j. Raptor Res. 23(3):72-84 ¸ 1989 The Raptor Research Foundation, Inc.

CENSUSING OF DIURNAL RAPTORS IN A PRIMARY RAIN FOREST: COMPARATIVE METHODS AND SPECIES DETECTABILITY

JEAN-MARC THIOLLAY

ABSTRACT.--Noreliable method has ever beenproposed to censusa rain forestraptor community.I investigated4 methods in primaryforest of FrenchGuiana and compared results; 1) Mappingterritorial pairs, displayingover canopyand followedfrom dominantlookouts, gives the mostcomplete data on regularlysoaring species (Harpagus, Buteogallus, Spizaetus, Spizastur); 2) meaninstantaneous number of birdsflying over a definitearea in optimumconditions may be a reasonabledensity estimate for vultures (Cathartidae);3) mappingof individualsrecorded along a regularnetwork of trailsin the understory wasonly successful for the highlyconspicuous Red-throated Caracara (Daptrius americanus); 4) density estimatesfrom understorystrip transectswere consistentwith thoseobtained by othermethods for 6 of 8 species.Specific detectability, soaring behaviors and frequencies of displayflights varied widely among speciesand so did time required to assessthe existence of a territorialpair (1-7 d). Marking,radiotracking andplayback of vocalizations are promising techniques but are very time consuming and more appropriate for a detailedstudy of particularpairs than for surveyof a wholecommunity. Use of an abundanceindex, with a distincttechnique for eachspecies, may avoidbiases of densityestimates.

Birds of prey are notoriouslydifficult to surveyin many speciesare almost impossibleto find in pri- tropical forests,especially in tall, dense,large un- mary forest,except under extraordinarily lucky cir- broken tractsof humid lowland forest.No complete cumstances.In fact, for many species,the nest has censusof a rain forestraptor community,with den- neverbeen described. Several species, including some sity estimatesover a significantarea, has ever been of the mostcommon neotropical forest raptors, seem publishedand no appropriatemethodology has even to never soar(?) nor even fly over the canopy and been proposed.However, many rain forestraptors very rarely ventureoutside the understory.Although are now threatenedby habitat destruction,distur- not particularly shy, most speciesare very secretive bance or fragmentation(Thiollay 1985b). Raptors and spend long periods perched motionless.Many may be suitableindicators of optimum sizeof a forest are very vocal,but othersare usually silent. Density reserve,because raptors are likely to require areas is often low and distribution very patchy, which larger than most other species.Yet, there is still an further decreases rate of encounters. urgent need of basicdata on natural distribution and Thus, most raptor-specific censusmethods (see density of rain forest raptors becauseof a concern review in Fuller and Mosher 1981), devisedfor tem- about the suitability of many reservesor even na- perate species,cannot be applied directly to tropical tional parks which may well prove to be too small forest species.Nevertheless, I have tried to adapt for long term survival of someraptor speciessup- classicalconcepts to proposeempirical methodsthat ported originally. need further improvementand testingin similar sit- As part of a larger studyon designof a national uations. My attemptsare preliminary, and as such, park in French Guiana, I assessedthe distribution voluntarily unsophisticated.No single method can and relativeabundance of raptorsover the country's be appropriate for every species,and 2 or more dif- 80 000 km 2 rain forest area, and I estimated the ferent techniquesshould always be used concur- densityof everyspecies within a representative100 rently. km2 samplequadrat (Thiollay 1989). In this paper I concentrateonly on comparative Life history and behaviour of most rain forest resultsof complementarymethodologies. Biological raptor speciesare very poorly known, if at all (Thiol- and conservationsignificance of the data are devel- lay 1985a). Often only scantinformation comes from oped elsewhere(Thiollay 1989). marginal habitats,rather openwoodlands or edges where a species'biology may be atypical.After 20 STUDY AREA AND METHODS yrs of personal experiencein both New and Old After 5 yrs of raptor surveysthroughout French World tropical forests,I can testify that nestsof Guiana, I tried to obtain a quantitativeestimate of

