Paleobiology, 30(2), 2004, pp. 297–324

Diets, habitat preferences, and niche differentiation of Cenozoic sirenians from Florida: evidence from stable isotopes

Bruce J. MacFadden, Pennilyn Higgins, Mark T. Clementz, and Douglas S. Jones

Abstract.—Cenozoic sediments of Florida contain one of the most highly fossiliferous sequences of extinct sirenians in the world. Sirenians first occur in Florida during the Eocene (ca. 40 Ma), have their peak diversity during the late Oligocene–Miocene (including the widespread dugongid Me- taxytherium), and become virtually extinct by the late Miocene (ca. 8 Ma). Thereafter during the Pliocene and Pleistocene, sirenians are represented in Florida by abundant remains of fossil man- atees (Trichechus sp.). Stable isotopic analyses were performed on 100 teeth of fossil sirenians and extant Trichechus manatus from Florida in order to reconstruct diets (as determined from ␦13C val- ues) and habitat preferences (as determined from ␦18O values) and test previous hypotheses based on morphological characters and associated floral and faunal remains. A small sample (n ϭ 6) of extant Dugong dugon from Australia was also analyzed as an extant model to interpret the ecology of fossil dugongs. A pilot study of captive manatees and their known diet revealed an isotopic enrichment (␧*) in ␦13C of 14.0‰, indistinguishable from previously reported ␧* for extant medium to large terrestrial 18 mammalian herbivores with known diets. The variation in ␦ OV-SMOW reported here is interpreted to indicate habitat preferences, with depleted tooth enamel values (ഠ25‰) representing freshwater rivers and springs, whereas enriched values (ഠ30‰) indicate coastal marine environments. Taken together, the Eocene to late Miocene sirenians (Protosirenidae and Dugongidae) differ significantly in both ␦13Cand␦18O from Pleistocene and Recent manatees (Trichechidae). In general, Protosiren and the fossil dugongs from Florida have carbon isotopic values that are relatively positive (mean ␦13C ϭϪ0.9‰) ranging from Ϫ4.8‰ to 5.6‰, interpreted to represent a specialized diet of pre- dominantly seagrasses. The oxygen isotopic values (mean ␦18O ϭ 29.2‰) are likewise relatively positive, indicating a principally marine habitat preference. These interpretations correlate well with previous hypotheses based on morphology (e.g., degree of rostral deflection) and the known ecology of modern Dugong dugon from the Pacific Ocean. In contrast, the fossil and extant Trichechus teeth from Florida have relatively lower carbon isotopic values (mean ␦13C ϭϪ7.2‰) that range from Ϫ18.2‰ to 1.7‰, interpreted as a more generalized diet ranging from C3 plants to seagrasses. The relatively lower oxygen isotopic values (mean ␦18O ϭ 28.1‰) are interpreted as a more diverse array of freshwater and marine habitat preferences than that of Protosiren and fossil dugongs. This study of Cenozoic sirenians from Florida further demonstrates that stable isotopes can test hy- potheses previously based on morphology and associated floral and faunal remains. All these data sets taken together result in a more insightful approach to reconstructing the paleobiology of this interesting group of ancient aquatic mammalian herbivores.

Bruce J. MacFadden, Pennilyn Higgins, and Douglas S. Jones. Florida Museum of Natural History, Uni- versity of Florida, Gainesville, Florida 32611. E-mail: bmacfadd@flmnh.ufl.edu, E-mail: loligo@flmnh. ufl.edu, and E-mail: dsjones@flmnh.ufl.edu Mark T. Clementz.* Department of Earth Sciences, University of California, Santa Cruz, California 95064.

Accepted: 15 August 2003

Introduction waters throughout the subtropical and tropi- Extant sirenians are fully aquatic mamma- cal Pacific and Indian Oceans (Walker 1975; lian herbivores that constitute the Order Sir- Husar 1978a). The second family, the Tri- enia. One of the two modern sirenian families, chechidae, consists of a single genus, Triche- the Dugongidae, with an extant monotypic chus, the manatee, a mixed-feeding herbivore genus of hypergrazer, the dugong (Dugong du- with three species distributed throughout riv- gon), is widely distributed in coastal marine ers, estuaries, and coastal marine areas of the tropical and subtropical Atlantic Ocean. Of relevance to the present study, the manatee * Present address: Smithsonian Marine Station at Fort Pierce, 701 Seaway Drive, Fort Pierce, Florida 34949. native to Florida is referred to T. manatus lati- E-mail: [email protected] rostrius, the West Indian Manatee (Walker

᭧ 2004 The Paleontological Society. All rights reserved. 0094-8373/04/3002-0009/$1.00 298 BRUCE J. MACFADDEN ET AL.

