Progress Toward a Practical Skywave Sea-State Radar

Home , Skywave

EEE TILANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. “-28, NO. 6, NOVEMBER 1980 75 1 Progress Toward a Practical Skywave Sea-State Radar

T. M.GEORGES

Absrracr-Recent advancesin propagation modeling, ionospheric theNational Oceanic and Atmospheric Administration’s diagnostics,and signalprocessing have helped overcome the lim- (NOAA’s) Coastal Ocean Dynamics Radar (CODAR) system. itations the ionosphere imposes on sea-state measurements with HF Apair of these compact transportable HF radarsproduces skywave radar.Wind-direction fields in tropicalstorms can be surface-current-vector maps out to 70 km from shore in near- routinely mapped under mostionospheric conditions, but waveheight real-time [6]. Othersurface-wave radars have beenused to andwave-spectrum extractionismore sensitive to ionospheric measure rms waveheight, as well as the scalar and directional distortions and requires care in signal processing and in selecting an ocean-wavespectra [ 71. Becausesurface-wave radars present ionosphericpath. Spot measurements with a high-resolutionradar have verified its ability to measure (in order of increasing dif€iculty) fewremaining problems and are now virtually operational, wind-direction fields,rms waveheight,and the scalar ocean-wave onlythose problems unique to skywave systems will be spectrumat rangesup to 3000 km using one ionospherichop. addressed here. Althoughsuch a radar can in principle map thesequantities over It is the increase in single-radar coverage area from thou- millions of square kilometers of an ocean area, the time requiredto do sands to millions of square kilometers that prompts the use of so under various ionospheric conditions remains to be determined. A ionosphericreflection with its attendant contaminations. minimumobjective of onemap of rms waveheight per day seems Except for the effects of the intervening ionospheric reflec- attainable. tions: the sea echoes received by skywave radars are identical tothose obtained with surface-wave illumination. Initially, I. HISTORICAL PERSPECTIVE researcherswere optimistic that the effects of imperfect ionospheric reflection could be easily predicted and controlled KYWAVEradars sense natural and man-made targets andthat vast oceanareas could be monitored with a few beyond the optical horizon by bouncing HF radio waves S judiciously located skywave radar stations [ 8 I , [ 91 . Some of off the ionosphere. The first skywave radars were research in- this optimism dissipated when they discovered that, although strumentsdeveloped in the late forties and early fifties to wind-directionfields can usually be mapped, ionospheric study the structure and motions of the ionosphere itself, using degradation of thesea-echo spectrum often prevented the echoesbackscattered from the earth. Many researchers tried extraction of other important sea-state information, such as to developground-backscatter radar into anoperational the rms waveheight. Therefore, progress in skywave sea-state diagnoseic toolfor ionospheric mapping or HF frequency radar development in the lastfive years might be characterized management,but because of problems in uniquelytrans- as a reassessment of the ionospheric problem and a marshalling formingradar signatures into ionospheric structure: such of forces to deal with it. effortshave been largely abandoned. Consequently, the re- Renewed optimism about skywave radar now seemswar- views of suchapplications by Hayden [ 11 andCroft [2] ranted because of four developments: 1) improved understand- remain reasonably up-to-date. ing of the spectral contamination caused by the ionosphere, In the fifties, the prospect of detecting and tracking mili- b) an ability to derive optimum radar propagation paths from tarytargets (missiles, aircraft, and ships) with an over-the- real-timeionospheric soundings, c). the availabilityof high- horizon (OTH) radar -spurred a distinct and largely classified resolution radar facilities built with defense funds, and d) the line of research that led to the construction in the sixties and evolution ofhigh-speed signal processing strategies to deal with seventies of large-aperture HF-OTH radar arrays by the major contaminated data. powers [31, 141, [511. The subject of this review is the third and most recent ap- 11. THEORETICAL BASIS plicationof skywave radar-sea-state monitoring-in which information about the wind and wave fields at the sea surface Crombie’sobservations [ 51 stimulated wave theorists to is derived from the sea-clutter echo that is regarded as noise constructbetter models of the interaction of HF electrc- in surveillance systems. magnetic waves withgravity waves on the sea surface. Their A sea-state radar need not employan ionospheric (skywave) radar-cross-sectionmodels had to explainnot only the first- path.Indeed, the whole field of HFradar oceanography orderor Bragg-backscattering process that produces an echo evolved from surface-wave experiments by Crombie [ 51 who with two sharp spectral lines but also the spectral structure of first demonstrated HF Bragg scattering by sea waves of half an observed weaker second-order continuum of echoes from the radar wavelength. Refinements in the mathematical model doublescattering by the sea waves. Barrick [lo], [ 11 ] has of HF scattering from the sea and the evolution of high-speed produced the most comprehensive second-order model for the signal-processingsystems have rapidly led tothe present HF seaecho spectrum and its dependenceon sea state, and his stateofsurface-wave radar development represented by modelhas withstood numerous comparisons between ex- perimentallymeasured sea-echo spectra and in situsampling ofocean-wave parameters. According to his model,two Manuscript received April 1, 1980; revised July 10, 1980. mechanismscontribute to the second-order continuum: one T. M. Georges is with the Wave Propagation Laboratory, National Oceanicand Atmospheric Administration, Environmental Research is actuallysingle scattering from evanescent ocean-wave Laboratories, Boulder, CO 80303. componentsthat arise from nonlinear hydrodynamic inter-

