J. Field Ornithol., 69(2):201-208

A TECHNIQUE FOR SAMPLING FLYING D^VIDJ. FLAS•OHI.ER Departmentof Wildlife Ecology A229 Russell Labs Universityof Wisconsin-Madison Madison, Wisconsin 53706 USA

Abstract.--I describea procedure for samplingflying insects.Using binoculars,a stopwatch, and a hand-held counter, an observer counts insectspassing through a measurablefocal volume for a set time. No identification skills are needed. I tested the accuracyand repeatabilityof the procedure under controlled conditionsand found that with known lim- itations, it is a reliable way to sample flying insect abundance. I used the procedure to describe daily activity pattern of flying insectsusing a clearing adjacent to a neotropical lowlandforest reserve.While flycatcherand flying insectactivity patterns were not strongly correlated, similaritiesin activitywere noted.

TP•CNICA PARA MUESTREAR INSECTOS VOLADORES Sinopsis.--Sedescribe un procedimientopara muestrearinsectos voladores. Utilizando bin- oculares,un cron0metroy contadormanual, un observadorpuede contar los insectosque pasana travis de un volumen focal medible pot un periodo de tiempo. Para utilizar esta t•cnica no es necesariotenet destrezaspara identificar los insectos.Puse a prueba la exac- titud y repetividaddel procedimientobajo condicionescontroladas y encontr• que, aunque con limitaciones,es un m•todo confiable para medir la abundanciade insectos.Utilic• el m•todo para describir los patrones de actividaddiaria de insectosvoladores utilizando un claro adyacentea una reservaforestal de un bosqueneotropical bajo. Aunque el halconeo (flycatching)y el patr6n de actividadde insectosvoladores no se correlacionaron,se notaron similaridadesen el patr6n de actividad. Foraging ecology and habitat partitioning have been well studied in Tyrant flycatchers (Davies 1977, Fitzpatrick 1980, 1981, Hespenheide 1971, Robinsonand Holmes 1982, Sherry 1984, Traylor and Fitzpatrick 1982). Many of thesestudies documented diet compositionand foraging behavior, but few have quantified resource availability.Similar research on dragonfly foraging behavior also does not addressthe relationship between prey availability and predator activity (Corbet 1963; Higashi 1973, 1978; May 1980). One reasonfor this is the lack of a suitablesam- pling method for quantifyingaerial insectabundance. Therefore, the goal of this studyis to developa new method to quantify daily fluctuationsin flying insect numbers.

METHODS Samplingprocedure.--The sampling procedure requires a pair of bin- oculars,a hand-held counter,and an alarm stopwatch.The observerlays againsta horizontal surface(such as the ground) for a set period of time, looks up at the sky (whether cloudy or clear) through binoculars,and counts insectswith the hand-held counter. Insects are counted as they pass though the focal volume visible through the binoculars (Fig. la). Observer concentrationand hence visual acuity may decline if the sam- pling period exceedsapproximately 3 min. With the binocularsset to a