1975; Husar 1978b). Despite this limited mod- Eocene of Florida (Ivany et al. 1990), demon- ern diversity and disjunct distribution, which strating very well preserved seagrass beds is further threatened by human impact, sire- (with Thalassia-like taxa similar to those of the nians have had a rich and fascinating macro- Caribbean basin today) associated with fossil evolutionary history of diversification and ad- sirenian bones referable to Protosiren sp. aptation over the past 50 million years since (Domning et al. 1982; Hulbert 2001). their split from the basal Tethytheria (a clade In contrast to the specialized seagrass graz- also including proboscideans [McKenna and ing of dugongs, manatees are considerably Bell 1997]). more generalized mixed-feeders, with Triche- One of the richest and most extensive fossil chus manatus latirostrius reported to consume records of sirenians occurs in the nearshore, over 60 plant species in Florida (Ames et al. shallow marine and estuarine sediments ex- 1996). In addition to seagrasses, the diet of Tri- posed throughout Florida (Hulbert 2001), and chechus can consist of an array of freshwater this sequence is critical to understanding the and marshy vegetation, including a consider- diversification of this order. The systematics, able proportion of true grasses (Poaceae) such morphology, and adaptive hypotheses con- as Distichlis, Spartina, Paspalum,andPanicum cerning these sirenians have been treated ex- (Domning 1982, 2001a). The rostral deflection tensively in earlier descriptions (e.g., Simpson angle of Trichechus is relatively weak (Ͻ40Њ; 1932; Reinhart 1976) and in a series of more Fig. 1F; Domning 2001a). With the exception recent papers by Domning (e.g., 1988, 1989a,b, of the earliest trichechid Potamosiren magdalen- 1990, 1994, 1997a). sis from the middle Miocene of Colombia In addition to developing a modern system- (Domning 1997b), other manatees have rela- atic framework for the Sirenia, Domning (e.g., tively short-crowned teeth that are shed and 1981, 2001a) has developed hypotheses about continually replaced ‘‘conveyor-belt’’-style the diet and ecology of extinct dugongs and from back to front throughout ontogeny. This manatees based on those of modern sirenians, continuous source of teeth is interpreted as an morphological correlates in the fossil record, initial adaptation to feeding on abrasive grass- and associated geological and paleontological es (Domning 1982, 1997b, 2001b), although evidence. Domning (2001a) defines the more recently manatees have adapted to a ‘‘aquatic megaherbivore adaptive zone’’ and more diverse herbivorous diet. explains how different sirenians partition Within the past decade, stable isotopes have food resources and other aspects of individual become increasingly useful in reconstructing species niches. Modern dugongs are predom- the paleoecology and paleoenvironment of ex- inantly grazers, feeding on seagrasses (Hy- tinct vertebrates. The original literature on drocharitaceae and Potamogetonaceae), al- this subject is expanding rapidly (see, e.g., though they also eat algae when seagrasses Koch 1998 for recent review). These geochem- are limited or unavailable. These sirenians are ical studies have allowed (1) testing of previ- characterized by bunodont (bulbous-cusped) ously existing hypotheses based on more tra- cheek teeth with limited tooth replacement ditional (i.e., morphological) data, and (2) re- and a high degree of rostral deflection (Fig. construction of various aspects of aut- and 1E), which indicates the use of the down- synecology of vertebrate species and com- turned snout for digging up seagrasses, in- munities not previously discernable with con- cluding their rhizomes. This high degree of ventional data. Studies of stable carbon iso- rostral deflection (up to 70Њ [Domning 1977]) topes preserved in tooth enamel carbonate can be traced back into the fossil record to the have mostly concentrated on terrestrial ver- middle Eocene, and for extinct dugongs is tebrates and communities in an attempt to re- likewise interpreted as an adaptation to feed- construct ancient herbivore diets and their ing on ancient seagrasses. Paleontological ev- feeding adaptations on C3 or C4 plants. Stable idence suggesting the long-standing sea cow oxygen isotopes, which have been used exten- and seagrass plant-animal coevolutionary in- sively in ancient marine ecosystems, have terrelationship first comes from the middle been more difficult to apply to terrestrial set- FLORIDA CENOZOIC SIRENIANS 299

FIGURE 1. Comparisons of degree of rostral deflection in extant and fossil sirenians. A, Primitive morphology in Prorastomus sirenoides from the middle Eocene of Jamaica (adapted from Savage et al. 1994). B, Protosiren fraasi from the middle Eocene of Egypt (adapted from Domning 2001b). C, High degree of rostral deflection in Metaxytherium floridanum from the early Miocene Lower Bone Valley Formation, Florida (adapted from Domning 1988). D, Rela- tively little rostral deflection seen in Metaxytherium crataegense from the middle Miocene of Florida (adapted from Simpson 1932). E, Extant Dugong dugon. F, Extant Trichechus manatus.

tings because of the high degree of variability 1. The relative degree of rostral deflection in- of meteoric waters and the many factors influ- dicates dietary adaptations, with increased encing the ultimate oxygen isotopic compo- flexure indicating a specialized seagrass sition of tooth enamel carbonate. Some recent diet. studies have used stable oxygen isotopes to 2. Like modern dugongs, extinct Florida pro- interpret evolutionary transitions between ter- tosirenids and dugongs inhabited relative- restrial versus marine habitats, e.g., the adap- ly marine waters, whereas Florida mana- tive radiation in Eocene whales (Roe et al. tees have inhabited a broader range of 1998). aquatic environments, including coastal Current hypotheses about the evolution of marine as well as freshwater, inland rivers. diets and related functional adaptations of ex- tinct sirenians have been extensively de- Sequence of Cenozoic Sirenians from scribed in the existing literature (see Domning Florida references cited above). Of particular rele- Taken together, all four sirenian families are vance here, stable isotopes of carbon and ox- represented in the nearshore marine and con- ygen can be used to test the following hy- tinental deposits of Florida (Table 1). Al- potheses about the diets and habitats of ex- though mention will be made of other sireni- tinct sirenians of Florida: ans, this discussion will principally focus on 300 BRUCE J. MACFADDEN ET AL.

TABLE 1. Fossil and extant Sirenia known from Florida. Taxa analyzed during the present study are presented in boldface, with the number of individual bulk samples analyzed in parentheses (also see Appendix). (Compiled from Domning 1994, 2001a and Hulbert 2001).