72 AUTUMN 1989 CENSUSING RAPTORS IN RAIN FOREST 73 the densityof every raptor specieswithin as large a crossedby foot at least twice, including 35 outside representativearea as possible. Zone II. Around the Nouragues field station, in north-cen- Mapping of SoaringBirds. One groupof species tral French Guiana (4ø05'N, 52ø41'W), a 100 km 2 regularlysoar above canopy and may then be easily quadrat of primary forestwas chosenwithin a much detectedfrom outside the forest or through large larger expanse of similar unbroken lowland rain openings:Gray-headed Kite (Leptodoncayanensis), forest. The area was hilly, crossedby small rocky Hook-billed Kite (Chondrohieraxuncinatus), Ru- watercourses,and ranged in altitude from 27-413 fous-thighedand Double-toothedKites (Harpagus m. Also includedwere sizeabletracts of every forest cliodonand H. bidentatus),Tiny, Bicoloredand Gray- type encounteredin Guiana, exceptcloud forest. The bellied Hawks (Accipitersuperciliosus, A. bicolorand quadrat was centeredaround 2 large natural open- A. poliogaster),White Hawk (Leucopternisalbicollis), ings, with small field stations nearby: a medium- Great Black Hawk (Buteogallusurubitinga), Crested s•zedriver in the southand a large graniticinselberg Eagle (Morphnusguianesis), Black and White Eagle in the northern part provided the most convenient (Spizastur melanoleucus),Black and Ornate Hawk observationsites (Fig. 1). Eagles(Spizaetus tyrannus and S. ornatus).Soaring The surveytook place in the driest season,from behavior performed by adults on their breeding early Septemberto mid-October1986 and 1987 over groundshas mainly a territorial function (surveil- a total of 73 d of intensivefield work with unusually lance and maintenanceof pair bond or territorial fine weather conditions. From the little information limits, Newton 1979). Soaringis often accompanied available (Haverschmidt 1968 and pets. obs.), most by loud calling, nuptial display or instraspecific raptors were either breedingor ending their repro- aggression.I assumedthat adults, especiallywhen ductivecycle (feeding fledgedyoung). All were as- displaying,flew mostly,if not only, over their own sumedto be sedentarywith an extendedperiod of territory. breeding activity and a permanent pair bond and A large, bare, rocky outcrop protruding 200 m territory occupancy. over the surroundingforest offered an ideal vantage Different censusmethods have been adapted to point. Four convenientlookouts were chosenon the the behaviorof individual species.Methods are here outermostparts of the inselberg,each offering an proposedas preliminary suggestions.Comparative unrestricted180 ø view. A total of 167 hr was spent results will be used as a test of reliability. As the overlookingthe forest in fine weather and mainly coveragewas better, more intensive and frequent during the optimal morning hours (0900 H-1200 around field stations and large openings,2 better H). The marked relief facilitated the location of known areas were distinguished(Fig. 1). Zone I (6 flying birds. km2) included the inselberg and its lookouts, the All raptors seenflying over the forest (or perched main field station and its clearing as well as a dense on emergent dead trees) were followed and their networkof trails. Every speciespresent was assumed itinerary carefully mapped (1/50000 scale). Day to be recordedand reasonablywell mapped(no new after day, the data were superimposed,soon giving data during the last 30 d.). ZoneH (42 km2 including a pictureof clearly separatedranges for mostspecies Zone I) was centeredaround the inselbergand cov- So-calledterritories were derivedby the minimum ered a 3-kin tad area around the 4 outmost lookouts convex polygon method (Ford and Myers 1981; of the mountain top from which only soaringspecies Southwood 1966) connectingthe outermostpoints were mapped. A 3-kin tad was the maximum dis- reachedby birds under observation.Territory size tance at which large raptors were accuratelyiden- was determinedusing a planimeterand correcting tified and located with 10 x 40 binoculars and 11- for small samplesize biasesby the methodof Jenn- 33 x telescope. rich and Turner (1969). Adjacent territories were The entire studyarea (100 km2) was visiblefrom discriminatedby simultaneousobservation of the 2 the top of the inselbergbut only the largestraptors pairs involved. could be seenbeyond Zone II. However, most outer Over 20 completeflight circuits were drawn for regionswere in sight abovecanopy within 2.5 km at least 1 pair of the 5 mostcommon and conspicuous of 9 additionallookouts (riverbanks and large treefall species(H. bidentatus,L. albicollis,B. urubitinga,S. gaps on ridges or steep slopes).The quadrat was melanoleucusand S. ornatus).Each pair yieldedrare- divided into 1 x 1 km squares,of which 70 were ly more than 1 circuit/d. No new information was 74 JEAN-MARCTHIOLLAY VOL. 23, NO. 3

+ + : ."..../7 + ß I :':...,

+ + + +

/ '; '.3 " :"".... +++I i.. + + +l + :.....j + + +

! + + + + + + + + + + + {../'....: •...... t..+ ñ.....['

