2_ooli?g? 6(0 55/36 q iii'7 _"!-1'7 2 _ Sea Ice Remote Sensing Using Surface Reflected GPS Signals Attila Komjathy', James Maslanik t, Valery U. Zavorotny 2,Penina Axelrad tand Stephen J. Katzberg a _CCAR/University of Colorado, CB 431, Boulder, CO 80309 2NOAA/Environmental Technology Laboratory, 325 Broadway, Boulder, CO 80303 3NASA Langley Research Center, Code 328, Hampton, VA, 23681-0001 Phone: (303) 492-4829; Fax: (303)-492-2825; Emaih komjathy @colorado.edu ABSTRACT This paper describes a new research effort to suppressed at delays of more than + 1 chip (300 m for the extend the application of Global Positioning System (GPS) civilian (C/A) code). To acquire and track a signal, a signal reflections, received by airborne instruments, to conventional GPS receiver generates a local replica of the cryospheric remote sensing. Our experimental results code for the particular satellite signal to be tracked, indicate that reflected GPS signals have potential to provide compensating for the expected Doppler shift and computing information on the presence and condition of sea and fresh- the cross-correlation between the replica and the incoming water ice as well as the freeze/thaw state of frozen ground. codes. In this paper we show results from aircraft experiments over the ice pack near Barrow, Alaska indicating correlation When a GPS signal encounters an ideal, smooth reflecting between forward-scattered GPS returns and RADARSAT surface, specular reflection occurs at a single point. The backscattered measurements. reflected signal code structure remains, but the polarization of the wave is reversed to left-hand circular polarization INTRODUCTION (LHCP), and the signal power is decreased. If, however, the surface is rough relative to the GPS wavelength of 19 cm, In 1996, Drs. Katzberg and Garrison of NASA Langley reflections are produced by multiple facets on the surface. Research Center (LaRC) developed the idea of using Around the ideal specular point, this creates a so-called reflected GPS signals for remote sensing applications, and glistening zone within which there is a distribution of varying published a number of papers describing the theory and ranges and Doppler shifts, as shown in Fig. 1. To measure mechanisms for this technique [e.g., 1-3]. Since then, GPS signals reflected from land and sea surfaces, Garrison research has advanced our understanding of reflected GPS and Katzberg [1] modified a typical receiver to measure signals and experimentally applied those techniques to ocean correlation power at these offset values of delay and Doppler remote sensing and mapping [4-101. The current investigation using a nadir-pointing LHCP antenna. extends this work into cryospheric applications of GPS reflections based upon a proposal made by Drs. Katzberg and In Fig. 2, we illustrate the type of swath coverage that GPS Garrison, to collect GPS signaIs reflected by ice surfaces. provides. Because of the nmltiplicity of GPS transmitters the Results are shown from the first GPS measurements from sea receiver can observe a number of simultaneous footprints on ice in the Beaufort Sea in April 1998. the surface. This illustrates that while the system does not provide imaging in a standard sense (like a cross-scanning MEASUREMENT METHODS system, for example), it does provide more spatial coverage than from a sensor that is only able to view one ground point The use of GPS in a bistatic radar configuration to measure at a time. surface properties relies upon our ability to extract information from the reflected signal. For standard GPS navigation purposes, the GPS receiver measures the signal delay from the satellite (the pseudorange measurement) and the rate of change of the range (the Doppler measurement). For remote sensing applications, the primary measurement is the received power from the reflected signal for a number of delays and Doppler values. This measurement and its sensitivity to the surface conditions are discussed in the following paragraphs. The GPS satellites transmit two right-hand circularly polarized (RHCP) L-band signals at 1.57542 GHz and 1.2276 Figure !. Illustration of glisiening zone (adapted from GHz. These carrier signals are modulated by unique Komjathy et al., 1999). pseudorandom noise (PRN) codes with an autocorrelation function similar to the ideal autocorrelation function. The autocorrelation power has a triangular shape, which