Order Sirenia †Family Prorastomidae Genus and species indeterminant †Family Protosirenidae Protosiren sp., middle Eocene (Clairbornian); ?early Oligocene (4) Family Dugongidae †Subfamily Halitheriinae Metaxytherium sp., very late Oligocene Metaxytherium crataegense, late early to middle Miocene (3) Metaxytherium floridanum, middle to late Miocene (23) Subfamily Dugonginae Undescribed small dugongid, middle Miocene Crenatosiren olseni, very late Oligocene to very early Miocene Dioplotherium manigaulti, very late Oligocene to middle Miocene Dioplotherium allisoni, early Pliocene Corystosiren varguezi, early Pliocene (1) Undescribed genera and species, Pliocene Family Trichechidae Trichechus sp., ?late Pliocene, early to late Pleistocene (45) Trichechus manatus latirostris, West Indian Manatee, the late Pleistocene (Rancholabrean) to Recent (18 wild, 6 captive) those taxa for which stable isotopic results are Formation of central Florida (Fig. 2). These presented below. This discussion is separated seagrasses consist of Thalassodendron and Cy- into temporal intervals that represent the modocea, genera believed to be very closely re- available isotopic samples, and not natural lated to the common Florida seagrass, Thal- phylogenetic groups. assia (Ivany et al. 1990). Sirenian fossils re- Eocene to Early Oligocene (40–35 Ma). The ferred to Protosiren sp. are found in direct as- earliest and most primitive sirenian group, sociation with these seagrasses and a Family Prorastomidae, although better known plant-herbivore interaction is implied. In con- from the early to middle Eocene of Jamaica trast to the more primitive condition seen in (Savage et al. 1994; Domning 2001c), is rep- Prorastomus, Florida Protosiren sp. is character- resented by extremely fragmentary material ized by a rostral deflection of 35–40Њ (Domn- (vertebrae) from the late Eocene of Florida that ing et al. 1982) (Fig. 1B), implying increased was not amenable to stable isotopic analysis. foraging on seagrasses. Prorastomidae are, so far as is known, char- It also should be noted that during the late acterized by a relatively flat rostral morphol- Oligocene through early Miocene, several du- ogy (Fig. 1A) indicating little of the deflection gongid species were present in Florida (Hul- seen in other sirenians. Savage et al. (1994) bert 2001), including three (Metaxytherium sp., speculated that Prorastomus was a selective Crenatosiren olseni,andDioplotherium mani- browser feeding on floating and emergent gaulti) that coexisted at some localities (Domn- aquatic plants, and to a lesser degree, sea- ing 1989a). Although diverse, teeth for these grasses. sirenians are both rare and difficult to identify The Family Protosirenidae is somewhat bet- taxonomically. Consequently these extinct du- ter represented in Florida relative to the Pro- gongids were unavailable for isotopic analy- rastomidae, and an intimate association with sis. fossil seagrasses has been observed. Ivany et Middle to Late Miocene (ca. 17–8 Ma). By far al. (1990) described a diverse assemblage of the most common Miocene sirenian in the well-preserved seagrass communities from world was Metaxytherium (Fig. 1C,D), and this Dolime Quarry, middle Eocene Avon Park also was the case in Florida (Domning 1988). FLORIDA CENOZOIC SIRENIANS 301

FIGURE 2. Map of localities discussed in text, including those from which specimens were sampled for this study.