mereram <'.'.7 AUTUMN 1989 GENSUSING RAPTORS IN RAIN FOREST 75 obtainedafter the 5-10 circuitsplotted initially. Thus, rad on a 180ø field, the largest area manageablefor samplesize was probably large enoughto avoid an this purpose), from the western lookout of the in- underestimationof territory size (Ford and Myers selberg.I divided daylight hours (between sunrise 1981). and sunset)into 4 periods.Each period was covered Population Estimatesof SpeciesHunting Over during 17-20 hr. (fine weatheronly). Within each the Forest. Another group of raptors fly abovecan- hour observation,the following parameterswere re- opyto searchfor carrion(vultures) or hunt for insects corded: 1)minimum number of different individuals (kites) and birds (falcons).Each group is easily de- seen; 2) total time (rain) where at least 1 bird was tected but wide ranging and may be occasionally flying over the area; and 3) behavior of each indi- gregarious. vidual (soaring, hunting, displaying,... ). Besides The most common speciesin the group is the vultures, whosehome rangeswere difficult to define, Greater Yellow-headed Vulture (Cathartes rnelam- all 12 soaringspecies were representedby 1 resident brotus)which wandersthroughout the area and tem- pair on the 10 km2 area (either > 50% of their ter- porarily concentratesaround carcasses.Particular ritory includedin the area or a larger territory cov- pairs could not be separated.For want of a more ering >50% of the area). accurate estimate, I have recorded during 20 hr in The number of flying raptors slowly increases the late morning (0900 h-1200 H) period,the mean from sunriseto 0900 H, then quicklyreaches a max- number of individualscrossing a 10 km2 area/hr. imum between 1000 and 1100 H, and decreases The 20-hour data, extrapolatedto the larger study almost continouslyafter noon. However, vultures area, are in good agreement with the largest con- remained very active up to about 1600 H and kites centration of birds seen during the course of the were even most active in mid and late-afternoon study. when no rain had occurred. A similar estimatehas been computed for the King Understory Censuses.The Forest Falcons(Mi- Vulture (Sarcoramphuspapa) and the Swallow-tailed crastur), the Black-faced Hawk (Leucopternisme- K•te (Elanoidesforficatus), both of which wander lanops)and, to a lesserextent, the Red-throated Ca- great distances;the former solitarily or in pairs and racara, rarely fly high over the canopy. Caracaras the latter in flocks. Additionally, resident families are very noisybut ForestFalcons usually call mostly were recognizableand their roostsites located. at dawn and Leucopternisis not often heard. The The PlumbeousKite (Ictiniaplumbea) is well dis- Harpy Eagle (Harpia harpyja) is only conspicuous tnbuted over the area in isolated pairs which oc- when sun-bathingon emergentdead trees in early casionallyjoin nomadicgroups of Elanoidesbut are morning. Since it was necessaryto scour the forest otherwise rather territorial. Plumbeous Kites have to find these species,I tried to use sightings in 2 beenmapped according to the previousmethod and different ways. For soaringspecies understory rec- a small floating population, inferred from rare con- ordswere comparedto resultsof the abovemapping centrations observed,has not been considered.One method. pair of Orange-breastedFalcon (Falcodeiroleucus) I first walked along a network of trails in the was attachedto the only suitablecliff. understory, 100 km long, designedto pass within Specific Daily Soaring Activity. Each species 500 m of every point. Thirty km were already laid had its own pattern of soaringbehavior and hence out through the undergrowth, mostly in Zone I. a differentdetection probability. To assessdaily flight Remaining "trails" were only outlined by a white pattern and time during which each specieswas threadat breastheight. I movedslowly, focusing my msible, I define a fixed 10 km 2 area (i.e., a 2.5 km attentionexclusively on raptors and recordingevery

