J. Raptor Res. 25(3):63-67 ¸ 1991 The Raptor ResearchFoundation, Inc.

BIASES IN DIETS DETERMINED FROM PELLETS AND REMAINS: CORRECTION FACTORS FOR A MAMMAL AND -EATING RAPTOR

R.E. SIMMONS • Departmentof Zoology,University of the Witwatersrand,Johannesburg 2050, SouthAfrica

D.M. AVERY AND G. AVERY SouthAfrican Museum,P.O. Box 61, Cape Town 8000, SouthAfrica

ABSTR•½T.--Numerousstudies of predatorybirds worldwide report dietary proportionsbased on analyses of large numbersof pellets or prey remains. Such analysesare often severelybiased, hence strictly unquantifiable,because some prey remainsare moreconspicuous or persistentthan others.We investigated this bias for the bird- and micromammal-eatingAfrican Marsh (Circus ranivorus),using an essentiallyindependent measure of diet, observedprey deliveriesto the nest.Comparisons of the frequency of occurrenceshowed that bird prey, particularly large wetland species,were over-representedalmost threefold among remains. Micromammals were under-representedabout 1.5-fold, while fish, frogs and eggswere marginallyover-represented. Analyses using pellets were alsobiased but in the oppositedirection to that of remains.We show that by combiningpellets and prey remains (collectedwith equal effort), accurateestimates of overall diet can be achieved.This was verified using month by month comparisons of micromammals,in which proportionsderived from pellets and remains never differed by more than 10% from those established from direct observations.

Parcialidaden los resultadospara dietasdeterminadas por egagr6pilasy residuos:factores de correcci6n para el casode raptorasque se alimentande mamlferosy aves EXTR•½TO.--Numerososestudios sobre aves de presa, en todo el mundo, informan sobre proporciones de dietasbasadas en el anilisis de un gran nfimero de egagr6pilas,o en el de residuosde presas.Tales anilisis confrecuencia resultan muy parcializados,por tanto no cuantificables,debido a que algunosde los residuosde presa son mils conspicuoso persistentesque otros. Hemos investigadoesta parcialidad, para avesde rapifia de la especieCircus ranivorus, las que sealimentan de avesy mamlferosmuy pequefios. Hemos usadouna medida de dieta esencialmenteindependiente, tal comola observaci6ndel acarreo de presasal nido. Comparacionesde la frecuenciade ocurrenciasmostr6 que las presasconstituldas por aves,particularmente especies grandes de zonaspantanosas, fueron sobre-representadaspor los residuos casien el triple. Mamlferos muy pequefiosfueron sub-representadosen aproximadamente1.5; mientras que peces,ranas y huevosfueron marginalmente sobre-representados. Los anilisisque usaronegagr6pilas han sidotambign parcializados pero en sentidoopuesto al de losresiduos. Demostramos que combinando egagr6pilasy residuosde presa,colectadas con igual cuidado,estimaciones precisas de la dieta general puedenser 1ogradas. Esto ha sidoverificado usando, mesa mes,comparaciones de dietasconstitu•das por mamiferosmuy pequefios,en las que las proporcionesderivadas de egagr6pilasy de residuosnunca difieren en mils de 10% de las establecidaspor observaci6ndirecta. [Traducci6n de Eudoxio Paredes-Ruiz]

It is probablethat the majority of predator studies 1982, Korpim//ki 1985) and North America (review rely on prey remainsin someform to determinethe by Marti 1987). Indeed, for one group, owls, there diet of their subject.This is particularly sofor wide- is rarely any other way of assessingdiet but from rangingor elusivebirds suchas raptorsboth in Af- pellets(Jaksifi and Marti 1981). While most studies rica (e.g., Steyn 1982, Tarboton and Allan 1984, acknowledgethat prey remainsat nestsor feeding Boshoffet al. 1990), Europe(Newton and Marquiss sitesmay not be representativeof what is actually taken (Newton and Marquiss 1982), quantitative estimatesof the biasesinherent in suchanalyses are • Presentaddress: Ministry of Wildlife, Conservationand almost non-existent for wild . Several studies Tourism, Private Bag 13306, Windhoek, . have, however, attemptedthis by feeding captive