is Presented at IEEE International Geoscience and Remote Sensing Symposium, Honolulu, Hawaii, 24-28 July 2000 --¸ ........... measurements of reflected signals using a downward-pointed LHCP antenna. DMR makes measurements of the correlation between the reflected signal and shifted versions of the local signal replica. This provides a trace of the reflected signal correlation as a function of code delay, thus mapping the return from annular regions of the glistening zone. The first measurements of reflected GPS data from sea ice were collected by Dr. Maslanik in the Beaufort Sea in April 1998. During the experiment, the GPS system was tested ,%f with antennas mounted on a boom and held over various ice surfaces at different heights ranging from 30 cm to 20 m. Each antenna was mounted on a small 20 cm by 20 cm Figure 2. GPS swath coverage. aluminum ground plane, with the two plates then attached to Results and observations suggest that the reflected GPS provide the up-looking and down-looking mount for the antennas. A second reflected GPS data set was collected signal is very sensitive to the presence of sea ice, and furthermore, contains information related to the reflection north of Barrow, Alaska in April 1999, using the same system coefficient of different ice conditions. but in this case mounted on a Cessna 185. Approximately 2 hours of data were collected in conjunction with a National The reflection coefficient of a frozen sea surface is Ice Center (NIC) reconnaissance flight and nearly coincident determined by the effective dielectric constant of ice and, in time with aRADARSAT overpass. under some conditions, by the dielectric constant of the Our experiment to observe GPS reflections from Arctic sea underlying water. The latter is important for thin ice ice showed that for the moderate altitudes of the airborne conditions where a significant portion of the radio wave energy may reach the second interface, between the ice and GPS receiver, the reflected signal shape was fairly consistent water. The effective dielectric constant of ice depends on (sharp, narrow waveform throughout the flight) indicating various factors, such as an ice composition, salinity, that ice surface roughness variations were not significant at L-band. On the other hand, the peak power was found to temperature, density, age, origin, morphology, etc. [11]. change significantly along the flight track. This behavior of Because first-year ice has needle-like inclusions of brine (predominantly oriented along the vertical direction), this the signal is a clear indication of the sensitivity to ice variety of ice is notably anisotropic. Therefore, the dielectric reflection coefficients. Correlations seen in our comparison between RADARSAT backscatter and GPS forward scattered constant is a tensor rather than a scalar constant. This makes data indicate that the GPS signal may indeed provide useful the problem of modeling and interpreting the GPS signal information regarding ice conditions, in addition to the basic scattering from ice rather challenging. At the same time, the ability to detect the presence of sea ice. sensitivity of the GPS scattered signal to these complex ice characteristics makes it a particularly attractive remote Fig. 3 shows the peak return power from GPS satellite sensing tool. PRN 30, collected during the flight. Reflected power is Given the unique potential offered by reflected GPS signal plotted versus RADARSAT backscatter along the flight leg, sensing (e.g., L-band observations, forward-scattering as a function of along track epochs. The two sets of curves sensing, low-cost and simple devices useable on any type of indicate a positive correlation (see Labels 1to 6) between the aircraft, etc.) and the geophysical significance of additional forward-scattered GPS peak power and the backscattared sea ice and permafrost data, research to investigate and RADARSAT data for locations of high reflection exploit GPS returns from ice and frozen ground is warranted. coefficients, as well as some other variations in the GPS data However much remains to be done to determine the that differ from the RADARSAT backscatter. Observation of consistency of the signal return from ice surfaces; to positive rather than negative correlation requires further investigate the information content of the signals; and to investigation. develop simplified retrieval algorithms to