On the accuracy of HF radar surface current measurements: Intercomparisons with ship-based sensors

Transcription

1 JOURNAL OF GEOPHYSCAL RESEARCH, VOL 102, NO C8, PAGES 18,737-18,748, AUGUST 15, 1997 On the accuracy of HF radar surface current measurements: ntercomparsons wth shp-based sensors R D Chapman, L K Shay, 2 H C Graber, 2 J B Edson, 3 A Karachntsev, 3 C L Trump, 4 and D B Ross 5 Abstract Hgh-frequency (HF) radar systems can provde perodc, two-dmensonal, vector current estmates over an area approachng 1000 km As the use of these HF systems has ganed wder acceptance, a number of attempts have been made to estmate the accuracy of such systems However, comparsons of HF radar current estmates wth n stu sensors are dffcult to nterpret snce HF systems measure currents averaged over an area of-1 km 2 and to a depth of only -50 cm whle n stu sensors measure currents at a pont and somewhat greater depths (-1 to 10 m) Prevous studes of the accuracy of HF radar technology have thus attrbuted the dfferences observed between HF radar and n stu sensors to an unknown combnaton of vertcal shear, horzontal nhomogenety, n stu nstrument errors, and HF radar system errors Ths study examnes the accuracy of HF radar current measurements usng data from the 1993 Hgh Resoluton Remote Sensng Experment, conducted off Cape Hatteras, North Carolna Data from four shpborne n stu current meters are compared wth data from an Ocean Surface Current Radar (OSCR), a commercal current-measurng radar We attempt to dscern the predomnant sources of error n these data by usng multple smultaneous measurements from dfferent sensors and by examnng the varaton of observed current dfferences as a functon of locaton The results suggest an upper bound on the accuracy of the OSCR-derved radal currents of 7 to 8 cm/s 1 ntroducton Near-surface ocean currents play a varety of roles n coastal envronments Physcally, wnd stress s mparted through an upper surface boundary layer These upper layer stresses play an mportant role n the development and mantenance of the mxed layer Bologcally, near-surface currents dstrbute and dsperse both plankton usually suffer from motons of the moored body Furthermore, the nstruments are typcally suspended below a surface float, whch physcally restrcts the mnmum depth of the current measurement to a few meters Currents measured from shp-mounted acoustc Doppler current proflers (ADCPs) can be useful, but the frst measurement bns are necessarly below the draft of the shp and and fsh eggs Ecologcally, many pollutants, such as ol, are sur- shp tme s expensve For these reasons, a land-based method for face borne The dsperson of such pollutants depends crtcally on contnuously measurng surface currents n the coastal regon has the near-surface current structure Thus, fro multple vewponts, consderable mert the measurement of near-surface currents n the coastal envron- The measurement of near-surface ocean currents usng electromagnetc backscatter measured from a hgh-frequency radar was ment s an mportant problem frst demonstrated over 20 years ago [Stewart and Joy, 1974; Barrck Unfortunately, conventonal methods for measurng near-surface et al, 1974] (Generally, the term hgh-frequency, or HF, refers to currents (depths-1 m) are somewhat problematc Whle careful systems wth carrer frequences between 3 and 30 MHz) Ths desgn of surface drfters can mnmze problems due to wndage technque, based on the results of Crombe [1972], utlzes the fact and wave nteracton, sngle drfters are lmted and uncontrollable that HF backscatter from the ocean at near-grazng angles s pren ther spatal and temporal coverage, especally n convergent and dvergent flow regmes Multple drfters can be prohbtvely ex- domnantly due to Bragg scatterng Thus the Doppler spectrum of the HF radar return contans sharp peaks assocated wth those pensve when used n suffcent quanttes to adequately characterze the current wthn a large regon over a sgnfcant perod of tme surface gravty waves wth a wavelength of half of the radar Moored nstruments can be used to obtan long tme seres, buthey wavelength that ar ether approachng toward, or recedng from, the radar ste [Crombe, 1955] The frequency of these Doppler peaks reflects the phase velocty of the scatterng waves By subl Appled Physcs Laboratory, Johns Hopkns Unversty, Laurel, Maryland tractng the known phase velocty of' the scatterng wave n the 2 Rosenstel School of Marne and Atmospherc Scences, Unversty of absence of current, Cp = x/g / k, from the measured phase Mam, Mam, Florda an estmate of the component of the surface current n the look 3 Woods Hole Oceanographc nsttuton, Woods Hole, Massachussetts 4 Naval Research Laboratory, Washngton, D C drecton of the radar can be obtaned Two or more statons, 5 vy nc, Mam, Florda vewng a partcular measurement pont from dfferent drectons, Copyrght 1997 by the Amercan Geophyscal Unon Paper number 97JC /97/97JC ,737 can then be used to estmate surface current vectors velocty, Several practcal radar systems for measurng ocean currents n the near-coastal envronment have been developed and commercalzed At least two such systems are currently avalable: the

2 18,738 CHAPMAN ET AL: ACCURACY OF HF RADAR CURRENT MEASUREMENTS Coastal Ocean Dynamcs Applcaton Radar (CODAR), manufactured by CODAR Ocean Sensors, Ltd, and the Ocean Surface Current Radar (OSCR), manufactured by Marcon, nc Whle we utlze an OSCR system, owned and operated by the Unversty of Mam, for ths study, we beleve that the results presented here would, n prncple, be the same for a CODAR system The two systems dffer n beam-formng technology but rely on fundamentally smlar physcs and Doppler-processng algorthms for ther operaton An nformatve sde-by-sde comparson of an OSCR and several CODAR systems has just been publshed by Fernandez and Paduan [1996] These commercal systems are lmted to near-coastal envronments because of ther desgn and the nature of ground wave HF rado propagaton (Attempts have been made to operate OSCR from both an anchored and a movng shp [cf Skop et al, 1994; Peters and Skop, 1995]) Most HF communcaton systems rely on reflectons off the onosphere to acheve over-the-horzon operaton Unfortunately, motons n the onosphere, whch are responsble for the fadng typcally observed n HF communcatons, act to blur the Doppler sgnature of the Bragg scatterers Whle some encouragng progress has been made on long-range current estmates usng HF radars [Trzna, 1982; Georges and Harlan, 1995], the most common HF current-measurng systems utlze ground wave propagaton to elmnate onospherc effects Losses n ground wave propagaton subsequently lmt CODAR and OSCR systems to effectve ranges of less than 100 km, although we note that addtonal power could be combned wth hgher gan antennas to extend ths range to several hundred klometers Snce the development of HF radar for current measurement, there have been a number of experments to evaluate the accuracy of ths technology The frst experments [Stewart and Joy, 1974; Barrck et al, 1977; Frsch and Weber, 1980] compared HF radar estmates of current to those derved from drftng buoys These studes reported dfferences