JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 95, NO.

C6, PAGES 9679-9688, JUNE 15, 1990

Measurements of Sea Spikes in Microwave Backscatter

at Moderate Incidence

A. T. JESSU},

Ralph M. Parsons Laboratory, MassachusettsInstitute of Technology, Cambridge

W. C. KELLER

U.S. Naval Research Laboratory, Washington, D.C.

W. K. MELVILLE

Ralph M. Parsons Laboratory, MassachusettsInstitute of Technology, Cambridge

Previous authors have associatedlarge events, or "sea spikes", in the radar cross section of the sea
surface with steep or breaking waves. Here we present a method based on an intensity threshold and
the mean Doppler frequency to identify and quantify sea spikeslikely to be associatedwith large-scale
wave breaking. This method is applied to microwave measurementsof K u band made with a
dual-polarized Doppler scatterometerat a moderate incidence angle (45ø) from a fixed platform in the
NorthSea.For VV andHH polarization
themeannormalized
radarcrosssection
rrø, thefrequency
of sea spike occurrenceN, and the sea spike contributionto the mean normalized radar cross section,
rrsøs,
arecorrelated withdirectmeasurements of thefrictionvelocity,u,, in therange13-< u, -<49
cm s-1 The frictionvelocityexponents for bothN and rrsøs indicatean approximatelycubic
dependence, whiletheresidual radarcrosssectionrrrøes
= rrø _ rrSS is approximately
proportional to
frictionvelocityfor u. greaterthanapproximately20cms-1. At highfrictionvelocities,(u* • 40-50
cms-l) thesedatashowthatthecontribution of largeseaspikesto themeanradarcrosssection for
measurementsbetween 25øand 45ørelative to the wind is in the range of 5-10% for VV polarization and
10-20% for HH polarization.

1. INTRODUCTION Duncan et al. [1974] observed scattering at moderate

Spiky fluctuations in radar return at grazing incidence incidence angles due to wave breaking in the laboratory
angles have long been associatedwith steep and breaking which was independent of polarization and was a major
waves [Katzin, 1957; Long, 1974, 1983; Kalmykov and contributor to the return at high wind speeds. Kwoh and
Pustovoytenko, 1976] and have been generally referred to as Lake [1984, 1986] measured the relative contributions to X
"sea spikes." Simultaneousvideo recordingswere used by band microwave return from mechanically generated water
Lewis and Olin [ 1980]to show that high-amplitude sea spikes waves due to specular and nonspecular reflections. The
correspond to the development and decay of whitecaps, specular contribution was attributed to either the turbulent
while Ewell et al. [1984] tracked sea spikesthat moved with wake of breaking or the steep capillary waves generated in
a speed approximately equal to the phase speed of the the process, while the nonspecularreturn was attributed to
dominant ocean waves. wedgelike diffractionfrom the small radius crestsof breaking
Sea spikes are also evident in moderate incidence angle or near-breaking waves. More recently, Kwoh et al. [1988]
measurements and have been attributed to various mecha- presented field measurementssupporting the association of
nisms associatedwith the steep crests of breaking waves specular return with wave breaking. Banner and Fooks
[Alpers et al., 1981]. Keller et al. [1986] presented field [1985] made X band microwave measurements of stationary
observationswhich correlated Doppler spectra with break- small-scalebreaking waves generated in a flume and found
ing events, indicating that the scatterer speed during break- that high levels of backscattered power were associatedwith
ing increasedtoward the phase speedof the dominant waves the breaking region. They concluded that the measured
and that the Doppler bandwidth was greatly increased.They backscatter was consistent with Bragg scattering from hy-
suggestedthat the scatteringmay include contributionsfrom drodynamic disturbancesgeneratedjust ahead of the break-
specular surface reflections and that a coherent microwave ing crest. A number of the proposed scattering mechanisms
radar may be the ideal instrument for quantitative study of associatedwith sharp-crestedand breaking waves have been
wave breaking. A limited set of video recordings made used to improve predictions based on the composite model,
simultaneouslywith moderate incidence angle scatterometer
especially for horizontal polarization [Lyzenga et al., 1983;
measurements have shown that sea spikes are consistently Donelan and Pierson, 1987].
associatedwith visual evidence of wave breaking [Jessup,
Phillips [ 1988] has recently considered the friction velocity
1988].
dependenceof the frequency of occurrence of sea spikesand
Copyright 1990 by the American Geophysical Union. their contribution to the radar cross section. He proposes
Paper number 89JC03347. that for moderate incidence angles the mean normalized
0148-0227/90/89J C-03347 $05.00 radarcrosssectioncrø is the sumof separate
contributions
9679
9680 KELLER ET AL.: SEA SPIKES IN MICROWAVE BACKSCATTER

