J. Field Ornithol., 61(3):331-338 PREY DELIVERED TO ROSEATE AND COMMON TERN CHICKS; COMPOSITION AND TEMPORAL VARIABILITY CARL SAFINA, RICHARD H. WAGNER1, DAVID A. WITTING 2, AND KELLYJ. SMITH2 National Audubon Society 306 South Bay Ave Islip, New York 11751 USA Abstract.--We observedprey deliveriesto Roseate(Sterna dougallii) and Common (S. hi- rundo)tern chicks.Roseate Terns were highly specializedand fed their chicksmostly sandeels (Ammodytesamericanus), whereas Common Terns delivered a diversity of prey. Species compositionof deliveredprey varied annually and weekly. Of four major prey species,the proportionof deliveriescontaining anchovies, juvenile bluefish, and herring varied with the time of morning, whereasthe proportionmade up of sandeelsremained generally constant. ENTREGA DE ALIMENTO A POLLUELOS DE STERNA DOUGALLH Y S. HIRUNDO: COMPOSICION Y VARIABILIDAD TEMPORAL Sinopsis.--Obscrvamosla cntregade alimentoa polluelosde Sternadougallii y S. hirundo. Los adultosdc S. dougalliialimcntaron a suspolluclos muy particularmentecon individuos dc Ammodytesamericanus, micntras quc la otra cspccicdc gaviotaalimcnt6 a suspichoncs con una gran divcrsidaddc prcsas.La composici6ndcl alimcntocntrcgado a los pichoncs, result6variable a travis de1cstudio. Dc las cuatrocspccics dc pcccsutilizadas principalmcntc comoalimcnto para lospichoncs, la proporci6ncntrcgada quc contcnlaanchoas, pcz azulado y arcncavari6 a travfisdc lashoras dc la mafiana,micntras quc la quc contcnlaA. americanus pcrmancci6gcncralmcntc constantc. Ever since Volterra (1931) and Gause (1934) formulated what has becomeknown as the competitiveexclusion principle (Hardin 1960), therehas been interest in resourcepartitioning by closelyrelated sympatric species.The Common (S. hitundo) and endangered(Federal Register 52 FR 42064) Roseate(Sterna dougallii) terns often breed and forage together in the northeasternUnited States,and occasionallyhybridize (Hays 1975). But their nest site microhabitatdiffers (Burger and Gochfeld 1988 and refs.therein), they exploit different foragingconditions (see Nisbet 1989, Safina 1990), and they are generallyreproductively isolated. In this paper we report interspecificdifferences and temporalvariability in prey deliv- ered to Roseate and Common tern chicks. METHODS We observedprey deliveriesto chicksof Roseateand Commonterns at Cedar Beach,New York (40øN, 73øW). During the breedingseasons of 1984-1987, we watched4, 7, 6, and 3 RoseateTern broodsand 4, 6, 9, and 9 Common Tern broodsin each year, respectively.Dates varied • Currentaddress: Edward Grey Instituteof Field Ornithology,South Parks Rd., OxfordOXl 3PS, England. 2 Currentaddress: Rutgers University, Marine Field Station,Box 278, Tuckerton,New Jersey 08087 USA. 331 332] C. Safinaet al. J.Field Ornithol. Summer 1990 from year to year becausethe terns bred later each year. Prior to or immediately after hatching, nestswere fenced with 0.5 m high, 2.5 cm hexagonalmesh wire to keep chicksnear the nestsite. A 30 cm high band of fine meshfiberglass cloth around the inner lower portion of the fences helpedboth to retain small chicksand preventinjury to chicks.All fenced compoundswere 2-5 m in diameterand includedadequate natural or added cover to allow chicks to hide and find shade. Prey deliverieswere observedwith binocularsfrom blinds between 0500 and 0900 E.S.T. 3-5 daysper week during the chickrearing period. We observed nests for a total of 1359 h for Roseate Terns and 2193 for CommonTerns, during which 863 fishtransfers occurred to roseatechicks and 2369 fish transfers to common chicks. We were able to record all prey deliveriesduring observationperiods. Upon transfer of a fish to a chick,we notedfish species,tern species,and hour. We are confidentthat our fish identificationsare accurate,Cezilly and Wallace's (1988) skep- ticismnotwithstanding. Observers were very familiar with the fish species in the area, many of which were quite distinctivein shapeand color,and numeroussimultaneous observations by two or more of us in the field showedvirtually completeagreement among observers, except when ob- serversagreed that positiveidentification could not be made. We avoided using judgment in identifying fish; if the prey's specificidentity was unclear, it was listed as unidentified.Observation blinds were positioned to take the best advantageof morning light in helping make prey iden- tification. Identificationwas further facilitatedby the fact that terns often circledrepeatedly before attempting to transferprey to chicks,and chicks' prey handlingtime, especiallywhen young,often allowedexcellent views of prey. Alosaand Clupeaherring couldnot be identifiedto speciesduring deliveryand were clumpedin the data. These two specieswere identified as being presentbecause uneaten specimenswere recoveredadjacent to nests. Fish size relative to terns' bill length was recordedin 1984 and 1985, but not in 1986 or 1987 becausewe feared that a changeof observersin the latter two years might introduce inconsistencyinto this subjective measure.We trappedthe adult terns and measuredtheir bill length. Data were analyzed using SAS softwareat Rutgers University. RESULTS Inter-speciesdifferences in chickdiets.