its phasespeed,we normalized
U•,, by Cp•ak,thephasespeed
corresponding
to the peak of the surfacedisplacementspectrum.
When the surface displacementspectrum was bimodal, the
higher-frequency
peak_was usedtodetermineCp•ak.
Thedependence
of U•,,/Cp•akonfrictionvelocityis shownin 30 I I I I
Figure9 for all wave conditions. The data show a moderate
-500 0 500
_amount
of scatter,with the values generally clusterednear
Um•,,/Cp•k
= 0.25. Considering
thatthecrestvelocityof a break- f/Hz
ing waveshouldb_eroughlyequalto its phasespeed,_wemight Fig.10. Dopplerspectrum
S(f) showingthe signature
of a fast-
expecta valueof Um•,,/Cp•ak
closertounity.However,
Uma,,
can- movingsplashcausedby a breakingwave.The frequencyof the fight-
not be interpretedas the averagevelocity of the breakingcrests. mostpeaknear400 Hz corresponds
to a line-of-sightvelocityof roughly
Becausethe Doppler spectrumreflectsthepower-weighteddistri- 4ms-1. Thesplash hasa negligible effect onthemeanDoppler fre-
bution of scatterer velocities within the illuminated area, the quency,which is less than 100 Hz and is indicatedby an X on the
abscissa.
influenceof the whitecapvelocityon the meanfrequency,or first
moment, depends on the location of the breaking crest with
respect
to the illuminated
area.Otherfactorsaffectingtherela- This spectrum includesthe Dopplersignature
of a fast-moving
tionshipbetweenthe first momentand the whitecapvelocity splashthat occurredwhenthe crestof a breakingwave was
includethe crest'sorientationand the targetstrengthof the thrownforwardanddownintoitsforwardface.Theline-of-sight
breakingregionrelativeto that of othersourcesof scattering.velocitycorresponding
to thefrequencyof thesmallpeakon the
ThusUm•,,
doesnotnecessarily
reflect
thewhitecap
velocity. rightisroughly4 m s-1. Thistranslates
intoa horizontal
velocity
Examples ofDoppler
spectra
fortimesofmaximum bandwidthof 5.7m s-1, whichis nearthephasespeed of roughly
8 m s-•
(Figures9 and 10 in Part1) showedlargevelocitiesassociatedbasedon the peak frequencyof the wave height spectrum.
with thecrestregionof breakingwaves.Becauseof thespectra's Becausetherelativetargetstrengthof the splashis solow, it has
large'bandwidth,
however,the meanDopplerfrequency
was a negligibleeffecton the meanDopplervelocity,whichis
significantlylessthan the maximumDopplerfrequency.The approximately 0.7 ms-'.
Dopplerspectrum in Figure10 of thispaperdramatically
illus- Thebandwidth of theDopplerspectrum indicatestherangeof
tratesthatthewhitecapvelocity,although measured
by thescat- scatterer
velocitieswithintheilluminatedarea. In Part1, large
terometer,may not be reflectedin the mean Dopplerfrequency. jumps in the bandwidthwere found to be associatedwith the

u,/cm s-1 U, / cm s-1

20
30 40 50 20 30 40 50
1.0 I I I 1.0 I
X:•t

•:6
Y:9
•:tl xX Y
N

(D:l

4>:5
•:6
Y:9
•:il
I I I X:i2
1 2 3
I I I
d.
1 2 3
Fig.9. AveragemaximummeanDopplervelocityUma,•(seeequation
(10))associated
withdetected
seaspikes
normalized
byCp•k,thephase
speed_•orrespondingto the peakof the surfacedisplacement spectrum. Fig. 11. Averagenormalizedbandwidth(seeequations(11) and (12))
ThatUm•xis about
25%of Cp•kindicates
thatthemean
Doppler
fre- associatedwith detectedsea spikes. The valuesindicatethat, on aver-
quencyof detectedbreakingeventsis not dominatedby the largeveloci- age, the maximum bandwidth associatedwith the detected events is
tiesassociated
with whitecaps. roughly50%-75% of themaximummeanDopplerfrequency.
20,568 JESSUP
ETAL.'BREAKING
WAVES
AFFECTINO
MICROWAVE
BACKSCATtER,
2