fromBraggscattering
tr•øandfromseaspikes
trsøs
associatedis counted correspondsto identifying breaking events asso-
with localized breaking events: ciated with wave numbers below the threshold wave number
k 1. If sea spikes are causedby scatteringfrom whitecaps,
trO=trg+ trsøs (1) then (8) is consistent with theoretical modeling which indi-
The small perturbation model for backscatteringfrom the catesthat the whitecapcoverageshouldvary as u.3 [Wu,
sea surface[Wright, 1966]was combinedwith an expression 1979] and with correspondinglylarge wind speed exponents
for the ocean gravity wave spectral density ½(k) under for various measurementsof whitecap coverage [Ross and
equilibrium conditions [Phillips, 1985]: Cardone, 1974; Wu, 1980; Monahan and O'Muircheartaigh,
1986].
½(k)= illcosqo[l/2tt,g-l/2k
-7/2 (2) Most recently, Melville et al. [1988] have reported mea-
where fl is a constant,k is the ocean wave number, qois the surementsof microwave scattering and sound generation by
angle between the wind and wave propagationdirection, and controlled breaking events in the laboratory. They found
g is the gravitational acceleration. The Bragg contribution that the dissipation due to breaking correlated almost lin-
rr•ø wasthengivenby early with both the backscatteredmicrowave power and the
radiated acoustic power. If these results carry over to the
try0
- field, microwave observationsof breaking may yield impor-
2•/•Icos
•pl
1/2
sin
1/2
0cot
40ml(O)(tl2,t(/g)
1/2(3) tant dynamic and acoustic information on the wave field and
the upper ocean.
where 0 is the incidence angle and •( is the electromagnetic In order to investigatethe statisticsof sea spikesand their
wave number. The functionFi(0) in generalwill dependon contribution to the mean normalized radar cross section,
the transmitted and received polarization as well as 0. The measurements of microwave backscatter, friction velocity,
form of •(k) given by (2) is valid for wave numbersin the and wave height made during a 2-week period in May 1987
gravity wave range whose phase speed is large compared from the German research platform NORDSEE were ana-
with u,. Thus k must satisfy the condition lyzed. The platform is located in 30 m of water approxi-
k0<<k< min{(g/y)1/2,g/u,2} (4) mately 40 n. mi (76 km) off the German peninsula in the
North Sea [see Keller eta!., 1989]. These measurements
where k0 is the wave numberof the peak of the wave spectral were performed in conjunction with the installation of a
density and 3' is the surface tension of water divided by its system for studying the long-term variation of the mean
density. This condition and the Bragg condition, radar cross section; thus the experiment was not specifically
k = 2t( sin 0 (5) designed to study the sea spikes associated with wave
breaking, and the data set is of limited duration. Neverthe-
lead to the following constrainton the electromagneticwave less, these measurementsprovide the first quantitative anal-
number:
ysis of sea spikes over a range of environmental conditions
2•( sin0 < min{(g/•,)1/2,
g/it,2} (6) by presenting the friction velocity dependence of their
frequency of occurrence and their contribution to the mean
Equation (3) predicts that the contribution from Bragg scat- normalized radar cross section. A more detailed account of
tering varies linearly with friction velocity u, for a given these results and subsequentrelated measurements from the
radar frequency, incidence angle, and azimuth angle •o. SAXON-CLT experiment is given by Jessup [1990].
Measurements of radar backscatter in the frequency range
0.4 to 9 GHz from Guinard et al. [1971] were usedby Phillips
2. EXPERIMENTAL PROCEDURE AND DATA ANALYSIS
to support the frequency and friction velocity dependenceof
(3). Strictly speaking,the form of the wave spectraldensity The microwave backscatter measurements were made
(equation (2)) is valid only in the gravity wave range. Thus using a coherent, continuous-wave,dual-polarized scatter-
the comparison of (3) with radar measurements in the ometer operatingat 14 GHz (K u band, wavelengthof 2.14
centimeter range of wavelengths may be questioned. cm) with a transmit power of 200 mW. The instrument was
On the basis of previous work concerning the energy designed, built, and calibrated at the U.S. Naval Research
dissipation due to wave breaking [Phillips, 1985], Phillips Laboratory, Washington, D.C. The instrument incorporates
derived anexpression fortrsøs,
thecontribution of seaspikes linear dual-polarized antennas which allow simultaneous
to the mean normalized radar cross section:
vertical (VV) and horizontal (HH) like-polarization mea-
trsOs • F2(O, X)[ll2,t(/g]
3/2 (7) surements. In-phase and quadrature channels at base band
for both polarizationsprovide a complex time seriesoutput.
where X is the angle between the radar look direction and the The envelope of the signal is proportional to the received
wind. Equation (7) predicts that for a given measurement power, and its rate of change of phase is the Doppler shift
geometry and radar frequency, the contributionof sea spikes frequency. The complex output schemepermits discrimina-
to trø is proportional to thefrictionvelocitycubed. tion of positive and negative target velocities. The instru-
Finally, a cubic dependenceon friction velocity was found ment was mounted 31 m above the mean sea surface, aimed
for the frequency of sea spike occurrence per unit area, with an incidence angle of 45ø and pointed northwest at
v(kl)o•g-lk14u3
, (8) 315øT.In this configurationthe measurementwas well within
the far field of the antennas,and the two-way 3-dB elliptical
wherek l is a thresholdwavenumbersignificantly largerthan illuminationarea on the sea surfacewas approximately 1.8 x
that of the spectral peak. Phillips asserts that breaking 2.5m2.Theradarwascalibrated witha swinging sphere, and
events associated with large wavelengths may be expected the absolute normalized radar cross-section values quoted
to produce more intense returns and that setting a radar here are accurate to within _+1 dB. The sea spike results are
cross section threshold of intensity above which a sea spike dependent not on the accuracy of the measurement but
 KELLERET AL.' SEASPIKESIN MICROWAVEBACKSCATTER 9681

u* - 33 cm/s

0 I I I I I I I I I I I I I I I I
0 4 8 12 16 20 24 28 52 56 40 44 48 52 56 60 64

0 4 8 12 16 20 24 28 $2 56 40 44 48 52 56 60 64
TIME (sec)

,, V
"'T ojv, , , 'v' '"' 'v'
Fig.1. Simultaneous
timeseries
(64s)ofthenormalized
radar
cross
section
&ø(t)
forVVandHHpolarization
and
meanDoppler
frequency
forVV.Theseaspike atapproximately
23s,witha large
jumpin&ø(t)coincident
witha
Dopplerfrequency
maximum
is typicalof thoseassociated
withbreaking
events.

ratheronits precision,or the relativestabilityof thecalibra- an effortto avoidconfusion,

we will usecrø to denotethe
tion over time. Experiencewith systemsof similardesign time-averaged normalized
radarcross-sectionand&ø(t)to
indicates that the relative measurement error is less than be thetimeseries fromwhichcrø is computed,
thatis,
+'0.5 dB.
Direct measurementsof friction velocity were providedby
RiscNationalLaboratory,Roskilde,Denmark,usinga sonic
anemometer(Kaijo-Denki,modelDAT-300with probehead
•ro=1LT
T•'O(t)
dt (9)

typeTR-61-B)mounted ona boomextending 20m duewest for a time series of length T.

