ALTERNATIVE FOODS OF A DIET SPECIALIST, THE SNAIL KITE STEVEN R. BEISSINGER Yale University,School of Forestryand EnvironmentalStudies, New Haven, Connecticut06511 USA, • and Departmentof ZoologicalResearch, National ZoologicalPark, SmithsonianInstitution, Washington,D.C. 20008 USA ABSTRACT.--AlthoughSnail Kites (Rostrhamus sociabilis) in Florida and Venezuelafed mostly on Pomaceasnails, I documentedthree alternativefoods. In Florida, kites fed on five species of smallturtles, but especiallyon Sternotherusodoratus and Kinosternonbauri. During the height ofa drought,one kite in Floridafed on a smallaquatic snail (Viviparus georgianus) for5 weeks.' In Venezuela,freshwater crabs (Dilocarcinus dentatus) made up on average10% of the Snail Kite's diet, but more than 25% during Septemberand October. Kites consumedalternative foodswith techniquesthat resemblesnail-eating behavior, such as enteringa turtle'sbody cavityby piercingthe only leg shapedlike a snailoperculum. Handling times for turtles(76 rain) and crabs(5.4 rain) were much longer than for Pomaceasnails (1.5 rain). Viviparussnails requiredapproximately one-third less handling time but containedone-fifth less body mass than Pomaceasnails. Handling time (30 s) for V. georgianusdid not differ betweenthe Snail Kite and the Boat-tailedGrackle (Quiscalus major), a diet generalist. Thesefindings are related to the factorsthat reinforcediet specialization and the ecological conditionsthat promote diet diversificationin specialists.Like Pomaceasnails, alternative foodshave shells or carapacesand moverelatively slowly. Alternative foods are probably lessprofitable than Pomaceasnails, except for large crabs.Although crabs were regularly eatenby kitesin Venezuela,turtles and Viviparussnails were eatenin Floridaonly during times of food scarcity.Received 9 June1989, accepted 6 November1989. DEPENDENCEon a single prey type is unusual Marisa snails for short periods (Mader 1981, among vertebrates,but may be more common Snyder and Kale 1983). in invertebrates (Bristow 1988). A notable ex- The evolutionary pathway to specialization ception, the Snail Kite (Rostrhamussociabilis), is often difficult to identify from ecological feedsalmost exclusively on one genusof fresh- studies(Futuyma and Moreno 1988). However, water snails (Pomacea).In Florida the kite eats someunderstanding of the factorsthat promote only Pomaceapaludosa (Howell 1932,Snyder and and maintain diet specialization may emerge Snyder 1969, Sykes 1987), while in Central and by examining the causesand consequencesfor SouthAmerica several Pomacea species are eaten specialistswhen they select alternative foods. (Hayerschmidt 1962, 1970; Beissinger 1983; For instance,two morphological specializations Snyder and Kale 1983; Bourne 1985). Alterna- allow Snail Kites to efficiently catch and eat tive foods are taken on rare occasions. These snails (Snyder and Snyder 1969). These are a include turtles (Sykes and Kale 1974, Woodin long, slender, hooked bill to sever the snail's and Woodin 1981, Takekawa and Beissinger columellar muscleand detach the snail's body 1989)and a smallmammal (Sykes and Kale 1974). from its shell, and long toesfor seizing the snail Other less carefully documentedobservations by the shell. These morphological specializa- of nonsnail foods include a "small yellowish tions could be expected to carry a cost of re- serpent"(Slud 1980),a ring-neckedsnake (Dia- ducedefficiency for feeding on alternativefoods dophispunctatus), and a deadAmerican Coot (Fu- (MacArthur 1972, Benkman 1988). lica americana)(Sykes 1987). In Venezuela and Here I documentalternative foods regularly Colombia, kites fed on freshwater crabs and eaten by Snail Kites in Florida and Venezuela. I report behavior associatedwith diet general- ization by a specialistand the ecologicalcon- Addressto which reprint requestsshould be sent. ditions that favored diet diversification. 327 The Auk 107:327-333. April 1990 328 STEVENR. BEISSINGER [Auk,Vol. 107 METHODS Turtles were eaten most frequently when I observed the food habits of Snail Kites in Florida water levelswere low during drought(n = 34 during studiesof their reproductivebehavior (1979- turtles;71%) or in yearsfollowing drought(n 1983). Studies were conducted on Lake Okeechobee = 10 turtles; 21%). Kites also ate turtles in winter (Glades, Hendry, and Okeechobee counties), State months(n = 4 turtles;8%) during or just after Water Conservation Areas 2 and 3 (Dade and Broward the passageof cold fronts. The seasonaldistri- counties), and Lakes Tohopekaliga and Kissimmee bution of turtle predation was as follows: 2 in (OsceolaCounty). Beissinger(1986) and Beissinger January,4 in February,9 in March, 28 in May, and Snyder (1987) describedthe study areasand cli- and 5 in July. matologicalconditions. Fieldwork