The oldest records of this genus analyzed here ent from the Bone Valley Metaxytherium that are from the early part of the middle Miocene, Domning (1988) refers to M. cf. floridanum. Me- including (1) the La Camelia Mine, Haw- taxytherium floridanum is characterized by a thorne Formation in the Florida panhandle relatively high degree of rostral deflection (Bryant 1991); and (2) lowermost units of the (ഠ75Њ, Fig. 1C), suggesting specialization on Bone Valley phosphate mines of central Flor- seagrasses (Domning 1988). ida (Fig. 2). Both of these localities are Barsto- Late Miocene/Early Pliocene (7–4 Ma). The vian land mammal age (sensu Tedford et al. Miocene/early Pliocene was a pivotal time in 1987; this North American Land Mammal Florida sirenian evolution. Unfortunately the ‘‘Age’’ [NALMA] is estimated to range from fossil record of this interval is notably lacking. about 16 to 11.5 Ma). These oldest specimens The diversity seen during the late Oligocene are referred to Metaxytherium crataegense (Fig. and early Miocene drops, and dugongs are 1D) (Domning 1988; Bryant 1991; Hulbert apparently not well represented after the mid- 2001). Thereafter, the extensive open-pit phos- dle Miocene. The latest Miocene–early Plio- phate mines of central Florida contain abun- cene (late Hemphillian) Upper Bone Valley dant sirenian remains in the Lower Bone Val- Formation (ca. 5 Ma), which produces exten- ley Formation, which Domning (1988) con- sive interbedded terrestrial and marine mam- cluded are all referable to Metaxytherium flor- malian faunal assemblages (the Palmetto Fau- idanum (Fig. 1C). These fossils come from the na [Morgan 1994; Hulbert 2001]), demon- later part of the middle Miocene Agricola Fau- strates a conspicuous absence of dugongs, na and, according to land mammal biochro- probably representing a local near extinction nology, are early Clarendonian NALMA in of sirenians. Two notable exceptions to the to- age (Morgan 1994), ca. 12 Ma. Thereafter, du- tal extirpation of Florida sirenians during the gongs occur rarely in younger deposits in the late Miocene include an early Hemphillian oc- Bone Valley district, from localities such as currence of Corystosiren varguei (also see Nichols Mine, which are probably early Hem- Domning 1990) and undescribed late-surviv- phillian in age (Hulbert 1988). In north Flori- ing dugongids (Domning 2001a; Hulbert da, various creeks in the Gainesville area con- 2001) from the late Miocene and early Plio- tain dugong remains of a form slightly differ- cene. It is indeed unfortunate that the fossil re- 302 BRUCE J. MACFADDEN ET AL. cord of late Miocene–early Pliocene dugon- horizontally replaced teeth that are continu- gids is not better represented, because this is ally shed during ontogeny. These teeth are the time that most dugongids became extinct very abundant in river localities from north- in the Atlantic and Caribbean, and the tri- ern Florida of this age. Domning (1982, 2001a) chechids presumably radiated from South suggests that late Pleistocene Trichechus sp. America into this biogeographic province. from Florida does not represent temporally This poor record of sirenian evolutionary his- successive populations, but may instead rep- tory in the Atlantic/Caribbean region has re- resent a sequence of genetically independent, sulted in speculative hypotheses either about multiple dispersal events into this region. the competitive superiority of manatees rela- Modern Florida manatees, Trichechus manatus, tive to dugongs or, if they did not indeed co- are characterized by a moderate degree of ros- exist and compete, about the manatees filling tral deflection (Fig. 1F), indicating some sea- the aquatic megaherbivore niche vacated by grass preference, although this species is the regional extinction of dugongs in the At- known to have a generalized diet (Ames et al. lantic Ocean (Domning 1982, 2001a). 1996; Domning 1982, 2001a). Modern mana- Early Pleistocene Through Recent (1.5 Ma to the tees are known to exist in a wider range of Present). So far as can be determined from habitats than dugongs, with the former rang- the available fossil record, the Trichechidae ing from full marine on one hand to fresh- probably arose in South America and is first water rivers on the other (Husar 1978b). represented by Potamosiren magdalensis from In summary, the sequence of fossil sirenians the middle Miocene La Venta Fauna of Colom- in Florida, although discontinuous, covers bia (Domning 1997b). The dental morphology some very important stages in the evolution of of P. magdalensis indicates that the teeth were this order. As will be demonstrated below, iso- not replaced continually during ontogeny as topic analysis of specimens from this se- in more-derived manatees. Domning (1982, quence offers some interesting results and in- 1997b) suggests that the diet of this most sights relative to previous hypotheses about primitive known manatee consisted of softer ancient diets and habitat preferences. vegetation but did not include the abrasive true grasses (Poaceae) known to be part of the Stable Isotopes, Plant Foods, and Habitat diet of the extant manatee Trichechus manatus. Reconstruction Although there have been suggestions that Within the past few decades, stable isotopes manatees are first known in the middle Plio- have become widespread as proxies for inter- cene (Blancan NALMA) of northern Florida preting and reconstructing the paleoecology about 2.5 million years ago (Domning 1982), of extinct vertebrates (e.g., Koch 1998). Carbon for example from the Santa Fe River, more re- and oxygen isotopes will be used here to re- cent interpretations suggest that this faunal construct, respectively, the diet and habitat of assemblage is mixed with Pleistocene (Ran- sirenians from Florida. The conventional delta cholabrean) mammals; therefore, a definitive (␦) notation is used here for carbon (␦13C) and older, middle Pliocene, date cannot be deter- oxygen (␦18O), where ␦ (parts per mil, ‰) ϭ 13 mined unambiguously (R. C. Hulbert person- [(Rsample/Rstandard) Ϫ 1] ϫ 1000, and R ϭ C/ al communication 2001). The first definitive, in 12Cor18O/16O. The data from the unknown situ record of Florida manatees is represented (fossil) are compared relative to internation- by Trichechus sp. from the 1.5-Ma Irvingtonian ally accepted standards, which for ␦13CisV- Leisey Shell Pit south of Tampa (Morgan and PDB (Pee Dee Belemnite) and for ␦18O report- Hulbert 1995). Having been derived from a ed here relative to V-SMOW (Standard Mean South American ancestor (Domning 1982, Ocean Water, both following Vienna conven- 1997b), the extinct Florida fossil manatees are tion, hence ‘‘V-’’ [e.g., Coplen 1994]). morphologically close to both T. manatus and Carbon (␦13C) and Ancient Diets. Plants frac- the West African T. senegalensis, and may rep- tionate carbon with one of three different pho- resent a distinct species. Pleistocene and Ho- tosynthetic pathways. Most plants use the locene Florida manatees are characterized by Calvin Cycle, in which carbon is incorporated FLORIDA CENOZOIC SIRENIANS 303 into three-carbon chain compounds, hence the mostly falls into a narrow range between term ‘‘C3.’’ C3 plants include trees; shrubs; Ϫ10‰ and Ϫ11‰ (Boutton 1991; Hemminga cool-growing-season, high-altitude and high- and Mateo 1996), and Hydrilla, with a ␦13Cof latitude grasses; and many aquatic plants, and Ϫ13‰ to Ϫ26‰ (Ames et al. 1996). It is im- represent about 85% of the world’s plant bio- portant to note here that Hydrilla is a non-na- mass. Tropical and temperate grasses, which tive species in Florida, introduced in the 1950s account for about 10% of terrestrial plant bio- (Langeland 1990), and was not a major dietary mass, fractionate carbon into four-carbon component for any of the sirenians examined chain compounds (as an intermediate product in this study. The remaining plant species in of photosynthesis), and hence the term ‘‘C4’’ the diet of T. manatus from Florida include plants. Of relevance here, grasses that live (Ames et al. 1996) terrestrial C3 and C4 grass- along the land/water interface photosynthe- es, macroalgae, mangroves, cattail (Typha sp.), size carbon by using either the C3 or C4 path- and water hyacinth (Eichhomia), which togeth- way depending upon species. The third pho- er range widely in their ␦13C values (about tosynthetic pathway, CAM (Crassulacean Ϫ5‰ to Ϫ35‰), but all of these are in rela- Acid Metabolism), exemplified by Cactaceae tively minor quantities. (cactus) and Euphorbiaceae (euphorbia), con- When isotopes of carbon are incorporated stitutes the remaining percentage of plant into skeletal tissues, there is an isotopic en- photosynthesis (Ehleringer et al. 1991). In richment (␧*) relative to the plant foods (i.e., view of the fact that CAM plants do not con- 13C is selectively incorporated into skeletal tis- stitute a part of the known diet of sirenians, sues, resulting in a higher ␦13C value). For me- particularly in Florida, this photosynthetic dium to large terrestrial modern mammals, ␧* pathway is not considered further. In terres- is 14.1‰ (Cerling and Harris 1999), although 13 trial ecosystems, C3 plants have a mean ␦ C little was previously known about this enrich- value of Ϫ27‰ and a broad range of Ϫ32‰ ment in marine mammals (also see discussion 13 to Ϫ23‰, whereas C4 plants have a mean ␦ C below). of Ϫ13‰ with a narrower range of Ϫ15‰ to Oxygen (␦18O) and Habitat Preference. In Ϫ10‰ (Dienes 1980; Farquhar et al. 1989; contrast to the relatively straightforward in- Boutton 1991). Most fully marine plants (in- terpretation of carbon uptake and fraction- cluding seagrasses) fractionate carbon by us- ation, and corresponding ␦13C values of tooth ing the C3 pathway, although because the car- enamel carbonate, understanding the array of bon source in the oceans can be derived from factors affecting the isotopic fractionation of bicarbonate as well as atmospheric CO2,the oxygen in the environment and how oxygen 13 ␦ CvaluesofC3 marine plants can be more is then metabolized by mammals is more com- enriched than their terrestrial counterparts plex. Thus, the inherent assumptions about in- (Boutton 1991; Keeley and Sandquist 1992; terpreting the source and history of oxygen, Hemminga and Mateo 1996). The ␦13Cof and how it can be used to interpret ancient en- coastal aquatic plants is more complicated be- vironments, requires multiple assumptions cause of the multiple sources of carbon from that are difficult to unambiguously control. It which plants draw and fractionate (Boutton is well known that the fractionation of oxygen 1991; Keeley and Sandquist 1992). isotopes in the environment is a function of As mentioned above, the modern manatee temperature, and thus the classic use of ␦18O Trichechus manatus eats more than 60 species of as a paleothermometer. However, because aquatic (marine/freshwater) plants and these mammals maintain a constant body temper- span a wide range of ␦13C values, approxi- ature, variation in enamel ␦18Ovaluesisnot mately Ϫ3‰ to Ϫ35‰ (Ames et al. 1996). just a result of temperature, but is influenced Nevertheless, the diet of Trichechus manatus by physiological processes within the body from Florida consists predominantly of sea- and the ␦18O composition of various environ- grasses (mostly Thalassia, Syringodium, Halo- mental sources of oxygen. dule, Halophila, Diplanthera,andRuppia [Ames In terrestrial mammals, the two principal et al. 1996]), which have a mean ␦13C that sources of oxygen are contained in the ani- 304 BRUCE J. MACFADDEN ET AL. mals’ drinking water and in plant water. These modern mammals therefore represents the two categories of ingested water frequently end result of all of these potential sources of have different isotopic values within the same variation. A similar interpretation is made for ecosystem that can introduce a significant fossil teeth, with the additional consideration amount of variation in ␦18O values among in- of possible secondary alteration of the initial dividuals in a population. Likewise, a variety ␦18O signature during fossilization. Despite of physiological processes, including meta- the possibility of diagenesis and earlier claims bolic rate, sweating, panting, and excretion, to the contrary (particularly for fossil bone can alter the magnitude and fractionation of [Schoeninger and DeNiro 1982]), tooth enam- oxygen fluxes entering and leaving a mam- el ␦18O has been used for paleoenvironmental mal’s body water pool (Bryant and Froelich interpretations (e.g., Bocherens et al. 1996; 1995; Kohn 1996). These factors further in- Cerling and Sharp 1997; and MacFadden et al. crease the ␦18O values among individuals and 1999). potentially complicate intraspecific and inter- Mean tooth enamel ␦18O values of modern specific comparisons of ␦18O values. and fossil sirenians are used as a proxy for re- Fortunately, among aquatic species, the fac- constructing their habitat preferences through tors affecting body water ␦18O values are time, based upon the following model for ␦18O greatly reduced. Observations of captive ma- of environmental water. Modern seawater has 18 rine species suggest that the majority (Ͼ98%) a global average ␦ OV-SMOW value of 0.0‰ (by of oxygen flux is the result of passive diffusion definition) and is relatively homogeneous of environmental water across the skin (Hui worldwide (Hoefs 1997). Nevertheless, some 1981; Andersen and Nielsen 1983). Therefore, of the greatest variations in seawater ␦18O are the ␦18O value of an aquatic mammal’s body observed in marginal marine settings where water should correlate strongly with the ␦18O they result from the combined effects of fac- value of the environmental waters in which it tors such as evaporation, continental runoff, lives (Yoshida and Miyazaki 1991), and pop- upwelling, and ocean currents. Except in es- ulations of aquatic mammal species living in tuarine environments, measurements indicate waters of relatively homogeneous isotopic that the shallow marine coastal waters of composition should yield low ␦18O variation coastal Florida are slightly enriched in 18O among individuals. The ␦18O value of an compared with the global mean (Swart et al. aquatic mammal’s body water is therefore la- 1996). Seawater from St. Augustine Beach beled by the ␦18O value of the water in which (used as an internal standard in the Stable Iso- it lives, and populations migrating between tope Laboratory of the Department of Geolog- waters of different ␦18O values should exhibit ical Sciences, University of Florida) has a higher variation in ␦18O values than popula- mean ␦18O of 0.7‰. This supports the obser- tions remaining in a single, localized water vation of Swart et al. (1996) that the ␦18Oof body (Clementz and Koch 2001). Florida ocean water lies between 0 and 1‰ Ultimately, the ingested oxygen is in isoto- and is strongly influenced by the strength of pic equilibrium with body tissues, including the Loop Current output from the Mississippi the hydroxylapatite of teeth. Tooth enamel ac- River as well as by evaporation (Ortner et al. cretes during a small portion of an animal’s 1995). life span, on the order of months (depending In contrast, Florida freshwater settings gen- upon factors such as size). Enamel ␦18Ovalues erally exhibit relatively depleted ␦18O values. thereby serve as an archive of fluctuations in a Most rivers display a range of ␦18O values that mammal’s body water ␦18O values during the generally decrease with increasing distance time of tooth mineralization (Bryant and Froe- inland from the coast as a result of evapora- lich 1995; Kohn 1996). This temporal variation tion and fractionation during precipitation within a tooth can introduce additional dif- (Coplen and Kendall 2000). The isotopic com- ferences among individual ␦18O values based position of precipitation, a major contributor upon the amount of enamel that is sampled. to the riverine water budget, is highly variable The ␦18O measured from tooth enamel of across Florida but generally averages about FLORIDA CENOZOIC SIRENIANS 305