extract the cryospheric conditions. In Fig. 4we present a photo taken from the aircraft during the flight. The picture were taken when the aircraft flew by Label EXPERIMENTS AND DATA ANALYSIS 5. The correspondence of the reflected peaks with the occurrence of large leads in the same area suggests that the Most of the GPS reflection experiments to date have been high signal returns may have originated from patches of new conducted with the Delay Mapping Receiver (DMR) or young ice as observed from the aircraft. In Fig. 4, we also designed by Drs. Katzberg and Garrison [i], based on the indicated an illustration of the possible specular reflection GEC Plessey (now MITEL) GPSBuilder-2 [12]. The DMR track crossing a region of recently formed ice resulting in has two antenna inputs to allow standard tracking of direct high GPS signal returns shown as gray line. Low signal signals using a zenith-pointed RJJCP antenna and correlation returnasroundthisregionmayhavebeenduetothickerf,irst REFERENCES yearicecondition(sdarkline). [1] J.L. Garrison, S.J. Katzberg, and C.T. Howell, "Detection 0,7 ......................................................................................................... _........................... of ocean reflected GPS signals: theory and experiment," Corresponding peak_ . In the Proceedings of the IEEE Southeastcon 1997, Blacksburg, USA, pp. 290-294, 12-14 April 1997. i 0.5-----i--_2_ P 3 .___ _ -_ [2] J.L. Garrison, S.J. Katzberg, and M.I. Hill, "Effect of sea _ 0.4 6 --i roughness on bistatically scattered range coded signals from the Global Positioning System," Geophys. Res. Letters, Vol. 25, No. 13, pp. 2257-2260, 1998. [3] SJ. Katzberg and J.L Garrison "Utilizing GPS to determine ionospheric delay over the ocean," NASA 4113 (-_ 8(]0 I000 1200 I400 1600 IN(X) Tech. Memo.- 4750, 1996. ,tdong Trek I_'lx_, [4] S.J Katzberg, J.L. Garrison, N.C. Coffey, and H.R. Figure 3. Normalized signal power for PRN 30. Kowitz, "Method and system for monitoring sea state using GPS," Application filed with the U.S. Patent and Trademark Office, 8January, 1998. [5] B. Lin, S.J. Katzberg, J.L. Garrison, and B. Wielicki, "The relationship between the GPS signals reflected from sea surface and the surface winds: Modeling results and comparisons with aircraft measurements," J. Geophys. Res., Vol., 104(C9), pp.20,713-20,727, 1999. [6] S.F. Clifford, V.I. Tatarskii, A.G. Voronovich, and V.U. Zavorotny, "GPS sounding of ocean surface waves: theoretical assessment," In Proc. Int. Geosci. and Remote Sens. Syrup. (IGARSS), Vol. IV, Seattle, USA, pp. 2005-2007, 1999. [7] V.U Zavorotny and A.G. Voronovich, "Scattering of GPS signals from the ocean with wind remote sensing application," IEEE Trans Geosci. Remote Sens., Vol. 38, No. 2, pp.951-964, 2000. [8] V.U. Zavorotny and A.G. Voronovich, "Bistatic radar Figure 4. Photo taken from the aircraft in the vicinity of Label scattering from an ocean surface in the small-slope 5 showing black and darker gray areas of newly formed approximation." in the Proceedings of the IEEE leads, new and young ice. International Geoscience and Remote Sensing Symposium: Remote Sensing of the System Earth - A CONCLUSIONS Challenge for the 21st Century, 99CH36293, pp. 2419- 2421, IEEE, Piscataway, N. J., 1999. Field experiments, including data acquired from aircraft [9] A. Komjathy, V.U. Zavorotny, and J.L. Garrison, "GPS: flights over the ice pack near Barrow, Alaska, suggest that the a new tool for ocean science," GPS World, Vol. 10, No. reflected GPS signals contain useful information over sea ice. 4, pp. 50-56, 1999. Given this new potential application for GPS remote sensing, [10] A. Komjathy, V.U. Zavorotny, P. Axelrad, G.It. Born an effort to investigate reflected GPS signals over sea ice is and J.L. Garrison, "GPS signal scattering from sea discussed in the paper. A combination of modeling surface: wind speed retrieval using experimental data and considerations, in-situ measurements and aircraft theoretical model," J. Remote Sens. Env., 2000 (in observations have been presented to quantify the theoretical press). and observed relationships between reflected GPS signals and [11] M.E. Shohr, "Field observations and model calculations cryospheric conditions. of dielectric properties of Arctic sea ice in the microwave C-band," IEEE Trans. on Geosci and Remote Sens., Vol. ACKNOWLEDGMENTS 36, No. 2,pp.463-478, 1998. [12] GEC Plessey