New Techniques for Time-Activity Studies of Avian Flocks in View-Restricted Habitats

j. Field Ornithol., 60(3):388-396

NEW TECHNIQUES FOR TIME-ACTIVITY STUDIES OF AVIAN FLOCKS IN VIEW-RESTRICTED HABITATS

MICHAEL P. LOSITO,• RALPH E. MIRARCHI, AND GUY A. BALDASSARRE 1 Departmentof Zoologyand WildlifeScience AlabamaAgricultural Experiment Station Auburn University,Alabama 36849-5414 USA

Abstract.--Focal-animal sampling was comparedto a newly developedtechnique, focal- switchsampling, to evaluatetime budgetsof Mourning Doves(Zenaida macroura) in habitats where the field of view was restricted.Focal-switch sampling included a formula employed to weighthabitat use and testfor restrictionsof habitatstructure on behavior,and a standard waiting period to decidewhen to end sampling or continue pursuit of lost flocks. Focal- animal samplingbiased estimates of activebehavior downward, whereas estimates of inactive behaviorwere similar for both methods.Focal-switch sampling increased research efficiency by 12%. The standardwaiting period saved24% of samplesfrom prematuretermination and reducedobserver bias by maintaining equal sampling effort per flock. Weighting of habitat use also reducedsampling bias. Focal-switch sampling is recommendedfor use in conjunctionwith focal-animalsampling when samplingin view-restrictedhabitats.

NUEVAS TP•CNICASPARA EL ESTUDIO DE ACTIVIDADES DE CONGREGACIONES DE AVES EN HABITATS CON VISIBILIDAD RESTRINGIDA Resumen.--Lat6cnica de observaci6ndirecta de animales(focal-animal) es comparada con un nuevom6todo en dondese evaluan los presupuestosde actividadesen Zenaidamacroura en habitatsen dondeel campode visi6nestaba restringido. La nuevat6cnica (cambio-focal) incluyeuna f6rmulapara "pesar"la utilizaci6nde habitaty medirlas restriccionesque imponenla estructuradel habitaten la conducta.Incluye ademfis,un periodode espera estandarizadoque permite al observadordecidir cuando terminar susobservaciones o conti- nuar las mismas.E1 muestreo de un soloanimal (focal-animal)tiene un sesgopara estimar la conductaactira, mientras que los estimados para conducta inactiva rueton similares para ambosm6todos. E1 nuevo m6todo increment6 la eficienciadel estudioen un 12%.E1 periodo de esperaestandarizado evit6 que se terminara prematuramente con el 24% de lasmuestras observadasy redujo el sesgoal mantenetel mismoesfuerzo de muestreopot agregaci6nde aves.Se recomienda la utilizaci6ndel nuevo m6todo en conjunci6n con el previamentedescrito cuandose muestreeen habitatdonde la visibilidadpueda set obstruida.

There are many effectiveand widely used methodsto study activity budgetsof avian flocks (e.g., Altmann 1974, Baldassarreet al. 1987). Most studieshave focusedon speciesthat congregatein open habitats (e.g., shorelineand open water) where large numbersof individualscan be sampledreadily (Maxson and Bernstein1984, Paulus 1984a, Quinlan and Baldassarre1984, Tamisier 1976). However, problemsof visibility biasarise in studiesof speciesin densehabitats (Bradley 1985), because the investigatoris confrontedwith samplingconstraints that often cause data variability and biases(Kessel 1976, Verner 1965).

• Currentaddress: Environmental and ForestBiology, College of EnvironmentalScience and Forestry,State University of New York,Syracuse, New York 13210.