Figure 1. Topography of the 100 km2 study area: main watercourses(unbroken lines), 100 m contourlines (dotted lines), maximum and minimum altitudes (meters asl), little woodedpart of the rocky inselberg(hatched contour). The circle featuresthe 42 km2 area coveredby a 3 km-rad around the 4 outmostobservation points around the inselberg'ssummit (asterisks).The hatched rectangle is the 6 km2 most intensively surveyedarea. Starsindicate the field stationsand the rectangularnetwork is the intensivestudy trails near the southernstation. Peripheral stippled squareshave been crossedby foot only once,if ever. 76 J•^N-M^I•C THIOLLAY VOL. 23, No. 3 observationon a map at the nearest 100 m. The area move, fly or call and thus can be detected.Large to be coveredwas too large and the density of birds terrestrialgame birds and macaws(Crax, Psophza, was too low to obtain enoughrecords and an accurate Ara spp.)were alsoincluded in the surveyand were mapping of territorial pairs outsidethe most inten- almostalways first detectedby calls. Flushing or sively searchedarea. detectiondistance is a critical parameter which has The second method was to use the data for a proved,from my experiencein Guiana, to be rather tentativedensity estimate of all individuals.The most constantfor a given speciesin primary, not hunted, rigorous technique shouldhave been a line transect forest. Detection distance is also lower than maxi- sampling [i.e., recordingalong a straight line sight- mum distance(both verticallyand horizontally)at ing and perpendiculardistances, as well as sighting which birds are visiblein undergrowthof high pri- angles,of every bird seen].A densityestimate may mary forest, ensuring that most birds do not flee then be derived through a probability function at when out of sight. Any departure from basic as- zero distancefrom the line. This is obtainedby fit- sumptionsleads to an underestimationof density. ting, to the perpendicular distance data collected, Such specificdetection distance (d) is used here Fourier series models which best meet the estimator as the radius of a circle moving with the observerat shape and goodnessof fit criteria (Burnham et al. its center and whose dia is the effective minimum 1980; Brennan and Block 1986). Unfortunately, the width of the strip transect.Sighting angle or per- first assumptionof line transecttheory, that all birds pendiculardistance are no longerinvolved since birds on the line must be seenwith probability 1, is not becomeconspicuous in any quarter as soon as the always met sinceraptors initially perchedjust along observeris closerthan d. Then densityestimate is the transectcan escapeundetected well ahead of the observer. Most of all, no reliable density can be D= n/2dL statisticallyestimated without at least 40-60 obser- vations, a sample size which for most raptors can where n is the number of birds detected within the hardly be obtainedwithin severalmonths (see Re- 2 d-wide strip and L is the length of the transect. suits). Probability of detectinga bird could not be From all individual detection distancesobtained, estimated.Hence any techniquebased on such an d will be the shortest distance at, or interval within estimation (Cochran 1977) could not be used. In- which the largest number of birds were recorded dividuals were not distinguishableand time was too [i.e., width maximizing the densityestimate (highest shortto make repeatedcounts on mosttransects, thus n/d ratio)]. Thus, detectionsfurther than d are not preventinganother method for estimatingprobabil- used. Sampling variance of the density estimateis ity of detection(Seber 1982) to be used.My survey dependenton sampling variance of n which could was not devisedto calculate proportion of the area be estimatedfrom replicatedcounts. However, sam- occupied(Geissler and Fuller 1986), a promising ple sizewas sosmall that the variancewas inevitably technique especially if birds are detectedby re- large and not very meaningful. sponsesto playbacksof their calls. Data Collection on Strip Transect. Field count- Therefore, the alternativemethod used was a strip ing procedure,or searchmethod, was carefully de- transectcensus which gave the best compromisebe- vised to meet as much as possiblethe prerequisite tween efficiencyand biases(Burnham et al. 1985; assumptionsof the strip census(all objectsmust be Verner 1985) and used the traditional conceptsof detected within the limits and have a fixed initial detectabilityand effective area surveyed (Emlen 1971; position; all sightings must be independent events Ramsey and Scott 1981). Strip transect censusis and their distancemust be accuratelymeasured). I equivalentto a long narrow quadrat within which walked very slowly (<1 km/hr), making as little all birds are assumed to be recorded. The transect noiseand movementas possible,along narrow trails was unbounded,drawn randomly,and transectwidth or throughundisturbed undergrowth and randomly was not adjusted to varying density of vegetation crossingevery forest type. Attention was focused which changedconstantly but within rather narrow exclusivelyon raptors and large game birds (Crax, limits at mid- or upper-levels.Daily sectionscovered Psophia,Ara). Only periods between sunrise and were of unequal size and many were traced 2 or sunset,without rain, fog or strongwind, were taken more different d. There is a distance from the ob- into account. serverunder which the speciesstudied almost always I recordedevery bird either sitting, walking or AUTUMN 1989 CENSUSING RAPTORS IN RAIN FOREST 77 flying, from groundlevel to top of the canopy,but rare that visual marking and banding are of limited not above. Distances between the observer and first interest for secretivespecies. sightingwere measuredwith a range finder (Opti- Vocalisations.Fortunately, several species have loud meter 620) then controlledwith the number of steps calls which helps in detectionand location. First of and rounded to the upper 5 m-interval. Although all, the very noisygroups of Red-throatedCaracara notused subsequently, angle deviation from the tran- and Black Caracara (Daptriusater) are easyto follow sectwas read throughthe sightingmirror of a liquid and cannot be missed if one moves a few hours filled compass.For eachobservation I alsorecorded through their home range. Around their nest, the the number of individuals, their age, sex, height, Orange-breastedFalcon and Bat Falcon (Falco ru- behavior,location on the map, directionof flight, as figularis) are aggressiveand vocal even when not well as habitat type and vertical structureof the breeding but may be silent and unobtrusivea few vegetation(estimated index of density,ranked from hundredmeters away. Black and the Ornate Hawk 1-4 in the 0-2, 3-14, 15-25, 26-36 and >36 m Eagles, as well as, to a lesserextent, the Great Black strata). Hawk, rarely fail to perform display flights above Birds were often detectedwhen taking flight or canopy with loud calling nearly every day if the alighting. Those calling were recordedonly if they weather is fine, but usually during a short time (5- were within sight distance.Only Caracaras,which 15 min) and only once-a-dayin mid- or late-morn- almost always utter their loud alarm calls when ing. Thus, territorial pairs can be locatedfrom inside seeing an observer,were sometimesnoted further the forest. away, but neverat more than 100 m. Somebirds The 4 sympatric Forest Falcons (Micrastur ruff- were seen in or from a tree-fall gap and their de- collis, M. gilvicollis, M. mirandollei and M. semitor- tectiondistance may havebeen higher than in closed quatus)are all very vocal but mainly during a short understory.Transect length was measureddaily on time at dawn when it is usually too dark to seethem a 1/50000 scalemap and monitoredusing a pe- in the understory.Most stop calling before sunrise dometer. Total distancewalked in good censusing and in Guiana during the dry seasononly M. miran- conditionsthrough the 100 km2 studyarea in 1986- dollei and M. semitorquatusare likely to be heard 1987 was 517 km in 498 hr. with somefrequency later in the morning and around Similar censuseswere performedfrom 1981-1986 sunset. However, calls (which are probably true in 8 other study areas throughoutFrench Guiana songs)are quite variable, rather similar to eachother (total: 1188 km in 1135 hr). Being often associated and not always easy to determine.Unfortunately, with the surveyof otherspecies, these censuses could playbackexperiments conducted so far rarely elicited not be as accurate(uneven speed, attention distract- a response(outside the dawn chorus) and failed to ed) and resultsmay sometimesunderestimate raptor attract birds.Many more attemptsare still necessary densities. to definewhich part of their call repertoiremay be Additional Techniques. Other methods are most efficiently broadcasted,at what time and at promisingfor applicationto the mostsecretive species what distance.Population density of the 6 km2 cen- but are time consumingwith limited results(Thiol- tral zone has beendrawn from distributionof calling lay and Tostain unpubl. data), as far as population birds, assumingthat both sexeswere vocal (often 2 estimateover a large area is concerned. birds close to each other seemed to call in turn or Trapping and marking. Forest Falcons and Bi- together). coloredHawk are not infrequentlycaught in mist Other techniquessuch as audio-luring utilizing nets or traps baited with live birds, but few other prey calls are being developedin the northern Neo- specieshave been caughtin this way. Yet, only an tropicsand have already producedsome very prom- unknown fraction of the population is likely to be ising results(J. Vannini, pets. comm.). Playback of captured,which does not fit our purpose.Subsequent conspecificcalls to attract birds or elicit a vocal re- radiotrackingof taggedbirds has provedto be in- sponseis certainly worth additional study although valuable to assesshome range and foraging patterns some speciesmay not be lured. Flushing canopy of individual speciesbut can hardly be considered birds or searchingfor nestsusing a low flying he- for a relatively short period multispeciessurvey. licopter has been attemptedrepeatedly in French Radiotracking could also be an elegant method to Guiana with disappointingresults. Specific trapping find occupiednests. Sightings and recapturesare so methodsalso remain to be investigated. 78 JEAN-MARCTHIOLLAY VOL. 23, No. 3