Semiconductors, Global Positioning The research was funded by NASA Langley Research Center Products Handbook, 1996. (LaRC) under grant no. NAG-I-1927. Thanks are due to A. Ulak of NIC for provision of ice reconnaissance information. RADARSAT data were obtained from NIC and the Alaska SAR Facility. A Comparison of GPS and Scatterometer Sensing of Ocean Wind Speed and Direction Michael Armatys 1, Attila Komjathy 1, Penina Axelrad I, Stephen J. Katzberg 2 1CCAR/University of Colorado, Campus Box 431, Boulder, CO 80309 2NASA Langley Research Center, Code 328, Hampton, VA 23681 303-492-3489 (P) / 303-492-2825 (F) / [email protected] ABSTRACT - Initial estimates of ocean surface wind estimates. QuikSCAT observations provide a snapshot of speed and direction based on observations of reflected the ocean surface conditions at the time of the satellite GPS signals are presented and compared to QuikSCAT overflight. The backscatter measurements at Ku-Band wind fields. The two wind speed estimates are generally are sensitive to rapid changes in the surface conditions in agreement to within 2-3 m/s, and under favorable con- making this a powerful sensor for wind retrieval. Because ditions of well developed seas and stable winds, direction of the substantially different mechanisms for forward scat- estimates agree to within 10 deg. An overview of the GPS tered L-Band reflections, the measurement and retrieval technique is presented as well as a presentation and dis- of wind vectors from GPS may provide a very useful com- cussion of these first results. plement to scatterometer observations in improving global wind observations and models. INTRODUCTION In 1996, NASA researchers Stephen J. Katzberg and GPS-BASED Wi-SFD ESTIMATION James L. Garrison identified the potential application of the Global Positioning System (GPS) to remote sensing As presented in [1, 2, 3], GPS-based wind retrievals rely of ocean surface conditions [1, 2]. Their concept was on the modification of the correlation function for signals to use GPS in a bistatic radar configuration with the reflected by rough surfaces. The basic observable is not GPS satellite transmitting an L-Band spread spectrum the pseudorange or carrier phase measurement used for signal, and the receiver on an aircraft or spacecraft plat- most GPS applications, but rather a measure of the cor- form measuring the reflected signal characteristics. Since relation of the reflected signal with a delayed replica of then, Katzberg and Garrison have continued to investi- the PRN code. The distribution of this function over de- gate the properties of the ocean-reflected GPS signals and lays (and Dopplers) provides a mapping of the reflecting have developed a specialized Delay-Mapping GPS receiver surface. In particular we use a model developed by Za- (DMR) to measure the reflected signals [1, 3]. Using the vorotny and Voronovich (Z-V) [5, 6] for prediction of the DMR and geophysical models to predict the interaction reflected signal structure. The Z-V model embodies the of the GPS signals with sea waterl researchers have been forward-scatter radar equation with the geometric optics able to estimate wind speeds on the ocean surface with limit of the Kirchhoff Approximation. The distribution of an accuracy of 2 m/s [1]. Recently we have developed a ocean surface slopes is assumed to be Gaussian with vari- new approach that extends the GPS-based wind retrieval ances determined by a wave spectra model such as that to determination of wind direction as well as wind speed, developed by Elfouhaily et al. [7]. These slope statistics based on measurements from two or more reflected GPS define the shape and orientation of the glistening zone signals [3]. To validate the performance of GPS for wind over which GPS signals are reflected toward the receiving vector retrieval we have previously relied on buoy data antenna. The models currently used predict the maxi- and wind speed estimates from TOPEX\POSEIDON [4]. mum variance in the up/down-wind direction and mini- mum variance in the cross-wind direction. This allows for This paper describes our first comparisons to wind vec- tors derived from the SeaWinds scatterometer onboard the identification of wind direction with a 180 deg ambi- guity. More advanced models that include up/downwind QuikSCAT. Ku-Band scatterometers represent the cur- rent state-of-the-art in sea surface wind remote sens- asymmetry