388 Vol.60, No. 3 NewTime-Activity Methods [389

Focal-animalsampling is the observationof an individualfor a pre- determinedperiod of time (Altmann 1974). It may be the best method for detailedanalysis of behaviorand the only methodavailable to sample small numbersof individuals.Focal-animal samplingis thereforeappro- priate for studiesof speciesin densevegetation where the number of potentialsample individuals is frequentlyreduced by vegetation. A major, unresolvedproblem with focal-animalsampling occurs when focal individualsdisappear from view (hereafterreferred to as "out-of- view";see Lehner 1979). For example,Verner (1965) arbitrarilyrecorded Marsh Wren (Cistothoruspalustris) out-of-view data as feeding,whereas Maxson (1977) broadlymeasured Ruffed Grouse(Bonasa umbellus) activity by radiotelemetry.Other observers(e.g., Paulus 1984b, Rave 1987) had biasedresults because they were unableto focal-sampleflying birds. Suchlimitations may discouragetime-activity studies of flockingupland species.Accordingly, techniques must be developedto addressproblems associatedwith: (1) focal individuals disappearingfrom view; (2) the amount of time a focal individual should be observed;(3) the spatial distributionof flock individualsamong heterogeneoushabitats; and (4) the eliminationof observer-expectancybias when a sampleflock departs beforethe terminationof the sampleperiod. We reporta new technique,focal-switch sampling, developed to address thesesampling problems during an activity-budgetstudy of Mourning Dove (Zenaida macroura) flocks in northern Alabama. We also report someresults of the activity study by relating behavioralfrequencies to broad habitat categorieswhile employingboth focal-animal and focal- switch sampling.

METHODS Focal-switchsampling was testedagainst focal-animal sampling using radio-fitted hatching-year(HY) Mourning Doves to lead observersto dove flocks on the Swan Creek Wildlife Management Area, Wheeler National Wildlife Refuge, and surroundingareas in northern Alabama. Dove flockswere locatedfrom July to September1987. Radio-marked individualswere not observedfor behavioraldata to eliminate potential bias inducedby the radio transmitter.Flocks were sampledduring three observationperiods (OPi = sunriseto 3.5 h after sunrise,OP2 = 3.5 h after sunriseto 3.5 h before sunset,OP3 = 3.5 h before sunsetto sunset) to block-out variation causedby differing periods of Mourning Dove activity (Deuver and Fatora 1968). The closestreadily visible individual in the flock was selectedas the first bird to be sampled(Altmann 1974). Two observers(one per method) collecteddata on eachsample flock. Each methodwas usedby all four observersduring the study.The total data include30.5 h (OP1 -- 10 h, 10 min; OP2 = 10 h; OP3 = 10 h, 20 min). Behavioral categorieswere: feeding,locomotion, comfort, resting, alert, and out-of-view. Focal-switchsampling, unlike focal-animalsampling, allowed switching to a new focal individual (nearestneighbor) if the original focal 390] M. P. Lositoet al. J. Field Ornithol. Summer 1989

T^BLE 1. Demonstrationof weighting focal samplesaccording to habitat use (Y•) where nA = 50, nB = 25, nc = 25, N = 100, T = 60 min, and f= 5 min.a

Focal-unit (sample {s}; no sample {--}) Habitat 1 2 3 4 5 6 7 8 9 10 11 12

A s -- -- s -- -- s -- -- s s s B -- s -- -- s -- -- s .... C -- -- s -- -- s -- -- s ------

a y, = (n,/N) x (T/J) where: Y, = numberof focal-units(f) sampledin habitat i; n, = number of individuals in habitat z; N = total number of individuals in flock; T = duration of sampleperiod;f = duration focal-unit. Then YA= n,•/N x T/f = 50/100 x 60/5 = 6, YB=3, and Yc = 3. individual went out-of-view. To avoid observer-expectancybias (Balph and Romesburg1986), two consecutiveout-of-view recordingswere allowed before focal-switching.Focal-switching also was restrictedto individuals in the same habitat as the out-of-view individual. This controlled the effectsof habitat structureon behavior.For example, if a focal individual in habitat A moved to habitat B and was lost from view, focal- switchingwas allowed only on individualsin habitat B. The null hypothesistested was that the two methodsresulted in the collectionof the sameamount of data per behavioralcategory during a given unit time. We used Hotelling's multivariate t-tests (SAS Institute, Inc. 1985) to determinethe overall differencein resultsobtained by use of the two methods.Because the data were normally distributed,we used pairedt-tests (SAS Institute,Inc. 1985) to comparebehavioral frequencies betweenmethods for each observationperiod. A probability level of 0.05 was used for all statistical tests. The sampleperiod (T), the maximum time limit for samplinga flock, was dividedinto focal-units(f) where: T = 60 min, andf = 5 min. The focal-unitwas the predeterminedtime within which behaviorof the focal individual(s)was sampledinstantaneously every 20 s (Altmann 1974) as dictatedby a tape recorderused as a metronome(Paulus 1984a, 1984b). The number of completefocal-units (focal-switching not needed)versus the number of incompletefocal-units (focal-switchingnecessary) was testedfor equalityusing paired t-tests(SAS Institute, Inc. 1985). To beginsampling a flock,the observersestimated visually the number of individualsusing major habitat types (e.g., upland field, forestcanopy, willow thicket) and focal-unitswere distributedproportionately (Table 1). Habitat weightingwas testedby comparingactivities sampled in perch (e.g., forestcanopy + willow thicket canopy)versus non-perch habitats (e.g.,upland field + mudflat). Unpaired t-tests(SAS Institute,Inc. 1985) were usedto testthe null hypothesisthat the two habitat structuresaffected behaviorequally. When a sampleflock departed before termination of the sampleperiod, we allowed a standardperiod of 10 min for a new samplingsituation to Vol.60, No. 3 NewTime-Activity Methods [391