U.S. Government work not protected by U.S. copyright. 752 IEEE TRANSACTIONSON ANTENNAS AND PROPAGATION,VOL. AP-28, NO. 6, NOVEMBER 1980 action of two water waves; the second contribution is from radar waves twicescattered from the water surface before returning to theradar. Interpretingthe properties of thefirst-order or Bragg- scatter features of the Doppler spectrum permits measurement of surface wind direction and surface currents, whereas analysis ofthe second-order contributions to the echo spectrum hasled to methodsfor extracting rms waveheight and both the scalar and directional ocean-wave spectrum from measured -1 HZ 0 +lHZ sea-echo spectra. When waveheight measurements are coupled Doppler Shlt with suitable models for wind-wave generation, surface wind speed can also be estimated. A - Short wlnd-wave drrectlon (surface wnd d~recbon). B - Surfacecurrent. 111. MEASUREMENT HIERARCHY D/c - Rrns wave heght Some of the sea-statequantities just mentioned are rela- E - Scalar ocean-wave spectrum tivelyeasy to measurewith an HF radar,whereas others, namelythose derived from second-order echoes, require F - Frequency of domlnant Dcean waw more care in removing noise and interference from the radar Fig. 1. Many quantities that indicate conditions at the sea surface echo. Here we summarize the particular oceanographic quanti- can be extracted from the spectrum of HF radar sea echoes. Those ties that an HF sea-state radar can measure, briefly describe derived from the fiist-order spectral features (A and B) are easier howeach quantity is derived fromthe sea-echo spectrum, to measure in the presence of noiseand contamination than are and assess theparticular problems that arise in making that those derived from second-order features (D/C, E, and F). measurement with a skywave radar (seeFig. 1). B. Surface Currents A. Wind-Direction Fields Becausesurface currents simply transport gravity waves The short ocean waves (a few meters in length) that pro- (withminor complications introduced by current shear), duce first-order, Bragg-resonant echoes at HF respond quickly the Bragg lines of the Doppler spectrum are simply shifted in to the driving wind, so that the resonant waves of maximum frequency (from symmetry about zero Doppler) by an amount intensitytravel in the direction of the surface wind. If the proportional to the radial(from the radar) current com- directional pattern of these waves’ response to the wind can ponent[20]. Thus, mapping currents is relativelystraight- bemodeled, then the wind direction can be determined forward with a pairof surface-wave radars [61, [ 21 ], [ 531. (exceptfor a left-right ambiguity with respect to the radar With a skywave radar, one kind of ionospheric contamina- beam direction) by matching that pattern to relative strengths tion is a frequency shift of the seaecho spectrum caused by of the Bragg echo from waves advancing and receding along upwardor downward motion of the reflecting layer, Such theradar beam. (Brag-line echoes are virtually always ob- shifts are indistinguishable from a current-induced shift. Thus, tained even from visually calm seas.) The first demonstrations unlessthe target sea areaincludes a zero-Doppler reference of skywaveradar for ocean monitoring produced maps of (such as land or a moored platform), surface currents cannot wind-directionfields derived fromsuch measurements of beextracted from skywave sea echoes. Maresca and Carlson Bragg-line ratios [12], [131, [141, [151, [271. Theleft-right [22] havereported near-shore cument’measurements with a ambiguity usually is easily resolved for tropical storms, whose skywaveradar, but measurements of the spatial variability circulation is known in advance [ 161, [ 171 ; for other weather ofionospheric Doppler shifts indicate that surface currents patterns,the ambiguity can be resolved byenforcing con- probably cannot be measured more than a few tens of kilo- tinuityon the overall circulation pattern. Alternatively, two metersfrom a zero-Doppler reference. Further experiments radars can resolve the ambiguity by measuring two wind-direc- are needed that use land echoes to measure the spatial vari- tion components. ability of theDoppler shifts introduced by each kind of Because the ratio of advancing to recedingBragg-line power ionospheric layer (particularly sporadic-E). remains relatively undisturbed by ionospheric multipath, such C. Waveheight contamination is seldoma problem. With ahigh-resolution rrns radar,wind-direction fields can be mapped in suchdetail Thesingle most useful descriptor of sea state .is the rms thattropical storms can be tracked with an accuracy that waveheight,the integral over all wavenumbersof the wave- competes with satellite tcacking methods but at a fraction of height spectrum. But because typical first-orderHF radar echoes their cost [ 181, [ 191. are from 5- to 15-m ocean waves whose amplitude can reach Inaccuraciesin these wind-field maps come mainly from only a few centimeters before saturating, information aboutthe threesources: a) the difficulty of specifying an appropriate higher (and longer) waves must come from the twice-scattered modelfor the directional spectrum of shortwind-driven echoes.The double-scattering process includes the influence waves inany given meteorologicalsituation, b) occasional of ocean waves of all wavelengths, because the condition for problems resolving the left-rightambiguity inherent in the double-interaction Bragg resonancerequires only the sum of directionmeasurement, and c) inaccuracies in locating the the two scattering wavenumbersto be a certain value [23]. radarcell caused by ionospheric tilts. Each of theseprob- Several methods are now used to derive the rms height of lemscontributes only minor errors, however, and skywave wind waves fromthe Doppler spectrum. Each explicitly radarcan routinely map wind-direction fields to ranges be- identifies the first- and second-order portions of the Doppler tween 1000 and 3000 km. spectrum and computes their power ratio. (Such waveheight GEO RGES: PRACTICAL SKYWAVE SEA-STATE RADAR SEA-STATE SKYWAVE PRACTICAL GEORGES: 753 algorithmsare self-normalizing, in thatthey do not require sharp Bragg lines.Although Bragg-line width appeared to be measurementsof absolute scattering cross section in the correlatedwith surface-wind speed, the empirical formulas presence of unknownpath losses.) The spectral resolution expressingthe observed relation proved inconsistent with requiredtodistinguish first- from second-order spectral subsequentmeasurements. Furthermore, Bragg-line width features normally requires50-100 s coherent-integration times, responds as much to ionospheric effects as to the state of the whereasmuch shorter integration may be used to measure sea. Bragg-line ratios. Subsequent theoretical calculations of the response of the Onemethod uses weightinga function computed by -10dB width to variousassumed ocean-wave spectra have Barrick [ 241 to removeelectromagnetic and hydrodynamic shown that the relation between spectral width and surface- singularities from the second-order spectrum before computing wind speed depends on the details of the assumed spectrum the power ratio. This model uses either or both sidebands of and its directional distribution (J. W. Maresca, personal com- thestronger Bragg lineand predicts a dependence of rms munication).For example, rms waveheight can increase waveheight on the square root of the (weighted) power ratio. without affecting Bragg-line width. Anothermethod, devised by Maresca andGeorges [25], Wind speeds in tropicalstorms have been estimated by assumes a power-law relation of the form combiningwaveheight measurements (described in the pre- vious section) with models of storm fetch and duration, and b koh 0.2 koh =aR , > goodagreement with in situ windmeasurements has been claimed [ 171, [ 521.Accurate wind speed measurements in where ko is the radar wavenumber, k is the rms waveheight, other environments have not yet been demonstrated. and R is the (unweighted) ratio of second- to first-order spec- tralpower in the stronger halfof thespectrum. The coef- E. The Ocean-Wave Spectrum ficients a and b have been determined empirically by comput- The most difficult information to extract from HF radar ing R from Barrick’s integral-equation model for the Doppler seaechoes concerns the frequency and directional spectrum spectrum,assuming many different ocean-wave spectra. The ofthe ocean waves in aradar cell. According to Barrick’s best overall power-law fit (+15 percent) was for a = 0.8 and model,the shape of the second-order sidebands that flank b = 0.6. Such a power-law model appearsto beeasier to use in the first-order Bragg lines approximately replicates the shape the presence of ionospheric distortions of the spectral shape of thescalar ocean-wave frequency spectrum [23], [ 291, thanthe weighting-function method. It is alsorelatively [30]. Therefore, to measure the scalar ocean-wave spectrum, insensitive to variationsin the angle between the wind and aradar must accurately recover the detailed structure of at radar look direction and in the directionality of the wind-wave leastone of thesesecond-order sidebands. Again, such re- spectrum [ 25 I. covery is straightforward in narrow-beam surface-wave meas- Thispower law applies only to locallywind-driven waves urements, [3 11 but skywave measurements must avoid even andnot to highlydirectional swell. Lipa et al. [26] give a smallamounts of ionosphericcontamination. Maresca and similar power law for swell waveheighth, Georges [ 251 have obtained scalar wave spectra with a high- resolutionskywave radar via F1 and E, propagation,but a synoptic program of such observations is needed to establish howfrequently the ionosphere will permitsuch measure- where r is acoupling coefficient that depends on the swell ments. wavelengthand direction. This formula is normallyapplied When a single second-order peak can be identified, even if only to a swell whose period exceeds 9 s. the details of the rest of the spectrum are obscured, its fre- Applyingsuch waveheight algorithms to surface-wave quency separation from the nearest Bragg peakcan be used spectraseldom presents any difficulty, Skywave spectra, on as an estimate of the dominant ocean-wave frequency present the other hand, often contain distinct multipath images of the in the radar cell. sea-echo spectrum or are simply broadened to an extent that Lipa [321, [33] first showed that directional information the necessary distinction between first- and second-order fea- about ocean waves can be derived from the second-order spec- tures is not possible, During the last few years, therefore, most tral structure, if the directional and frequency spectra can be of theadvances in skywave sea-state radar have focused on represented as separate multiplicative factors. Lipa and Barrick ways to increase signal-to-noise ratio and minimize the iono- 1341 and Lipa et al. [261 have subsequently shown that swell sphericdegradation of sea-echo signals thatprevents the direction,height, and period can be derived from both sky- distinctionbetween the first- and second-order portions of wave andsurface-wave measurements when distinct long- the echo spectrum. These advances have come from improved period (swell) peaks appear in the Doppler spectrum (Fig. 6). ionosphericmonitoring and signal processingalgorithms Atpresent the measurement accuracy required to extract described in later sections. wind-wavedirectional information cannot be obtained with anyregularity over skywave paths; only the most stable D. Surface- Wind Speed ionosphericlayers, such as sporadic-E, would permit such Because the height of short wind waves responds quickly to accuracy. surfacewinds, measurements of therms waveheight can in principlebe related to surface-windspeed. The earliest at- IV. THE NATURE OF IONOSPHERIC CONTAMINATION tempts to establish such a relation compared direct wind speed The phenomenology of sea-backscatter spectral contamina- measurements with skywave radar measurements of the -10 tionand its dependence on ionosphericconditions do not dB width of the sea-echo Bragg lines [ 271, [ 281. With 13-s seem to differmuch from those of point-to-point transmis- coherentintegration, waveheight increases that increase the sions, which have been much more extensively studied [ 351, second-orderspectral energy merely broaden the normally 1361,[371. The earliest observers of sea-echospectra with 754 IEEETRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-28,NO. 6, NOVEMBER 1980