between the radar- and drfter-derved currents of 15 to 27 cm/s Paduan and Rosenfeld [1996] present a more recent study usng the same technque, reportng 13 cm/s rms dfferences n current magntude Later studes, comparng HF data wth current measurements from moorngs or bottom-mounted ADCPs [Holbrook and Frsch, 1981; Lese, 1984; Porter et al, 1986; Matthews et al, 1988], reported dfferences rangng from 9 to 17 cm/s Whle the level of agreement found n these studes s encouragng, all acknowledged the dffcultes n comparng HF radar-derved currents, whch are area-averaged estmates made at the surface, wth n stu measurements, whch are essentally pont measurements made at some fnte depth Prandle [1991] ncludes an overvew of a seres of studes comparng tdal ellpses determned from HF radar and conventonal current meters Usng HFradar-derved surface currents from the Hgh Resoluton Remote Sensng Experment, Shay et al [1995] found rms dfferences of cm/s between surface and moored subsurface current mea- surements Ths study focused on understandng these dfferences wthn the context of bulk vertcal current shears of about 1-2 cm/s dfferences and nstrument errors The frst technque nvolves the smultaneous comparson of estmates derved from multple nstruments, ncludng the HF radar We hoped that the relatve magntude of the observed dfferences would lead to an apportonment of the errors The second technqu examnes the dependence of the observed current estmate dfferences wth locaton By combnng the data wth a smple model of the spatal dependence of errors n the HF radar system, we estmate upper bounds on the accuracy of that system 2 Data Sources The data used n ths study were obtaned durng the Hgh Resoluton Remote Sensng Experment [Herr et al, 1991] Ths experment, jontly supported by the Offce of Naval Research and the Naval Research Laboratory, was desgned to nvestgate the detaled processes nvolved n radar magng of submesoscale features The experment took place 10 to 50 km offshore from Cape Hatteras, North Carolna, from June 10 through June 26, 1993 The reader s referred to Shay et al [1995] for a descrpton of the oceanography of the regon durng the perod of the experment The experment was supported by a wde varety of research platforms, ncludng two research vessels feldng n stu sensors, several arcraft wth real and synthetc aperture radars, the ERS-1 satellte, and a land-based Ocean Surface Current Radar (OSCR) Both of the research vessels, USNS Bartlett and R/V Columbus seln, also deployed small towed platforms contanng current sensors as well as other nstrumentaton Our study compares data from four separate current sensors on the research vessels and towed platforms to those obtaned from the OSCR n order to evaluate the accuracy and lmtatons of OSCR Each of the systems nvolved n the comparson s descrbed n the paragraphs below 21 OSCR OSCR s a dual-ste pulse-doppleradar operatng at a frequency of 254 MHz The transmt antenna for ths system s a fourelement Yag confguraton wth a front-to-back power rato of 6 db and a 90 ø wde beam pattern A phased-array recevng antenna, wth a beam wdth of approxmately 6 ø, s used to azmuthally scan the ocean regon llumnated by the transmtted beam The radar estmates near-surface currents wthn each range and azmuth cell by dentfyng and trackng frequency shfts n the peaks of the Doppler spectra of the ocean backscatter correspondng to the advecton of the Bragg wave Snce an ndvdual staton s only senstve to radal Doppler veloctes, two separate statons are used n the system to obtan vector current estmates Durng the Hgh Resoluton Remote Sensng Experment the two statons of the OSCR system were deployed on the shore n the towns of Avon and Waves, North Carolna These two stes were separated by a dstance of 24 km [Shay et al, 1995] Whle ntermedate data products were recorded from the OSCR system, the fnal data products, used n ths study, were maps of surface current estmates These maps dsplay estmates of the nearsurface current everywhere wthn a crcle of about 35-km dameter centered about 25 km off shore and wth a mean horzontal reso- per meter by decomposng the observatons nto varous frequency bands More recently, Paduan and Rosenfeld [1996] reported on extensve ntercomparsons of HF radar data wth n stu sensors n Monterey Bay, ncludng a rather complete dscusson of the luton of approxmately 12 kn Such maps were produced every dffcultes and uncertantes n such a comparson 20 ran durng nearly the entre perod of the experment One such Ths study s an attempt to better estmate the accuracy of map s shown n Fgure 1 current estmates made by an HF radar system Our approach s to Several aspects of the OSCR system, some of whch are mporcompare shpborne surface current measurements to those obtaned tant for ths analyss, are worth notng by HF radar over the 2 weeks of the Hgh Resoluton Remote 1 n contrasto the n stu sensors, whch essentally measure Sensng Experment We utlze two technques n an attempt to currents at a pont, the OSCR current estmates are based on averages apporton the observed current estmate dfferences to physcal taken over an area of between 25 and 56 km 2 More precsely,

3 CHAPMAN ET AL: ACCURACY OF HF RADAR CURRENT MEASUREMENTS 18, June 27, Z ste operates for 5 mn Both stes are then slent for the remander of the 20-mn cycle durng whch tme data are exchanged between the stes and processng s performed Ths sequencng of measurements means that the two radal measurements used for each vector estmate are not taken smultaneously Errors due to ths temporal undersamplng are unlkely, as ths would requre a feature wth a phase velocty of the order of 4 m/s (= 12 km/5 mn) 6 The phase velocty of the Bragg waves s altered n shallow water, whch drectly affects the surface current estmates Fortunately, at the wavelength at whch OSCR operates, ths effect only becomes sgnfcant for water depths less than 3 m, and no depths ths shallow were encountered n ths study 7 The expected precson of a radal current estmate made by OSCR s 22 cm/s Ths fgure s based on the length of the analyss records for the specfc setup utlzed durng ths experment and represents a lmt n the system's ablty to locate a partcular Doppler peak To avod nterference wth each other, the OSCR master and slave stes are not operated smultaneously but are nstead operated sequentally The sequence begns wth the master ste operatng for a perod of 5 mn The master ste then goes slent and the slave 22 R/V Columbus seln R/V Columbus seln was equpped wth a hull-mounted 300-kHz narrowband ADCP Ths ADCP produced current estmates startng at a depth of 46 m n the shallower waters, from the center of the 352 OSCR measurement doman to near shore, bottom trackng and the shp's gyrocompass were used to estmate true near-surface currents n less than 5% of the data, the bottom track sgnal became Fgure 1 Typcal surface current map produced by OSCR Durng too weak to use, so the ADCP and gyrocompass data were comthe course of the experment, over 1100of these maps were produced bned wth data from the shp' s standard Global Postonng System (GPS) to estmate currents n bottom track mode, the theoretcal standard