TABLE2. Number of instantaneousrecordings (mean + SE) of each Mourning Dove behaviorcollected using focal-animaland focal-switchmethods simultaneously during eachobservation period, northern Alabama, 1987.

Method b Observation Behavioral category perioda Focal-animal Focal-switch

Locomotion 1 4.5 + 1.1' 9.5 + 2.3* 2 2.6 + 1.3' 5.2 + 1.9' 3 6.9 _+ 1.4' 11.9 _+ 2.4* Feeding 1 21.0 + 5.0* 32.3 + 8.5* 2 1.4 + 0.5 4.5 + 2.3 3 38.6 + 8.0 49.0 ___11.0 Resting 1 13.8 + 3.1 12.3 + 2.2 2 30.1 ___6.4 30.3 + 6.4 3 18.8 + 2.4 21.4 ___ 6.7 Comfort 1 21.8 + 6.7 28.7 + 7.6 2 39.2 + 7.2 47.5 ___ 7.5 3 27.6 + 4.2 27.5 + 6.3 Alert 1 35.8 + 6.4 40.4 + 6.0 2 26.0 + 5.2 31.3 + 5.4 3 31.2 + 3.3 36.6 + 2.5 Out-of-view 1 55.0 + 8.4* 26.8 + 5.0* 2 49.7 + 9.4* 30.1 + 6.4* 3 45.9 + 4.6* 22.3 + 2.7*

Samplesize (no. of flocks)for OP1 = 12, OP2 = 14, and OP3 = 11. Asterisksindicate differences(P < 0.05) between methods. ariseonce all individualsin the sampleflock had left the area, had been sampled,or wereout-of-view. Once the standardwaiting period expired, a new flock was selectedfor sampling. The standardwaiting periodwas usedto avoid observer-expectancy bias(Balph and Romesburg 1986). For example,if a flockdeparts before the end of the sampleperiod, the investigator,upon decidingwhether to end samplingor to continuepursuing the flock,could induce bias by chasingflocks he knowsare easyto locateand ignoringthose difficult to locate(observer-expectancy bias). The standardwaiting periodprovided a constant for these decisions. We recorded the number of times the standardwaiting period was needed.

RESULTS The amount of behavioral data collectedby the two methodswas significantlydifferent. Focal-switch sampling resulted in an averageof 11.7% more data per sampleperiod than focal-animalsampling. More (P < 0.05) locomotiondata were collectedwith focal-switchsampling than focal-animalsampling during all observationperiods (Table 2). Feeding time wasgreater (P < 0.05) with focal-switchsampling than focal-animal samplingduring OP1, but not duringOPs 2 and 3 (P > 0.05; Table 2). There were no differences(P > 0.05) between methods during any 392] M. P. Lositoet al. j. Field Ornithol. Summer 1989

TABLE3. Comparisonof the number of completevs. incompletefocal-units (mean _ SE) collectedon Mourning Doves(methods pooled) during eachobservation period, northern Alabama, 1987.

