S Guide to Radar Ornithology?

A Peterson's Guide to Radar Ornithology?

Some day it may be possible to identify a number of species by their radar "signatures."A state of the art report.

Timothy C. Williams and Janet M. Williams

NTHE PAST TWO DECADES radar has made some of the greatest contribu- visual observation show that birds are tions to the study of bird migration; this technique has revealed the able to change their wingbeat rate by general patterns of avian migration at night when most migrants make their 50ø70on take off and landing and, thus, journeys. A powerful search radar can detect the movementsof birds over it is not surprisingto find great variation in the radar signature of a bird flying thousandsof square kilometers, and some radars which can determinethe closeto the ground or recentlydropped altitude of migrants have revealedmass migrations at altitudes of 6000 meters from a balloon or helicopter. Vaughn, above the earth; tracking radars have followed singlebirds for more than 100 using the superb facilities of NASA at kilometers. The advanceof radar ornithology, however, has been stymied by the Wallops Flight Center has gone far- an apparent paradox: nocturnal migrants detectedby radar have not been thest in analyzing these data, and he identified as to species,and accurately identified birds have not been fol- believesit is possibleto removemost of lowed in their migrationsby radar. In this paper we would like to discussthe the observed variation by making very problem briefly in the hope of stimulatingsome of the readersof this journal precisecorrections for rate of climb or to report observationsthey may havemade under particularlyfavorable con- fall. These corrections will, however, ditions. answer only part of the question; one still must know how variable is the radar signature of a single bird, or, more ex- IDENTIFICATION OF MIGRANTS Almost any object which is not a per- actly, a speciesof birds. BY RADAR fect spherewill changeits effective echo- We approached this problem by ex- ing area, or radar cross section, as it amining the variability in the radar sig- moves. Radar signatureshave long been natures of unidentified nocturnal HEREARE MANY DIFFERENT types of used to identify aircraft and missiles. migrants tracked over unusually long radar used for observingmigration Several papers in the last decadereport distances. For this analysis we used (Eastwood, 1967). All of them indicate progressin identifying birds and insects tracks of birds obtained with the missile that there are small, slowly moving ob- by means of radar signatures(Konrad et tracking radars at Wallops Flight Center jects aloft which are not simply drifting al., 1968; Schaefer, 1969, 1976; Williams (NASA) between 1969 and 1973. Infor- with the wind; these we assume to be et al., 1972; Vaughn, 1974; Emlen, mation on the radars and our tracking birds or insects. If we can determine the 1974). These studies report that radar techniques may be found in Williams direction, speedand altitude of theseob- signatures of many different birds are and Williams (1972) and Williams et al jects (as with a weather radar), and can recognizably different. The obvious next (1977). measurethe velocity of the wind at that step is to make a catalogof radar signa- CLASSES OF RADAR SIGNATURES altitude with a weather balloon, we can tures of known birds and then use this as then calculate the flight speed (or air- a guide for the identification of birds LL RADARSIGNATURES obtained speed)of the bird. (The movementof a detectedwith radar when they cannot be with the radars at Wallops Island bird over the surface of the earth is the visuallyidentified. Emlen (1974) tracked fall into one of the four categories vector sum of airspeed and the speedof birds released from a balloon. Vaughn shown in Figure 1. The birds giving the wind.) Both visualobservations and (1974, 1978) released birds from a signatures illustrated in Figures la and aerodynamic calculations (Pennycuick, helicopter, and Bruderer et al. (1972) lb were identified, in diurnal observa- 1969) indicate that certain flight speeds and the authors identified birds flying tions (with a 1 m focal length telescope are charactertisticof many if not all bird during the day with telescopesmounted mounted on the radar antenna), as soar- species and, thus, calculated airspeed on the radar antenna. The radar signa- ing birds (la) or a small flock of birds should be helpful in identifying tures obtained in all these studies were (lb). Neither of these types of signature migrants. usuallymuch more variable than the sig- shows periodic fluctuations. The soaring Potentially the most useful informa- natures of birds flying on long straight records however, do illustrate the very tion for identification is the radar signa- flights at night. great change in average radar cross- ture of an object being followed by a section with aspect:the strengthof the trackingradar. This is thechange in ef- HEMAJOR PERIODIC fluctuations in echo from the Turkey Vulture tracked in fective radar echoingarea of the object radar signatureare almost certainly Figure la varied by more than 40 db (a with time and is obtained from tracking due to movement of the wings in birds factor of 10,000) as it circled on a ther- radars which follow a singleobject at a and insects (Eastwood, 1967, and mal updraft. time, plotting its path accuratelyin three Schaefer, 1976). Both high speed Only the signaturesshown in Figures dimensions. photography (Ruppel, 1978) and direct lc and ld, termed "continuous" and