due to effects such as Bragg scattering are currently under development.. ing, providing highly accurate measurements with good temporal and spatial reso]ut]0n on a global scale. The To estimate wind vectors from GPS observations we first QuikSCAT science team provides a variety of wind prod- determine the surface slope statistics in the model which ucts including wind vector maps at 25 km resolution. best fit the data. The second step is to use an ocean model Nominal accuracy for tile QuikSCAT wind vectors is 2 to retrieve a wind speed and direction estimate from the m/s and 20 deg for wind speed and direction, respectively. slopes. Two satellites are necessary to recover the direc- Performance is somewhat worse for wind speeds below 3 tional information lost in measuring the return from the m/s, for grid points within 25 km of land, and for points entire annulus of the glistening zone with the DMR. In very close to nadir or at large look angles. Algorithms are the current implementation, we try to use satellites with included in the QuikSCAT data reduction that resolve relatively high elevation angles to insure that the spec- wind direction ambiguities and corresponding wind speed ular points are nearly co-located and that the elevation angle-dependesnlotpestatisticsarenearlythesamefor speeds(< 6m/s)throughouta,ndwinddirectionsthat bothsatellites.A degeneratceaseoccurswhenthean- varyover70deg. Furthermorew,indsalongtile flight glebetweetnheincidenpt lanesofthetwoGPSsatellites pathoriginateoverthelandareas.SurprisinglyG, PS isverycloseto90deg.In thissituationonlythemag- windspeedestimatefsallwithin2m/softheQuikSCAT nitudeoftheanglebetweentheincidentplaneandthe estimatesd,espitethelowwindspeedasndalargediffer- up/downwinddirectioncanbedetermineda,ndanaddi- enceinsatelliteelevationsT.heGPS-derivewdinddirec- tionalambiguityexistsintilewinddirection. tionisverystablei,n therangeof80-90degthroughout theflight,notat allreflectingtherapidvariationsseen Thecorrecwt indspeedischosebnyinterpolatingthees- intheQuikSCATestimatesT.herearenumbeorfpossi- timatedslopestatisticsagainsttheslopestatisticsfrom bleexplanationfosrthis,mostimportantofwhichisthe theElfouhailymodel[7]. Underlyingthewavespectra lowwindspeedw, hichmayleadtounreliabledirection andZ-Vmodelsareassumptionasbouttheoceancondi- estimatefsorbothmethods.A secondpossibilityisthe tionsthataffecttheperformanceexpectedIn. particular, longeraveragintgimeinherenitntheGPSL-Bandobser- thecurrentmodeal ssumewselldevelopesdeasw, ithno vationsanddatagrouping.Hourlydatafromallthree swell.Coastael ffectsarenotincluded.Thesesimpli- buoysshowsconsistenwtinddirectionintherangeof90- fyingassumptionlismit theapplicabilityofthecurrent 110degovera twohourperiodsurroundingtheflight implementatiounndercoastaalndhighlyvariableocean tests.AthirdconsideratioisnrelatedtotheGPSsatellite conditions. geometryT.hroughoutthisflight,theazimuthaslepara- tionofthetwoGPSsatelliteswasverycloseto90deg. INITIALRESULTS Asdescribepdreviouslyth,isdegenerate case leads to an OnFebruary28andMarch11,2000D, r. Katzbergcon- ambiguity in the wind direction related to the angle be- ductedtwoflight testsin a B200aircraftofftheVir- tween the wind and the incident plane. For a true wind ginacoasttogatherGPSreflectiondatacoincidenwtith direction of 90 deg, at the start of the experYment (Sample QuikSCAToverflights.Fig. 1 andFig. 3 showthe 1), the ambiguity is such that the possible wind directions QuikSCATwindfieldsandaircraftflighttracksforMarch are only 90 deg and 270 deg; however, for sample 30, the 11andFebruary28,respectivelyT.heFebruary28map two satellite azimuths have shifted by 30 deg, resulting in alsoshowsthelocationof threebuoysforwhichwind possible wind directions of 90, 150, 270, and 330 deg. speedanddirectiondatawasloggedbytheNationaDl ata BuoyCenter(NDBC).