sumptionunderlying computationof density using RESULTS the strip method and probably leads to the overes- I am only concernedhere with applicability of timation found here. censusmethods actually used. More details about Mean Density of Aerial Hunters. Raptorshunt- densitiesand distribution patterns are given else- ing overthe canopyare usuallyseen on line transects where (Thiollay 1989). only in flight and where the canopyis rather open. Understory Mapping. The Red-throatedCara- Suchrecords would violatethe main assumptionsof cara aloneprovided enough records to constructcom- the strip censusmethod and thus cannot be used prehensivehome ranges,not only from > 200 sight- except to map speciesdistribution. Mean number of ingsbut from the movementsof noisyflocks followed individuals seen per hour over a given area has a out of sightand the locationof roostingplaces (Fig. high varianceand can rarely be extrapolatedto a 2). Limits between group territories have been muchlarger area. Only instantaneousdensities would checkedby simultaneousrecords of contiguousflocks. be reliable but are feasibleonly if almostall birds Outsidethe 6 km2 intensivelysurveyed core area, all are in flight at the sametime over a large area. This other specieswere seenor heard too infrequently to assumptionis met only for vulturesduring the best accurately map territories, and no individual was hours and it has been used as a minimum estimate marked. Nevertheless,data were useful in assessing of the total population (Table 2). Kites, on the other overall distribution of non-soaring speciesand to hand, are not on the wing for such long and pre- check from the ground the exact location of birds dictable periods. Therefore, the basic estimate of displayingover the canopyfar from any lookout. mean number of birds seen in flight over a given Strip Census.Notwithstanding the low number area may be biasedand must be complementedby of records on strip transects,density estimatesob- maximum flock size and the localisationof pairs. tained on strip censusthrough the 100 km2 study Mapping of Soaring Birds. Mapping is by far area are remarkably closeto thosederived from map- the easiestand probablythe mostaccurate technique ping of soaringbirds on the restricted42 km2-circle. when suitablelookouts are available(Fig. 3). How- The 2 estimatesare within a range of +40% for 6 ever, any method relying on occurrenceof raptors of the 8 speciesavailable for comparison(Table 1). soaringover the forestis very sensitiveto activity Significantly,lower densitiesare givenby strip cen- patternof the speciesinvolved. Detectability of most susfor specieseither very shy,secretive and unevenly forest speciesmay be very lqw, evenwhen the ob- distributed(A. bicolor)or restrictedto tree topsand serveroverlooks the wholeterritory of a pair in fine openings (L. albicollis).At least 2 pairs of Lined weather. The probabilityto contacta speciesat least Forest Falcon (Micrasturgilvicollis) were found on onceduring any 1-hr observationbout is <50% for 300 ha in the Centralzone; the strip censusslightly 10 of 14 speciesand -<30% outside the 0900-1200 underestimatesthe overall density of this secretive H period (Table 3). There is >85% chanceof en- species. counteronly for 2 speciesand during the midday The Red-throatedCaracara, accordingto our re- period. At the other extreme, 5 of 14 specieshave sults,is the mostabundant raptor. On the studyarea been recordedin < 15% of any hour. the Caracara lives in flocks of 4-7 birds. Mean num- If one takesinto accountthe actualtime spentin ber of individuals recordedper flock encounteron flight, 11 of 14 speciesare visible< 5%of the daylight the -<200 m wide strip transect (3.43 + 0.91) is period whereas the 2 most conspicuous(Cathartes lower than actual size of groups occurring on the and Ictinia) were seenan overall 14-16% of the time. studyarea (5.50 + 0.90, N = 12). Yet overalldensity From mid-morningto mid-afternoon(peak of activ- obtained by the strip censusmethod is markedly ity), the last 2 specieswere visible during 20-25% higher than that derivedby plottingthe flock home of the time against <10% for all others and even ranges(Table 1 and Fig. 2). Frequencyand length <1% for 5 species. of the Caracaras' movements violate the basic as- It is not known how the time budget of these

Figure 2. Distribution of groupsof Red-throatedCaracaras (Daptrius americanus) over the 100 km2 study area Dashed lines:see Figure 3. Open limits: uncertainboundaries due to lack of recordsbecause of an insufficient survey in marginal areas. AUTUMN 1989 CENSUSING RAPTORS IN RAIN FOREST 79

II ! I

I

I

I '• •

I I I I 80 JEAN-MARCTHIOLLAY VOL. 23, NO. 3

Table 1. Comparativeresults of quantitativesurveys of territorial raptorsin primary rain forestsof French Guiana Zone I: intensivesoaring bird censusand understorymapping on 6 km2 (seetext). Zone II: mappingof soaringbirds on 42 km2. Total area:minimum population estimate, combining both methods, on the 100 km2 study area. Data are numbersof individuals(i.e., •- 2 times the number of pairs). Strip censuses' N = number of birds recorded within 100 m on either side of cumulative line transects (517 km for the main studyarea and 1188 km for otherlocalities). d = detectiondistance (m) maximizingthe densityestimate 1• (in numberof individuals/100km2).