skywave radars discovered that the lowest ionospheric layers provide the basis for other decontamination strategies, which providethe most stable reflections, with sporadic-E (E,) will be described in Section VI. layers giving spectra of quality (that is, sharpness) comparable The strategy used to correct or avoid such contamination withsurface-wave spectra, and F2-propagated spectra ex- depends on whetherthe ionospheric conditions that cause periencing the most distortion. Normal4 and F1 layers pro- itcan be diagnosed in timeto adjust the radar's operating videreflections of intermediate stability, but their presence parameters.Ionospheric motions occur both on scales too cannot be relied on. small to bediagnosed directly and also on largerscales that Thespectral width of groundechoes or sea-echo Brag submit to real-timemonitoring. The following two sections lines (which are quite sharp in the absence of contamination) describe thedifferent strategies that this dual view ofthe is used toquantify ionospheric broadening. Using aradar contamination makes necessary. witha 10" beam, Earl and Bourne [38] foundthat spectral widthsof ground-backscattered signals weretypically in the V. IONOSPHERIC DIAGNOSTICS AND range 0.2 to 0.3 Hzfor both E,- andF-propagated echoes. PROPAGATION MANAGEMENT With an 0.5" beamwidth radar, however, Georges and Maresca Multipathpropagation caused by large-scale ionospheric [39] found average widthsof about 0.1 Hz for F-layerstructure can be partly but not completely avoided in advance propagatedechoes, and about 0.06 Hzfor E,-propagated byjudicious selection of the radar's operatingfrequency, echoes.Thus, one observed difference between backscatter range gate, and look azimuth. The optimum range-frequency- contamination and that measured on a point-to-point path is azimuthcombination is notnecessarily the one that gives a dependence on the radar's beamwidth. maximum echo strength, but is instead the one that permits Because no usefuloperational radar can rely on thepre- only one (preferably stable) ionospheric path. In general, paths sence of the lower, more stable ionospheric layers whenever that reflect near the bottom of an ionospheric layer, rather than sea-statemeasurements are needed, techniques are being penetrate deeply into it, produceless Doppler spreading.If more developed to utilizeF2-layer reflections much of the time, than one pathis possible (for example,via both E and F layers, particularlyfor propagation to the longer ranges (ZOOO- high- and low-angle rays, one and two hops), the number of 3000 km). Dealing effectivelywith F2-layer contamination two-way path combinations (each imposing a different Dop requires a sound conceptual and mathematical model of the pler shift)multiplies rapidly. With real-time knowledge of contamination process. the ionospheric layer structure, however, high-low-ray multipath and multiple-layer multipath can usually be avoided, but A. Models ordinary-extraordinarymultipath (magnetoionic splitting) At least two approaches to modeling F2 ionospheric spec- usually remains. tralcontamination have been taken. In one, a deterministic Periodicionospheric soundings (near-vertical-incidence model of wavelike ionospheric irregularities was constructed, ionograms) above the radar cangive a first-order picture of the andthe Dopper shift of ground-backscattered echoes was layerstructure, but such soundings apply exactly to the calculatedby ray tracing 1401. Theother approach models obliqueradar path only if horizontalvariations can be ne- theionosphere as anundulating perfect-reflecting surface glected. Optimum radar frequencies can be derived from such whoseproperties are only statistically specified [ 391. Both soundings with the aid of special transmissioncurve ionogram modelsexplain an observed increase in spectralbroadening overlays similar to those used by HF communicators for many with an increase in either the radar's azimuthal beamwidth or years,but modified tobe parametric in radarrange and to the coherent-integration time usedto compute a spectrum. account for any obliquity in the sounding geometry (Fig. 2). Initially, the contamination process was represented mathe- In principle,such overlays can transform ionograms made matically as a frequency-domain convolution of the sea-echo over any oblique path to propagation conditions applying to powerspectrum with some simple frequency-domain path any other range. function,usually a Gaussian broadening or multiplediscrete As asupplement to vertical-incidence ionograms, direc- spectral lines to represent discrete multipath. This path func- tional sweep-frequency-ground-backscatter soundings indicate tion was equivalent tothe power spectrum of amonochro- ionospheric tilts, traveling disturbances and sporadic-E patches matictransmission over the same ionospheric paths. The over the radar path, and permit adjustmentof radar frequency, contaminatedsea-echo spectrum is thuseither a frequency- range,and azimuth toadapt to actual oblique propagation smeared version of the original or a sum of frequency-shifted conditions [421. Inone research radar system [SRI's Wide attenuated images of the original. (Although frequency smear- ApertureResearch Facility (WARF)] , steep-incidenceand ing anddiscrete multipath seem qualitatively distinct, the oblique-backscatterionograms areproduced alternately convolutionmodel merely regards smearing as unresolved every five minutes during sea-state measurements (Fig.2). multipath.) Atpresent, propagation management is amanual process Unfortunately,a closer look at this model reveals thata thatdemands real-time human decisions. Although diag- convolutionof amplitude, not power, spectra is amore ac- nosticprocedures can bespeeded up and somewhat auto- curate representation [ 41 I, [ 561. Phase interference between mated,skilled interpretation of ionosphericsoundings will multiplepropagation paths gives rise to crossterms in the probably be required for some time. power spectrum (the square of a convolution) that preclude Besides describingthe ionospheric transmission path in asimple deconvolution approach, unless care is takento real time,a skywave radar system must also monitor con- average overasufficient number of samples of the same templatedradar frequencies to avoidinterference to and pathconfiguration. Furthermore, nonstationarity in time of from other services. thepath spectrum contributes spurious spectral structure that cannot be treated as simplemultipath. These complica- VI. SIGNAL PROCESSING tionsof the model explain why early attempts at decon- Most strategiesfor dealing with the ionosphere's small- volution of theionospheric contamination failed, but also scaleor random effects on the sea echotake advantage of GEORGES: PRACTICAL SKYWAVE SEA-STATE RADAR 75 5 lonogram Looking Above the Radar I Transmission-Curve lonogram Overlay v) E Secdicorrected)lsecb,85km z