(Unfortunatelys,imilarbuoydata CONCLUSIONS forMarch11,2000hasnotbeenarchived.)DMRdata The results presented here represent the first compar- loggedat10Hzweregroupedintoone-minutseegments isons between GPS-derived ocean surface wind vectors numberesdequentiallaylongtheflighttracks.Resultsof and QuikSCAT wind fields. We plan to continue these theGPS-basewdindvectorretrievalsforeachoneminute comparisons with additional flight experiments and anal- sampleonthetwodaysareshowninFigs.2and4. ysis under both favorable and challenging conditions with TheMarch11QuikSCATwindfieldshowsverystable the goal of exploring the possibility of using GPS as a com- windsintherangefrom8.5-10m/s,duenorth(within+/- plimentary sensor to airborne and spaceborne scatterom- 10deg)overtheentireaircraftflightpath.GPS-derivedeters. The use of GPS for wind retrievals, is an emerging winddirectionestimatess,howninFig.2barealsocon- technology with considerable opportunity for further de- sistentlywithinabout5degofNorth.GPS-basewdind velopment of the theory and instrumentation, and signif- speedfsorthefirst10sampleasrequitelow,reportingval- icant potential for contributing to our understanding of uesaslowas5m/sascomparetdotheconsisten1t0m/s regional and global ocean surface conditions. QuikSCATresult.Beginningatsample18,theGPSwind speedestimatemsovetowardtheQuikSCATvalueswith ACKNOWLEDGMENTS mostofthepointsfallingwithinthe2m/serrorrange. The authors would like to thank Dr. Zavorotny for his ThisdatasetpresentasveryfavorablesituationforGPS technical discussions on forward scatter signal models. We versusQuikSCATcomparisonsT.hewindfieldisquite would also like to thank Dr. Elfouhaily for his insight into stableanddoesnotoriginateovertheland;thusthewave wind direction ambiguity resolution. The research was spectramodelassumptionsshouldbevalidunderthese funded by NASA Langley Research Center (LaRC) under conditionsandgoodagreemenistto beexpected.The grant no. NAG-l-1927. discrepancinywindspeedmayberelatedtoarelatively largedifferencbeetweetnheelevationosfthetwosatellites REFERENCES atthebeginningoftheflight.Initiallyonesatelliteisat anelevationof44degwhilethesecondisat63deg.By themiddleoftheflight,thefirsthasrisento53deg.We [1] J. L. Garrison, S. J. Katzberg, and M. I. Hill. "Ef- expectthatamoreaccuratewindspeedestimatecanbe fect of Sea Roughness on Bistatically Scattered Range obtainedusingonlyoneofthetwosatellitesorbymodi- Coded Signals fl'om the Global Posltioning System". fyingouralgorithmtoaccommodadteifferenetlevations. Geophysical Resarch Letters, 25:2257-22601 1998. TheFebruar2y8QuikSCATwindfieldpresenteindFig.3 [2] J. L. Garrison, S. J. Katzberg, and C. T. Howell, III. showsquiteadifferenstituationwithrelativelylowwind "Detection of Ocean Reflected GPS Signals: Theory , i , j_ 1 2[ *l*l**i*e* "** * It i • • • • 38° _',_ ........................ • |-,0 o 1o _o 3o io 5o Figure 2: Wind speed (a) and Wind direction (b) estimates 37_ / from GPS observations for 11 March, 2000 GPS data. 3O e 7s.w _ _ i _ 75°W 74°W Figure 1: Quiver plot of QuikSCAT data for 11 March, 2000. Also shown is the aircraft flight path. and Experiment". In Proceedings of the IEEE South- eastcon '97: Engineering the New Century, pages 290 294, Blacksburg, VA, 1997. [3] "--.x'.,. M. Armatys, D. Masters, A. Komjathy, P. Axelrad, and J. L. Garrison. "Exploiting GPS as New Oceano- graphic Remote Sensing Tool". In Proceedings of the 2000 National Technical Meeting of the Institute of Navigation, Anaheim, CA, 26 28 Jan 2000. In press. [4] A. Komjathy, V.. Zavorotny, P. Axelrad, G. Born, and J. Garrison: "G}'_ Signal Scattering from Sea Surface: Wind Speed Retrieval Using Experimental Data and Figure 3: Quiver plot of QuikSCAT data for 28 February, Theoretical Model". Journal of Remote Sensing o] 2000. Also shown are the aircraft flight path,'and the Envrironment, 2000. In press. locations of the Delaware Bay (DB), Chesapeake Light (CL), and Virginia Beach (VB) buoys. [5] S. F. Clifford, V. I. Tatarskii, A. G. Voronovich, and V. U. Zavorotny. "GPS Sounding of Ocean Surface Waves: Theoretical Assessment". In Proceedings of the IEEE International Geoscience and Remote Sensing , ......,.;.. : Symposium: Sensing and Managing the Environment, 4 • *,. ....$........ _ -..... :.............. :..... VB pages 2005-2007, Piscataway, NJ, 1998. [6] V. U. Zavorotny and A. G. Voronovich. "Scatter- ing of GPS Signals From the Ocean With Wind Re- mote Sensing Application". IEEE Transactions on Geoseienee and Remote Sensing, 38(2):951-964, Mar 2000. [7] T. Elfouhaily, B. Chapron, K. Katsaros, and D. Van- demark. "A Unified Directional Spectrum for Long and Short Wind-Driven Waves". Journal of Geophys- Figure 4: Wind speed (a) and wind direction (b) estimates ical Research, 102(C7):15781-15796, 15 Jul 1997. from GPS observations for 28 February, 2000 GPS data. Also shown on the i_iots are the Delaware Bay (DB) and Virginia Beach (VB) buoy estimates at the positions in the flight track closest to the buoy positions.