63 64 SIMMONS ET AL. VOL. 25, NO. 3 birds known dietsand subsequentlyidentified what mons 1989). Collectionsvaried, however, and generally occursin the pellets(Yalden and Yalden 1985, Vil- yieldedlittle at the start of breeding,increased for nests at which adults were incubatingor brooding,and de- lage1990). Alternatively,wild birdscan be observed creasedas large or flying youngfound and consumedall or photographedwith time-lapsecameras for long prey left by the adults.Remains were sortedto identify periodsat the nestto determinewhat is delivered; individualsand subsequentlyremoved. Extensive data on these data may then be comparedwith simulta- delivery patternsand prey types were concurrentlycol- neouslycollected remains and pelletsto determine lectedat 19 nestsover a three-yearperiod (Simmons 1989). These observations,totalling 2200 h, began about 1 mo the biases(e.g., Jarvis et al. 1980, Collopy 1983, beforebreeding each year and were continuedthroughout Marti 1987). The ultimate goal is to determinewhat breedingfor 6 mo. Nest watcheswere alsoevenly spaced an individual is eating using a correctionfactor for over the daylight period and resultspresented here are eachtype of prey categoryassessed. based on 701 observed deliveries to nests observed from hides placed60-100 m away. Our purposehere is to quantifythe biasesinherent As a third measure of diet composition,we simulta- in dietary analysesbased on remainsand pellets neouslycollected pellets from the same pairs. To both alone. We examine the diet of an avian predator, concentratepellet collectionsand protectthem from ubi- the African (Circusranivorus), which quitousmammalian carnivores,we providedperch posts consumesboth birds and mammals ranging in size for territorial males,and oncebirds beganto regularly use them, wound chicken-wirebaskets around poststo catch from tiny mice of 7 g to game birds up to 700 g. all regurgitatedmaterial. The basketswere high enough Specifically,we ask: are birds over-representedin and affordedsufficient protection to thwart mongoosesthat remainsand if soby how much do large birdspre- visited such areas. We do not claim that all pellets cast dominate over small birds, are micromammals un- were collected,but birds preferredour poststo their pre- der-representedand do somemice predominateover viouslyused ground roosts, thereby allowing a larger than usual sample.Harrier pellets,notoriously difficult to an- others?We subsequentlyshow that by combining alyzebecause few bones,only teethand skull partsremain data from pellets and remains, an accurateassess- undigested,were analyzedusing extensivemuseum ref- ment of diet can be achieved. erencematerial. We thus had three partly independent We use the term "micromammals"in preference measuresof diet: remains, pellets and direct nest obser- vations.Direct observationsare judged the bestindicator to the more usual "small mammals" to emphasize of diet (Marti 1987) becausethey representa deliveryby that only the extremelower massrange of mammals deliveryaccount of what breedingharriers brought to their on the African continentwere capturedby the har- nests,and they also representthe largestmost uniform riers studied.Furthermore, we are solelyconcerned sample. Naturally, this method is itself not completely representativeof what each bird catches;not all of the with the frequencyof occurrenceof prey in the diet, largestor smallestitems may be broughtto nestsbecause not the biomass.The biomassconsumed by an in- of foraging constraints(Simmons 1986b), and small or dividualis itselfstrongly biased when computed from partly dismembereditems were not alwaysidentifiable in averageprey weights as advocatedby some (e.g., the graspof flyingbirds. We donot believe these are serious Craigheadand Craighead1956, Steenhof 1983). This biases,however, because large items(3200 g), thosemost likely to bemissed because they are not carriedto the nest, arises becausemammalian prey of lower than av- comprisedonly 3% of all 707 identifiedprey, and very eragemass may be taken more frequentlythan pre- small items were frequently delivered. dicted(e.g., MacWhirter 1985), bird prey in harrier dietsare typically juveniles (e.g., Barnard et al. 1987, RESULTS this study) and smallerraptors rarely consumeall Diet from Direct Observations. Of 701 prey of the prey they capture(leaving major bones),again items delivered to 19 harrier nests between 1984 and biasing upwards the biomassestimates computed. 1986, 374 could be identified. Of these, 74% were Ways of avoidingor alleviating these biasesin the micromammals (rats, mice and shrews), and 23% laboratory (Wijnandts 1984), field (Masman 1986, passerinesand waterbirds. The remainder com- Simmons1986a) or via statisticalprocedures (Marti prisedsmall frogs(2%) and fish (1%). Considering 1987), have been discussedelsewhere. only the micromammals(N = 326), 89 could be identifiedto species;51% were Rhabdomyspurniho, STUDY AREA AND METHODS and 43% Otomysirroratus. The remaining 6% were In a 3-year study of African Marsh Harriers on the shrews (Table 1). Micromammals, therefore, pre- southerncoast of (34ø00•S22ø40•E), we col- dominated in the diet of these marsh harriers. lected and identified remains at nestsand known feeding sitesregularly from aboutone month before breeding until Bird Prey Biasesin Remains and Pellets. Of the young becameindependent (6-7 mo eachyear, Sim- 82 remainscollected at harrier nestsor feedingareas, FALL 1991 BIASES AND CORRECTIONS FOR HARRIER DIETS 65