devaton for the OSCR current estmates are based upon all of the spectral horzontal current from a sngle png from ths ADCP s 57 cm/s nformaton obtaned wthn the resolved area Horzontal vara- (10-m bn, +30 ø beam angles) Wth a png every 15 s, 20 mn of tons n the current feld wth scale lengths less than 1 km wll result averagng should reduce ths standard devaton to 2 mm/s Unforn broadened or multple Doppler spectral peaks The software must tunately, ths calculaton fals to take nto account longer-term then choose between vald Bragg pars Thus there wll be tmes bases caused by flow dstorton about the vessel, the effect of when the OSCR-estmated current, though vald, wll dffer from bubbles entraned near the sensor head, compass errors, and varaan n stu measurement Due to the complexty of the current regme tons n flow durng the averagng nterval These other sources of off Cape Hatteras, these averagng effects may contrbute substanerrors, whch are dffcult to estmate, are lkely to domnate n our tally to the observed rms dfferences measurements Experence suggests though that the combned errors 2 The wavelength of the Bragg scatterers for OSCR s 59 m can be expected to be on the order of 1 to 5 cm/s The same As such, OSCR responds to surface currents ntegrated over apconsderatons hold true for the other ADCPs used n ths study proxmately the upper 05 m (= X/8, where Xr s the radar wavelength) of the water column [Stewart and Joy, 1974] The n stu data used n ths study were obtaned at depths of 1 to 10 m Vertcal 23 LADAS nhomogenety n the near-surface current over these scales con- LADAS was a towed catamaran system developed by Erk Bock trbutes to dfferences n the current estmates at the Woods Hole Oceanographc nsttuton for measurng the 3 The OSCR system utlzes smple crtera to elmnate bad structure of centmeter-scale surface waves and the modulatons of data These crtera are based on the strength of the backscatter and these short waves by long waves (We use the past tense here snce the form of the resultng Doppler spectra n partcular, current LADAS was lost at sea durng a recent experment on the West estmates are not recorded for those ponts and tmes where nether Coast) ts prmary nstruments were a scannng laser slope gauge the approachng nor the recedng Bragg peaks are greater than 6 db [Bock and Hara, 1995], a sute of meteorologcal nstruments, a above the background nose floor Thus the current maps produced dfferental GPS (DGPS) recever, a sx-degree-of-freedom, strapby OSCR have some gaps n them Ths explans some of the holes down moton sensng package [Edson et al, 1996], and a threen the coverage shown n Fgure 1 These gaps are not common, axs ultrasonc current meter Data from ths current meter and the representng at most a few percent of the data wthn the nteror dfferental GPS (DGPS) recever were utlzed n ths study of the measurement doman The three-axs ultrasonc current meter was a UCM 40 Mk, 4 The data qualty crtera used by OSCR are not perfect, and manufactured by NE Sensortec, Norway Ths nstrument measures some current estmates that are wldly erroneous can stll be found currents over a set of orthogonal 10-cm paths by measurng the tme n the data Such wld ponts were observed to occur less than 1% of the tme wthn the data of flght of a seres of hgh-frequency acoustc pulses The nstrument has a resoluton of 1 mm/s and an accuracy of 3% of the measured value + 5 mm/s wth an ntegraton perod of 1 s The nstrument utlzes an nternal three-axs flux gate compass and a two-axs tlt sensor to compensate the data for the nstrument orentaton The system also ncludes bult-n temperature, conduc-

4 18,740 CHAPMAN ET AL: ACCURACY OF HF RADAR CURRENT MEASUREMENTS tvty, and pressure sensors Ths nstrument was mounted forward on the body of the catamaran, between the two hulls n ths mountng poston the sensors were outsde the wakes of the pontoons when the catamaran was beng towed forward The current sensors were located at a mean depth of 10 m The data from the LADAS UCM were combned wth the dfferental GPS data to estmate true currents The predomnant errors n ths combned data set are lkely due to bases and nose ntroduced by ptch and rollng motons of the catamaran n the wave feld Whle to lowest order, these hgh-frequency motons should average out, they may ntroduce bases nto the estmate snce the depth of the UCM measurements vared coherently wth the surface waves as the platform ptched up and down 24 USNS Bartlett USNS Bartlett was equpped wth a 300-kHz hull-mounted, narrowband ADCP n the shallower waters, throughout most of the OSCR measurement doman, bottom trackng and the shp's gyrocompass were used to estmate true currents n approxmately 3% of the data the bottom track sgnal became too weak to use, so the ADCP and gyrocompass data were combned wth data from a realtme dfferental GPS system to estmate currents 25 TOAD are less than a few seconds, snce all of the data streams ncluded tmng nformaton traceable to GPS tme An equvalent seres of surface current data were then constructed by selectng the OSCR data closest to the n stu data n tme and space Note that whle the data were n the form of a tme seres, ths tme seres was obtaned at the tme-varyng poston of the platforms as they moved about the expermental area Thus conventonal tme seres analyses cannot be usefully appled to these data nstead, we restrct ourselves to nontemporal statstcal comparsons Our ntal analyss s based on smple scatterplots between pars of related estmates of the north and east components of the surface velocty The rms dfference n the velocty estmates, whch we refer to as o- a, s used to characterze the comparsons Ths s the most commonly used statstc n prevous ntercomparson studes, makng our results drectly comparable to those studes The reader s cautoned though not to attrbute all of the rms dfference to errors n the HF radar current estmate The reader should note that throughouths paper we are careful to dstngush between the terms "dfferences" and "errors" n our usage, subtractng two current estmates results n a current estmate "dfference" Ths current estmate dfference s lkely due to a combnaton of dfferences n the quantty measuredue to physcal processes and to nose n the nstruments themselves We refer to nose wthn an nstrument, causng the nstrument to read other than a true value, as an "error" Thus we are tryng to estmate HF radar errors by studyng observed dfferences The Towed Acoustc Doppler (TOAD) platform s a small tube wth stablzng fns that s desgned to hold an ADCP as t s towed on the surface, outboard of a vessel [Marmorno and Trump, 1996] 4 Results and Dscusson Ths system, developed by the Naval Research Laboratory, was Before presentng the results, t s nstructve to consder the desgned to keep the ADCP away/¾om the nfluences of the shp possble causes of dfferences n surface current estmates made by whle stll provdng a relatvely stable platform for makng current the OSCR system, an ADCP, and the ultrasonc current meter measurements Durng the Hgh Resoluton Remote Sensng Ex- The frst possble cause s the comparson of dssmlar quanttes perment, a 600-kHz broadband ADCP was deployed