Focal-unit (type)b Observationperiod a Complete Incomplete

1 4.1 + 0.5* 6.1 _+ 0.6* 2 4.7 _+ 0.6* 5.2 _+ 0.5* 3 4.2 _+ 0.4* 7.1 _+ 0.6*

Sample size (no. of flocks)for OP1 = 24, OP2 = 28, and OP3 = 22. Asterisks indicate differences (P < 0.05) between focal units. observationperiod for resting,comfort, or alert behavior(Table 2). Fre- quenciesof out-of-viewdata were greater (P < 0.05) with focal-animal samplingthan with focal-switchsampling during all observationperiods (Table 2). The numberof incompleteand completefocal-units collected per sample period did not differ (P > 0.05) betweenmethods. Overall, sample periodscontained more (P < 0.05) incompletefocal-units than complete focal-units(methods pooled) for all observationperiods (Table 3). We maintained observationson individual Mourning Doves for complete focal-units during 48% of the sample periods, so focal-switchingwas needed more often than not. Regardlessof methodemployed, feeding time was greater (P (0.05) in non-perch habitats than perch habitats (Table 4). Locomotionwas greater (P (0.05) in non-perchhabitats than perch habitats for focal- switchsampling, but wassimilar (P > 0.05) betweenhabitats using focal- animal sampling(Table 4). Restingand comfortbehavior were greater (P (0.05) in perchthan non-perchhabitats for both methods(Table 4). Alert behaviorwas similar (P • 0.05) betweenhabitat structuresfor both methods (Table 4). The standardwaiting period was usedin 21% of the samples(n -- 37) and savedsample periodsfrom premature termination 24% (n = 9) of the time by allowingsampling to continueon flocksthat occasionallyhad all individuals out-of-view.

DISCUSSION The major problem we encounteredwith focal-animalsampling was losingvisual contactwith the focal individual beforethe focal-unit ended (as per Lehner 1979). Thus, focal-animal sampling was inefficientbe- cause no data were collected from active birds in dense habitats. This problemwas alleviatedby focal-switchingto visiblebirds when the focal individual was lost from view. Both methodswere equally effectivein collectingdata on comfort, resting,and alert behaviorbecause these were generallystationary states. Individualsengaged in suchbehavior were lesslikely to disappearfrom Vol.60, No. 3 NewTime-Activity Methods [393

TABLE 4. Number of instantaneousrecordings (mean +_ SE) of each Mourning Dove behavior collectedusing focal-animal (FA) and focal-switch(FS) methodssimulta- neouslyin perch and non-perchhabitats, northern Alabama, 1987.

Habitata, • Behavioralcategory Method Perch Non-perch Feeding FA 1.9 +_ 1.3' 33.0 +_21.0' FS 1.6 +_ 0.8* 55.0 +_ 26.0* Locomotion FA 1.7 _+ 0.6 7.5 +_ 6.5 FS 2.4 _+ 0.7* 12.5 _+ 5.5* Resting FA 25.4 _+7.1' 0.5 _+ 0.5* FS 23.2 _+ 6.8* 5.5 _+ 0.5* Comfort FA 41.7 _+ 9.3* 3.5 +_ 3.5* FS 50.6 _+ 9.9* 3.0 +_ 2.0* Alert FA 33.1 _ 6.1 39.0 _+ 37.0 FS 36.1 +_ 7.1 33.5 ñ 25.5