Table 1. A comparison of the diet of African Marsh Table 3. A comparison of the seasonalproportions of Harriers determinedfrom direct observations,pellets and micromammalsderived from pelletsand remains,relative prey remains. to thosefrom direct observations(N = sample size).

DIRECT DIRECT OBSER- PELLETS AND OBSERVATIONS VATIONS PELLETS REMAINS MONTH REMAINS(N) (N) PREYT¾?E (374)a (251)b (82) July 85% (26) 75% (4) Micromammals 74% 83% 48% August 88% (65) 80% (71) 51% 38% 47% September 77% (70) 84% (110) Otomys 43% 44% 39% October 70% (44) 66% (105) Shrews 6% 8% 11% November 60% (53) 67% (67) Birds 23% 14% 40% December-January 51% (41) 53% (19) Frogs 2% 0% 4% Fish 1% 2 % 5 % Eggs 0% 1% 4% identifiedin pellets,included coots (Fulica sp.), var- Number of itemswithin eachcategory. iousrallidae and flufftails (Sarothrurasp.). Suchprey Numberof individualsidentified from 156 pellets. were, however,present in prey remains, and com- prised30% of the 40% total birds (Table 2). Birds 40% were birds.Observed deliveries and pelletsboth that could be identified in pellets were typically showedthat considerablyfewer birdsoccurred in the smallerspecies (doves, warblers and weavers).Hence, diet, than found in remains (Table 1). Thus, remains large avian specieswere more likely to be found in at harrier nestsor pluckingareas over-estimated bird remainsand smallerspedes in pellets.We couldnot prey 1.7-fold accordingto direct observations(the quantify the bias becauseof the large number of most accuratemethod) and 2.9-fold accordingto unidentifiedavian prey in pellets. pellets. Accordingto remains, avian prey becameprom- In pellets,however, the frequencyof individual inent from October(67%) and predominatedthere- birds was under-estimatedaccording to direct ob- after (Table 2). Since harriers partially switch to servations.Only 14% of all prey identifiedin pellets young avian prey as mouseabundance and vulner- were birds, whereas direct observationsshowed that ability declines(Simmons 1989), this was not un- birdscomprised 23% of the diet, a 1.6-fold difference. expected.However, accordingto directobservations, Thus the two methods either over-estimated (re- birds never exceeded small mammals in the diet of mains) or under-estimated(pellets) the proportion theseharriers. Once again, therefore, birds were over- of birds in the diet. Pellets were marginally more estimated in remains. accurate. Micromammal Prey Biases in Remains and BiasesAmong Bird Prey. Wetland birds which Pellets. Of the 82 prey remains only 48% were were commonlyseen in the area but never found or micromammals--mainlyRhabdomys pumilio (46%)

Table 2. Summary of monthly differencesin Harrier prey compositionassessed from remains found at nestsand plucking sites, ! 984-1986.