from TOAD The OSCR measures surface currents averaged over a 15-km 2 area The TOAD platform was deployed perodcally from USNS Bartlett and to an effectve depth of about 50 cm The ADCPs measure over throughout the experment to track near-surface current features an area of at most a few square meters n the bns nearesthe surface Because of ts small sze, the TOAD platform experences more and measure at an effectve depth of several meters, dependng on transent acceleratons than the towng shp These acceleratons the system and how t s mounted The ultrasonc current meter adversely affected the onboard atttude sensors, makng the absolute water veloctes measured from TOAD noser than those from measures over a scale length of 10 cm at a depth of 1 m One should the shpboard system To mnmze ths problem the shpboar data not expect such measurements to agree for a varety of physcal were merged nto the TOAD data by offsettng the TOAD data so reasons For example, dfferences could be due to horzontal nthat the average absolute water velocty between 10 and 20 m homogenety caused by local eddes and fronts Lkewse, vertcal measured by TOAD was dentcal to the value measured by the shears arsng from the near-surface boundary layers or nternal shpboard system The result of ths mergng process s that the waves can also lead to dfferences [Shay et al, 1995] Here, and absolute water veloctes reported by both systems between 10 and elsewhere throughouthe paper, by vertcal shear we mean those 20 m wll be vrtually dentcal The advantage of the TOAD data bulk vertcal shears remanng after 20 mn of averagng les n ts smaller bns (05 m versus 10 m) that measure closer to The second possblty s temporal or spatal msalgnments of the surface (16 m versus 88 m) than the Bartlett ADCP the measurement data sets Any msalgnments of the data sets, ether temporally or spatally, wll cause a decorrelaton of the results Gven the relatvely hgh accuracy of the data set tmng, 3 Methodology we do not expectemporal decorrelaton to be a problem Lkewse Ths study compares the OSCR current maps wth the nearsurface current measurements made from three separate ADCPs (seln, Bartlett, and TOAD) as well as data from the three-axs we beleve that the spatal msmatch was also small The n stu sensors utlzed dfferental or standard GPS data for spatal postonng, systems wth accuraces of better than 100 m The other current meter mounted at 1-m depth on LADAS The methodology possble uncertanty was the postonng of the OSCR grd locaused to compare these data s relatvely straghtforward Data from the n stu sensors were combned wth the bottom track veloctes, or the dfferental GPS navgaton data where the depth of the water precluded bottom trackng, and platform orentaton data to obtan estmates of the Earth-fxed currents at the platform locaton These tons n the radal drecton, these postons are determned qute accurately by the system range gate tmng n the azmuthal drecton, these postons are determned by the physcal and electrcal algnment of the OSCR phased-array antenna The estmated physcal accuracy of the antenna algnment was less than 05 ø, whch would tme seres were reduced to a sample every 20 mn by averagng bound the errors n cell postonng to less than 1/3 of the cell sze those values correspondng to each OSCR observaton perod We estmate that the errors n algnment of these data-averagng perods Electrcal msalgnment of the phased array, whch could take the form of ncreased sdelobes n the antenna pattern, s a concern

5 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS 18,741 wth ths type of system No sgnfcant correlaton was observed between antenna look angle and radal current estmates dfferences made from each OSCR ste and from the Bartlett and seln ADCPs Ths suggests that phase msalgnments of the antenna are at least unformly varyng across the antenna aperture, makng substantal errors due to ths source unlkely A thrd possblty s systematc bases or nose n the OSCR data Any bases or nose wthn the estmates derved by OSCR wll contrbute to the dfferences n the comparson wth other nstruments These are the errors we are tryng to evaluate The fourth possblty s systematc bases or nose n the n stu data Lkewse, any bases or nose n the n stu data wll also contrbute to dfferences n the comparsons Such errors could arse from errors n the sensors themselves or the platform moton and orentaton correctons Hence these data depend on the accuraces of the current sensors, bottom-trackng algorthms, or DGPS velocty estmates, as well as compass sensors An apprecaton for the possble sources of dfferences led to our general approach of comparng these data streams n ths approach, to the extent that we could, we utlzed multple data comparsons n an attempt to apporton the observed dfferences to one of the above causes Ths approach s explaned n detal below 41 OSCR Versus Columbus seln Fgure 2 presents the comparsons between OSCR and the seln ADCP data The north and east current component comparsons are shown n the upper panels (Fgures 2a, 2b, and 2c) and lower panels (Fgures 2d, 2e, and 2f), respectvely The leftmost two panels (Fgures 2a and 2d) compare OSCR and the shallowest ADCP bn at 46-m depth The mddle two panels (Fgures 2b and 2e) compare OSCR and the next deepest ADCP bn at 56-m depth The rghtmost two panels (Fgures 2c and 2f) compare the data from the two adjacent ADCP bns The dotted lne, whch has been placed on each panel for vsual comparsons only, ndcates the lne where the two current estmates are equal Ths lne does not represent a least squares ft to these data The rms dfference of each of the comparsons, O',l, s gven n the upper left corner of each panel n Fgure 2a at least two erroneous ponts n the OSCR estmates are evdent These wld ponts can easly be assocated wth errors n the OSCR estmate snce the OSCR-estmated current component value of 90 cm/s s beyond any others observed by ether nstrument A thrd wld pont s evdent, although ts source s not so clear Other than these three wld ponts, the remanng 401 ponts are ncely clustered about the lne of equal velocty The standard devaton of the dfference s 148 cm/s for all 404 ponts f these dfferences were drawn from a Gaussan dstrbuton, the error bounds on ths estmate of O'd, to the 95% confdence level, are estmated to be _+10 cm/s Fgure 2b shows a smlar groupng of ponts about the equal velocty lne, but wth a slghtly hgher standard devaton of the dfference of 151 cm/s One pont, whch s not vsble n ths fgure, has been excluded from the computaton of o'a n Fgures 2b and 2c because the OSCR velocty estmate was greater than 15 m/s Gven the error bars of _10 cm/s at the 95% confdence level, the dfferences observed between the OSCR and ADCP data at 46-m and 56-m depth are on the border of statstcal nsgnfcance At the same tme, as should be expected, the agreement between the 46-m and 56-m ADCP bns (Fgure 2c) s much better than the agreement between ether ADCP bn and OSCR (a) d = 148 cm/s (b) (Jd = 151 cm/s (C) (Jd: 32 cm/s,,,;,', ',,, ", :-, 1 l :3' '' c' OSCR North Vel (m/s) ',;' :; ;/,' " : [,' 7' ':' ";';'¾' Deleted 1 wld pont l ''" from OSCR data set OSCR North Vel (m/s) ': ed 1 wld pontl from OSCR data set se/n ADCP (56 m) North Vel (m/s) 10 - (d)(7d= 145 cm/s ': (e) (Jd: 153 cm/s 10- (f) (::Td: 30 cm/s , 1' ;;' ";,'!',- "ae ', B' ' B 05-,' '" '-'"" 1," : a ;' 5' ['"" "2" " ½ , OSCR East Vel (m/s) OSCR East Vel (m/s) -05 -,' seln ADCP (56 m) East Vel (m/s) Fgure 2 Comparsons of the surface current component estmates made from the seln ADCP and OSCR The dotted lne ndcates the lne of equal currents t does not represent a ft to the data The rms dfference between the estmates, o'a, s gven n the upper left corner