MAIN STUDY SITE 8 OTHER LOCALITIES

STRIP CENSUS STRIP CENSUS SPECIES a ZONE ZONE TOTAL I II AREA N D •) N D I•)

Leptodoncayanensis 0 0 >2 1 30 3.2 1 25 1.7 Harpagusdiodon -<1 +2 >2 1 25 3.9 2 25 3.4 Harpagusbidentatus -<2 +6 >8 3 20 14.5 9 25 15.1 Accipiterbicolor <2 +4 >6 I 25 3.9 2 20 4.2 Leucopternismelanops -<2 -- >2 I 25 3.9 8 30 11.2 Leucopternisalbicollis + 3 + 10 15 5 50 9.7 10 35 12.0 Buteogallusurubitinga +3 -<12 -<20 10 45 17.2 9 40 9.5 Morphnusguianensis -<1 2 •_2 2 40 4.8 3 35 3.6 5)Oizasturrnelanoleucus <2 <4 •-6 3 30 6.4 5 30 7.0 5•izaetusornatus <2 4 -•10 9 45 12.9 15 50 10.1 Micrastursemitorquatus <2 -- >10 3 25 11.6 10 25 13.4 Micrastur ruficollis -<2 -- >8 2 20 9.7 3 20 6.3 Micrasturgdvicollis >4 -- >30 22 20 72.5 31 20 63.0 Micrastur mirandollei <2 -- >6 1 25 3.9 4 25 5.0 Daptriusamericanus _ 6 -- •- 66 206 100 199.2 395 100 166.2 Additionalspecies: Accipiter superciliosus, A.poliogaster, andS•pzzaetus tyrannus (one pair of each recorded on the main study area) and Itarpia harpyja(3 records,only outsidethe main studyarea, D = 3.1).

Table 2. Populationestimate of vulturesand kiteshunting in flight abovethe canopy.Number of individualscrossing a 10 km2 samplearea (meanof 20 hr between0900 and 1200 H); numberof territorialpairs or families settledwithin the 100 km2 studyarea; highest number of birdsseen together; total population(i.e., resident adults + fledgedyoung + estimatedadditional birds).

MEAN NUMBER OF NUMBER OF HIGHEST AVERAGE BIRDS/HR/10KM 2 RESIDENTPAIRS CONCENTRATION POPULATION _+S.D. LOCATED/100KM 2 RECORDED (IND/100 KM2)

Cathartes melambrotus 1.95 + 1.50 • 12 19 Sarcoramphuspapa 0.90 _+0.96 2 7 9 Elanoidesforficatus 0.60 + 1.09 2 30 >-10 Ictinia plumbea 1.90 + 0.97 3 7 9

Figure 3. Distribution of territorial pairs of the Great Black Hawk (Buteogallusurubitinga) over the 100 km2 study area of primaryrain forest.Home rangesshown here are mostlyareas covered by displayingadults. Dashed lines are provisionallimits basedon too small a numberof recordsor on observationsthat couldnot be accuratelylocated. AUTUMN 1989 CENSUSING RAPTORS IN RAIN FOREST 81

%-%

I /

,.%

I I I

.e,t

o /"..% 82 JEAN-MARCTHIOLLAY VOL. 23, No. 3

Table 3. Occurrenceof soaringraptors over a 10 km2-areaof primary forest.Proportion of time 1 or morebirds spentin flight, then probabilityto recordthe speciesduring 1 hr and numberof birdscounted within each hour. Observationperiods: A = 0630-0900 H; B = 0900-1130 H; G -- 1130-1530 H; D = 1530-1800 H Sample sizes(hours): A = 17; B = 20; G = 17; D -- 17.

PERCENT OF OBSERVATION PERCENT OF 1-HR TIME WHERE >--1 BIRD PERIODS WHERE THE MINIMUM NUMBER OF WAS FLYING OVER SPECIES WAS INDIVIDUALSSEEN/HR THE FOREST RECORDED (RANGE) SPECIES A B C D A B C D A B C D

Cathartes melambrotus 1.3 27.9 22.1 5.1 18 85 100 47 1-2 1-6 1-6 1-5 Sarcoramphuspapa 0.7 12.3 4.5 2.1 6 55 24 30 1 1-3 1-2 1-2 Elanoidesforficatus 5.9 5.5 6.5 10.6 24 30 30 30 1-7 1-3 1-15 1-17 Ilarpagus diodon -- 0.7 0.9 -- -- 5 6 Harpagusbidentatus 1.0 3.3 -- -- 12 40 -- -- 1 1-3 -- -- Ictinia plumbea 9.5 22.8 17.1 17.2 47 100 77 71 1-4 1-4 1-3 1-7 Accipitersuperciliosus -- 0.3 ------5 -- Accipiterbicolor -- 0.9 ------10 -- Leucopternisalbicollis 0.1 7.7 0.2 -- 6 45 6 -- 1 1-4 1 -- Buteogallusurubitinga 0.7 9.2 4.5 0.1 41 75 47 6 1-2 1-4 1-2 1 Morphnusguianensis -- 1.3 ------15 -- Spizasturmelanoleucus 0.4 0.8 -- 1.8 12 10 -- 12 1 1 -- 1 Spizaetusornatus -- 3.6 0.6 -- -- 25 6 -- -- 1-2 1 -- Falco deiroleucus 1.3 4.3 1.6 5.1 30 50 24 24 1-2 1-2 1-2 1-2