ot the platformat a heightof 33 m abovethe seasurface. In a typicalopenoceanenvironment, 10-15%or fewerof
Wave heightmeasurementsfrom the Baylot gaugeperma- thewavecrests passing a fixedlocationmightbe breaking
nently installedon the platform were used to obtain a [Holthuijsen
andHerhers,
1986].
Thusarelativel•
longtime
phasespeedwith whichto normalizethe record must be used in order to provide enoughsamplesto
cha?acteristic
Dopplervelocity measurements. ensurethe statisticalsignificanceof an analysisbased on
•The radaroutputandwave heightdatawere recordedon counting
events.
Ontheotherhand,thetimerecordshould
a Hewlett-Packard HP3968A FM analog tape recorder not be so long that the environmentalconditionshave
(bandwidthof 625Hz), whilethe dataacquisition systemfor drasticallychanged.For our analysisa record lengthof
the sonic anemometerproduced 10-rain averagesof the 1-hourwas chosenover which to computethe averagesof
friction velocity u. in real time. The microwavesignalwas interest.Out of a totalof approximately
80 hoursof'analog
digitallysampledat 2 kHz andreducedto timeseriesof the taperecordings,
48hours
ofdatahavebeenselected
for'thi•
receivedpower and the meanDopplerfrequencywith an analysis.
averagingtime of 0.25 s. The meanDopplerfrequencywas Figure1 is an exampleof a 64-stime seriesof processed
computedusing an autocovarianceestimationtechnique data which includesa typical sea spike. The two top traces
commonlyused for Doppler weatherradar [Doviak and showthevariation
of &ø(t)ona linearscalefor VV andHH
Zrnid, 1984].Comparisonof thistechniquewith calculation polarization.The meanDopplerfrequencyshownin the
of the first momentof the Doppler spectrumusingFourier bottom trace is proportionalto the line-of-sightcomponent
transform methods showed excellent agreement. of the surfacescattervelocityaveragedover the illumination
The normalized radar cross sectionis proportionalto the area,whichis dominatedby the orbitalvelocityof the long
backscatteredpower averagedover the illuminationarea; surfacewaves [Plant and Keller, 1983]. For both polariza-
the notationcrø oftenimpliesan additional
timeaverage.In tionsthe seaspikeat approximately
23 s is characterized
by
9682 KELLER ET AL.' SEA SPIKES IN MICROWAVE BACKSCATTER

VV POLARIZATION
u* - 16 crn/s

0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640

•. TIME (sec)

ot VV
•
o
POLARIZATION u* - 31 cm/s

0,, F1-i
I J ,•-- 11-

0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 S40
•. TIME (sec)
vv P( '.ATIOt
cm/s

o
ldi,,IL,,l,
•qlWv'-rw-'•
0 I I • •
40 I I 80 20 60 200 240 280 320 I
360 I
400 I
440 I
480 I
520 I
560 I
600 640
TIME (sec)
Fig. 2. Comparison
of timeseries(10min)of radarcross-section,
•ø(t), for VV polarization
showing
thequalitative
difference in frequency of occurrenceof sea spikeswith increasingfriction velocity u,.

a large excursion in backscattered power and a local maxi-
mum in the Doppler frequency.
Comparison of the time series of radar cross section for
increasingvalues of friction velocity illustrates qualitatively
how the frequency of occurrence of sea spikes increases.
Figure 2 shows three representative time series of length 10
min of VV radar cross-section data for different values of u,.
Thetoptraceisfor a lowvalueof 16cms-•, for whichlittle
or no breaking would be expected. In addition to an increase
in the mean radar cross section, the number of sea spikesis
increased in the second and third traces, for which the function of thepeakvalueof 6•ø(t)for twodifferentfriction
frictionvelocityis 33 cm s-• and46 cm s-• respectively. velocities are shown in Figure 3. The Doppler frequency is
Wave breaking is a processthat occurs over a wide range
of scales, from small-scale or microscale breaking [Phillips
and Banner, 1974] through intermediate-sized breaking
events, which may not produce discerniblefoam patches, to
larger events that generate whitecaps. Figures 1 and 2
illustrate
thattheamplitudes
of spikesin •røalsooccurover
a wide range of scales. As was noted already, previous
authors have associated large sea spikes with breaking
waves, that is, events that produce whitecaps. The correla-
tion of lesser-amplitudespikeswith smaller-scalebreakingis
unclear without supplementaryinformation such as a video
image. We also note that Phillips' expression for the fre-
quency of sea spike occurrence (equation (8)), was derived
for "larger-scale breakers producing the more intense re-
turn" [Phillips, 1988, p. 1070]. To avoid the ambiguity
associatedwith smaller sea spikesand for comparisonwith
(8), we will concentrate on detecting sea spikes that are
likely to be associated with larger-scale breaking events.
Furthermore, we will take larger-scale breaking events to be
those occurring at or near the long wave crests. Thus our
detection schemeshould identify large sea spikesassociated
with the long wave crests.
From comparison of the time series of the radar cross
section and the mean Doppler shift (see Figure 1), large
spikesin the return power are clearly associatedwith large
positive surface velocities, correspondingto measurements
at or near the long wave crests. Scatter plots of the distri-
bution of the measured velocity for VV polarization as a
proportionalto the areal average of the line-of-sightcompo-
nent of the surface scatterer velocities, which we assume to
be nearly horizontal at a wave crest. The measured velocity
values in Figure 3 have been resolved to be horizontal,
which is appropriate for the larger sea spikesassociatedwith
large velocities (the lower velocity values associated with
parts of the wave other than the crest are corrupted). Since
the crest velocity of a breaking wave is expected to be of the
order of its phase speed, the measured velocities have been
normalized by the phase speed correspondingto the wave
spectral density peak.
Figure3a is for a frictionvelocityof 28 cm s-• while
Figure3b is for a highervalueof 49 cm s-• . In bothcases,
low values of O'l•ea
~ 0 k correspond
to negativevelocities,which
indicate return from a location away from a wave crest.
Above some threshold, only positive velocities are found,
and the more intense returns are associated with larger
positive velocities. Scatter plots for each hour of data
 KELLER ET AL' SEA SPIKES IN MICROWAVE BACKSCATTER 9683