6 18,742 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS E 10- (a) Od= 157 cm/s 10- (P) c a: 'ø om/s 10- (C) Od= 144 cm/s o 05- z o v 00- O Q / / /- t',',/ ' lj/d',: :' OSCR North Vel (m/s) l'0-(d) (:5 d 112cm/s 05 -," 05- z oo- (D -o5-1, ½ ',',:,;::,--', ; 72'" OSCR North Vel (m/s) (e) Od: 126 cm/s 05 - ":'; - -'" ¾:,: 00- ' ' ' ø,:4 '"' ',v,4r'; seln ADCP (56 m) North Vel (m/s) (f)(:td: 135 cm/s,:;,-:,: 00- :"" -1 /'/ OSCR East Vel (m/s) ; oo- Q j-" ' OSCR East Vel (m/s) (D 03 < -05- // ;, ;, :,: /",= 'Y,,' seln ADCP (56 m) East Vel (m/s) Fgure 3 Comparsons of the surface current component estmates made from the LADAS UCM, seln ADCP, and OSCR 42 OSCR Versus LADAS Comparsons of the OSCR, UCM, and ADCP data from LADAS are shown n the sx panels n Fgure 3 Whle the total number of ponts s much smaller for these comparsons because of the lmted amount of tme that LADAS was deployed durng the experment, all of the comparsonshow a smlar trend about the lne of equal current The rms dfferences for these comparsons range from 11 to 16 cm/s The LADAS comparsons are partcularly nterestng due to the presence of three dstnct types of sensors f the dfferences n the OSCR comparsons were predomnantly due to mean vertcal shear, we would expect to fnd that the agreement between OSCR and the LADAS UCM would be substantally better than the agreement of ether nstrument wth the deeper ADCP Ths does not appear to be the case f, on the other hand, the dfferences were predomnantly due to the dfferences between pont measurements and areaaveraged measurements, we would expect that the agreement between the LADAS UCM and seln ADCP would be superor to ether n stu sensor compared to OSCR Agan, ths does not appear to be the case 43 OSCR Versus USNS Bartlett Fgure 4 contans the data comparsons between OSCR and the Bartlett ADCP Note that a sngle wld pont, wth a current estmate exceedng 15 m/s, was deleted from the OSCR north component data set pror to estmatng the rms dfference show good agreement between the data sets, wth rms dfferences rangng from 9 to 16 cmjs 45 Comparson Summary The results of these comparsons are summarzed n Table 1 The second and thrd columns presenthe number of ponts used n the comparson and the lnear correlaton coeffcent for those comparsons The rms dfferences for each comparson are gven n the fourth column, along wth error bounds for these estmates These bounds were computed at the 95% confdence level, assumng that the component dfferences are Gaussan-dstrbuted random varables The estmates of the varance from a gven sample sze are thus assumed to be X 2 dstrbuted, and the error bounds are easly computed The last two columns contan depth correctons as descrbed below Several nterestng observatons can be made from ths data summary 1 All of the comparsons wth OSCR show roughly comparable dfferences of between 9 and 16 cm/s Ths places an absolute upper bound on the errors of the OSCR estmates, whch s comparable to that reported n prevous studes [Shay et al, 1995] 2 There s a statstcally sgnfcant dfference between the north and east rms dfferences for only two cases, the LADAS UCM and the TOAD 2-m ADCP These are also the two shallowest nstruments More wll be made of ths pont n the followng secton 3 The lowest correlaton coeffcents are assocated wth the comparsons of OSCR wth the deepest nstruments Ths suggests 44 OSCR Versus TOAD that near-surface vertcal shear s an mportant, but not domnant, Our fnal set of data, contanng comparsons between OSCR component n the dfferences observed between OSCR and the n and the ToAD ADCP, s shown n Fgure 5 Agan, the comparsons stu sensors

7 CHAPMAN ET AL: ACCURACY OF HF RADAR CURRENT MEASUREMENTS 18,743 _>, 10 1 a d = 162 cm/s 05,_,, d' 4', '' ',' ',,,' /,, " -05-" leted 1 pt from OSCR North OSCR North Velocty (m/s) 10 between the n stu and OSCR data n the same manner as fluc- tuatng dfferences or nose Fgure 6 ndcates that the rms dfferences grow quckly n the upper 4 to 5 m, and then more slowly below that depth The last column n Table 1, labeled "TOAD rms Shear," contans estmated rms dfferences taken from Fgure 6 at the depths of the nstruments beng compared n makng these estmates we assumed that there was no vertcal shear n the near-surface layer above the bns sampled by the TOAD ADCP The reader s cautoned that these results cannot be drectly appled to compensate for shear n Table 1, because many of the measurements were taken at dfferent locatons and because the analyss does not separate the shear effects from depth-dependent nose n the TOAD ADCP We have ncluded these values solely for the purpose of comparson Despte these caveats, the data do suggesthat the effects of vertcal shear are largest on the dfferences observed between the deepest and shallowest nstruments Furthermore, t appears that the dfferences due to shear are probably comparable to the dfferences due to other sources 47 Calculaton of Geometrc Dluton of Precson for OSCR C)5- (b) 3 m/s,,,;' d 1' [ ql L,,![?_ 'P " _ ' ;' '"%,-, l" '',' [11 ql,,,/ OSCR East Velocty (m/s) Fgure 4 Comparsons of the surface current component estmates made from the Bartlett ADCP and OSCR 4 The seln ADCP shows better agreement durng the perod when LADAS was deployed then for the experment as a whole There are two lkely explanatons for ths Frst, the seln was lmted n speed when LADAS was deployed Second, LADAS was only deployed n low-to-moderate sea state condtons These effects combned to reduce ADCP nose durng LADAS deployment relatve to the experment as a whole,',,',,,,,,,,,,, The comparsons we have presented to ths pont have been lmted to scatterplots and frst-order statstcs The next step n ths analyss s to examne the spatal dependence of the observed dfferences Durng the 2 weeks of the experment, the two research vessels traversed most of the OSCR measu'ement doman numer- ous tmes Fgure '7 s a map of the locatons of the research vessels durng the experment Ths wde range of measurement locatons, combned wth a smple theoretcal model predctng the response of OSCR, can be used to examne the spatal dependences of errors n the OSCR estmates Smple fgures of mert for OSCR's ablty to estmate surface current components can be derved (The reader s referred to Lpa and Barrck [1983] for an alternatve dervaton) Fgure 8 presents a sketch of the geometry pertnent to current component determnaton wth OSCR Wthn ths dagram, ste 1 and ste 2 are the locatons of the OSCR master and slave stes, respectvely At each pont n the OSCR measurement doman, the radal veloctes, Vl and V2, are estmated The n-lne and orthogonal veloctes, V