738 AmericanB•rds, September 1980 "bursting" are likely to lead to lden- tlhcation of nocturnal migrants. Even thesebroad categoriesare not mutually exclusive as occasionally bursting records show 20 to 30 seconds of con- nnuous activity, and continuousrecords are at times interrupted by brief periods of inactivity. This seemsreasonable for a) Soaring the wing-beatpatterns of ascendingor descendingbirds; but such changesare also seen in records from birds flying straight and level paths (although the b) FJock radar track would probably not show deviations of lessthan 10 m). Thus, even at this stageof analysisthere is evidence that migrantsmay alter flight behavior c) Continuousbeating significantly.

VARIABILITY IN THE RADAR SIGNATURES OF SINGLE BIRDS d) Bursting

NEMIGHT ASSUME THAT the shape I. I of the signaturerecords--curved, f second sharp-pointedor rather flat--might be useful, but inspectionof even a few such Figure1. Examplesof four typesof radarsignatures. a and b werefrom diurnalmigrants ]den- titled with a telescopeduring the day. c and d are examplesof signaturesfrom nocturnal records(see Figure 2) clearlyshows great migrants.Some signatureslike c and d have been obtainedfor 2 to 3 hourswith little change changesin this parameter in a single record. These changes are due to two From all available tracks we selected those than 20 db. We were interested only in Nrds factors: first, the radar signaturerecords most likely to give consistent signature with significant aspect changes in their records. The criteria for selection were as tracks, and rejected all tracks with aspect are taken from the Automatic Gain follows: all tracks had to be straight and changesof less than 45ø . Fundamental fre- Control (AGC) in logarithmicform and, relatively level (less than I m/sec. rate of quencyof the radar signaturewas determined thus, a sinusoidalrecord such as that in climb or fall) and taken during periods of by visual inspectionin those portionsof the Figure lc does not indicate sinusoidal moderateor heavy springor fall arian migra- 20-second records showing the clearest pat- changesin the radar reflectanceof a tion as determinedby radar (see Williams et tern. The distance between peaks at thmr al., 1977). All tracks had to be more than 15 sharpestpoints was measured with dividers bird; and second,small changesin the minutes long, have a low noise level, show and converted into an instantaneousfrequen- aspectof the bird relative to the radar clear, regular fluctuationsin the radar signa- cy. The measurementerror of 0.2 mm was the (whetherthe bird is viewedhead on, tail ture, and have a signal-to-noiseratio greater on, or broadside) produce very large changesin the radar signature(see East- wood, 1967).We are, thus, left with the primary frequencyof oscillationor fundamental frequency of the radar signature as the most important parameter and will concentrate the remainder of this discussionon that topic.