VV POLARIZATION
VV POLARIZATION z• : u*=46 cm/s
x x
x

x
x
• [] : u*=31
cm/s
E) : u*=15 cm/s
x x
x
x xXX
x
xX x
•x

)o(x
x
u* = 28 crn/s
x

x x

-12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2

-13 -12 -11 -10 '9 -'8 -'7 -'6 25 -4
O•hreshold
(d B)
Fig. 4. The number of sea spikes, N, in a 1-hour record as a
function of radar cross-section detection threshold over a wide
VV POLARIZATION
range of friction velocities, u,. The relatively constant slope and
spacingof curves indicates that sea spike statistics are not strongly
x x x
xx xX
dependent on threshold over a range of several decibels.

xx x:•
x Xx x

velocities. Examples of this analysis for three different

Xxx velocities are given in Figure 4, showing that for a given u,
u* = 49 crn/s the number of events decreases as the threshold increases.
xX
x x The relatively constant slope and spacing of these curves
b
over a threshold range of several decibels indicates that the
sea spike statistics may not be especially sensitive to the
-12-'11-'10-9 -8 -'7 -'6 -'5 -4-5 -2 choice of threshold. That is, the relative change between
•';ook (dB) curves of different u, for a given threshold is not a strong
Fig. 3. Scatter plots showing the distribution of normalized function of that threshold. To quantify the threshold depen-
surface scatterer velocity for peak radar cross section(VV) for two dence of the sea spike statisticspresented below, the detec-
differentfrictionvelocities: (a) u, = 28cms-1 and(b) u, = 49 cm tion threshold was varied over a range of approximately 3
s-1 . The detectionthresholdof -7.2 dB has been chosento detect dB, centered at the chosen level of -7.2 dB.
mainly sea spikes associatedwith a positive velocity, corresponding
The threshold of-7.2 dB is supported by previous anal-
to a location near a long wave crest.
ysis of video recordings taken during the experiment [Jes-
sup, 1988]. One hour of simultaneous radar and video
measurements
(u, = 26 cm s-1) were madeby visually
aligning the radar antennas and video camera to be aimed at
processed havebeenanalyzed,anda thresholdof (rø(t) = approximately the same location on the sea surface. A
-7.2 dB has been chosen to eliminate most large sea spikes spectrum analyzer connected to one channel of the radar
associated with negative velocities for both VV and HH output was then used in conjunction with a video monitor to
polarizations over a friction velocity range of 15 -< u, -< 49 optimize the alignment. Although there was no way to
-1
cm s . measure the accuracy of the alignment, the presence of a
The use of a fixed thresholdfor analyzing data over a wide whitecap on the video monitor was repeatedly associated
range of friction velocity may at first seem inappropriate, with a large jump in received power and increased Doppler
sincefrom Figure 3 the O'l•ea ~0 k valuesat which negative frequency. Unfortunately, the quality of the video recording
velocities are eliminated appears to be a function of u,. was seriously degraded by a mismatch in video formats
However, consider the implication of Phillips' assertion between the camera and recording unit.
(noted in section 1) that sea spike intensity should increase Despite the alignment uncertainty and the poor recording
with the wavelength or scale associatedwith the breaking quality, the 1-hour video tape was carefully reviewed in
event. If we accept this assertion and assume that the sea conjunction with the simultaneous microwave measure-
spike intensity correspondingto a particular scaleof breaker ments to yield useful but limited quantitative information.
is not itself a function of u,, then a fixed detection threshold The radar data were played into a spectrum analyzer to
is appropriate for detecting events in the same range of scale produce time histories of Doppler spectra in a waterfall or
regardless of u,. spectral map display. A total of 82 events that exhibited a
The usefulnessof a counting technique would be enhanced microwave signature characteristic of breaking events were
if the sea spike properties were relatively insensitive to the identified from the Doppler maps. The video tape was
choice of detection threshold over some reasonableintensity independently viewed and a total of 71 whitecaps were
range. In order to investigate this threshold dependence,we counted. Because of the poor quality of the video, no
have counted the number of sea spikes in 1-hour records as attempt was made to classify the size or scale of the
a function of the cross-section threshold for different friction whitecaps identified.
9684 KELLER ET AL..' SEA SPIKES IN MICROWAVE BACKSCATTER

Although the number of events counted independently in

the radar and video recordings were roughly the same, only
43 breaking events were simultaneously found in both the
radar and video recordings. Nonetheless, approximately
70% of the events which simultaneouslyappearedas white-
caps on the video monitor and were identified in the Doppler
• Method 1
spectral maps were subsequentlydetected by the intensity
threshold of-7.2 dB. The discrepancybetween the number
of independently counted and simultaneously occurring
events may be due to (1) misalignmentof the radar and video
spots, (2) the radar's detecting breaking events that do not I

16 18 20 22 24 26 28 30 32
produce a whitecap, and (3) whitecap events that do not
TIME (sec)
produce a distinctive microwave signature. A breaking event
which does not produce a noticeable whitecap may corre-
spond to the intermediate scale breaking mentioned previ-
ously, while a whitecap whose propagation direction is
oblique to the radar look direction may not produce a
distinctive microwave signature. The uncertainty in align-
ment of the radar and video spots and the poor quality of the
video recording have frustrated attempts to extract further Method 2
meaningful results from these simultaneousmeasurements.
By analogywiththe computational
definition
of •rø given
by (9), the mean normalized sea spike radar cross section
•rsøs
corresponding
to theseaspikecontribution
to •rø is I I I I I

16 18 20 22 24 26 28 30 32

TIME (sec)