--The compositionof chickdiets (Table 1) differed between tern speciesin each year and in all years combined(X 2 = 499.29, P ( 0.0001, df = 5; prey other than sandeels, anchovies,butter fish, bluefish, and herring were clumpedas 'other'). The mostapparent difference was the preponderanceof sandeels(Ammodytes americanus)in RoseateTern dietscompared to a greater diversityof prey speciesbrought by Common Terns (Table 1). We analyzedthe proportionof each of five major prey speciescom- prisingprey fed to chicksof eachtern species(Fig. 1). Roseatesbrought a higherproportion of sandeelsto chicksthan did CommonTerns in each Vol.61, No. 3 PreyDelivered to TernChicks [333 TABLE1. Prey broughtto tern nestsunder observationat Cedar Beach;number of indi- viduals observed,percent of each species'deliveries, and niche breadth values(B = 1/Zp•2; Levins 1968). Common Terns Roseate Terns Per- Per- Species n cent n cent American sandeel Amrnodytesamericanus 780 35.5 751 72.6 Bay anchovy Anchoa rnitchilli 168 7.7 42 4.0 Bluefish Pornatornus saltatrix 267 12.2 137 13.4 Butterfish Peprilus triacanthus 139 6.3 5 0.4 Herring Clupea and Alosa 251 11.4 70 6.8 Commonpipefish S?2gnathusfuscus 214 9.7 0 0 Atlantic mackerel Scornher scornbrusa 108 4.9 16 1.6 Long-tinnedsquid Loligopealez 1 0.05 0 0 Round herring Etrumeus teres 17 0.8 4 0.4 Scup Ste,2otornuschrysops 1 0.05 0 0 Killirish Fw2dulusspp. 16 0.2 1 0.1 Atlantic menhaden Brevoortiatyrannus 4 0.2 0 0 Flatfish, probably Winter flounder Pseudopleuronectesamericanus or Windowpane Scophthalrnusaquosus 1 0.05 0 0 Jack Caranx spp. 1 0.05 0 0 Threespined stickleback Gasterosteus aculeatus 1 0.05 0 0 Goosefish Lophius americanus 1 0.05 0 0 Unid. hake Merluccius bilinearis? 36 1.6 0 0 Common shore shrimp Palaemonetesvulgaris 59 2.7 0 0 Moth unidentified insect 53 2.4 0 0 Isopod unidentified 12 0.6 0 0 Unidentified fish 35 1.6 9 0.9 Niche breadth B = 5.63 B = 1.81 year, and for all yearspooled (X 2 -- 387.54, df = 1, P < 0.0001). Common Terns deliveredbutterfish to chicksin higherproportion than did roseares in each year and all years pooled (X2 = 56.44, df = 1, P < 0.0001). Anchovieswere deliveredin greaterproportion by CommonTerns than by rosearesfor all years pooled(X 2 = 14.91, df = 1, P (0.0001). This differencewas most pronouncedin the first two years. Both tern species deliveredherring (Clupeaharengus and Alosaaestivalis) to observedchicks in three of the four years. In two of thesethree years there was virtually no differencein the proportionwith which eachspecies delivered herring to chicks,but in 1987 commonsdelivered a significantlyhigher proportion (X2 = 33.80, df = 1, P (0.0001), and this resulted in a significant differencebetween species for all years pooled(X 2 -- 17.09, df -- 1, P ( 0.0001). Deliveries of juvenile bluefish (Pomatomussaltatrix) were ob- servedin 1985-1987. Roseatesdelivered a significantlyhigher proportion of bluefish than commonsin 1985 and 1986 (respectively,X2 = 7.10, df = 1, P ( 0.01; X2 = 13.41, df = 1, P < 0.0001), but commonsdelivered a higher proportion of bluefish than rosearesin 1987 (X2 = 5.10, df = 1, 334] C. Safina et al. j. Field Ornithol. Summer 1990 1984-1987 80 [] Roseate Terns [] Common Terns 60 40 20 Sandeels Anchovies Butterfish Bluefish Herring Other FIGURE1. Diet compositionof Roseateand Commonterns, 1984-1987. P < 0.02), resultingin no interspecificdifference for pooledyears (X2 = 0.76, df = 1, P < 0.4). Temporaldifferences in chickdiets.--From year to year, diet composition showedconsiderable variability. Prey compositionof chick dietsdiffered among yearsfor RoseateTerns (X2 = 264.82, df = 15, P < 0.0001) and CommonTerns (X2 = 816.58, df = 15, P < 0.0001; Fig. 2). The proportion of dietscomposed of eachof severalprincipal prey speciesvaried among yearsfor both terns (Table 2). Prey compositionof chickdiets was highly variable among weeks during the breedingseason in each year and for all years pooled(Table 3), for both Roseate Terns (X2 = 179.06, df = 24, P < 0.0001) and Common Terns (x 2 = 1051.14, df = 32, P < 0.0001). Prey compositionof chick diets differed among hours of the morning for RoseateTerns (X2 = 47.47, df = 16, P < 0.0001) and Common Terns (x 2 = 140.49, df = 16, P < 0.0001) for all years combined.Of several major prey species,the proportion of deliveriesmade up of anchovies, juvenile bluefish, and herring showedincreasing