MICRO- PROPORTION OF MAMMALS BIRDS FROGS/FISH EGGS MAMMALS:BIRDS July-August 16 2 2 0 79 %:11% September 12 6 2 2 57%:26% October 3 10 [2]a 1 I 20%:67% November 7 8 [4] I 0 44%:50% December-January 0 7 [1] 1 0 0%:88% Totals 39 33 [7] 7 3 48%:40% No. of juvenilesin the total. 66 SIMMONS ET AL. VOL. 25, NO. 3 and Otomysirroratus (38%). Since micromammals possiblecorrection factor we undertooka seasonal comprised73% of the diet from direct observations assessment,combining remains and pelletsby month. (Table 1), remains under-estimatedthis prey type Again, the correspondencebetween proportionsof 1.5-fold. However, for the two main speciesof mi- micromammals derived from remains and pellets cromammals,Rhabdorays and Otomys,the relative were almostidentical to that found by direct obser- differencein their proportions(8%) was identical vation (Table 3). In any one month,the proportion from both methods. of dietary micromammalsfound by pelletsand re- Pellets appearedto be the least accuratemethod mains, did not differ by more than 10% from that of assessingwhich micromammalsoccurred most of- foundby directobservations. The differencesranged ten in the diet, sinceour analysessuggested that more from 2-10% (for the lowestsample size) and aver- Otoraysthan Rhabdorayswere eatenby harriers. By aged 6.3%--a difference small enough to be ex- providingcertain harriers with supplementaryfood plained by chance. (Simmons 1989), we could determine one reason why the large (50-200 g) Otomyswere more likely DISCUSSION to be found in pelletsthan the 30-80 g Rhabdorays; This studyquantifies what many avian research- Rhabdorayswere typically completelyeaten except ers have often suspected--thatbirds are seriously for a small sectionof the skull, including the jaw, over-representedin the prey remainsof mammal/ which was often discarded. This was not so for Oto- bird-eatingraptors. This is the first study,however, rays. Hence the skeletalelements that providedthe which attemptsto both quantify and subsequently mostreliable means of identifyingthis specieswould rectify suchbiases for a free-rangingraptor taking not always appear in the pellet. Under-representa- micromammals.That bird prey can be over-esti- tion of Rhabdoraysin pelletswas thus explicable. mated almostthreefold was unexpected,and shows Greater Musk Shrews (Crocidura ftavescens), the value in expendingconsiderable time in deter- strongly scented30 g insectivores,were observed mining diet from direct observation.Since avian re- beingcaught but discardedby African Marsh Har- searchersseldom have the time or perhapsthe in- riers, presumablybecause of their strongmusk and clinationto sit for hundredsof hourswatching their taste (cf. Smithers 1983). For example, in one case quarry, we have provideda much simpler method a completespecimen lay untouchedat an activefeed- of determiningdietary intake. By combiningpellets ing sitefor 4 days.According to remainsthey should, and prey remainswe showthat for any one month, therefore,be over-representedin the diet. This was proportionsof micromammalsdiffer by an average so (Table 1), but samplesare very small. of 6% (and no more than 10%) from that actually Other Prey Types. In general,other prey taken observed.This considerabletime-saving finding may by harriers were more likely to be found in remains also allow much greater accuracyfor diet determi- than either pelletsor directobservation. Hence frogs nation of raptors that are known (or suspected)to and eggswere apparentin remainsbut were rarely switch prey at certain seasons.That the method of recordedin pellets.The value of studyingremains, combiningpellets and remains gives accurateesti- therefore, lies in exposingthe more unusual items matesfor harriers (this study) and eagles(Collopy rarely recordedby other methods. 1983), suggeststhat it may have a more universal Correcting Biases:Pellets and Remains Com- application for mammal/bird-eating raptors than bined. When diet compositionfrom prey remains presentlyappreciated. and pelletswere combined(N = 333), the proportion For thesemethods to be applicablein other studies of micromammalsin the diet (74%) was exactlythat it is necessaryto determinethe number of pellets recorded from direct observations.Similarly, pro- andprey remainsrequired, and in what proportions, portionsof bird prey from pelletsand remains(20%) for an accurateassessment of diet. In this study,we were nearly identicalto that found from directob- collectedremains and pelletswith equal effort on an servation (23%). Hence it seemsthat for harriers, approximatelymonthly basis. That pelletsoutnum- remains and pelletscan be combinedto increasethe bered remains (about twofold) was a natural phe- accuracyof prey analyses.Similar conclusionswere nomenonattributable to the harriersand