I - - - - 3 4 56 810 15 20 1 25 30 1-1 1-1 I---- lonogram frequency. MHz f,pdex Frequency - I

~ ~~ Backscatter lonogram Looking Overthe Radar Path 1 1 Real-Time Facsimile Output I....

Frequency 4 Fig. 2. To help select the best radar path through a constantlyI varying ionosphere, the SRI-WARF radar displays two kinds of FMCW ionospheric soundings alternately every five minutes. At the upper leftis a near-vertical-incidence ionogram received over the 185-km path between the radar transmitting and receiving sites. This ionogram is interpreted using a transmission-curve overlay, like the one shown at the upper right, as though the ionosphere were horizontally uniform. Intersections of curves of constant radar delay with the ionogram indicate possible ionospheric reflection heights and the layers involved, thus revealing possible multiple paths to the same range. The effect of varying the radar frequency (fob) is obtained by sliding the entire set of curves horizontally, while keeping the ionogram traces stationary.Radar frequency can be read on the bottom scale at the indexarrow. The second diagnostic sounding, shown at the lower left is a sweep-frequency ground-backscatter echo obtained over all the oblique paths in the radar beam. Ionospheric features, such as tilts, sporadic E and traveling disturbances, not visible on the overhead sounding, can be diagnosed with the aid of this display. The photograph at the lower right shows the WARF receiving station with the ionograms displayed in real-time by a facsimile recorder.

differencesin the spatial and temporal scales that charac- singlesample radar record contains large fluctuations and terizechanges in the sea surface and the ionosphere. Often therefore little information about the average spectrum of the theaverage condition of the sea surfacethat we call the processfrom which it is drawn. Incoherent (power) averag- sea state remains unchanged in the open sea for several hours ingover a number of independent sea samples (having the in time and many tens of kilometers in space. On the other same sea state or average properties) is required to reduce the hand, the configuration of ionospheric ray paths can change variance of spectral estimates to usable levels [43], [54]. With in a few minutes in time and a few kilometers in space. There- surface-wave measurements, this averaging process is straight- fore, a radar can see different distorted pictures of the same forwardbecause sequential time records of therequired average sea state. There are two kinds of strategies for dealing number and length can easily represent independent samples with the distortion: one is to avoid it, for example, by search- of the same sea conditions. The short-term variations of the ing for the least distorted sample, the other strategy attempts ionospheremake the incoherent averaging process more dif- tofind the common feature in all thepictures, namely the ficult for skywave measurements because adjacent radar cells unchanging sea state. (in space or time) may include different spectral contaminations from the ionosphere. If such spectra are indiscriminately A. Reducing Sampling Errors averaged,a low-variance spectrum may be obtained, but it will notaccurately represent the average sea-echo spectrum Because the ocean surface and thus the radar echoes from it or any simple contamination thereof. are random processes, the seaecho spectrum derived from a The problem is to find away to obtain several independent 756 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION,VOL. AP-28, NO. 6, NOVEMBER 1980

samplesof the sea echo over adjacent(in space or time) of the contamination process (SectionIV) have so far not been ionospheric paths that all introduce essentially the same form verysuccessful. Standard deconvolution methods work only of spectral contamination (in amplitude but not phase). Then, if the convolving function (the path contamination spectrum deconvolutionand other multipath-removal schemes can be in this case) can be guessed accurately and if the signal is not successfullyapplied tothe average spectrum. Even better, degradedby noise. When simple two-path contamination is if a small ensemble of adjacent radar cells can be found (with evident, for example, the path spectrum can be successfully automaticsearching schemes, for example) in which the approximated by two impulses, but how well deconvolution ionospheric contamination happens momentarily to be neglig- works depends on how many samples of the same contamina- ible, then the resulting average spectrum accurately represents tion can be incoherently averaged to reduce the random spec- the sea-echoprocess. Acquiring many such adjacent samples tral fluctuations. simultaneouslyrequires very highradar resolution (cells no Moresophisticated “homomorphic” deconvolution meth- largerthan a few kilometers) in range and azimuth and a ods [44], in whichthe convolving function need not be method for distinguishing changes in path contamination from known, have been applied to skywave spectra, but with little randomspectral fluctuations. Such methods are now under success, again primarily because of spectral fluctuations. The development. success of suchdeconvolution schemes seems to hingeon finding ways to form low-variance estimates of the sea-echo B. Slowly Varying Frequency Shifts spectrum as contaminated by the same path function. Onekind of contamination, in whichthe ionospheric frequencyshift varies slowlywith space or time, can be E. The Minimum-Envelope Method toleratedwithin an ensemble of spectrato be averaged, be- Asimplified form of deconvolution can beapplied if cause such relative shifts are readily removed, as long as they several low-variance spectra can be formed, each representing do not change significantly over the spectral integration time. adifferent multipath contamination but the same sea state. Ionospheric frequency shifts are typically within f0.5 Hz for If fluctuations caused by phase interference are reduced to a F-layer propagation and iO.l Hz for sporadic4 propagation negligiblelevel byaveraging, then multipath contamination [38]. As long as ocean currents are not of interest, the wave appears entirely as unwanted power added to the uncontami- and wind information to be extracted from a sea-echo spec- natedspectrum. Power appears in all the spectra at the fre- trum is unaffected by such spectral shifts. The absolute shift quencies corresponding to the average sea-echo spectrum, but correction is thereforeunimportant, as long asall thecor- poweradded by multipath appears at different frequencies responding Bragg linesare shifted to the same valuebefore in each multipath realization. Therefore, after suitably super- averaging. Usuallythe peak Bragg frequencies of individual imposing the ensemble of spectra, a minimum envelopecan be spectra accurately indicate the spectral shift, as long as enough constructedthat approaches the common spectral features spectral smoothing (such as provided by a time-series window) (i.e., the sea-echospectrum) and rejects the features that is used.Experiments show significant sharpening of spectra differamong the ensemble (i.e., the contamination). This averaged after shifting, compared with those averaged without methodhas also been called a “cookie cutter,” because the shifting. effect is the same as though each average spectrum were made C. Spectral Renormalization intoa cookie cutter and successively applied to the same piece of dough. The resulting “cookie” represents the decon- Becauseunknown path-loss fluctuations accompany the taminatedspectrum. This method appears to producesatis- echoes from different radar cells, the corresponding Doppler factoryresults under some experimental conditions. Its spectraexhibit spatial and temporal fluctuations in overall reliabilitydepends on the character of thecontamination level. If such spectra were averaged without renormalization, and the number of realizations of a given multipath configura- thecontributions of thecomponent spectra to the average tion that can be averaged and thatis currently being analyzed. would be randomly weighted, possibly permitting the average to be dominated by a few strong spectra. Since the purpose F. Sorting of averaging is to reduce the variance of the spectral estimates withinsome band of interest, arenormalizing method is The first attempt to deal with ionospheric contamination needed by which the unknown path-loss factors can be esti- by sorting appears to be Stewart and Barnum’s [27] method mated and removed, using only information available from the of selecting the sharpest spectra from a daily ensemble, onthe Doppler spectra themselves. assumptionthat broader spectra were ionospherically con- Estimatingpath-loss factors using peak spectral values or taminatedand that only the sharpest ones accurately re- totalspectral power can introduce even more fluctuations flectedsecond-order broadening. Although many more so- into the spectral weighting than no renormalization. In a rela- phisticatedmethods for modeling and removing the iono- tivelysimple but effective method called logarithmic renor- sphericcontamination have been investigated since then, malization,the relative path losses areestimated by the suchsorting procedures, based on avoidingrather than re- geometric mean of all the power spectral values within some moving the contamination, still appear to be among the most band (usually about 20.4 Hz) about the dominant Bragg line. effectivesignal-processing strategies. Refined sorting schemes Thevariance reduction achieved by logarithmic renormaliza- now employan objective sharpness index [39] thatpermits tion is only slightly less than that achieved by a much more automaticselection of datasubsets least likely to becon- complicated iterative variance-minimization scheme[ 4 1 ] , [ 55 ] . taminated (Fig. 3). The principal problem with sorting is the possibilityof bias in whateverquantity (waveheight, for D. Deconvolution example)is estimated from the data subset, so thatcare is Attemptsto model explicitly and remove multipath con- necessarywhen that quantity is correlatedwith spectral tamination from spectra, based on a simple convolution model sharpness. GEORGES: PRACTICAL SKYWAVE SEA-STATE RADAR 757