1• •Ti&•s(t)
o'sOs
=•i=1 dt (lO) Fig. 5. A schematic diagram showing the two methods used to
compute the sea spike contribution to the normalized radar cross-
section. The horizontal line labeled •rø is the normalized radar
where&•s(t)isthecontribution
ofanindividual
seaspike,N cross-section for a one-hour record. The contribution of an individ-
is the numberof sea spikesin a record of length T, and Ti is ual sea spike correspondsto the shadedarea. The duration of the sea
the duration of the ith event. spike in method I (Figure 5a) is taken as the time during which b(t)
The two methods we have chosen to define an individual is elevatedabovethemean•rø. In method2 (Figure5b)the spike
duration is taken as the time between local minima on either side of
sea spike intensity and duration are illustrated in Figure 5.
the sea spike peak.
For method 1, Ti is the time over which the radar cross
section
is elevated
abovethemean,•rø, and&sOs(t)
is that
portion of the total cross-section which is above the mean
during Ti: data is 20ø or less. The data at higher friction velocities in
Figure 6 appear to be highly correlated, but the scatter at
&•s(t)= &ø(t)- •rø (11) lower values of u. suggeststhat a linear fit over the entire
The duration Ti for method 2 is the time between local range of u. is not appropriate.
minimaoneithersideof theseaspikemaximum,
and&•s(t) The cross section of an individual breaking event is
is given by expected to be a function of the orientation of its crest with
respectto the radar look direction. A reasonableassumption
&•s(t)= &0(t
) -- O'mi
~0n (12) is that the directional distribution of breaking wave crests is
symmetric about the wind. Then upwind measurements
wherefirømi
n is the lesserof the localminimadefiningTi. made over a restrictedrange of the angle qobetween the radar
While other methodsare possible,thesetwo provide reason- look direction and the wind should provide a valid sampling
able upper and lower bounds on the contribution of an of breaking events, even if that angular range is not centered
individual sea spike. at •= 0.
The seaspikecontribution
•r•s as definedby method1
3. RESULTS (equations (10) and (11) and Figure 5a) is shown versus
The measuredmean normalized radar cross section •rø is friction velocity u. on a log-log plot in Figure 7 for the
cross-section threshold of-7.2 dB. For the relation
plotted versus friction velocity u. on a log-log scale for VV
and HH polarization in Figure 6, where each point repre- •r•s= C•u• (13)
sents the mean for a 1-hour record. The radar antennas
remained fixed during the experiment, pointing in the NW we find the exponent a of the least squareslinear fit to be 3.3
direction of 315øT. All data analyzed were for wind direc- and 3.5 for VV and HH polarizations, respectively. The
tions in the angularbands 340o-360øor 270o-280ø.That is, the exponent a doesnot vary greatly over a thresholdrange of 3
angle qobetween the radar look direction and the wind was dB for both methods of defining the sea spike and is listed
between 25 ø and 45 ø for all measurements. These data are with 95% confidence levels and correlation coefficients p
presented
together,sincethe variationof •rø is expectedto [Bendat and Piersol, 1986] in the upper portion of Table 1.
be symmetric about (p= 0 and the range of (pvariation for the The relative importance of the sea spike contribution to
 KELLER ET AL ' SEA SPIKES IN MICROWAVE BACKSCATTER 9685

VVPOLARIZATION

x
xx
xxxx•
x
• xX•X•
•
VV
POLARIZATIO
xXxJX
t XX

x
x

x/ xxxx
a

,o io ;o
u* (cm/s) 10 2'0 ,3'0 40 50
u* (cm/s)
o

, ,

HH POLARIZATION

x xW•
xXXXx x
• xXxXx HH
POLARIZATION
x•xx
x•xX
x.
xX X x•x x
x
x

b x x x

o io io ,;o so
,.,. (cn-,/•)
1o 20 30 40 50

Fig. 6. The normalized radar cross-sectionversus friction ve- u* (crn/s)

locity u,, for (a) VV and (b) HH polarization(each point represents
a l-hour record). Note the relatively large amount of scatterfor low Fig. 7. Theseaspikecontribution
O-•sto thenormalized
radar
u, values comparedwith the high degreeof correlationfor large u,. cross-sectionversus friction velocity u, based on method I and a
threshold of &(t) = - 7.2 dB. The slope of the least squaresline is
3.3 for VV polarization (Figure 7a) and 3.5 for HH (Figure 7b),
indicating a roughly cubic dependence.
the mean radar cross section is of interest for improving
currentmodelsfor O-0.Figure8 showsthefractionalcross-
section
duetoseaspikes,
O-•s/o-
ø,versus
friction
velocity
for for the threshold of -7.2 dB as the number of events N in
the cross-sectionthreshold of-7.2 dB. Clearly, sea spikes each 1-hour record for which at least two events were
contributea greaterfraction of the return power for HH than counted and the sea spike contribution was at least 1%, that
for VV polarization. In general, the average polarization isforN > 1 andO-•s/o-
ø ->0.01.Theleastsquares
fit,
0 0
ratio O-VV/o-HH is greater than unity [Wright, 1966], but
during a breaking event the instantaneouspolarization ratio N= C2u,
s (15)
can approachor equal unity [Kwoh and Lake, 1986]. There- finds friction velocity exponentsof 2.8 for both VV and HH
forethelargercontribution
of seaspikes
to O-]/H
isexpected. polarization, for the threshold of -7.2 dB. The exponents/5
The value of the fractional radar cross section is of course a
and correlation coefficientsp for a threshold range of 3 dB
function of the cross-section threshold. For the threshold of
are given in the lower portion of Table 1. As with the friction
-7.2 dB, the percentage of the return power due to sea velocitydependence
of O-SS,
0 theexponent/5
doesnotsignif-
spikes with 25ø _< •0 -< 45ø at high friction velocity is icantly change over this threshold range.
approximately 5-10% for VV polarization and 10-20% for The friction velocity exponents computed for both the sea
HH polarization. spike contribution to the mean cross section and the fre-
The residualradarcrosssection,O-ies,
0 given by quency of occurrence are all close to 3 (see Table 1). This
similardependence on u, impliesthat O-•sandN maybe
0
O'res = O- -- 0 O-•s (14) linearly related. Therefore in Figure 11 we have plotted N,
with method 1 for the thresholdof-7.2 dB is plotted versus thenumber
of eventsin each1-hourrecord,against
O-•sfor
friction velocity in Figure 9. Since the maximum sea spike the threshold of -7.2 dB. The exponents •/for
contribution for this threshold is 20% (approximately 1 dB
for HH polarization at large u,), Figure 9 is similar in N-- C30-• (16)
appearance
to theplotof O-0in Figure6. As withFigure6, are 0.82 ___
0.06 (p = 0.98) and 0.77 ---0.05 (p = 0.98) for VV
the scatter of the data in Figure 9 at lower values of u, does and HH polarization, respectively.
not suggesta least squarefit over the entire range of u,. The
fit for u, -> 23 cm s-• shownin Figure9, for whicha high 4. DISCUSSION
degree of correlation is apparent, gives exponents very
nearly equalto unity: 1.1 - 0.2 (p = 0.93) for VV polarization The argument may be made that sea spikes defined by a
and 1.1 --- 0.2 (p = 0.92) for HH polarization. rationally chosenyet somewhat arbitrary intensity threshold
Figure 10 showsthe frequency of occurrenceof sea spikes may not necessarilybe due to breaking events. For instance,
 9686 KELLERETAL.' SEASPIKES
IN MICROWAVE
BACKSCATTER