Ϫ3‰ (Swart et al. 1989). Springs represent an- other important source of water for many Florida rivers and are frequently sought out by manatees during the colder months of the year. Year-round monitoring of ␦18O during 1998–99 at a typical spring (Hart Springs) lo- cated in north-central Florida recorded rela- tively depleted values; mean ␦18O ϭϪ4.0‰, with monthly values ranging between Ϫ3.6‰ and Ϫ4.4‰ (Jones and Quitmyer unpublished data). A U.S. Geological Survey analysis of six Florida rivers from different geographic re- 18 gions within the state revealed mean ␦ O val- FIGURE 3. Simple box model of potential ␦13Cand␦18O ues ranging from Ϫ0.1 to Ϫ3.9‰ (Coplen and end-member values of tooth enamel used to interpret, Kendall 2000). In a seventh river system, the respectively, the diets and habitats of extinct Florida si- renians. Econlockhatchee from central Florida, the mean ␦18OwasϪ2.0‰ (Gremillion 1994). Not all of Florida’s freshwater environ- water ␦18O curve (Lear et al. 2000; Billups and ments, however, are characterized by low ␦18O Schrag 2002), and can be used in the recon- values. For instance, the large standing body struction of ancient habitat preference. We of water that makes up the Everglades is iso- therefore use it to differentiate marine and topically enriched in 18O (0 to 2‰) as a result freshwater habitats for fossil sirenians from of sustained evaporation (Meyers et al. 1993; Florida. Swart et al. 1996). Fortunately, such exceptions A simple conceptual model with four the- do not appear to play a major role in the en- oretical end-members (Fig. 3) is used below to vironmental life history of Florida sirenians, interpret the carbon and oxygen isotopic data past or present. for the fossil sirenians from Florida. These It is well documented that the ␦18O value of data are also compared with those from mod- global seawater has varied throughout the Ce- ern sirenians with known diets and habitat nozoic from Ϫ1.2‰ during the ‘‘greenhouse’’ preferences. The abscissa is calibrated by end- world of the Paleocene and early Eocene, to members ranging from a specialized seagrass modern values consistent with the late Ceno- feeder (tooth enamel ␦13C ഠ 4‰) to a predom- 13 zoic ‘‘icehouse’’ (Zachos et al. 2001). New geo- inantly C3 feeder (tooth enamel ␦ C ഠ chemical proxies for seawater paleotempera- Ϫ13‰). The ordinate is calibrated by end- tures have allowed investigators to assess the members ranging from individuals living in relative contributions of temperature and ice relatively freshwater habitats (tooth enamel volume to the global ␦18O record of marine ␦18O ഠ 25‰) to those inhabiting marine sea- carbonates and thereby reconstruct seawater water (␦18O ഠ 30‰). Thus, starting in the up- ␦18O evolution during the Cenozoic (Lear et al. per right-hand corner of the model, individ- 2000; Billups and Schrag 2002). On a regional uals with relatively positive ␦13Cand␦18O scale, the local ␦18O variations along Florida’s end-member values are interpreted to feed ancient coasts and rivers remain poorly con- mainly on seagrass and live in principally ma- strained throughout the Cenozoic. However, rine habitats. A modern analog for this is the these factors do not diminish the robustness of dugong (Dugong dugon) from the Pacific the present-day model; i.e., the ␦18Ovaluesof Ocean (Husar 1978a). In the lower right-hand freshwater are depleted relative to seawater, corner, individuals with relatively positive and hence enamel ␦18O values of freshwater si- ␦13C and negative ␦18O values would be inter- renians (see below) are lower than for their preted to feed mostly on seagrass and mainly marine counterparts. We assert that this mod- live in freshwater habitats. This combination el can be applied throughout the Cenozoic, of diet and habitat is rare in nature because particularly when anchored to the global sea- seagrasses are primarily marine, but perhaps 306 BRUCE J. MACFADDEN ET AL. was represented in the past. In the lower left Indian Manatee, Trichechus manatus, from Flor- hand corner, individuals with relatively neg- ida cataloged in the skeletal collection of the ative ␦13Cand␦18O values are interpreted to Florida Museum of Natural History Mammal feed principally on C3 aquatic plants and live Collection (abbreviation ‘‘UF M’’; Appendix) in predominantly freshwater habitats. Popu- were analyzed in accordance with the U.S. lations of the extant West Indian manatee (Tri- Fish and Wildlife Service permit. With the ex- chechus manatus) from Florida are known to ception of one specimen from La Venta, Co- trend towards this kind of diet and habitat lombia (see further discussion below), all iso- (Husar 1978b; Ames et al. 1996). In the upper topic analyses of fossil sirenians are from Flor- left-hand corner, individuals with more neg- ida, with 67 specimens from the Florida Muse- ative ␦13C and more positive ␦18Ovalues um of Natural History Vertebrate Paleontology would be interpreted to have a diet of mainly Collection (abbreviation ‘‘UF’’) and nine speci-