not to any reachedby Collopy(1983) studyingGolden Eagles differential collectingeffort. We also took care to (Aquila chrysaetos). increaseour efficiencyin pellet collecting,by pro- As an additional checkon the accuracyof this vidingperch posts within the territoriesof eachhar- FALL 1991 BIASES AND CORRECTIONS FOR HARRIER DIETS 67 rier, thereby minimizing the numerouspossible ar- Food of Crowned Eaglesin the Cape Province.Ostrich easin which birds might castpellets. As previously 51:215-218. stated,these collections were also protectedagainst KORPIMXKI,E. 1985. Prey choicestrategies of the Kestrel Falco tinnunculus in relation to available small mam- ubiquitousmammalian predators. In conclusion,it seemsthat the value of studying mals and other Finnish birds of prey. Annalesof Zool- ogicaFennici 22:91-104. diet from direct observationslies in the consistency MACWHIRTER, R.B. 1985. Prey selectionand provi- and accuracyof sucha method.On the other hand, sionins strategiesof breeding Northern Harriers (Cir- studyingpellets allows a more accurateassessment cuscyaneus). B.Sc. thesis.Mt. Allison University, NB, of speciescomposition, particularly small micro- Canada. mammals seldom recordedby direct observation. MARTI, C.D. 1987. Raptor food habits studies.Pages Lastly, the value of studyingremains, while biasing 67-80 in B.A. Giron Pendleton,B.A. Milsap, K.W. the more common remains in favor of birds, allows Cline and D.M. Bird [EDS.],Raptor managementtech- us to determinemore unusualprey suchas eggsand niques manual. National Wildlife Federation, Wash- fish. Each method,therefore, has its advantages.The ington, D.C. mostimportant point, however,is that it is possible, MASMAN,D. 1986. The annual cycleof the Kestrel Falco tinnunculus,a study in behavioral energetics.Ph.D. at least for harriers, to circumvent biases inherent thesis,University of Groningen, The Netherlands. in collectingjust remains or pellets by combining NEWTON,I. AND M. MARQUISS.1982. Food, predation them.Dietary proportionswithin about10% of the and breeding seasonin the Sparrowhawk (Accipiter "true" diet are then possible. nisus). J. Zool. Lond. 197:221-240. ACKNOWLEDGMENTS SIMMONS,R. 1986a. Ecologicalsegregation of the Red- breastedSparrowhawk (Accipiterrufiventris) and six Thanks are due to staff of the Cape Department of Nature and Environmental Conservation at Rondevlei for co-existingaccipitrine raptors in southernAfrica. Ardea 74:137-149. logisticalsupport to R.E.S. throughout,and the National Parks Board for accommodation in 1984. Phoebe Barnard 1986b. Food provisionins,nestling growth and madevaluable comments on a draft of this paperand Marc experimentalmanipulation of broodsize in the African Bechard, Paul Steblein and Guy Cowlishaw aided pre- RedbreastedSparrowhawk (Accipiterrufiventris). Or- sentation.Funding for D.M.A. was providedby the Foun- his Scandinavica 17:31-40. dationfor ResearchDevelopment (FRD) and the South 1989. Adaptationand constraintin the breeding African Museum and R.E.S. was funded from Wits Uni- of subtropicalharriers and eagles.Ph.D. thesis.Uni- versity and the FRD. versityof the Witwatersrand,Johannesburg, South Af- LITERATURE CITED rica. SMITHERS,R.H.N. 1983. The mammals of the southern BARNARD,P.E., R.B. MACWHIRTER, R.E. SIMMONS,G.L. African subregion.Pretoria University Press,Pretoria, HANSENand P.C. SMITH. 1987. Timing of breeding South Africa. and the seasonalimportance of passerineprey to breed- STEENHOF,K. 1983. Prey weightsfor computingpercent ing Northern Harriers (Circuscyaneus). Can. J. Zool. biomassin raptor diets. Raptor Research17:15-27. 65:1942-1950. STEYN,P. 1982. Birds of prey of southernAfrica. Groom BOSHOFF,A.F., N.G. PALMERAND G. AVERY. 1990. Re- Helm, Beckenham, U.K. gional differencesin the diet of Martial Eagles in the TARBOTON, W. AND D. ALLAN. 1984. The status and Cape Province,South Africa. S. Aft. J. Wildl. Res. 20: conservationof birds of prey in the Transvaal. Trans- 57-68. vaal Mus. Monogr. No. 3, Pretoria, South Africa. COLLOPY,M.W. 1983. A comparisonof direct obser- VILLAGE,A. 1990. The Kestrel. T&A.D. Poyser,Berk- vationsand collectionsof prey remainsin determining hamsted, U.K. the diet of Golden Eagles.J. Wildl. Manage. 47:360- WIJNANDTS,H. 1984. Ecologicalenergetics of the Long- 368. eared Owl (Asio otus). Ardea 72:1-92. CRAIGHFAD,J.J. AND F.C. CRAIGHEAD.1956. Hawks, YALDEN,D.W. AND P.E. YALDEN. 1985. An experi- owls and wildlife. StackpoleCo., Harrisburg, PA. mentalinvestigation of examiningKestrel diet by pellet JAKSI•, F.M. AND C.D. MARTI. 1981. Trophic ecology analysis.Bird Study32:50-55. of Athene owls in mediterranean-type ecosystems:a comparativeanalysis. Can. J. Zool. 59:2331-2340. JARVIS,M.J.F., M.H. CURRIEtdqD N.G. PALMER. 1980. Received26 October 1990; accepted9 February 1991