and Vo, can be estmated n terms of the sums and dfferences of the radal veloctes: v = < + vgv2cos o, v o =(v] - v2)/2sno, (1) where 0 s half of the angle between the ntersectng beams The n-lne and orthogonal veloctes can then be rotated to obtan estmates of the north and east velocty components: V n = V sn o + V o cos a, V e = V cos o - V o sn a, where o s the mean look angle as defned n Fgure 8 Substtuton yelds (2) 46 Estmates of Shear To better examne the effects of shear, we computed the rms dfferences between measurements at varous depths and the measurement at 16-m depth made by the TOAD ADCP Fgure 6 shows these rms dfferences as a functon of depth for the north and east current components We are usng the rms dfferences, whch nclude the mean bases between the measurement sets, because any mean shear wll contrbute to the observed dfferences = + V 1 + Vn 2cosO 2snO 2cosO 2snO ' (cosc 2snO sng)(cosa + 2cosO + 2snO sn )V 2 ' V = 2cosO These equatons are of the form z = a V] + bv2, where V1 and V2 are random varables representng the radal current measure- (3)

8 / 18,744 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS 10- (C) (7d: 116 cm/s _,, 4; ", 1', - ', OSCR North Vel (m/s) (d) (7d = 86 cm/s = : {;?':: ' OSCR East Vel (m/s) / (e) (Td: 159 cm/s, ",, 4',! OSCR East Vel (m/s) TOAD ADCP (10 m) Noah Vel (m/s) _ 10- (f) d = 55 cm/s /, 1 ', ' ", ' ', TOAD ADCP (10 m) East Vel (m/s) Fgure 5 Comparsons of the surface current component estmates made from the TOAD ADCP and OSCR ments f we assume that the nose n each radal measurement, o-, s dentcal, then the nose of the scaled sum s gven by o- z -o- /a2+ b 2 Applyng the relatonshp (u + v) 2 + (u-v) 2 = 2(u 2 + v 2) to (3) yelds estmates for the errors n the north and east components: Table 1 Summary of the rms Dfferences Obtaned from Comparsons of Dfferng Estmates of Surface Current Component Estmates Measurement Number p rms TOAD rms Dfference, (cm/s) Shear, (cm/s) Lveln 46-m ADCP versus OSCR rms(u(depth) - u(16)) --- rms(v(depth) v(16)) Lveln 56-m ADCP versus OSCR Lveln 46-m ADCP versus Lveln 56-m ADCP LADAS 1-m UCM versus OSCR Lveln 56-m ADCP versus OSCR (only dur ng LADAS deployment) LADAS l-m UCM versus lseln 56-m ADCP Bartlett 10-m ADCP versus OSCR rms Velocty Dfference (cm/s) Fgure 6 The rms dfferences observed n the TOAD ADCP data between current measurements made at varous depths and those made at 16-m depth TOAD 2-m ADCP versus OSCR TOAD 10-m ADCP versus OSCR TOAD 2-m ADCP t06 versus TOAD 10-m ADCP TOAD 2-m ADCP versus Bartlett ADCP The upper and lower values on each lne are the north and east current component statstcs, respectvely The last column, representng the magntude of the vertcal shear observed by the TOAD ADCP durng the experment, s based on an analyss of Fgure 6

9 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS 18, o OSCR stes [] seln [] Bartlett - 1 North GDOP --- East GDOP B 357 -" -20 B [] [] [] [] :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: : : : : : : : : : : : : : : : : : : : : : : ':-:' u :::::::::::::::::::::: :::::::::::,:: :: ======================================== :--:-:::--:-:,::u:-el ',:-:--:- :-:::,:' ::-: , ::::::-:-:::-:::::-:: --::E3 :-:,'" :-:,::::::::-: :,::: :_,- '" j' '-', ; : :?' ' ' ::::: : ':'':' ': :'"':'':E}''' :'' r':''"':'':' -' --' ::,'? ':: ':'':'':'' ''"':'':'" '-:: ':'' : :'!:':': ': ::':: o::!'tb!::,¾::': ::" :' ::::::::::::::::::::::: o c O:'" ": :'":' ""-'' ':' " '"':'"'0! ':':::':::' l'":' ]:::': :'::"'"':"l :::C?::":"::: ::'::,:" ' -':' o : ; m [] ''::::::'':: ::-:::-: E ::: -'m W oe,-, mm o ' '"'":- ':::':::':::m::::': l m r O 3 oo ¾ []', o o o Fgure 7 Map of research vessel locatons durng the experment relatve to the OSCR measurement doman Fgure 9 Map of the north (sold lnes) and east (dashed lnes) geometrc dluton of precson (GDOP) for the OSCR measurement doman O' n = 2 sn2 o sn 2 O + cos 2 o cos 2 O (4) O'e= 2 cøs2asn20+sn2acøs20 Borrowng from the termnology of the GPS navgaton system [Wells et al, 1986], we wll refer to the ratos Crn/cr and ere/or as the north and east geometrc dluton of precsons or GDOPs These GDOPs can be thought of as multplers of the nose assocated wth the geometry of the HF radar measurement Applyng these formulae, we can derve the map of the north and east GDOPs for the OSCR measurement doman shown n Fgure 9 The map n Fgure 9 ndcates that the north GDOP vares from 1 to 25 n the OSCR doman, wth the largest errors occurrng at the farthest range cells The east GDOP vares from 075 to 15, wth the largest errors occurrng at the northern and southern extremes of the doman That the east GDOP falls below 10 at the farthest Ste 2 ½ Ste 1 Fgure 8 Geometry of OSCR current determnaton

10 18,746 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS range cells should cause no surprse At these locatons, the radal velocty estmates are nearly parallel and so the estmate of the east component of velocty s nearly the average of the two radal velocty estmates n the lmtng case of a pont nfntely dstant toward the east, the error n the east velocty estmate would go to tr/ 4 due to the averagng of two, lke random varables The apparent lack of symmetry n the map n Fgure 9 s due to the fact that the stes are not algned along a north-south axs t s mportant to note that the GDOP does not take nto account the effects of reduced sgnal to nose that mght be expected to decrease the accuracy of the current estmates n the farthest range bns ts sole purpose s to predct the purely geometrc component of the errors expected n the OSCR current estmates We also note that the expressons n (4) are apparently not consstent wth the smlar but more general expressons derved by Lpa and Barrck [1983, equatons (41) and (42)] We attrbute these dfferences to mnor typographcal errors n the paper by Lpa and Barrck A reanalyss of ther least squares approach to current determnaton n overdetermned systems shows that ther corrected expressons reduce to our formula n the case of two stes Fnally, our results are n dsagreement wth those of Prandle [1991, equaton (2) and Fgure 3] We fnd that there s a typographcal error n equaton (2): the terms should be subtracted, not added Furthermore, Fgure 3 was apparently obtaned by takng the absolute value of each of the terms n equaton (2) Ths approach cannot be reconcled wth our results, whch correspond to addng the two terms of hs equaton (2) n quadrature (Strctly speakng, ths result only holds n the lmt of large mean current magntude n general, the standard devaton of current magntude depends on true mean current magntude n a complex fashon, makng t n an napproprate method for comparson of current measurng devces) 48 Apportonment of Errors where z s the observed component dfference at the th pont, The theory of the prevousecton shows that the OSCR-dependent GDOP s the GDOP at that locaton, x s the equvalent error n errors are poston dependent t s reasonable to assume that the the OSCR radal veloctes at pont, and Y s the error n the other errors due to other sources wll not be poston dependent Ths nstrument Squarng and takng the expected value yelds dstncton can be used to apporton the