C3 plants (i.e., excludes seagrasses) and live in mens from the private fossil collection of John marine habitats. A potential sub-Recent ana- Waldrop (abbreviation ‘‘JW’’) of Lake Wales, log within the Sirenia for this marine C3 feeder Florida. Residual sample powders of the JW end-member is the Steller’s Sea Cow (Hydro- specimens used for isotopic analysis are ar- damalis gigas) from the North Pacific Ocean chived in the UF collection. that principally ate kelp (mean ␦13CofϪ19‰ Tusk samples of six modern dugongs were Ϯ 1‰ [compiled from Hobson et al. 1994; Si- collected from populations within the Torres menstad et al. 1993]) and was hunted to ex- Straits off the coast of Australia by the School tinction in 1768 (Domning 1999). of Tropical Environment Studies and Geog- raphy at James Cook University in Queens- Materials and Methods land, Australia. This dugong population was Stable Isotope Sample Acquisition (Teeth, Food, selected as a modern analog because it has and Water). Six naturally shed teeth of live, been studied extensively by researchers at captive Florida manatees collected from the James Cook University, providing a long re- bottom of their enclosures were borrowed for cord of observational data on the ecology of analysis from the Mote Marine Laboratory, this population (Marsh et al. 1984; Nietsch- Sarasota (‘‘Hugh’’ or ‘‘Buffett’’; also see Ames mann 1984; Marsh 1986; Marsh and Saalfield et al. 1996), and from the South Florida Mu- 1991). These studies confirm that dugongs are seum (‘‘Snooty’’) in accordance with a U.S. strictly marine mammals specializing on sea- Fish and Wildlife Service permit. Hugh and grasses with only minor amounts of marine Buffett, two male manatees, were born at the algae in their diet. Most importantly, no re- Miami Seaquarium (Hugh, 28 June 1984; Buf- cords suggest any habitation of freshwater fett, 16 May 1987) and have spent their entire sites or inclusion of freshwater plants in their lives in captivity. Although both manatees are diets. nearly the same length (298 and 313 cm re- Isotope Sample Preparation, Analyses, Stan- spectively), Buffett is much heavier, weighing dardization, and Conventions. Bulk samples about 660 kg to Hugh’s 535 kg (D. Colbert per- were mechanically drilled from teeth, by us- sonal communication 2001). Because Hugh ing either a Foredom௢ variable speed drill or and Buffett share the same enclosure, it is not a Dremel௢ tool. Sirenians other than dugongs possible to determine from which individual (see below) were sampled by collecting a the shed teeth were derived. Snooty is another small sample of powdered enamel from some male manatee that has also spent his entire life point along the tooth. The exact sample loca- in captivity. He is the oldest manatee in cap- tion was based on tooth preservation, ease in tivity and was born 21 July 1948, at the Miami sampling, and minimizing destruction of di- Aquarium. Snooty weighs about 455 kg (C. agnostic morphology. Serial samples were re- Audette personal communication 2001). Man- moved from one Metaxytherium tooth, one atee food and water were collected from their captive Trichechus tooth, and two fossil Triche- tanks for isotopic analysis. chus teeth to assess the isotopic variation along Eighteen tooth samples of the extant West the length of a tooth. Dugong tusks (Dugong FLORIDA CENOZOIC SIRENIANS 307 dugon) from Australia were sectioned longi- Hugh and Buffett also are fed a large quantity tudinally and bulk samples were collected by of kale. Other foods added to supplement drilling the edge of the tusk across several their diet are relatively minor (Ͻ1% each). successive growth bands. This method al- Samples of all foodstuffs consumed by these lowed homogenization of several years of manatees were collected for isotopic analysis. growth and provides a robust estimate of the All samples were dried in an oven and then average isotopic values for each individual. each was pulverized. The carbon isotopes for Sample sizes prior to preparation were char- each of these organic samples were measured acteristically about 100–300 mg for off-line ex- with the Carlo Erba CNS Analyzer in the Uni- traction and 10–25 mg for on-line extraction. versity of Florida Department of Geological Following standard protocol (e.g., as de- Sciences. Water samples were collected by dip- scribed in MacFadden and Cerling 1996), sam- ping samples from the surface of the tank and ples were pulverized by using a mortar and also directly from the fill spigot for each of the pestle, then treated with either NaOCl or 30% two tanks housing captive manatees. Water