201 202] D. Flaspohler .1.Field Ornithol. Spring 1998

i 4O

[] Black beans 3O ITo•al objects

i 18 i i

14.0 rn • 14

ø,9.. 10

..u 6..5

3.4 m 2 • I • I • I

I / 0.5 •,1

Observer 0 20 40 60 80 100

% of objects seen

A B FIGURE1. Graphic representationof insectsampling method (a). Observerdetection rates will vary in accordancewith a detectionfunction within the focal volume.The detection function (b) wascalculated for total objectsand blackbeans (total objects:n = 1750, black beans:n = 350). Each point on the y-axisrepresents the mean distancethrough which a set of objectswere thrown (note that the spacingof mean distancesis roughly regular until the last two samplingdistances). Each point on the x-axis representsthe percent of objectsseen by an observer. Vol.69, No. 2 InsectSampling [203 known focal distance, all insectsdetected during the count period are tabulated on the counter and when the alarm sounds,counting stops. Because the magnification and focal distance of the binoculars used will determine the minimum size of insects seen, one must establish the range of distanceswithin which small objects,such as insects,are visible. This is done by choosingan object approximatelythe sizeof the smallest insectsone wishesto count. In this study,I wasinterested in knowingthe abundanceof the flying insectsmost likely to be preyed upon by medium- size flycatchingbirds. Therefore, when calibratingmy binoculars,I chose a 0.15 cm'• object to represent the smallestprey size normally captured by the flycatchersincluded in this study. I attached a 0.15 cm• circle of black paper to a pale vertical surface (such as an exterior white wall) and set my binoculars at their nearest focus. Starting close to the wall, I backed awaywhile watching the area on the wall until the black circle first came into view, at which point it wasblurry but clearlyvisible. The distancefrom the binocularsto the wall is the minimum distance at which insects will be detected. When deter- mining minimum and maximum distancesof visibility,it is useful to sim- ulate the motion of the insectsacross the field of view by moving the binoculars slightly from side to side. By doing this, one can accurately determine the distanceat which a moving object of a given size can be seen. I continued to back awayfrom the wall until the black circle came into sharp focus. As I continued to back away,the black circle remained in sharpfocus for a measurabledistance, then blurred and disappeared.The point at which the blurry black circle disappears is the maximum sam- pling distance.The minimum and maximum samplingdistances are rep- resentedby circular planes betweenwhich a truncated cone is formed. This cone representsthe samplingvolume (Fig. la), and is specificto the size object/insect you are observing. Insects of a given size passing through this volume will be seen and counted while those passingabove and below it will not be detected. Detectabilitywithin this volume varies as a function of distance from the observer and size of insect. The volume will vary with binocular magnification,aperture opening,focal adjustment and insect size. I restricted sampling to the 12 h period between 0600-1800 h which roughly correspondedto sunriseto sunset.While the method wasbeing used and later tested,conditions were mostlyclear with occasionalclouds passingacross the field of view. Sky conditionsdid not seem to alter my ability to see and count flying insectsbecause I could easilyfollow insects acrossa background of mixed cloudsand sky.Occasionally, a would fly through the field of vision either within the focal volume or far above. The motion of the bird's wings made it easyto differentiate it from an insect.Hummingbirds may be counted as insectsas they passthrough the observer'sview. A period of careful observationof hummingbird activity in the sampling area prior to samplingwould allow one to estimatethe potential for this type of sampling error. 204] D. Flaspohler j. FieldOrnithol. Spring 1998

In order to test the accuracyand repeatability of this sampling tech- nique, I simulated the technique using a known number of objects of known sizes.Objects were chosen to reflect the range of insect sizesthat might be encountered. The five objects were, in ascending order with approximatesurface area of their broadestprofile: brown rice (0.15 cm'•), lentils (0.25 cm2), black beans (0.5 cm2), red beans (1.25 cmS), and al- mond meats (2.25 cm'•). A single thrower threw 50 of each size object acrossthe field of view while the observer counted the objectsand noti- fied the thrower who tallied the objectsseen. Thus, 250 objects (50 of each of the five sizes) were thrown across seven zones, each a different distancesfrom the observer.The thrower randomized the objectsso that the observercould not anticipate the order in which objectswere thrown. Becausered beansand almond meatswere visibleat distanceswell beyond the other objects,I determined the distanceat which they cannot be seen by having the thrower send one through the field of view at increasing distancesuntil I could no longer see them. Care wastaken to ensure that all objectswere thrown at a relativelyconstant speed and that they passed through the viewingarea. Comparisonsof the speedof the thrown objects with that of insectspassing through the samplingarea showedthat they were similar. In order to generate a detection function for objectsseen, the number of each size object was converted to a percent of the total thrown. Zones of varying distancesfrom the observerwere chosenand the range within each zone was then averagedto give a mean distancefrom observerfor that zone. Insectactivity and flycatcherforaging.--I used the insectsampling meth- od describedabove to test the hypothesisthat aerial insect abundanceis correlated with flycatcher foraging activity throughout the day. I con- ducted this research at the La Selva Biological Reserve (10ø26'N, 83ø59'W), Costa Rica. Data were collected in a clearing (0.2 ha) along the entry road to the Reserve.During the three daysof data collection weather conditions were consistent,with clear to partly cloudy skies,no rain and wind speedszero to light. On 10-11 Mar. 1993, I measured the abundance of flying insects throughout the day at half-hour intervals at two points, one in the clear- ing and one at the edge adjacent to the clearing. I counted all insectsseen passing through a focalvolume of 28.7 m3 for a 3-minperiod every half hour from 0600-1800 h. For statisticalpurposes, I used the mean number of insectscounted at these two points as a measureof the relative abundanceof flying insects. On 9 Mar. 1993, total salliesof four flycatcherspecies (Tropical , Contopuscinereus; Gray-capped Flycatcher, Myiozetetes granadensis; Social Flycatcher, M. similis; and Tropical Kingbird, Tyrannus melancholicus) were counted continuouslyfrom 0600-1800 h and grouped into 23 30- min periods.I used correlationanalysis in SYSTATversion 5.01 (Wilkin- son 1990) to assesswhether the daily pattern of flying insectabundance Vol.69, No. '• InsectSampling [205 was correlated with the daily pattern of sallyingactivity for four flycatcher .