Sortinq by Sharpness spectral estimators, particularly the maximum entropy method I1 I! (MEM), have offered the promise of spectral resolution greatly 1 16 May 1978 ’I\ - N=W W = 0.047 HZ exceedingthat available from an FFT, for a given record length [45]. Consequently, a number of workers in sea-state radarhave investigated the properties of MEM spectra (all- pole models) derived from short radar data sets. Although Bragg frequenciescan be accurately measured with the MEM, attempts to recover accurately the shape of the first- and second-order spectra, or even their relative powers, havegenerally led to disappointingresults. Evidently, the method very accurately estimates the frequencies of multiple sinusoids in additive noise, but it fails to reproduce the spectral shape of broadenedsignals. In an all-pole model, the power attributed to each pole is closely related to the sharpness of the spectrum of that pole, so that power and sharpness cannot vary independently. If a spectral feature contains a relatively

I: I I large amount of power, the MEM will necessarily represent it + 1.25 0 - 1 25 by a very sharp spectral line, even though the actual feature Doppler Shift. Hz may be broader. Fig. 3. A quality index derived from equivalent spectral width can be Anotherproblem with MEM is that no oneyet under- used to automatically sort sea-echo spectra. Each spectrum shown standsthe relation between spectral confidence limits of here is an average of 84 individual spectra measured simultaneously MEM estimates and the number of polesused to model the over a 60 by 60 km radar grid. The upper spectrum is too broadened by ionospheric multipath to permit separation of first- and second- process (that is, the number of prediction filter coefficients). order features, whereas the lower spectrum, obtained 6.8 min later, So far, MEM doesnot appear to provide any processing can be successfullyanalyzed for rms waveheight. The difference advantage over an FFT for extracting spectral shape. It may, between-the two is quantitatively indicated by the averageequivalent however, prove a useful tool for accurately estimating Bragg width (W)of the 84 component spectra In general, if an ensemble of spectra has an average equivalent width less than about 0.03 Hz, frequencyshifts for current measurements. At present, the its average is likelyto be of sufficient quality for waveheight analysis. problemof ionospheric frequency shifts during an FFT in- tegrationtime appears to bebest solved by avoiding such times. C. Adaptive Searching Schemes Radar signal processingusually refers to what you do to VII. RADAR SYSTEM CONSIDERATIONS the receiver’s outputafter the radar’s operating parameters Radars with very narrow azimuthal beams now appear to have beendetermined. However, when the signal’s quality haveasignificant advantage in reducingF-layer multipath varies with these parameters, it is useful to provide real-time contamination [39]. Withbeams larger than about 2O, con- feedbackbetween a quality decision (which may be made taminated spectra will be obtained most of the time in which after some signal processing) and the choice of radar param- F-layer reflections are used, and beams as small as 0.5’ offer eters. In sea-state monitoring with skywave radar, a computed significant reduction of contamination. spectralsharpness index [39] canbe used to guide an auto- Frequency-modulatedcontinuous-wave (FM-CW) radars maticsearch over a limited range-azimuth-time space, until offer significant signal-to-noise improvements in sea-echo spec- enough data of sufficient quality are obtained. Without such trabecause of their100-percent duty cycle, compared with feedback, a radar operates blindly and can waste valuable time thetypical 1-2 percent of pulseradars. Other advantages in cells where the ionospheric contamination is momentarily includea relative immunity to interference from (and to) unacceptable or in cells where enough good data have already thecrowded HF spectrum [46] anda greater flexibility in beenobtained. However, designing a grid-scanning strategy selectingrange and Doppler resolution [47]. Adisadvantage requires advance knowledge of the spatial and temporal cor- is the requirementfor isolated transmitting and receiving relation of spectral quality under various ionospheric condi- stations. tions, and such measurements have yetto be made. Digitalrecording of demodulated radar data is nowpre- Such adaptive searching schemes, currently under develop- ferred over the analog tape recording sometimes used because ment, will permit the first valid assessment of the limitations wow and flutter of analog machines are seldom low enough thatthe ionosphere imposes on datacollection. A more to prevent Bragg-line broadeningand masking of second- elaborateoperational scheme might employ two or more order spectral structure, which may lie less than 0.1 Hz away independentradars to divide upthe searching and data- froma Bragg lineand be 10 dB down from it. For similar collection functions, reasons,master-oscillator phase stability has been found to bea critical factor in determiningthe frequency resolution H. Nonclassical Spectral Estimation in both surface and skywave radar systems. Therequirement that second-order spectral structure be resolved dictates a coherent integration time of at least 50 s, VIII. ACTIVE EXPERIMENTAL PROGRAMS whena classical spectralestimator such as the fast Fourier Inthe United States, the principal active skywave sea- transform (FFT) is used. Because ionospheric changes can be state-radarprogram is acooperative effort between the Na- significantover such an integration time, spurious spectral tionalOceanic and Atmospheric Administration and SRI structure can arise from the nonstationarity of the radar echo International. NOAA’s skywave radar on San Clemente Island (“temporalmultipath”). Recently developed nonclassical [431 hasbeen abandoned in favorof SRI’s Wide Aperture 758 TRANSACTIONSIEEE ON ANTENNASPROPAGATION,AND VOL. AP-28, NO. 6, NOVEMBER 1980