species,and hencetheir detectability,may vary over determinethe mostappropriate census methods (Ta- the seasons.Obviously, long periodsof time with ble 4). good observationaland weather conditionsare nec- Mapping of soaring birds was the most accurate essaryto assessthe presenceof somespecies, let alone technique. Territorial speciesregularly displayed their spatialdistribution. However, severalvery ter- abovethe canopyand easily provideda fairly con- ritorial species(H. biclentatus,B. urubitinga, the 2 venient set of data within a rather short time. For Spizaetusand, to a lesser extent, L. albicollisand many pairs, almostno additional informationwas Spizastur)performed their displayflight oncenearly accumulated after 2-3 d of fine weather. However, everyday, mostlyin late morningwhen soaringcon- most territory sizesmay be very conservativeesti- ditionswere suitable.Displays often lasted 5-15 min mates becauseof potential biases:1) some species but provedto be a fairly reliable indicationof an (e.g., Accipiter)perform relatively short flightsand occupied territory. Thus, concentratingmost re- may not displayover their entire homerange; 2) it searchduring the favorable hours may savemuch is not known whether the area flown coincides with time and, for suchspecies, most information may be defendedterritory and/or hunting range; 3) a few obtained within only 1-2 d. Conversely,to assess exceptionalflight circuits of individuals not surely confidentlythe presenceor absenceof the mostse- identified, or long pursuits of a neighbor, were ig- cretive species(H. cliodonand Accipitersp.), which nored;4) the exactposition of birdstoo far away was are more occasionallysoaring, it is necessaryto re- not precise and always conservativelyestimated. peat the aboveobservation from a vantagepoint for There was a substantial floating population, either at least6-7 d. (Table 3). Indeed,a completemapping transient birds or members of temporary or per- of their territory may require much more time. manent trios. Conversely,an accuratemapping of non-soaring DISCUSSION speciesis difficultwithout the help of time consuming Reliability of results rests with the detectability radio-tracking. First of all, unmarked birds cannot of eachspecies. Therefore, specificbehavioral traits be assignedto a particular pair or territory. More AUTUMN 1989 CENSUSING RAPTORS IN RAIN FOREST 83

Table 4. Behaviorand mostsuitable census methods of forestraptors in French Guiana. MS = mapping movements of soaringterritorial pairs;TP = minimum estimateof total population.Extrapolation of the mean number of individuals soaringover a given area; MU = mapping location and movementsof birds seenor heard in the understory;DE: mean density estimate from understorystrip census.

METHODS

BEHAVIOR SPECIES M S T P MU D E

Soaring--regular: Mid-late morning: loud calls rarely heard Buteogallus,Spizaetus + + + Leucopternisalbicollis + Harpagus bidentatus + + + Mid-morning to mid-afternoon Cathartes melambrotus Sarcoramphuspapa + Morning to evening Llanoidesforficatus + Ictinia plumbea + Falco (2 sp) + Soaring--Occasional: Mid-late morning Leptodon,Morphnus + + + Morning and afternoon Accipiter,tt. diodon, + Spizastur + + + Non-soaring--silent: Sometimesconspicuous Harpia harpyja + + secretive(undergrowth) Leucopternismelanops + + Non-soaring--calling: Early morning Micrastur (4 sp) + All day Daptrius (2 sp) +