errors between the two sources The GDOP predctonsuggest that the north component errors (z,2 > = (DO P, 2 x,2> + (y;> + 2 (DO P, x;y;> (7) n our OSCR data should be larger than the east component errors Note that the expectaton value s taken over the range of possble As we noted prevously, there are only two cases where the df- errors, whch s, as n Fgure 10, geometrcally orthogonal to the ferences between the north and east errors are statstcally sgnf- GDOP axs Thus we expect that GDOP (whch s not stochastc, cant, and these are for the two shallowest n stu measurements, but determnstc) and x are ndependent, an asserton we later made by the LADAS UCM at 1-m depth and the TOAD ADCP check We also assert that (GDOP }- GDOP 2, snce at 2-m depth These observed dfferences can be used to apporton the observed errors through the observaton that the total current dffertaton operator acts orthogonal to GDOP Fnally, GDOP and Y are certanly ndependent and x should be ndependent of Y, so (7) reduces to ence varance for each component can be splt nto the sum of the varances 2 of two terms: 2 = GDOp2rr( scr + O othe 2 r (S) O' a, Table 2 Analyss of Current Dfferences as a Functon of GDOP Platform Mean O'd, O'OSCR, O'othe r, Component GDOP cm/s cm/s cm/s LADAS UCM north LADAS UCM east TOAD 2 m north TOAD 2 m east lseln north Bartlett north TOAD data sets are 16 and 09, respectvely These average GDOP values are determned by the dstrbuton of locatons of the seln and Bartlett durng the experment The fact that they are the same for LADAS and TOAD s somewhat concdental An alternatve approach s to examne the dependence of dfferences on GDOP wthn a gven data set Fgure 10 contans plots of the square of current estmate dfferences from both the seln and Bartlett ADCP/OSCR comparsons as a functon of the square of GDOP (The reasons for ths partcular choce of parameters wll be made clear n the followng dscusson) Smlar plots for the TOAD and LADAS are not presented because of the lmted number of ponts avalable from these platforms Whle an ncrease n the varance of the dfferences as a functon of GDOP 2 s dffculto see, a consstent trend s ndeed present n the north component current data A statstcal method for estmatng the GDOP-dependent and -ndependent components of the varance s requred to examne ths trend We began by assumng that for each pont, the component dfference s gven by the twodmensonal stochastc equaton z = GDO P x + Y, (6) the expec- = + (8) Wth the proper substtutons, ths s exactly equaton (5) The mean error s very close to zero n these data, and more where O-OSCR s the effectve radal velocty varance for OSCR mportantly, we a pror expect the errors to be zero mean, so the and O'othe 2 r represents the varance from all other sources ncludng optmal estmator for varance s exactly the average of the varable the n stu measurement errors and the true dfferences n the y/2 Thus each y 2 can ndvdually be thought of as a sngle pont measured quanttes estmate of the local varance Wth ths as background, we chose 2 Gven a par of data sets wth dfferng mean GDOP and as- to perform a least squares ft of a lne to the squared errors Y sumng that the varances O'OSCR 2 and O'otherare 2 ndependent of plotted aganst GDOP/2 n the above notaton the slope and nter- 2 drecton or locaton, we can solve these equatons for O-OSCR and cept then are rroscr 2 =(x ) and trother 2 = (,2 3 ), respectvely 2 O'othe r usng the observed mean O'd for the data sets The results of Ths analyss depends on the ndependence of the radal errors such an analyss are shown n the frst two lnes of Table 2 Note n OSCR and GDOP Possble reasons for such a dependence that the average north and east GDOPs for both the LADAS and nclude reduced sgnal-to-nose rato (SNR) wth dstance or an-

11 CHAPMAN ET AL' ACCURACY OF HF RADAR CURRENT MEASUREMENTS 18,747 O (a) -, (b) (3 ' (North Geometrc DOP) 2 Q_ - o o15- < B, m, ' ': '" ' ' ; : / ' - oo5-;,,,, t- 't, ' "- ' '- t:: O _ - -4-r,k, _,,; ; _,w, m _;,,,,,-, _ z ooo- ;'r - :,,,, (North Geometrc DOP) 2 O ' 000 v 0 (c), -, 025- O ooo (d) l&' (East Geometrc DOP) 2 (East Geometrc DOP) 2 Fgure 10 Comparsons of the seln and Bartlett current component estmaton dfferences squared as a functon of GDOP 2 The dashed lnes n Fgures 10a and 10b are the least squares lne ft to the data, ndcatng the dependence of the error on GDOP tenna pattern effects Thus, before applyng the technque to the Bartlett and seln north data, the correlaton of the radal errors wth GDOP was examned Ths comparson was performed by computng the dfferences between the radal current estmates from each OSCR ste and the n stu currents projected nto these same radal drectons We were somewhat surprsed to fnd a statstcally sgnfcant correlaton between these dfferences and the GDOP Ths correlaton appears to be due to SNR reducton wth dstance, as the dfferences also correlated wth dstance from the ste, but not wth the antenna look angle We found that these correlatons were drven by measurements made at dstances greater than about 30 km and dsappeared when the data sets were lmted to ponts closer than 30 km Thus the least squares lne ft technque was appled to the subset of the Bartlett and seln north data that fall wthn 30 km of the central pont between the OSCR stes Ths analyss s somewhat senstve to wld ponts so we elmnated all data ponts from the calculaton where the total error exceeded 50 cm/s Ths elmnated only four ponts from the seln data set and two ponts from the Bartlett data set The results of ths analyss, shown n the last two lnes of Table 2, agree substantally wth the prevous estmates These results ndcate that the effectve radal velocty errors n the OSCR system are no worse than 6 to 8 cm/s, values comparable to those of all of the other combned errors 2 We refer to O-OSCR as the effectve radal velocty varance n order to dstngush t from the smple radal velocty varance prevously quoted as havng a value of (22 cm) 2 We feel that ths 2 dstncton necessary snce O-OSCR s the sum of the smple radal velocty varance plus any addtonal varance whch s correlated wth the GDOP For example, any varance due to measurement feld msmatch, such as that resultng from an azmuthal msalgnment of the antenna, s drectly correlated wth locaton Gven a drect correlaton wth locaton, ths varance wll also be ndrectly correlated wth the GDOP Thus the fact that we estmate an effectve radal velocty error of 7 to 8 cm/s s not necessarly n contradcton wth the manufacturer's asserton of radal velocty errors of 22 cm/s 5 Summary Ths study s an attempto assess the accuracy of remote surface current measurements usng HF radars n ths study we have compared concdent near-surface current data from four separate platforms wth data from a commercal HF radar system ntercomparsons of the data from the varous systems exhbt rms dfferences rangng from 9 to 16 cm/s At the very least, these comparsons provde an absolute upper bound on the errors assocated wth the current estmates from the HF radar system We recognze though that all of the observe dfferences are not attrbutable to errors n the remote current estmates Some of the dfferences are undoubtedly due to errors n the n stu measurements More mportantly, some of the dfferences are lkely due to physcal processes assocated wth dfferent effectve depths at whch the measurements are made (e, vertcal shear), as well as the dfference n temporal and areal averagng of the remote and n stu measurements An mprovement n the error estmate for the HF radar system can thus only be obtaned by some apportonment of the dfferences between these possbltes n order to further ths analyss, we developed a smple model of the effect of measurement locaton on the HF radar errors We utlze the fact that HF radar errors vary over the deployment area of our experment Ths allows us to decompose the dfferences