Research Facility, a bistatic FM-CW radar in central California 29 I I I withnominal 0.5O angularresolution. This facility is de- 2140 Z 31 AUGUST 1977 1 scribed by Washburn et al. [481.The combination of high WARF-DERIVED WARF angularresolution and FM-CW processing give thisradar a WINDDIRECTIONS nominal resolution cell of 3 km (radial) by 15 km (transverse). Maresca and Georges [ 251 and Lipa et al. [26] describe some recentWARF measurements of rms waveheightand ocean- wave spectra at 1600-km range in the north Pacific. Inaddition, some U.S. workcontinues at the Naval Re- searchLaboratory east-coast facility described by Headrick and Skolnik [3]. Future measurements will employ the ITT 1.24-km White House receiving array [ 21 for improved resolution in attempts to measure offshore currents using sporadic- I I I E reflections (D. B. Trizna, private communication, 1979). 95 94 93 92 91 90 LONGITUDE - ‘W In theUnited Kingdom, ajoint program between the UniversityofBirmingham and the Appleton Laboratory operates a monostatic pulse radar near Swindon in southwest England. This facility uses a single 49-element steerable array with nominal 6’ beamwidth that can look either eastward or westward.Recent results are described by Shearman et al. [491. A French program at the Universite de Paris uses a mono- staticpulse radar in southeastern France to look northward into the North Sea. The transmitting antenna has a nominal -t beamwidth of IO’, and a separate 32-element receiving array has a nominal 4.5’ beam (both measured at 10 MHz) steerable over k36’ (J. Delloue, private communication, 1979). 24

Innorthern Queensland, Australia, a skywave program at , 0 1aJ theJames Cook University uses mechanicallya steerable I I I 97 95 93 91 89 21.84-MHz Yagi array with a nominal beamwidth of 12’ [ 501. LONGITUDE - % Recent wind-field mapping results are described by Dexter and Fig. 4. When hurricane Anita moved into the Gulf of Mexico, the SRI Casey 1151. WARF radar began producing wind-direction maps like the one at Finally, the Australian government is constructing a large the top to monitor the storm’s circulation pattern and to track its andpowerful skywave defense radar near Alice Springs in center. At a range of about 3000 km from the California radar, the central Australia to monitor ship traffic in the waters north of size of each radar celi is about 15 X 25 km. Wind directions meas- the continent [4]. Known as Project Jindalee and directed by ured at NOAA data buoy EB-71 are shown for comparison, based on the assumption that simple storm translation permits the space- the DefenceResearch Centre Salisbury, the bistatic FM-CW time transformation shown. Storm centers derived from seventeen radar wiU closely resemble the SRI WARF system but will be such maps are compared at thebottom with the track of Anita esti- atleast ten times as powerful. Wind-direction measurements mated by the NOAA National Hurricane Center (NHC) using micro- have already been carried out with a prototype radar, and a wave radar, aircraft, and satellite sensors. The average difference between the two tracks is 19 km (after Maresca and Carlson 52). sea-state monitoring capability will be developed as a second- ary radar mission. Efforts to determine whether the Soviet Union is using any ofits HF-OTH radars for sea-state studies have not been B. Measuring rms Waveheight andScalar Ocean- WaveSpectrum successful. Waveheightand wave-spectrum measurements by skywave radarcan be verified by comparing them with simultaneous IX. RECENT SIGNIFICANT RESULTS buoy measurements of the same quantities. Fig. 5 shows such a comparison during the passage of a cold-weather front past The following results obtained with the SRI WARF system NOAA data buoy EB-20 in the north Pacific Ocean. At pre- demonstrate the measurement capabilities now possible with a sent,the estimated errors in both the buoy and the radar high-resolution skywave radar. measurementsare larger than the disagreement between the two.

A. Tracking Tropical Storms C. Swell Measurements Mapping wind-direction fields finds one of its most useful Swell is the very directional, long-period remnant of wind applicationsin locating and tracking tropical storms [52] waves generated by distant storms at sea. When the height of (Fig. 4). Many such wind-direction maps can be produced in a shortwind waves is small, and swell dominates, the second- 24-hourperiod, so thatthe storm center can be tracked in orderDoppler spectrum sometimes reveals distinct “swell near-real-time.Errors in locating storm’sa center seldom peaks”very close to the Bragg lines. Because the direction exceed 40 km and result mainly from inaccuracy in locating of swell travel is unrelated to that of the short locally driven the radar cell (caused by ionospheric tilts and correctable with wind-wavesin a radar cell, theamplitude ratio of innerto referencetransponders) and from uncertainty in locating outer second-order swell sidebands can be reversed relative to the centerof rotation of the vector field. thatpredicted for wind waves. This inverted ratio can be GEORGES: PRACTICAL SKYWAVE SEA-STATE RADAR 759

-- FROM AT 1800 2 42 I -

144 142 140 138 136 134136 138 140 142 144 LONGITUDE - OW

I I I I11111 I Ill I A EB20 ’. B EB 20 WARF

W 0

0.05 0.10 0.50 0.05 0.10 0.50 1.0 0.5 0 -0.5 -1.0 WAVEFREQUENCY - Hz DOPPLER - Hz Fig. 5. A joint NOAA/SRI experiment using the SRI WARF skywave radar in California measured these wave conditions 1600 km out in the North Pacific Ocean near NOAA data buoy EB-20. At the upper left is a map showing the location of EB-20, three radar cells, 1 labeled A, B, and C, and the 1800 Z location of a cold front movingeastward past the buoy’s location. The three Doppler spectra on the right show the differences in second-order spectral structure in the echoes from the three radar cells. The spectrum at the lower right appears to be contaminated (arrows) by an image of the Brag lines caused by multipath propagation. The two dotted ocean-wave spectra at the lower left were computed from the radar spectra obtained from points A and B and show an increase in long-wave energy (andrms waveheight) following the front’s passage. The solid lines show the spectrum measured at EB-20 at nearly the same time (after Maresca and Georges[ 25 1 ).