H2O2 overnight to remove any surficially samples were equilibrated with CO2 at 45ЊC bound organic compounds and then with 1 N for 5 hours before CO2 is drawn into and mea- or 0.1 N acetic acid (some were also buffered sured by the VG Prism mass spectrometer at following recommendations in Koch et al. the University of Florida. 1997). For off-line extractions (done at the Uni- In all cases, unknown sample analyses were versity of Utah), approximately 50–100 mg of calibrated to either internal laboratory stan- the treated powders was reacted with H3PO4 dards (TEC-CC, Carrera Marble, MEme, or for 40–48 hours at 25ЊC. The evolved CO2 sam- UCSC elephant enamel) or NBS-19, and ulti- ples were then purified and extracted follow- mately back to the V-PDB or V-SMOW stan- ing standard cryogenic laboratory procedures dards (Coplen 1994). Analytical precision for (McCrea 1950). Previous analyses of tooth each mass spectrometer is characteristically enamel carbonates have demonstrated the better than about 0.1‰ for ␦13C and 0.2‰ for 18 presence of a contaminant, probably SO2 (e.g., ␦ O. MacFadden et al. 1996), that will significantly affect the carbon and oxygen isotopic values Results and Discussion of the tooth enamel carbonate sample. Accord- Isotopic Fractionation in Modern Sirenian Food, ingly, all off-line enamel CO2 samples were Water, and Tooth Enamel Carbonate. Previous collected in sample tubes containing Ag wool studies have shown that tooth enamel ␦13C and then heated for at least 24 hours at 50ЊC. values of mammalian herbivores are enriched This essentially removes this contaminant relative to ingested plant foods. For mammals from the purified sample. Off-line extracted in general, this isotopic enrichment has been samples were measured on a Finnigan mass stated to be approximately 12–14‰ (e.g., Koch spectrometer in the Department of Biology et al. 1992) for medium to large mammalian SIRFER laboratory at the University of Utah. herbivores. Body size may be a contributing For on-line extractions, approximately 1 mg source of variation. Ambrose (1998) found that was weighed and loaded into an Isocarb ex- small mammals, i.e., lab rats, have a signifi- traction carousel (University of California- cantly lower mean isotopic enrichment of Santa Cruz) or individual reaction vessels on 9.4‰. Using extensive analytical data from a a multiprep device at the UF. The samples wide variety of extant medium to large mam- were reacted with H3PO4 (for varying times malian herbivores with known diets, Cerling and temperatures, depending on instrument and Harris (1999) calculated that the mean and inlet system used) and then purified and isotopic enrichment (␧*) for tooth enamel car- extracted for analysis on either a VG Prism bonate relative to diet is 14.1 Ϯ 0.5‰. As most (UF) or a Micromass Optima mass spectrom- previous studies have dealt with terrestrial eter (UCSC). mammalian herbivores, little is known about The diet of the captive manatees used in this the corresponding ␧* for marine mammals, in- study consists principally of romaine lettuce. cluding sirenians. Ames et al. (1996) analyzed 308 BRUCE J. MACFADDEN ET AL.