Samplesize (no. of flocks)for perch habitat = 10, non-perch= 2. Asterisksindicate differences(P < 0.05) between habitats. view into vegetationcompared to active individuals so focal-switching often was not needed. The most commoncauses of out-of-view data at feeding times were visual obstructionscaused by groundvegetation (visibility bias), and the focalbird becoming"lost" in the groupby intermixingwith flockmem- bers.While recordingout-of-view data on an individual the investigator can losefeeding data becauseof the increasinglikelihood that the flock will terminate its feedingbout. For example,by the time the observer finishesa focal-unit on an out-of-view individual, the entire flock may be engagedin a different activity (e.g., resting),and thus overall feeding time is underestimated. Feedingdata also may be under-representedwith focal-animalsampling through samplingbias, even though the majority of the flock may have been foraging. For example, using focal-animal sampling, if two flock individualsperched in a tree were sampled,and each received15 records"resting" and 0 records"out-of-view" their sum "resting"would equal 30 (15 resting x 2 birds -- 30). During the same sample,if 10 otherbirds in the sameflock were sampledfrom a field and eachreceived 1 record"feeding" and 14 records"out-of-view," their sum "feeding" would equal 10 (1 feeding x 10 birds = 10). Therefore the flock'sfeeding percentagewould equal 25% (10 feeding/40total = 10 feeding/30resting + 10 feeding)and restingwould equal75% (30/40); the overallestimation of feedingwould be too low. Focal-switchsampling eliminates this problem becauseit permitsswitching to in-view feedingbirds. Like feeding,locomotion probably was underestimatedwith focal-animal samplingbecause flying individualswere lostalmost instantaneously. Focal-switchsampling permitted continuoussampling of flying individ- 394] M. P. Lositoet al. J.Field Ornithol. Summer 1989 uals becauseit allowed use of the sky as a habitat. For example, if 50 birds on the groundincluding the focal bird took flight and the focal bird was lost, other flying individuals in the flock were sampledby focal- switchinguntil the last onewent out-of-view. Radio-markedHY Mourn- ing Dovesusually departed areas in smallgroups and oftenmade sustained flights(>2 km) to their next area; hence,by the time the last focal-switch was made on a group leavingan area, the original focal bird probably was still in flight. Although it is apparent from the out-of-view data that focal-switch sampling can be an efficient means of studyinghighly active speciesin densevegetation, it shouldonly be usedif the physicalobstruction between the observer and focal animal has no influence on the focal animal's immediate behavior. For instance,focal-switch sampling could not be usedeffectively with bay ducks(Anatinae) becauseindividuals cannot be followedunderwater; hence, feeding probably would be underestimated in this situation(Baldassarre et al. 1987). Also, being aboveand below the water surfacemay be considereduse of two different habitats. Incompletefocal-units (/) were more commonthan completef'sin this study.Had ourf duration of 5 min been lower, more completef'swould have been collected.If the objectivesof a study warrant comparisons among individuals in a flock, the investigatorshould strive to collectas many completef'sas possible,because each represents an individual bird (i.e., the samplingunit). Comparisonsbetween flocks can be made by poolingcomplete and incompletef's becausethe samplingunit would be the flock. The duration of the focal-unit (/) shouldbe correlatedinversely with relative flock size to sample more birds when flock size is large, and to accountfor behavioralvariation amongflock members. Theoretically, the ultimate duration of the f would allow for sampling the entire flock in order to removeassumptions about variability within, and among individuals.For example,in a 60 min sampleperiod, a flockof 12 individuals would requiref-- 5 min whereasa flock of 60 would requirer-- 1 min. However, since flock size inevitably will vary, the investigatorshould establishan f that will accountfor an averagesize flock for the species and studyobjectives. Using marked (e.g., wing-tagged)individuals, when feasible,could reduce potential bias of samplingthe sameindividual more than onceper sampleperiod. It is apparent that habitat structureaffects avian behaviorin different ways. In Mourning Doves, active behavior (e.g., feeding, locomotion) generallywas highestin non-perch habitats, stationarybehavior (e.g., comfort, resting) was highest in perch habitats, and alert behavior was independentof habitat structure.Use of the equationin Table 1 assures that sampleswill reflect the flocks'overall behavioralstate. The investigator should be cautiouswhen using this equation and not assumethat a groupof birds constitutea flock unlessthere is evidenceof flock unity. For example, birds using three habitatscan be considereda flock only if Vol.60, No. 3 NewTime-Activity Methods [395 there is a reciprocalmovement of individuals among the three habitats. The investigatormay find it usefulto divide major habitat types (e.g., forest)into subtypes(e.g., perchwith canopyvs. perchwithout canopy). Observer-expectancybias can be avoidedby incorporatinga standard waiting period. This would maintain an equal samplingeffort for each flock and eliminate bias from expectedpredictions by the investigator (Balph and Romesburg1986). Also, our data suggestpotential for in- creasedresearch efficiency using a standard waiting period to control premature sampletermination causedby observerbias. Duration of the standardwaiting period shouldreflect the behaviorof the speciesunder study.Those capableof large daily movementsand having a low probability of returningto view, shouldhave a shorterstandard waiting period than specieswith small daily movementsand a high probability of re- turning to view. Also, view-restrictedareas shouldhave a long standard waiting period to allow time for individualsto return to view. These factorsshould be evaluatedprior to data collection,and the techniques practicedon the speciesbefore selecting a durationfor the standardwaiting period.