regardedas a swellsignature (Fig. 6). Resolvingthese peaks 0 with a skywave radar usually requires very stable (E,) ionospheric reflections. -:o - A theoreticalmodel recently developed by Lipa and Bar- rick [34] predicts that swell period, height, and direction can beextracted from accurate measurements of thelocation 2 -20 - andpower in the four swell sidebands. For the caseillustrated 2 - bythe spectrum of Fig. 6, agreementof 10 percent is claimed 2 - between in situ andradar-derived period, height, and direction E -33 Y.. [IT]. B L -40 - X. OPERATIONAL CONSIDERATIONS w hat practical use would the kinds of sea-state meas-sea-state of kinds the would usepractical Of what I urementsthat ground-basedwouldHF And make? radars 1.0 c.5 0 -0.5 -1.0 measurements become obsolete when ocean-monitoring when obsolete become measurements DOPPLER-Hz satellitesbecome operational? Fig. 6. A sea-echo spectrum obtained with WARF via E, propagation Surface-waveradars provide detailed wave andsurface- revealssecond-order swell peaks, (approximately kO.1 Hz from the current maDSOver a fewthousand Sauare kilometers of sea Bragg lines) inwhich the ratio of inner-to-outer sideband amplitudes extending iPto km from shore~~Contemplatedapplica- is inverted, compared with that whichwould be obtained if the second-order echoes were from locally generated wind waves. Swell tionsinclude tracking surface-borne pollutants (such as oil), period, height, and direction can be extracted from the power and environmentalimpact assessment, platform-design studies, positions of the four sidebands (after Lipa e? al. [26] ). 760 TRANSACTIONSIEEE ANTENNASON AND PROPAGATION,VOL. AP-28, NO. 6,NOVEMBER 1980

monitoringthermal pollution bypower plants, and studies HF coastal ground-wave radars: A progress review.” Ocean Wave of fish-egg transport and beach erosion. Climate. Earleand Malahoff. Eds.. New York:Plenum, 1979. D. E. Barrick. “The use of skywave radar for remote sensingof sea Skywaveradars can illuminate millions of squarekilo- states,“ MTS J.. vol. 7, pp. 29-33. 1973. meters of openocean to watch for developing storms and R. S. Rhodes. “A previewofbenefits from skywaveradar to monitor regions of high seas, not only for the benefit of measurements of sea state,” MTSJ.. vol. 9. pp. 29-33. !975. fishingand shipping interests, but also to permit more ac- D.E. Barrick, “First-order theoryand analysis of MF/HF/UHF scatter from the sea.” IEEE Trans. Antennas Propagat.. vol. AP- curateforecasts of coastalstorm damage. For example, a 20. pp. 2-10, 1972. singleradar stationin Tennessee, with two antenna systems -. “Remote sensing of seastate by radar,” in RemoteSensing of for looking southward and eastward, could monitor and track theTroposphere. V. E. Derr.Ed. Washington. D.C.:U.S. most of the hurricanes that threaten the U.S. Gulf Coast and GovernmentPrinting Office. 1972, ch. 12. the eastern seaboard. A. E. Long and D. B. Trizna. ”Mapping of North Atlantic winds by HFradar backscatter interpretation.” IEEE Trans.Antennas Orbiting sea-state sensors, when they become available, will Propagat.. vol. AP 21. pp. 680-685, 1973. complement surface-based radars, in that each is designed for J. L. Ahearn, S. R. Curley. J. M.Headrick. and D. B. Trizna. different coverage. Restricted by their orbital dynamics, satel- “Tests of remoteskywave measurements of oceansurface con- lites provide coarse global coverage, often in areas inaccessable ditions.” Proc. IEEE. vol. 62. pp. 681487. 1974. tosurface sensors!whereas ground-based radars can provide J. R. Barnum. J. W. Maresca. Jr.. and S. M.Serebreny. “High- resolution mapping of oceanic wind fields with skywave radar.” detailedlocal coverage and continuously monitor areas of IEEE Trans. Antennas Propagat.. vol. AP-25. pp. 128-132. 1977. special interest. P. E. Dexterand R. Casey.“Ocean windfield mapping .. at long For severalyears, thecentral question about skywave ranges with an HF radar,” Austral. Meteorol. Mag.. vol. 26. pp. sea-state radar has been, “What fundamental limitations does 33-44,1978. the ionosphere impose on sea-state measurements?” or more J. W. Maresca. Jr. and J. R. Bamum. “Remote measurements of the position and surface circulation of humcane Eloise by skywave simply stated, “Will skywave sea-state radar work?” We now radar.” Monthly Weather Rev., vol. 107, pp. 1648-1652. 1979. knowthat wecan map wind-direction fields with whatever J. W. Maresca. Jr. and C. T. Carlson. “HF skywave measurement spatialresolution we want to build into our radars. On the ofhurricane winds and waves.” in Proc. 16th Conf. Coastal other hand, it appears that we will not soon be able to meas- Engineering, Hamburg.Germany. 1978. pp. 190-108. -. “HF skywrave radar track of tropical storm Debra.” Monthly ure full wave-directional spectra routinely. The usefulness of U’eather Rev.. 1980. to be published. skywave radar thus seems to hinge on its ability to routinely -. “Trackinghurricane Anita by HFskywave radar.” J. measure rms waveheight and, less frequently, the scalar ocean Geophys. Res.. 1980. to be published. wave spectrum. R. H. Stewart and J. W.Joy. “HF radio measurements of ocean Onlyfive years ago the prospects for routine waveheight surface currents.“ Deep-sea Res., vol. 21. pp. 1039-1W9. 1974. D.E. Barrick. M. W. Evans.and B. L. Weber.“Ocean surface measurementswere bleak, based on themeasurements of a currents mapped by radar.” Science. vol. 198, pp. 138-144. 1977. fewbroad-beam radars that used inadequate ionospheric J. W. Maresca. Jr. and C. T. Carlson. “A comment on ‘Longshore diagnosticsand primitive signal-processing methods. Today, currents onthe fringe ofhurricane Anita.”‘ by Ned Smith. J. theability to measure rms waveheight using F-layer reflec- Geophys. Res.. vol. 85. pp. I64gl64 I. 1980. D. E. Barrick. “HF radio oceanography-A review.“ Boundav tionshas been demonstrated in a number of experiments, LayerMeteorolog!. vol. 13. pp. 2313. 1978. so thatthe major remaining question is whatspatial and -. ”Extraction of wave parameters from measured HF sea-echo temporal coverage theionosphere will permit. In viewof Doppler spectra.” Radio Sci.. vol. 12. pp. 415424. 1977. presentspot-measurement capabilities, we can expect an J. W.Maresca. Jr. and T. M. Georges.“Measuring rms adaptivesearching strategy to fill a radar sector with wave- waveheight and the scalar ocean wave spectrum with HF skywave radar.” J. Geophw. Res.. vol. 85. pp. 2759-2771. 1980. heightmeasurements at least once a day. The experiments B. J. Lipa. D. E. Barrick. J. W. Maresca. Jr.. and C. C. Teague, nowbeing implemented to establish these coveragelimita- “HF radar measurements of ocean swell parameters.“ J. Geophys. tions more precisely could not be successfully performed with- Res.. 1980. to be published. out the fullcomplement of diagnostic and signal-processing R. H. Stewart and J. R. Barnum. “Radio measurements of oceanic tools that have marked the progress over the last few years. windsat long ranges: An evaluation.“ Radio Sci.. vol. IO. pp. 853-857. 1975. J. W. Maresca. Jr. and J. R.Bamum. “Measurement of oceanic ACKNOWLEDGMENT wind speed from HF sea scatter by skywave radar.” IEEE Trans. Antennas Propagat.. vol. AP-25, pp. 132-136, 1977. I thank the various skywave-radar teams mentioned earlier D. E. Barrick. “Ocean waveheightnondirectional spectrum from fortheir hospitality and cooperation in helping me prepare inversion of the sea-echo Doppler spectrum.” Remote Sensing of this review. Environment, vol. 6, pp. 201-227. 1977. K. Hasselmann.“Determination of ocean-wave spectra from Dopplerradio return from the sea surface.” NaturePhysical REFERENCES Science. vol.229. pp.16-17, 1971. [I] E. C. Hayden, ”Ground scatter in review.” inscatter Propagation D. E. Barrick and B. J. Lipa. ”A compact transportable HF radar of Radio Waves. AGARD Conf. Proc.. no. 37. 1968. system for directional coastal wave fieldmeasurements.“ Ocean [21 T. A.Croft. “Sky-wave backscatter:A means of observing our WaveClimate, Earleand Malahoff. Eds. New York: Plenum. environment at great distance.” Rev. Geophys. Space Phys.. vol. 1979. IO. pp. 73-155. 1972. B. J. Lipa.“Derivation of directionalocean-wave bpectraby [3] J. M. Headrickand M. I. Skolnik.“Over-the-horizon radar in the integral inversion of the second-order radar echoes.” Radio Sci.. HF band.” Proc. IEEE. vol. 62, pp. 664-673, 1971. vol. 12. pp. 425434. 1977. [4] D.Ball. ”Jindalee-Over-the-horizon radar in thedefence of -. “Inversion of second-order radar echoes from the sea,” J. Australia.” Pacific Defence Reporrer. pp. 17-22. Feb. 1977. Geophvs. Res.. vol. 83. pp. 959-962. 1978. [5] D.D. Crombie. “Doppler spectrum of sea echoat 13.56 Mc/s.“ B. J. Lipaand D. E. Barrick.“Methods for the extraction of Nature. vol. 175.pp. 681482. 3955. long-period ocean wave parameters from narrow beam HF radio sea [6] IM.W. Evansand T. M.Georges, “Coastal ocean dynamics radar echo.” Radio Sci.. vol. 15. pp. 843-853.1980. (CODAR). NOAA‘s surface currentmapping system.” in Pror. K. Daviesand D. M. Baker. “On frequencyvariations ofion- Oceans 79. IEEE Conf.. San Diego. 1979. ospherically propagated HF radio signals.” Radio Sci.. vol. I. pp. [7] D. E. Barrick and B. J. Lipa. ”Ocean surface features observed by 515-556. 1966. G EO RG ES : PRACTICAL GEORGES: RADARSKYWAVE SEA-STATE 76 1