METHODS

ETWEEN1969 AND 1973 we obtained over 500 tracks of free flying birds with the largetracking radars at Wallops Island, Va. The two radars used for the data reported in this paper werethe SPANDER: 10 cm wavelength, 3.4 MW peak power, beamwidth 0 39ø at the 3 db points, and the FPS-16:5 cm wavelength, I MW peak power, 1.2ø beamwidth at the 3 db points. We recorded radar signatures every 5 minutes for 20 secondson a paper chart recorder run at 4 ]n/sec. for the SPANDAR and at 10 cm/sec. for the FPS-16 radar. The signature in both second' • =One Cycle casesconsisted of peak detectedAutomatic Gain Control voltage. More details of the Figure2. Sixradar signatures from a singlebird taken at differenttimes. Elapsed time from the specificationsof the radarsused and the pro- startof the track is givenfor eachrecord. The bar indicatesthe lengthof a singlecycle of the fun- ceduresfollowed are given in Williams and damentalfrequency of the signature.The lengthof cycleshown by the bar remainsrelauvely Wdliams (1972). constantbut the shapeof the signaturechanges markedly from recordto record.

Volume34, Number5 739 major sourceof error, especiallyfor records showing high frequencies;at 17 Hz the error would be about 4ø7oand at 8 Hz, 2ø7o.

RESULTS o ß - SUCCESSIVE •, ß - •qlTHIN 20 SECONDS FTHE FUNDAMENTAL frequency ofa ra- • • - •qlTHIN 10 MINUTES dar signature is to be used in identifica- non •t must be a relatively unvarying charact- o eristic of the bird. Each of the points in o• lc F•gure 3 represents the standard deviation x (s d ) of ten measurementsof the fundamental frequency of a radar signature. A cross representsthe s.d. of ten points evenly dis- tabuted over all records available for that b•rd A triangle indicates the s.d. for ten points evenly distributed over a 20-second record of radar signature, and a circle shows the variation within ten consecutivecycles of a s•ngle record. The data for each bird are enclosedby a dotted line; in casesfor which data from two birds are enclosed in the same area, the data are distinguished by open vs sohd symbols. Figure 3. Standard deviation in the fundamental frequency of radar signaturesas a function of F•gure 3 shows that the variability in fun- the averagefundamental frequency.Symbols represent data from different samplingtechniques damental frequency for nocturnal migrants is (seetext). Data from one or two tracksare enclosedby a dotted line. Open and closedsymbols often very low (s.d. = 5ø7oin 20-minute distinguishdata from different birds. Data setsare identified by letters; when two birds are iden- fhghts). In no case did the signatures from tified by the sameletter, the one with the lower averagefundamental frequency(x) is referred to m•grants show the extreme changesrecorded as C 1 andthe otheras C 2. m released birds (Vaughn, 1974; Emlen, 1974). Changesin the fundamental frequency track. The relationship between rate of climb bursting signatures being less reliable w•thm a 20-secondrecord were rarely greater or fall and relative length of the active phase Vaughn (1978) using very long records of than changesrecorded for 10 sequentialwing- is apparent even for these very small rates of flights with a sophisticatedsystem for deter- beats. This conclusionwas tested by compar- altitude change. Percent activity within a mining frequency from tape recorded AGC ing the variance of the two sampling meth- signatureis, thus, a very labile aspectof the output shows correlations between rate of ods Only tracks (B and F) gave significant F radar signatureand might changeas birds en- climb or fall, airspeed,and the signaturefre- ratios (p < .01) and in both casesthe signifi- countered slight updraft or downdrafts on quencywhich may considerablyreduce varia- cant difference was entirely due to a single even straight and level flights. Emlen (1974) tion in the fundamental frequency that we aberrant point. reports similar findings for White-throated found. It thus appears that the radar signature of Sparrows (Zonotrichia albicollis). these birds is quite constant over relatively VARIABILITY IN THE OBSERVED short periods of time (20 seconds). Is the In conclusion,the fundamental frequency FLIGHT BEHAVIOR OF BIRDS same constancy true over periods of 20 of the radar signature appears to be the most m•nutes? To test this we subjected the reliable indicator for species identification, measurements taken from different records with class of signature (bursting vs con- F WEARE TO USE the radar signatures but for the same bird track to an analysis of tinuous) and percent time beating for describedabove to identify birds, we variance.Five of the records(B, D2, E, F, G2) showedno significantdifference in the +2 fundamental frequency over all the records for that track. One (A) showeda difference at the p< .05 levelof confidenceand four (C1, C2, Di, Gi) showeddifferences at the p < .01 z level of confidence. Thus, in most cases the variability in long records was similar to that •n short records, but in four cases the fundamental frequency was significantly different at various times in the same track. If we had been using a less powerful radar which could only track these birds for 5 m•nutes we would have concluded that the different parts of what were in fact one track were different birds or perhaps different radar species. For those tracks exhibiting a "bursting" type of signature (see Figure 1), the relative length of the active and inactive phases has been used to distinguish different types of s•gnatures (Bruderer and Steidinger, 1972). Although the tracks of birds we used were nominally straightand level therewere departures from absolutely level flight. In Figure 4 we plot the percent time in the active phase vs - 20 50 40 50 60 70 80 90 100 the rate of climb or fall. This figure contains PERCENT OF TIME BEATING data from those tracks with bursting signatures which showed more than .25 m/sec. Figure 4. Variation in the relative length of the active and inactive phasesof "bursting" radar changein the rate of climb or fall during their signature records with change in rate of climb or fall.