importantly, frequencyof encounterswith secretive ing in March 1989 in the study area where it was and/or rare speciesis extremelylow. Eventhe Lined never seen before. ForestFalcon, by far our mostcommon solitary rap- The strip censusmethod, or any techniqueadapt- tor, was sightedat bestonce every 2 or 3 d, whereas ed from it to the conditionsof the rain forest, remains mostother specieswere spottedas a mean only once to be encouraged,provided that underlyingassump- every13-56 full d (> 10 hr/d) spentactively search- tionsare carefully respected.Good resultshave been ing the forest.Only speciesoften flying low over the obtainedwith large terrestrial birds [the Trumpeter canopy (kites, vultures, Buteogallus,Leucopternis, (Psophiacrepitans ) and the Curassow( Craxalector) ]. Spizaetus)were seenonce every 3-10 d. I met with Each lives in small flocks, slowly moving in the 1-3 groupsof Caracarasnearly eachday along a 10 undergrowth and always giving a soft alarm call or more km line transect. A low rate of raptor without fleeing, allowing the observerto detecttheir encountersseems to be more a consequenceof very presence. low density,uneven distribution or too confidinga The overall consequenceof possiblebiases is an behaviorrather than of shyness,since some perched underestimationof the density of most species,es- birds allowed a surprisingly closeapproach or did peciallywhen largeareas (> 10 km2) and smallspecies not take flight when they were in the upper canopy. are involved. The density estimateof the 100 km2 The possibilitythat some speciesor pairs may be area is lower than that of the intensivelysurveyed overlookedis suggestedby the followingexamples. 6 km2 core area for 11 of 14 species(Table 1). Accipiterpoliogaster was identifiedhere for the first However, the greatest care must be taken in ex- time in French Guiana eventhough 10 other similar trapolating density of a small area to a larger area, areashad beencarefully surveyed in previousyears. becausemany speciesare patchily distributed and A pair of Chondrohieraxuncinatus was found breed- territoriesare far from being contiguous.An overall 84 JEAN-MARCTHIOLLAY VOL. 23, NO. 3 densityestimate is meaningfulonly if drawn from LITERATURE CITED an area including >2 pairs. Such an area is neces- BRENNAN,L. A. AND W. M. BLOCK. 1986. Line transect sarily large for most speciesand difficult to survey. estimatesof Mountain Quail density.J. Wildl. Manage 50:373-377. BURNHAM, K. P., D. R. ANDERSONAND J. L. LAAKE CONCLUSION 1980. Estimation of densityfrom line transectsam- pling of biologicalpopulations. Wildl. Monogr. 72. 202 To obtain a reliable densityestimate of breeding pp. raptors in a primary rain forest, the best strategy 1985. Efficiencyand biasin strip and line tran- shouldbe the designof a particular censusmethod sectsampling. J. Wildl. Manage.49:1012-1018. for almost each speciesseparately (i.e., a specific COCHRAN,W. G. 1977. Sampling techniques.Wiley, New York. mixture of severaltechniques complementing each EMLEN,J. T. 1971. Populationdensity of birds derived other).Moreover, following radio-tagged birds should from transect counts. Auk 88:323-342. be stronglyencouraged and is currentlythe only way FORD,R. G. ANDJ.P. MYERS. 1981. An evaluationand to confirm the actual territory size of most species. comparisonof techniquesfor estimatinghome range When the aim is only to comparethe raptor pop- and territory size. Cooper Orn. Soc. Studiesin Arian ulation of differentforest areas, it may be advisable Biol. 6:461-465. to use a specificabundance index. Such an index FULLER,M. R. ANDJ. A. MOSHER. 1981. Methods of may be the highestfrequency (e.g., mean number detectingand counting raptors: a review. Pages 235- of individualsrecorded/hour) given by the mostre- 246. In C. J. Ralph and J. M. Scott,EDS. Estimating wardingmethod for the speciesinvolved. Indeed, any numbersof terrestrial birds. Cooper Orn. Soc.Studies in Arian Biol. 6. index is comparablewithin but not betweenspecies. The sametechnique should be usedfor a givenspecies GEISSLER,P. H. AND M. R. FULLER. 1986. Estimation of the proportion of an area occupiedby an animal over different study sites(including season,time of species.7986 Proc. Sectionon SurveyResearch methods, day and sightingradius). So, the method would avoid Amer. Statis. Ass. 533-538. the biasesof densityestimates coming from different HAVERSCHMIDT, F. 1968. Birds of Suriman. Oliver & specificdetectabilities and the more difficult assess- Boyd, Edinburgh. ment of actual density. JENNRICH,R. I. ANDF. B. TURNER. 1969. Measurement Time is more accuratelyand easilymeasured than of non circular home range. J. Theor.Biol. 66:227-237 area or distance and is a more constant unit. I have NEWTON,I. 1979. Populationecology of raptors. T & found here better correlations between number of D. Poyser,Calton. birds detectedand time spent than with distance RAMSEY,F. L. ANDJ. M. SCOTT. 1981. Analysisof bird surveydata using a modificationof Emlen's method travelled(all specificr = 068-0.92; P < 0.01), either Studies Arian Biol. 6:483-487. inside or outside the forest. SEBER,G. A. F. 1982. The estimation of animal abun- Resultsare encouragingbut we are far from mas- danceand relatedparameters. MacMillan, New York tering reasonablyeasy and accuratecensus methods SOUTHWOOD,T.R. 1966. Ecologicalmethods. Methuen, appropriatefor mostraptor specieswithin largeareas London. of denserain forest. Always necessarywill be the THIOLLAY,J. M. 1985a. Falconiformsof tropical rain use of severalcensusing methodologies in conjunc- forest:a review. Pages155-165. In I. Newton and R tion, including audio-luring, in order to minimize D. Chancellor EDS. Conservation studies on raptors, biases inherent in all. To work out and test such vol. 5. ICBP Tech.Publication, Cambridge. methodologiesremains an urgent challengewith re- 1985b. Raptor communitystructure of a pri- gard to rapidly decliningtropical forestraptors. mary rain forestin FrenchGuiana and effectof human hunting pressure.Raptor Research18:117-122. 1989. Area requirements for the conservation of rain forestraptors and gamebirds in FrenchGuiana. ACKNOWLEDGMENTS Conservation Biol. 3:128-137. VERNER, J. 1985. Assessmentof counting techniques. I extendmy gratitudeto J. L. Dujardin and O. Tostain Curt. Ornithol. 2:247-302. who have beenvery helpful in the field, as well as to F. Thiollay, who alsotyped the manuscript,and to P. Char- Laboratoire d'Ecologie,E.N.S., 46, rue d'Ulm, 75230 lesdominique,director of the LaboratoireECOTROP and Paris, Cedex 05, FRANCE. of the field station.This studywas supportedby a grant from the French Ministry of Environment(SRETIE). Received6 June 1988; accepted15 July 1989