ACKNOWLEDGMENTS

Funded by the Alabama Department of Conservation and Natural Resources,Division of Game and Fish, Pittman-RobertsonFederal Aid in Wildlife RestorationProject No. W-44-10, Jobs XVA-XVB, and Alabama Agricultural Experiment Station Project No. 13-0065. Publishedas Alabama Agricultural Experiment StationJournal Series15-881758P. We thank: R. R. Hitchcock and C. M. Marn for their suggestionsand advice while we developedthese techniques;T. A. Givens, J. W. Tucker, Jr., R. M. Turnbull for field assistance;A. Appel, N. R. Holler, P. J. Trichilo, and M. C. Wooten for reviewingearlier drafts of the manuscript;T. S. Baskett and G. D. Schnell for helpful suggestionsthey provided as referees.

LITERATURE CITED

ALTMANN,J. 1974. Observationalstudy of behavior:sampling methods. Behaviour 49: 227-267. BALDASSARRE,G. A., S. L. PAULUS,A. TAMISIER,AND R. D. TITMAN. 1988. Workshop summary:techniques for timing activity of wintering waterfowl. Pp. 181-188, in M. W. Weller, ed. Waterfowl in winter. Univ. Minnesota Press, Minneapolis. BAL?H,D. F., AND H. C. ROMESBURG.1986. The possibleimpact of observerbias on some avian research. Auk 103:831. BRADLEY,D. W. 1985. The effectsof visibility bias on time-budgetestimates of niche breadth and overlap. Auk 102:493-499. DEUVER,M. j., ANDJ. R. FATORA. 1968. Daily and seasonalactivity patterns of mourning doves on the AEC Savannah River Plant. Proc. Ann. Conf. Southeast. Assoc. Game and Fish Comm. 22:181-189. KESSEL,B. 1976. Winter activitypatterns of Black-cappedChickadees in interior Alaska. Wilson Bull. 88:36-61. LEHNER,P.N. 1979. Handbook of ecologicalmethods. Garland STPM Press,New York & London. MAXSON,S. J. 1977. Activity patterns of female Ruffed Grouse during the breeding season. Wilson Bull. 89:439-455. , ANDN. P. BERNSTEIN.1984. Breedingseason time budgetsof the southernBlack- backed Gull in Antarctica. Condor 86:401-409. 396] M. P. Lositoet al. J.Field Ornithol. Summer 1989 PAULUS,S. L. 1984a. Activity budgetsof nonbreedingGadwalls in LouisianaßJ. Wildl. Manage. 48:371-380. ß 1984b. Behavioral ecologyof mottled ducks in Louisiana. Ph.D. dissertation. Auburn Univß, Alabamaß139 pp. QUINLAN,E. E., ANDG. A. BALDASSARRE.1984. Activity budgetsof non-breedingGreen- winged Teal on playa lakes in TexasßJ. Wildl. Manage. 48:838-845ß RavE, D. P. 1987. Activity budgetsand carcasscomposition of Green-wingedTeal wintering in LouisianaßM.S. thesis.Auburn Univ., Auburn, Alabamaß51 pp. SAS INSTITUTEINC. 1985. SAS user's guide: statistics,version 5 edition. Cary, North Carolina. 956 pp. TAMISIER,A. 1976. Diurnal activitiesof Green-wingedTeal and Pintail wintering in Louisiana. Wildfowl 27:19-32ß VERNER,J. 1965. Time budgetof the male Long-billed Marsh Wren during the breeding season. Condor 67:125-139. Received21 Jul. 1988; accepted12 Mar. 1989.