12 18,748 CHAPMAN ET AL: ACCURACY OF HF RADAR CURRENT MEASUREMENTS observed between the HF radar data and those obtaned from the n stu sensors Four such decompostonsuggesthat the effectve radal velocty errors n the HF radar system are no more than 7 to 8 cm/s, a value comparable to the total nose from all other sources of current dfferences Even gven HF radar errors as large as 7 to 8 cm/s, such systems can stll be extremely useful for a varety of research projects t s mportant to note that due to the nature of our analyss we are unable to examne the temporal correlaton of the nose f, as we expect, the nose n the OSCR system s not temporally correlated, ths nose wll be spread over a broad range of temporal frequences The nose wthn any partcular frequency band of nterest, say a tdal band, would thus be a small fracton of the values quoted here For example, the standar devaton wthn a sngle frequency bn of a spectral estmate obtaned from an 1100-pontme seres would be (8 cm/s)/ = 034 cm/s A realstc tdal surface current of 5 to 8 cm/s, as found by Shay et al [1995], could be easly resolved above ths nose level n summary, whle HF radar systems have been promoted for remote surface current measurements for over a decade, we beleve that ther acceptance wthn the general oceanographcommunty has been slowed by a lack of valdaton studes Ths study, along wth the related studes of Shay et al [ 1995] and Graber et al [ths ssue], s our attempt to fll ths vod These studes demonstrate that the remote sensng of surface currents n coastal areas usng HF radar systems s an accurate technology sutable for wder use wthn the oceanograph communty Acknowledgments The authors would lke to thank Danel Fernandez for pontng out the underlyng consstency of our work wth that of Lpa and Barrck We'd also lke to thank Erk Bock for use of LADAS as a platform for makng some of these measurements The nsghtful revews of Jeff Paduan and Davd Prandle were much apprecated Ths research was supported, n part, by the followng grants from the Offce of Naval Research: N C-0001 (Chapman), N J-1585 (Edson and Karachntsev), N (Shay), and N J-1775 (Graber) The OSCR measurements were supported by the ONR Remote Sensng program and Mnerals Management Servce through grant N J (Shay, Ross, and Graber) References Bartck, D E, J M Headrck, R W Bogle, and D D Crombe, Sea backscatter at HF: nterpretaton and utlzaton of the echo, Proc EEE, 62, , 1974 Barrck, D E, M W Evans, and B L Weber, "Ocean surface currents mapped by radar," Scence, 198, , 1977 Bock, E J, and T Hara, Optcal measurements of capllary-gravty wave spectra usng a scannng laser slope gauge, J Atmos Oceanc Technol, 12(2), , 1995 Crombe, D D, Doppler spectrum of sea echo at 1356 Mc/s, Nature, 175, , 1955 Crombe, D D, Resonant backscatter and ts applcaton to physcal oceanography, n Proceedngs qf leee Ocean '72 Con l%rence on Engneerng n the Ocean Envronments, , EEE Press, Pscataway, NJ, 1972 Edson, J B, J E Hare, and C W Farall, Drect covarance flux estmates from moble platforms at sea, J Atmos Oceanc Technol, n press, 1996 Fernandez, D M, and J D Paduan, Smultaneous CODAR and OSCR measurements of ocean surface currents n Monterey Bay, n EEE nternatonal Geoscence and Remote Sensng Symposum, pp , EEE Press, Pscataway, NJ, 1996 Frsch, A S, and B L Weber, A new technque for measurng tdal currents by usng a two-ste HF Doppler radar system, J Geophys Res, 85 (C1), , 1980 Georges, T M, and J A Harlan, Mappng surface currents near the Gulf Stream usng the Ar Force over-the-horzon radar, n Proceedngs of the EEE Ffth Workng Con l krence on Current Measurement, pp , EEE Press, Pscataway, NJ, 1995 Graber, H, B K Haus, L K Shay, and R D Chapman, HF radar comparsons wth moored estmates of current speed and drecton: Expected dfferences and mplcatons, J Geophys Res, ths ssue Herr, F, C Luther, G Marmorno, R Med, and D Thompson, Ocean surface remote-sensng program planned, Eos, 72, 214, 1991 Holbrook, J R, and A S Frsch, A comparson of near-surface CODAR and VACM measurements n the Strat of Juan De Fuca, August 1978, J Geophys Res, 86(Cll), 10,908-10,912, 1981 Lese, J A, The analyss and dgtal sgnal processng of NOAA's surface current mappng system, EEE J Oceanc Eng, 0E-9(2), , 1984 Lpa, B J, and D E Barrck, Least-squares methods for the extracton of surface currents from CODAR crossed-loop data: Applcaton at ARSLOE, EEE J Oceanc Eng, 0E-8(4), , 1983 Marmorno, G O, and C L Trump, Prelmnary sde-scan ADCP mea- surements across a shp's wake, J Atmos and Oceanc Technol, 13(2), , 1996 Matthews, J P, J H Smpson, and J Brown, Remote sensng of shelf sea currents usng a hgh-frequency ocean surface current radar system, J Geophys Res, 93(C3), , 1988 Paduan, J D, and L K Rosenfeld, Remotely sensed surface currents n Monterey Bay from shore-based HF radar (Coastal Ocean Dynamcs Applcaton Radar), J Geophys Res, 101(C9), 20,669-20,686, 1996 Peters, NJ, and R A Skop, VHF radar measurements of ocean surface currents from a movng shp, RSMAS Tech Rep , 56 pp, Unv of Mam, Mam, Fla, 1995 Porter, D L, R G Wllams, and C R Swassng, CODAR ntercomparson: Delaware Bay 1984, n Proceedngs qf EEE Thrd Workng Con l%rence on Current Measurement, pp 36-44, EEE Press, Pscataway, NJ, 1986 Prandle, D, A new vew of near-shore dynamcs based on observatons from HF radar, Prog Oceanogr, 27, , 1991 Shay, L K, H C Graber, D B Ross, and R D Chapman, Mesoscale ocean surface current structure detected by HF radar, J Atmos Oceanc Technol, 12(4), , 1995 Skop, R A, D B Ross, NJ Peters, and L Chamberlan, Measurements of coastal currents usng a shp based VHF radar system, RSMAS Techncal Report , 25 pp, Unv of Mam, Mam, Fla, 1994 Stewart, R H, and J W Joy, HF rado measurements of surface currents, Deep Sea Res, 21, , 1974 Trzna, D, Mappng ocean currents usng over-the-horzon HF radar, lnt J Remote Sens, 3(3), , 1982 Wells, D E, et al, Gude to GPS Postonng, Can GPS Assoc, Fredercton, N B, Canada, 1986 R D Chapman, Appled Physcs Laboratory, Johns Hopkns Unversty, Laurel, MD (e-mal: chapman@teslajhuapledu) L K Shay and H C Graber, Rosenstel School of Marne and Atmospherc Scences, Unversty of Mam, 4600 Rckenbacker Causeway, Mam, FL J B Edson and A Karachntsev, Woods Hole Oceanographc nsttuton, Woods Hole, MA C L Trump, Code 7340, Naval Research Laboratory, 4555 Overlook Ave, SW Washngton, DC D B Ross, vy nc, 3100 N Bay Road, Mam, FL (Receved March 21, 1996; revsed November 12, 1996; accepted November 22, 1996)