RESULTS AND DISCUSSION

Using size classesof a known number of objectsgenerated a detection function (Fig. lb). The shapeof the detectionfunction demonstrateshow distancefrom observer affects object detectability,with highest detect- ability for all size objectsat the center of the zone of sharp focus (ap- proximately5 m). Object sizealso affected the shapeof the function. The shapeof the detectionfunction for black beans,lentils, and rice wassim- ilar. Large objects (red beans and almonds), also had similarlyshaped functionswhich resembledthat for total objects;red beansand almonds accountedfor all observationsbeyond 20 m. Large objects/insectscan be seenwell beyond the zone of sharp focuswhile small objects/insectsrap- idly become invisibleto the observeras one movesaway from this zone. Becausesmall insectsare often more common than large insects,this characteristicof the samplingtechnique may prove beneficial by gener- ating adequate sample sizesof rarer large insectsand minimizing count- ing difficulty of abundant small insects.Because the focal volume varies with insectsize, data generatedusing this method shouldnot be used to comparethe relativeabundance of insectsof greatlydifferent sizes.How- ever,this technique doesprovide a way to estimateabundance of a given size class of insect across time and between locations. To explore how binocular magnification,aperture diameter and optical quality influence focal distance,! tested two binocularsof different mag- nifications: 10 x 50 and 7 x 42. With the 10 x 50 binocular set on the nearestpossible focus, ! was able to see a stationary0.15 cm2 black object betweenthe distancesof 3.4-14 m, with sharp focusbetween 5.9-9.5 m. This settinggave me a conically-shapedfocal volume of 28.7 ms (Fig. la). Using the 7 x 42 binocularson their nearestfocus setting, ! wasable to see the sameobject betweenthe distancesof 2.1-10.0 m with sharpfocus between 3.2-5.8 m. The focal volume of the 7 x 42 binoculars was 13.9 ms. This method can be usedto calculaterelative abundance. To generate an estimateof absolutedensity one would need to combine the detection function with the physicalcharacteristics of the detection volume. Becauselarge, distant insectscan be mistakenfor small near insects, use of this method for size-classspecific studies is not recommended. Duration of the sampling period should be adjusted relative to insect abundance.At low insect density,the period may need to be longer to reduce sampling error. One can choose an optimal period by plotting the coefficientof variation in abundanceestimates for a variety of sam- pling periods and choosinga period where variancedeclines or stabilizes. The data collectedvia the samplingprocedure describea common pat- tern of daily activityfor flying insects(D. K. Young,pers. comm.). Abun- dance of flying insectsdid not showa correlation with total sallyingfre- quencyfor the four speciesstudied (r -- 0.34, df = 1, P -- 0.11) (Fig. 2). 206] D. Flaspohler j. FieldOrnithol. Spring 1998

30

o o

o o o

20

o o

A

10 ø øoo

o

1 I I 0600 1200 18oo

Hour of day

50 i i

B

o o 10

I 0 o o o 0600 200 1800

Hour ofday FIGURE2. Abundanceof flyinginsects (a) and totalsallies of four speciesof (b) asa functionof time of day.Insect abundance is the mean of two estimatestaken from differentpoints 7 m apart.Bird acti•Styrepresents the numberof sallyattempts in a 30- min periodprior to the graphedpoint. I usedrobust locally weighted regression (lowess) (Gleveland1985) to generatea line through the data. Vol.69, No. '• InsectSampling [207