TABLE 1. Friction Velocity Exponents frictionvelocitypointsfromFigure10andalsofromFigure

Threshold, dB 7 mightbejustifiedonthegrounds thatvirtuallynobreaking
eventsoccur. However, the minimumfriction velocityfor
-9.0 -7.2 -6.0 theonsetof breakingis likely to dependon otherfactorsnot
availablein thisdataset.For example,longwaveslopehas
Method 1 beenshownto affecttrø at X band[Kelleret al., 1985],
VV 3.3 + 0.5 (0.91) 3.3 -+0.5 (0.90) 3.2 + 0.7 (0.85) especiallyat low wind speeds.Indeed, the increasedscatter
HH 3.1 + 0.5 (0.90) 3.5 + 0.6 (0.89) 3.4 + 0.6 (0.89) of trø for low u, in Figure6 maybe an indication of this
Method 2
effect.Removalof the low-velocitydata doesnot signifi-
VV 3.8 + 0.5 (0.93) 3.6 _+0.6 (0.89) 3.6 + 0.7 (0.88) cantlychangethecomputed
exponents,
andthuswe present
HH 3.3 _+0.5 (0.91) 3.5 + 0.6 (0.89) 3.5 + 0.7 (0.88)
all availabledatain the interestof completeness.
N = C2 u,• Theresidualradarcrosssectiongivenby (14)andplotted
VV 3.6 + 0.6 (0.89) 2.8 + 0.5 (0.89) 3.0 + 0.6 (0.88) in Figure9 (aswellasthetotalradarcross-section
in Figure
HH 3.0 + 0.6 (0.84) 2.8 + 0.4 (0.91) 2.7 + 0.5 (0.87)
6) showsmuchlessscatterfor the largervaluesof u,. This
Correlationcoefficients
p are givenin parentheses. observation
motivatedthe linearfit over the higherfriction
velocityvaluesshownin Figure9. The residualradar cross-
section
•rrøes
corresponds
toPhillips'
[1988]Bragg
contribu-
tion (equation(3)) which is basedon the wave spectral
the random nature of the distributionof scattererson the sea
surfacemayleadto constructive interference in a density•k) givenby (2). The derivationof the spectral
resulting
seaspike'shavinglittle to do with breaking.Thisargument densityassumes thatlossesdueto breakingareimportant in
is diminished
by thedistributions shownin Figure3, which the equilibriumrangeof validityof •k) [Phillips,1985].
indicatesthatlargepeakradarcrosssections aregenerally Sincelittle wavebreakingis expectedat low u., the friction
velocityexponents
associated with large surface scatter velocities. Further- closeto unityfor •rrøes
at higheru. might
more, the choiceof a relativelylargethresholdensuresthat beinterpretedassupportingPhillips'predictedlineardepen-
the sea spikescountedwill be associated with positive dence.Strictlyspeaking,
however,equation(3)for theBragg
surface scatterer velocities. contributionshouldnot be appliedto thesedata (Bragg
For frictionvelocitiesbelow20 cm s-] or so,fewerthan wavelength of 1.51cm),since•k) is derivedonlyfor gravity
waves.
10 events per hour were counted for the chosenthresholdof
-7.2 dB (see Figure 10). Melville [1977]suggested that a The scatterin the plot of N versustr•s (Figure11) is
minimumfrictionvelocityfor the onsetof breakingmay significantly less than that for each quantityplottedindi-
of 23 cm s- ]. Thusdeletionof low viduallyagainstu, (Figures7 and 10). The nearlylinear
occurin the neighborhood

VV POLARIZATION
o

1:13
VV POLARIZATION
v•l 3
T
ø•o x x
x x•
i xx x
x x)o(x x
x
i x
x

x x)o( x i
x

,'o..... 5O
lO 2'0 46 5O
u, (•../•)

!
x
HH POLARIZATION
x x

HH POLARIZATION x
x)o(x
x x x

x x

x x '
x
xN•
x x
xX X
X
X

0 10 20 •0 ,•0 50 b

10 2b 4b 50
u, (•../•)
Fig. 8. The fractional
radarcrosssectionrrsøs/rr
ø dueto sea
spikesversusfrictionvelocityu, basedon method1 anda threshold Fig. 9. The residualradarcrosssection,rrø - rrsøs, versus
of -7.2 dB. At highu, the seaspikecontribution
to the meancross frictionvelocityu, for (a) VV and(b) HH polarization.
For u, > 23
section measured between 25ø and 45ø relative to the wind is cms-l, thereisverylittlescatter
inthedata,andtheslope
ofthe
approximately
5-10% for VV polarization(Figure8a) and 10-20% leastsquarefit overthisrangeis 1.1for bothpolarizations,
indicat-
for HH (Figure 8b). ing an approximatelylinear dependence.
 KELLER ET AL.' SEA SPIKES IN MICROWAVE BACKSCATTER 9687

dependencebetween the sea spike contribution to the radar

cross section and the frequency of occurrence of sea spikes I I I l

VV POLARIZATION
implies that the average radar cross section of a sea spike,
i.e., oss,•,maybe:"
_o ,•, ,,,,.,,.p,.nd,.n,.
• • '• * of u,.
Thattrsøs/N
isindependent
of u, maybea consequence
of
the illumination area's being small compared with the dom-
inant surfacewavelength. For example, if specularreflection z

from the steep forward face of a breaking wave is important,

then a maximum detectableintensity would be reachedwhen
specular scattering occurs over the entire illumination area. a

Similarly, if the cross section increaseswith the size of the -'40 -'3s -'30 -'2s
active breaking region, then a maximum detectable return o
O'ss
might occur when the scale of the breaking crest reachesthe
dimensionsof the spot size. In both cases, the cross section
measurementwould be limited by spot size, and thus above I I l

some u, threshold the average sea spike return might be HH POLARIZATION

constant.