740 AmericanBirds, September 1980 must have information on the flight be- 100 different "radar" speciesof birds tative analysis of bird migration with a havior of identified birds under similar under the best of conditions with a con- tracking radar. In Animal Orientation circumstances.We first approachedthe servative estimate being about 30 radar and Migration, A Symposium.NASA-SP 262, U.S. Gov't Printing Office, Wash- problem by making high speedfilms of species.However, at presentwe lackthe ington, D.C. birds flying during the day near the information on the flight behavior of __ B. JACQUATund U. BR•OCKNER ground. Almost none of these films was identified migrants needed to make use 1972. Zur Bestimmungyon Fli•gelschlag- of birds making straight, level, long of these radar identifications. frequenzen tag- und nachziehender Vogelarten mit Radar. Orn. Beob 69 distance flights. As might have been Observations of the flight speed and 189-206. predicted, analysis of such films showed wingbeat rate of birds should also EASTWOOD, E. 1967. Radar Ornithology great variability (up to 100%0) in the specify the direction and speed of the Methuen and Co. Ltd., London.. 278 pp wingbeat frequency of the birds. We, wind and should assure that the tracks EMLEN, S. 1974. Problems in identifying therefore, are suspiciousof the values are straight and level for a distance of birds speciesby radar signatureanalysis Intraspecific variability. Proc. Conf on for both airspeedand wingbeat frequen- perhapshalf a kilometer or more. To be the Biological Aspects of the Bird/Air- cy (see Greenewalt, 1975 for review) sure, such observation conditions are craft Collision Problem. S. Gauthreaux which have been published for many not easy to come by and yet some (ed.), ClemsonUniv., Clemson, S.C speciesof birds. We then attempted to readersof this journal may well have en- __ and N. DEMONG. 1978. Orienta- tion strategies used by free-flying bird track migrants during the day with radar countered such an opportunity, perhaps migrants. In Animal Migration, Naviga- and identify them with large telescopes during the early morning reorientation tion and Homing, K. Schmidt-Koenigand as they flew near the radar. This proved flights of nocturnal passerinemigrants. W.T. Keeton (eds). Springer Verlag, most frustrating as the closest distance If such data are at hand or could be easi- Berlin, pp. 283-293. which a bird could be followed with ly obtained, they would be of great in- GREENEWALT, C.H. 1975. The flight of birds. Trans. Am. Phil. Soc. 65:4. terest to those of us who use radar. these large radars was very near the KONDRAD, T.G., J.J. H1CKS and E B greatest distances that a bird could be DOBSON. 1968. Radar characteristics of Identified with the best optical equip- ACKNOWLEDGEMENTS birds in flight. Science159: 274-280. ment. Reliable resultscould only be ob- PENNYCUICK, C.J. 1969. The mechanics tinned with diurnal migrants such as of bird migration. Ibis 111: 525-526 ETHANK THE MANY people on the RUPPEL, G. 1978. Bird Flight. Van Nos waterfowl. (The tabular results of this staff of Wallops Station, NASA Reinhold, New York. work will be sent to readerson request.) who assistedin this work. Many friends SCHAEFER, G.W. 1969. Bird hazard and its and graduate studentsassisted in obtain- radar detection. Proc. World Conf on Bird Hazards to Aircraft. Nat. Res Accurate airspeedsand wingbeatrates ing the radar signatures.In particular we Council Canada, Assoc. Comm. Bird