was conducted from During the height of a drought in Florida in Januaryto August each year, except 1980 (July to 1981, one banded 2-year-old male Snail Kite August)and 1983(March to June).Additional obser- (#230) fed almostexclusively on a very small vations were made during February and April 1986. In Venezuela, studies were conducted in the llanos aquaticsnail (Viviparusgeorgianus) at Lake Okee- chobee. The kite was observed 300-500 m north at the ranch "Fundo Pequario Masaguaral" (8ø34'N, 67ø35'W), 45 km south of Calabozo in Guarico, and of the Clewistonlocks on sixdifferent days dur- on other ranchesalong the main highway 30 km to ing a 5-week period (11 June-13 July). Lake the north and south. ! observed food habits of kites levels were so low (ca. 3.0 m above mean sea from July to December1985, and from July to No- level) that V. georgianussnails, normally found vember 1986and 1987,usually by watchingfood de- on the bottomof lakesand streams(Snyder and livered to nestlingsby parents.No observationswere Snyder1969), lay exposedon a shallowmudflat. madeduring the dry season(January-May) when most During 5 hoursof observations,male #230 rap- kites leavethe llanos.Eisenberg (1979) and Beissinger idly ate 46 Viviparussnails and captured12 oth- et al. (1988) describedthe study areas and climato- logical conditions. ersthat he discarded,possibly because they were I observed behavior with 10 x binoculars and 15- empty shells.Although no Pomaceasnails were 60 x spottingscopes. When kites were feeding,their capturedduring these observationperiods, he behavior was described and timed to the nearest sec- caughtone Pomaceasnail during the five weeks ond. Handlingtime is definedas the ti.ne requiredby that he foragedalong the mudflat. Remainsof a kite to extractand consumecompletely a food item 170 Viviparusgeorgianus, 4 PomaceapaIudosa, and after perching with it. Whenever possible,remains 1 unopened freshwater musselwere under male of nonsnail foods were recovered, identified, and #230's extractionperch on 3 July. measured.During this process,I often collectedother During thissame period, another kite foraged food items beneath perches where kites had eaten snails or nonsnail foods. near the mudflaton four occasions.Although Data were analyzed by SYSTAT microcomputer this otherkite frequentlyhad difficultyfinding programs.As the datawere normallydistributed and Pomaceasnails, it never fed on Viviparusgeor- varianceswere equal, I used Student'st-tests to com- gianuseven though it watched male #230 feed pare handling timesbetween Pomacea snails and var- on Viviparussnails. ious alternative foods. Means (g) are reported with On one other occasion (during the 1981 standard deviations (SD). drought), kites fed on Viviparusgeorgianus on Lake Okeechobee.Twelve empty snail shells RESULTS were found under an extractionperch used ex- clusively by kites. Selectionand occurrenceof alternativefoods.-- In the llanos of Venezuela, Snail Kites aug- Small turtles were the alternative food most fre- mented their diet with freshwater crabs (Dilo- quently eaten by kites in Florida (Table 1). I carcinusdentatus). Of 3,866 food items captured watched kites eat 9 turtles and found 39 turtles by kites, 82.9% were positively identified. dead under perchesthat kites used regularly Freshwatercrabs made up 10.3%of the diet and for extractingsnails. Musk turtles (SternotherusPomacea doIiodes the remainder; but, the con- odoratus)accounted for 58.3% and mud turtles sumptionof crabsby Snail Kites varied season- (Kinosternonbauri) 31.3% of the turtles eaten. Soft- ally (Fig. 1). Crabswere rarely eatenfrom July shelledturtles (Trionyx ferox), and peninsulaand through early August, the first half of the wet Florida cooters(Chrysemys neIsoni and C. fiori- season.The frequency of crabsin the diet in- dana)made up the remainder.Most turtles eaten creased in September and October, and crabs by kiteswere smallerthan full-sizedadults, es- constituted more than one fourth of the food peciallythe larger species(Table 1). items in some months. Crab-eating declined in April 1990] Whena SpecialistDiversifies 329 40 ß 81 0 0 1985 / •- -• 1986 / • 3o / •- ß- • 1987 / / m 25 ,, 20 III /•8o o 228 / / • "' 727 "/ / 'Z•...• '• 267 ,•- -:." -O ,,, 56,,,' ..K 307 / _. •'• 0 45;381_.,L--•'•'•,,:' ' ' ' ' 91 : : -:' 7 8 9 10 11 12 MONTH Fig. 1. Seasonalvariation in the diet of Snail Kites in the llanos of Venezuela during three years of study. Only two foods were identified: Pomaceadollodes snails and freshwater crabs (Dilocarcinusdentatus). The percentageof crabsin the diet and the total number of food items identified are given for each month that >20 food items were identified. November and Decemberin 1985,the only year kite pulled small piecesof meat