T. M. Georges,“HF Dopplerstudies of travelingionospheric [50] J. F. Ward and P. E. Dexter, “A long range ocean radar for ocean disturbances,” J. Atmos.Terrest. Phys., vol. 30,pp. 735-746, surfacestudies usingbackscatter via theionosphere,” Aust. J. 1968. Phys., vol.29, p. 183.1976. M. L. Heron and R. J. Rose. “A resolution limitation on sea-echo [51] 0. G. Villard. “Over-the-horizon or ionospheric radar,” QST. pp. radarspectra inferredfrom point to pointionospheric Doppler 39-43, April, 1980. broadening,” Geophys. Res. Letters, vol. 5. pp. 379-381. 1978. [52] J. W. Maresca, Jr. and C. T. Carlson,“High-frequency Skywave G. F. Earl and I. A. Bourne, “Spectral characteristics of HFground radarmeasurements of hurricane Anita,” Science. vol.209, pp. backscatter,”J. Atmosph. Terrest. Phys.. vol. 37. pp. 1339-1347, 1189-1 196, 1980. 1975. [53] J. W. Maresca, Jr.. R.R. Padden, R. T. Cheng.and E. Seibel, T. M. Georges and J. W. Maresca, Jr.. ”The effects of space and “HF radar measurements of tidal currents flowing through the San time resolution on the quality of sea-ech’o Doppler spectra measured Pablo Strait, San Francisco Bay,“ Limnology and Oceanography, with HF skywave radar,” Radio Sci., vol. 14, pp. 455-469, 1979. vol. 25. pp. 929-935,1980. G. F.Earl, “Synthesis of HF groundbackscatter spectral [541 D. E. Barrick,“Accuracy of parameterextraction fromsample- characteristics.” J. Atmosph.Terrest. Phys.. vol.37, pp. 1444. averagedsea-echo Doppler spectra.” IEEE Trans.Antennas 1562,1975. Propagat., vol. AP-28. pp. 1-1 1, 1980. T. M. Georges. J. W. Maresca. Jr.. C. T. Carison, J. P. Riley, and [55] -. “Logarithmicnormalization to remove unknown signal-gain R. M.Jones, “Recovering ocean waveheightfrom HF radarsea factorsbefore averaging.” NOAA Tech.Memo ERLWPL-56, echoes degraded by imperfectionospheric reflection,” NOAA 1980. Tech. Memo. 1980, tobe published. [56] T. M. Georges and R. M. Jones. “A mathematical model for how R. E. DuBroff. N. N. Rao, and K. C. Yeh. “Backscatter inversion ionosphericreflection degrades sea-echo spectra.” NOAA Tech. in sphericallyasymmetric ionosphere,” Radio Sci.. vol. 14. pp, Memo ERL WPL-60. 1980. 837-841,1979. D. E. Barrickand J. B. Snider,“The statistics of HF sea-echo Doppler spectra,” IEEE Trans. Antennas Propagat.. vol. AP-25, pp. 19-24, 1977. A. V. Oppenheim and R. W. Schafer, Digita! Signal Processing. Englewood Cliffs, NJ: Prentice-Hall, 1975. J. Makhoul, “Linear prediction: A tutorial review.” Proc. IEEE, T. M. Georges was born in Los Angeles. CA, on vol. 63, pp. 561-580,1975. August 8, 1938. He receivedthe B.S.degree G. H. Barry and R. B. Fenwick, “Extraterrestrial and ionospheric from Loyola University of Los Angeles in 1961 sounding with synthesized frequency sweeps,” HP J.. vol. 16. pp. andthe 1M.S. degreefrom University of Cali- 8-1 I. 1965. fornia. LosAngeles, in 1963.both in engi- D. E. Barrick. “FM/CW radar signals anddigital processing,” neering. In 1967. he received a Ph.D. degree in NOAA Tech. Rep. ERL 283-WPL 26. 1973. engineeringfrom the University of Colorado. T. W. Washbum, L. E. Sweeney. J. R. Barnum. and W. B. Zavoli, Boulder. “Developmentof HF skywave radar for remote sensing appli- Since 1963he has worked atthe Boulder cations.” in AGARD Conf. Proc., No. 263. 1979. Laboratoriesof the U.S. Department of Com- E. D. R. Shearman,W. A.Sandham. D. J. Bagwell. S. merce,first with theNational Bureau of Theodoridis. P. A. Bradley, E. N. Bramley. F. D. G. Bennett, and Standards, and now withthe Wave Propagation Laboratory, National C.Carter, “Sky-wave HF radar measurements of sea-stateand Oceanic and Atmospheric Administration. He is the author or co-author wave direction,“ in ACARD Conf. Proc., No. 263. pp.30-1-30- of over 40 paperson radio and acoustic-gravity-wave propagation, 1 I. 1979. ionospheric dynamics. meteorology. acoustics. and radar oceanography.