might be obtained from birds released wish to thank Carol McClintock for her Hazard to Aircraft. Ottawa. pp. 321-322 from helicopters or balloon and then assistance in analyzing the signatures 1976. Radar observations of insect tracked by radar. However, as we men- and her helpful comments on the manu- flight. In InsectFlight, R.C. Rainey(ed), tloned before, such birds often seem script. Charles Vaughn of NASA assist- Syrnp. Royal Ent. Soc. London, No 7 Blackwell, Oxford. disinclined to begin their migratory ed in elucidating many of the more tech- VAUGHN, C.R. 1974. lntraspecific wing- journeysafter this treatment.Emlen and nical aspects of this problem. This re- beat rate variability and speciesidentifica- Demong's(1978) work on White-throat- search was supported by the National tion using tracking radar. Proc. Conf on ed Sparrowsseems to be a stepforward, Science Foundation grants GB 13246 the Biological Aspects of the Bird/Air- as some of the many birds they released and 43252 to the Woods Hole Oceano- craft Collision Problem. S. Gauthreaux did appear to begin straight and level (ed.), Clemson Univ., Clemson, S.C graphic Institution and NASA contract __ 1978. NASA radar entomology program flight in an appropriatemigratory direc- NGR 33-183-003 with the State Universi- at Wallops Flight Center. In Radar, Insect tion, but the effort involved in obtaining ty of New York. Population Ecology and Pest Manage- these data for even a single specieswas ment. Proc. Workshop May 2-4, 1978 C very great and probably is not possible SUMMARY Vaughn, W. Wolf and W. Klassen(eds), NASA Wallops Flight Center, Wallops for a large number of species.Vaughn's Station, Va. approach of correctingobserved flight HEVARIABILITY OFTHE radar signa- WILLIAMS, T.C. and J.M. WlLL1AMS behavior for departures from straight ture of enroute nocturnal migrants 1972. Tracking radar techniquesfor stu- and level flight holds promisebut needs in nearly steady state flight was deter- dying migratory birds, July 1972, NASA CR-62082, NASA Contractor Report to be widely applied before it can pro- mined for flights of 15 to 20 minutes. Available from NASA, Wallops Station, vide a means of identifying unknown The fundamental frequency of the Wallops Island, Va. 23337. nocturnal migrants. radar signature was the only parameter __ and J.M. TEAL and J.W. KANWISH- judged to be a reliable discriminator of ER. 1972. Tracking radar studiesof bird Despitehigh technologyand the dedi- different signatures. migration. In Animal Orientation and cated work of many scientistswe still re- Navigation: A Symposium. NASA SP Although the standard deviation of main at the impassestated as we began 262. U.S. Gov't Printing Office, the radar signatureswas low (about 5%) this paper. Radar offers the possibility Washington, D.C. pp. 115-128. several signatures from single birds __ and L.C. IRELAND and J.M. TEAL of identifying nocturnal migrants flying showed significant changes over the 1977. Bird migration over the western beyond the limits of human vision. On North Atlantic Ocean. Am. Btrds period of observation. the basis of the observed variation in 31:251-267. airspeedsalone and the variations in LITERATURE CITED fundamental frequency we found, it seems likely that radar signature and BRUDERER, B. and P. STE1DINGER. --Department of Biology, Swarthmore airspeedmight reliably distinguish50 to 1972. Methods of quantitative and quali- College, Swarthmore, PA 19081.

Volume 34, Number 5 741