On the other hand, an average sea spike cross section

which is independent of u, might be consistentwith the fact
that breaking occurs over a wide range of scales. If the cross
section of a breaking wave is a measure of its size, then the
size distribution of breakers and the increase in the number
of breaking events at all scales might be such that the b
average sea spike return is invariant.
-'40 -'3s '30 -'2s -20
o
O'ss (dB)
5. CONCLUSIONS
Fig. 11. The frequency of occurrence of sea spikes, N, plotted
Field and laboratory measurements of microwave back- against
seaspikecontribution
crsøs
to the normalized
radarcross
scatter at moderate incidence in the presence of wave section. The slope of the least squaresfit is 0.77 for VV polarization
breaking have indicated that steep and breaking waves (Figure 11a) and 0.82 for HH (Figure l lb). The nearly linear
produce events referred to as sea spikes.We have presented dependencesuggeststhat the cross sectionof an average sea spike,
i.e., cr•s/N,maybeindependent
of u,.
quantitative measurements of the relationship between the

friction velocity u. and sea spikes for the Ku band radar

crosssection.The intensity threshold for the definition of the
sea spike contribution to the mean radar cross section was
VV
POL•RIZATIO'N
' xx
chosen to count large sea spikes associatedwith a positive
mean Doppler frequency.
The variationof the sea spikecontribution
rrsøs
with
friction velocity was found to be consistent with Phillips'
[1988] prediction of a cubic dependence (equation (7)). The
fractional power for high friction velocities (u. • 40-50 cm
s-1) measured
25øto 45ørelativeto thewindwasfoundto be
20 30 40 50
approximately 5-10% for VV polarization and 10-20% for
•, (oml•)
HH polarization with the procedure designated method 1
(equations (10) and (11) and Figure 5a). These findings
supportthe inclusionof breaking waves in scatteringmodels
for HH polarization.
HH POLARIZATION
The frequency of occurrence of sea spikes was computed
as the number of events counted in a 1-hour record for the
xx Mz x
chosenthreshold. The roughly cubic dependenceon friction
x •xx
velocity is consistent with theoretical modeling and field
x• x x
measurementsof whitecap coverage reported by other in-
vestigators and with Phillips' [1988] result (equation (8)).
Finally, the data suggestthat the contribution of an average
seaspiketo rrø maybe independent
of u..
10 20 30 40 50

u* (cm/s)
Fig. 10. The frequency of occurrence of sea spikes in terms of
number of events in a 1-hour record, N, versus friction velocity u,
for (a) VV and (b) HH polarization. The slopeof the least squaresfit
is 3.5 for VV and 3.1 for HH, indicating an approximately cubic
dependence.
Acknowledgments. We gratefully acknowledge the efforts of
Finn Hansen of RisO National Laboratory, Roskilde, Denmark, in
providingthe measurementsof friction velocity under ONR contract
N00014-87-G-0271. The use of video equipment p,eovidedby Peter
Lobemier of FWG, Kiel, West Germany was much appreciated.
We extend our thanks to the crew of the Forschungsplattform
9688 KELLER
ETAL.:SEASPIKES
IN MICROWAVE
BACKSCATTER