crestof breakingwaves.We computed bandwidth associated

a normalized with whitecaps.A normalizedbandwidthassociated
B,,to investigate
thescaling maximaassociatedwith the detectedeventsrangedbetween0.50 and 0.75 and
of thebandwidth
withbreaking waves.Foreacheventdetected, the showedno functionaldependence
wenormalized on frictionvelocity.
bandwidth maximum by the correspondingmaximum mean In conclusion,our investigationshowsthat the microwave
Dopplerfrequency
Fma,,associated
withtheevent: detectionof individual wave breakingeventscan be usedas a
meaningful
measure
of thedegreeof wavebreaking.
B lnax
B,• - (11)
/•'max Acknowledgments.In additionto the help acknowledged
in Part1, we
thank the anonymousreviewerwho pointedout the appropriateness of
For each 1-hourrecord,the averagenormalizedbandwidthB,• of orthogonalregressionover linear regression.
the detected events is then
REFERENCES

B mltx Casella, G., and R.L. Berger, Statistical Inference, p. 584, Wadsworth
(12) and Brooks/Cole Advanced Books and Software, Pacific Grove, Calif.,
1990.
Donelan, M., and W.J. Pierson,Jr., Radar scatteringand equilibrium
rangesin wind-generatedwaves with applicationto scatterometry,
whereN is the tot_al numberof eventsdetected.Figure11 shows J. Geophys.Res.,92(C5), 4971-5092, 1987.
the variation of B,• with friction velocity u. for all wave condi- Holthuijsen, L.H., and T.H.C. Herbers, Statisticsof breaking waves
tions. On average,the maximum bandwidthassociatedwith the observedas whitecapsin the open sea, J. Phys. Oceanogr.,16,
detected events is roughly 50-75% of the maximum mean 290-297, 1986.
Doppler frequency. The average (dimensional) bandwidths Jessup,A. T, W.C. Keller, and W.K. Melville, Measurements of sea
spikesin microwavebackscatterat moderateincidence,J. Geophys.
corresponding to the sea-spike maxima were 60-70 Hz. Res.,95(C6), 9679-9688, 1990.
Degradedazimuthalresolutionin syntheticapertureradarimages Jessup,A. T, W. K. Melville, andW.C. Keller, Breakingwavesaffecting
has been associated with localized scatterer coherence times of microwavebackscatter,1, Detectionand verification,J. Geophys.Res.,
this issue.
theorderof 10-2 s [Lyzenga
andShuchman,
1983].Therecipro-
cal of the average bandwidth associatedwith the detectedsea Kitaigorodskii,S. A., Fizika VzaimodestviyaAtmosferiI Okeana(Physics
of Air-Sea Interaction),Gidrometeorologicheskoe Izdatel'stvo,Len-
spikesis of this sameorder. ingrad, 1970. (Englishtranslation,Israel Programfor ScientificTrans-
lation, Jerusalem,1973.)
3. SUMMARY AND CONCLUSION Loewen, M.R., and W.K. Melville, Microwave backscatterand acoustic
radiationfrom breakingwaves,J. Fluid Mech.,224,601-623, 1991.
Thirty-eight hoursof data from the SAXON-CLT experiment Longuet-Higgins,M.S., and N.D. Smith, Measurementof breaking
were analyzedusing the microwave techniqueoutlined in Part 1 wavesby a surfacejump meter,J. Geophys.Res., 88, 9823-9831,
1983.
for detectingbreaking events. The dependenceof the detected
Lyzenga,D. R., andR. A. Shuchman,
Analysisof scattermotioneffectsin
events on wind and wave conditions was compared with the MarsenX bandSAR imagery,J. Geophys.Res.,88, 9769-9775, 1983.
dependence calculatedfor othermeasurements of wavebreaking Melville, W.K., M.R. Loewen,F.C. Felizardo,A.T. Jessup,and M.J.
and with analyticalmodels. The contributionof seaspikesto the Buckingham,Acousticand microwavesignatureof breakingwaves,
mean radar crosssectionwas computed. Finally, measurements Nature, 336, 54-59, 1988.
of the Dopplerfrequencyandbandwidthwereusedto investigate Monahan, E.C., Laboratorycomparisonof freshwaterand saltwater
whitecaps,
J. Geophys.
Res.,74,6961-6966, 1969.