However,a generalpattern of greater activityin the morning and evening for both insectsand flycatcherswas apparent. Fitzpatrick(1981) reported that neotropicalflycatchers studied in , Venezuela, and displayed morning and late afternoon activity peaks.In this study,flying insectsshowed peak abundanceearly in the morning and late in the afternoon. The four flycatchers studied also showeda pattern of higher activityin the morning and again in the af- ternoon. While insect activitypatterns were not strongly correlated with flycatchersallying activity, increases in sallyingfrequency in the afternoon may be related to increased activity of flying insects.Data on insect and flycatcheractivity were collected on different consecutivedays. However, becauseweather conditionsduring the three daysof data collection in CostaRica were virtually identical I believe that activitypatterns for both birds and insectswere similar for the entire samplingperiod. Flycatcher foraging activityis influenced by a number of factors including hunger, preference for favored insectprey, weather, and seasonalbreeding events, and some of these factors may exhibit a stronger influence on daily ac- tivity patterns than insectabundance alone. The daily activitypatterns of bird, mammal, and insectpredators that forage for flying insectprey are influenced by the availabilityof flying insects.The activityand, therefore, availabilityof flying insectsis affected by many factorssuch as ambient temperatures,wind, and precipitation. The sampling procedure presented here can be used to determine daily activitypatterns of flying insectsof specificsize ranges and at specific distancesabove the ground. This technique has applicationsfor ornitho- logical, entomological, and mammalian (bat) research. My method of indexing flying insectabundance provides a meansto further understand the relationshipbetween aerial foraging speciesand their prey.

ACKNOWLEDGMENTS

I thank J. Blake and N. Greig for assistancein the studydesign and data analysis,and S. Moegenburgfor field support in CostaRica. J. Flaspohlerand C. VanderVeenassisted with data collectionin Wisconsin.Useful commentson an earlier draft of the manuscriptwere providedby P. Arcese,R. Flaspohler,S. McWilliams,L. Payne,S. Temple, and D. Young.J. Cary provided insightful assistancewith data analysisand manuscriptorganization. R. Chan- dler, M. Woodrey,and an anonymousreviewer provided comments which greatlyimproved the manuscript.I am grateful to the Organizationfor Tropical Studies,the Univ. of Wisconsin Departmentof Wildlife Ecology,the Programfor ConservationBiology and SustainableDe- velopment, and the Max McGraw Wildlife Foundation for funding, and to the studentson OTS #93-1 for many illuminatingdiscussions.

LITERATURE CITED

CLEVELAND,W. S. 1985. The elementsof graphing data. Wadsworth,Inc., Monterey,Califor- nia. CORBET,P.S. 1963. A biologyof dragonflies.Quadrangle Books. Chicago, Illinois. D^WES,N. B. 1977. Prey selectionand the searchstrategy of the SpottedFlycatcher (Mus- cicapastriata): a field studyon optimal foraging. Pmim. Behav.25:1016-33. FITZPATRICK,J. W. 1980.Foraging behavior of neotropicaltyrant flycatchers. Condor 82:43-57. --. 1981. Search strategiesof tyrant flycatchers.Anim. Behav. 29:810-821. 208] D. Flaspohlo' J.Field Ornithol. Spring 1998

HESPENHEIDE,H. t. 1971. Food preferenceand the extent of overlapin someinsectivorous birds,with specialreference to the Tyrannidae.Ibis 113:59-72. HIGASHI,K. 1973. Estimationof the food consumptionfor somespecies of dragonflies.I. Estimationby observationfor the frequencyof feedingflights of dragonflies.Rep. Ebino Biol. Lab. 1:119-129. 1978. Daily food consumptionof Sympetrumfrequens selys ( Odonata:Libellulidae). JIl•pSynthesis 21. M•Y, M. L. 1980. Temporal activitypatterns of Micrathyria in Central America (Anisoptera: Libellulidae).Odonatologica 9:57-74. ROBINSON,S. K., ANDR. T. HOLMES.1982. Foragingbehavior of forestbirds: the relationships among searchtactics, diet, and habitat structure.Ecology 63:1918-1931. SHERRY,T. 1984. Comparativedietary ecology of sympatric,insectivorous neotropical flycatch- ers (Tyrannidae).Ecol. Monogr. 54:313-338. TRAYLOR,M. A., JR., ANDJ. W. FITZPATRICK.1982. A surveyof the tyrant flycatchers.Living Bird 19:7-50. WILtoNSON,L. 1990. SYSTAT:the systemfor statistics.SYSTAT, Evanston, Illinois. Received10 Oct. 1996;accepted 15 May 1997.