NORDSEE for their excellentsupport.We also wish to thank Keller,W. C., V. Wismann,andW. Alpers,Tower-based measure-
WilliamJ. Plantfor hishelpfulcomments on an earlydraftof this mentsof theoceanC bandradarbackscattering crosssection,J.
paper.Oneof us(W.K.M.) wishes to thankTomSpence (formerly Geophys.Res., 94(C1), 924-930, 1989.
of ONR) for hisearlysupportandencouragement of thiswork.The Kwoh,D. S. W., andB. M. Lake, A deterministic, coherent,and
participationof W.K.M. andA.T.J. wassupported by grantsfrom dual-polarized laboratorystudyof microwavebackscattering
the MIT SloanBasicResearch Fund,NSF (Physical Oceanogra- fromwaterwaves,1, Shortgravitywaveswithoutwind,IEEE J.
phy),andONR(Physical Oceanography) to W.K.M. Theparticipa- Oceanic Eng., OE-9, 291-308, 1984.
tion of A.T.J. wasalsosupportedby the NASA GraduateStudent Kwoh, D. S., and B. M. Lake, Microwave backscatterfrom short
Researchers' FellowshipProgram(grantNGT-50292).A prelimi- gravitywaves:deterministic, coherent,dual-polarizedstudyof
nary versionof theseresultswas first presentedat the 1988Fall therelationship betweenbackscatter
andwaterwaveproperties,
meetingof theAmericanGeophysical Union(Jessup et al., 1988). in WaveDynamics andRadioProbingof theSeaSurface,edited
by O. M. PhillipsandK. Hasselmann,pp. 443-447,Plenum,New
York, 1986.
REFERENCES Kwoh,D. S., B. M. Lake,andH. Rungaldier,Microwave scattering
frominternalwavemodulated surfacewaves:A shipboard real
Alpers,W., D. B. Ross,andC. L. Rufenach,
Onthedetectability
of aperture
coherent radarstudyin theGeorgiaStraitExperiment,
J.
ocean surfacewave by real and syntheticapertureradar, J. Geophys.Res., 93(C10), 12,235-12,248,1988.
Geophys.Res., 86(C7), 6481-6498, 1981. Lewis,B. L., and I.D. Olin, Experimentalstudyandtheoretical
Banner,M. L., andE. H. Fooks,Onthemicrowave reflectivity
of modelof high-resolutionradarbackscatter from the sea,Radio
small-scale
breakingwaterwaves,Proc.R. $oc.London,$er.A, $ci., 15, 815-828, 1980.
399, 93-109, 1985.
Long,M. W., On the two-scatterer
theoryof seaecho,IEEE Trans.
Bendat,J. S., and A. G. Piersol,Randomdata: Analysisand Antennas Propag., AP-22, 667-672, 1974.
MeasurementProcedures,2nd ed., Wiley-Interscience,New Long,M. W., RadarReflectivity of Land and Sea,pp. 502-505,
York, 1986.
Artech House, Dedham, Mass., 1983.
Donelan,
M., andW. J. Pierson,
Jr., Radarscattering
andequilib- Lyzenga,D. R., A. L. Maffett, and R. A. Shuchman,The contri-
riumrangesin wind-generated
wavewith application
to scatter- butionof wedgescatteringto the radarcross-section
of the ocean
ometry, J. Geophys.Res., 92(C5), 4971-5092, 1987. surface,IEEE Trans.Geosci.Remote$ens.,GE-21(4),1983.
Doviak, R. J., and D. S. Zrni•, Doppler Radar and Weather Melville,W. K., Windstressandroughness lengthoverbreaking
Observations, Academic,SanDiego,Calif., 1984. waves, J. Phys. Oceanogr., 7, 702-710, 1977.
Duncan,J. R.. W. C. Keller, and J. W. Wright,Fetchand wind Melville,W. K., M. R. Loewen,F. C. Felizardo, A. T. Jessup,and
speeddependence of Dopplerspectra,Radio $ci., 9, 809-819, M. J. Buckingham, Acousticandmicrowavesignatures of break-
1974.
ing waves, Nature, 336(6194),54-59, 1988.
Ewell,G. W., M. T. Tuley,andW. F. Horne,Temporal
andspatial Monahan,E. C., andI. G. O'Muircheartaigh,
Whitecaps
andthe
behaviorof highresolution
seaclutter'spikes,'IEEE Radar '84 passive
remotesensing
of theoceansurface,Int. J. RemoteSens.,
Conf., 100-104, 1984. 7(5), 627-642, 1986.
Guinard,N. W., J. T. Ransome,andJ. C. Daley,Variationof the Phillips,
O. M., Spectral
andstatistical
properties
of theequilibrium
NRCSof theseawithincreasing roughness,
J. Geophys.Res.,76, rangein wind-generated
gravity waves,J. Fluid Mech., 156,
1525-1538, 1971. 505-531, 1985.
Holthuijsen,
L. H., andT. H. C. Herbers,Statistics
of breaking Phillips,
O. M., Radarreturnsfromtheseasurface--Bragg
scatter-
wavesobserved
aswhitecaps
in theopensea,J. Phys.Oceanogr., ingandbreaking waves,J. Phys.Oceanogr.,18, 1065-1074,1988.
16, 290-297, 1986.
Phillips,O. M., andM. L. Banner,Wavebreaking in thepresence
Jessup,
A. T., Microwave
detection
andcharacterization
of deep- of winddrift andswell,J. Fluid Mech., 66, 625-640,1974.
water wave breaking, M.S. thesis, Mass. Inst. of Technol. Cam-
Plant,W. J., andW. C. Keller,Thetwo-scale radarwaveprobeand
bridge, Mass., May 1988. SARimagery of theocean,J. Geophys.
Res.,88(C14),9776-9784,
Jessup,A. T., Detectionandcharacterization
of deepwaterwave 1983.
breakingusingmoderateincidenceanglemicrowavebackscatter Ross,D. B., andV. Cardone, Observations
of oceanicwhitecaps
fromthe seasurface,Ph.D. thesis,360pp., JointProgramin andtheirrelationto remotemeasurements
of surface
windspeed,
Oceanogr.and Oceanogr.Eng., Mass. Inst. of Technol./Woods J. Geophys.Res., 79(3), 444-452, 1974.
Hole Oceanogr.Inst., WoodsHole, Mass., 1990. Wright,J. W., Backscattering
fromcapillary
waveswithapplication
Jessup,A. T., W. C. Keller, W. K. Melville, Microwavedetection to seaclutter,IEEE Trans.AntennasPropag.,AP-14, 749-754,
andcharacterizationof deep-waterwavebreaking(abstract),Eos 1966.
Trans. AGU, 69(44), 1245, 1988.
Wu, J., Oceanicwhitecapsand seastate,J. Phys.Oceanogr.,9,
Kalmykov,A. I., andV. V. Pustovoytenko,
On the polarization 1064-1068, 1979.
features of radio signal scatteredfrom the sea surfaceat small Wu, J., Wind-stress coefficients over sea surface near neutral
grazingangles,J. Geophys.Res., 81, 1960-1964,1976. conditions--Arevisit,J. Phys. Oceanogr.,10, 727-740,1980.
Katzin,M., On themechanism
of radarseaclutter,Proc.IRE, 45,
44-54, 1957.
A. T. Jessupand W. K. Melville, MassachusettsInstitute of
Keller,W. C., W. J. Plant,andD. E. Weissman,
Thedependence of Technology,Cambridge, MA 02139.
X bandmicrowaveseareturnon atmospheric stabilityandsea W. C. Keller, U.S. Naval ResearchLaboratory,Code8310-140,
state, J. Geophys.Res., 90(C1), 1019-1029, 1985. Washington, DC 20375.
Keller, W. C., W. J. Plant,andG. R. Valenzuela,Observations
of
breakingoceanwaveswith coherentmicrowaveradar,in Wave
Dynamics
andRadioProbingof theSeaSurface,editedby O. M. (Received April 19, 1989;
Phillipsand K. Hasselmann,
Plenum,New York, 1986. accepted August 31, 1989.)