the kinematicsof the breakingprocess. Monahan,E.C., Oceanicwhitecaps,J. Phys. Oceanogr.,1, 139-144,
The frequencyof seaspikes,N, wascomputedasthenumberof 1971.
eventsin a 1-hourrecord.The roughlycubicdependence of N on Monahan,E.C., and I.G. O'Muircheartaigh,Whitecapsand the passive
frictionvelocityfor unimodalwave conditionssupports
Phillips' remotesensing
of the oceansurface,
lnt. J. RemoteSens.,7, 627-642,
1986.
[1988] model and is consistentwith theoreticalmodeling and Monahan,E.C., and D.K. Woolf, Commenton "Variationsof whitecap
field measurements of whitecapcoverage.The dependenceof the coveragewith wind stressandwatertemperature,"J. Phys.Oceanogr.,
percentageof breakingcrestsproducingsea spikes,Pss,on a 19, 706-709, 1989.
roughness ReynoldsnumberRe. wasrelativelyinsensitiveto the Phillips,O.M., TheDynamicsof the UpperOcean,p. 161,Cambridge
shapeof the wave height spectrumand was consistentwith UniversityPress,New York, 1977.
Phillips,O.M., Radarreturns
fromtheseasurface--Bragg scattering
and
laboratoryandfield measurements. An Re. exponentof 1.5 was breakingwaves,J. Phys.Oceanogr.,
18, 1065-1074, 1988.
shownto be equivalentto a cubic friction velocity dependence Thorpe,S.A., andP.N. Humphdes, Bubblesandbreakingwaves,Nature,
usinga fetch-dependentscalinglaw. 283, 463-465, 1980.
The dependence on friction velocityof the sea-spikecontribu- Toba, Y., Local balancein the air-sea boundaryprocesses,1, On the
tion osø•
to the meanradarcrosssectionalsosupports
Phillips' growthprocess
of wind waves,J. Oceanogr.
Soc.Jpn.,28, 109-121,
1972.
[1988] model. The fractionalpower for high friction velocities Toba,Y., andM. Chaen,Quantitativeexpressionof the breakingof wind
(40-50 cms-•) wasbetween10% and 15% for VV polarization waveson the seasurface,Rec.Oceanogr.Work. l' Jpn.,12, 2-11, 1973.
and between20% and 25% for HH polarization. These findings Toba, Y., and H. Kunishi, Breaking of wind waves and the sea surface
supportthe inclusion of wave breaking in scatteringmodels, wind stress,J. Oceanogr.Soc.Jpn.,26, 71-80, 1970.,
Toba, Y., H. Kunishi, K. Nishi, S. Kawai, Y. Shimada, and N. Shibata,
especiallyfor HH polarizationandhigh frictionvelocities.The Studyon air-seaboundaryprocesses at the ShirahamaOceanographic
dataalsosuggest thattheaverageradarcrosssectionof an indivi- TowerStation,DisasterInstitute,(in Japanese with Englishabstract)
dual seaspikedoesnot dependon u.. KyotoUniv.Ann.,148,519-531, 1971.
The averageof the maximummean Dopplervelocitiesassoci- Weissman,M. A., S.S. Atakturk, and K.B. Katsaros,Detectionof break-
ated with the detectedsea spikeswasroughly25% of the phase ing eventsin a wind-generatedwave field, J. Phys. Oceanogr.,14,
1608-1619, 1984.
speedcorrespondingto the peak of the surfacedisplacement Wentz,F.J.,S. Peteherych,
andL. A. Thomas,A modelfunctionfor ocean
spectrum.This indicatesthat the mean Doppler frequencyof radarcrosssectionsat 14.6 GHz, J. Geophys.Res.,89(C3), 3689-3704,
detectedbreakingeventsis not dominatedby the large velocities 1984.
 JESSUP
ETAL.:
BREAKING
WAVESA•Ecrn•MICROWAVE
BACKSCATtER,
2 20,569

Wu,J., Oceanic
whitecaps
andseastate,
J.Phys.Oceanogr.,
9, Fishe,ry
Sciences,
University
ofWashington,
1013NE40thStreet,
Seat-
1064-1068, 1979. fie, WA 98 ! 05.
Wu,J.,Variations
of whitecap
coverage
withwindstress
andwatertem- W.C.Keller,
U.S,NavalResearch
Laboratory,
Washington,
DC20375.
perature,
J. Phys.Oceanogr.,
18, 1448-1453,1988. W.K. Melville,R.M. Parsons
Laboratory,
Massachusetts
Institute
of
Wu,J.,Reply,
J.Phys.
Oceanogr.,
19,710-711,1989. Technology,
Cambridge: MA02139.
(Received
January
3, 1991;
revisedJuly29, 1991;
A.T. Jessup,
AppliedPhysicsLaboratory,
Collegeof Oceanand acceptedJuly29, 19.91.)