13 TABLE 2. Stable isotopic composition (␦ CV-PDB) of manatee food and isotopic enrichment (⑀*, sensu Cerling and Harris 1999) for Hugh/Buffett and Snooty.

Sample number Material Amount ␦13C (‰) Mote Marine Laboratory, Hugh/Buffett PH 01–06 Romaine 96 heads/day/2 manatees Ϫ27.8 PH 01–07 Kale 12 heads/day/2 manatees Ϫ30.7 PH 01–08 Monkey biscuit 12–16/day Ϫ21.3 PH 01–09 Carrot K1% Ϫ27.8 PH 01–10 Apple K1% Ϫ24.7 PH 01–11 Beet K1% Ϫ26.7 weighted mean1 ϭ Ϫ27.9 South Florida Aquarium, Snooty PH 01–12 Romaine 200 lbs/day/2 manatees Ϫ27.8 PH 01–13 Romaine 200 lbs/day/2 manatees Ϫ28.1 PH 01–14 Monkey chow 3% Ϫ19.8 PH 01–15 Elephant chow K1% Ϫ26.5 PH 01–16 Apple K1% Ϫ24.4 PH 01–17 Broccoli K1% Ϫ25.7 PH 01–18 Sweet potato K1% Ϫ23.7 weighted mean2 ϭ Ϫ27.4 Isotopic enrichment between food and tooth enamel carbonate ␦13C: ␦13C values (Appendix) used to calculate mean enrichment ⑀* include Hugh/Buffett (5 samples) and Snooty (1sam- ple) each subtracted from weighted mean of food source above yields isotopic enrichment of 14.0‰.

1 Approximate ␦13C for average daily intake of Hugh and Buffett, where for total diet, romaine ϭ 85%, kale ϭ 11%, monkey biscuit ϭ 3%, and all other foods ഠ 1%. 2 Approximate ␦13C for average daily intake of Snooty, where for total diet, romaine ϭ 96%, monkey biscuit ϭ 3%, and all other foods ഠ 1%. the ␦13C values of known plant food and soft indicate that the mean ␦13C value of ingested tissues (liver, kidney, blubber, and skin) of foodstuffs was Ϫ27.9‰ for Hugh/Buffett and Hugh and Buffett, then at the Lowry Park Zoo, Ϫ27.4‰ for Snooty, both of which are close to 13 Tampa (later moved to Mote Marine Labora- the mean ␦ C value for C3 terrestrial plants tory), but no teeth were analyzed during that (Dienes 1980; Farquhar et al. 1989; Boutton study. 1991). Using five teeth from Hugh/Buffett Given the lack of existing data from which and one tooth from Snooty (Appendix), and to model the ␧* of modern sirenians, and then with the known isotopic composition of the by inference, fossil sirenians from Florida, food, we report here the calculated isotopic known food and naturally shed teeth of enrichment (␧*) of 14.0‰. This is essentially Hugh/Buffett and Snooty were analyzed. A the same value determined for terrestrial potential source of error is introduced from mammalian herbivores (14.2‰ [Cerling and teeth that were shed and collected from the Harris 1999]), and is the model used to inter- bottom of the manatee’s tanks prior to the time pret the isotopic enrichment between fossil si- that we actually collected the plant foods for renians from Florida and their ancient diets. isotopic analysis. Therefore, the assumption is Eight water samples, representing either that the carbon isotopic values of the diet in- tap fill or tank water, were analyzed from the corporated into the teeth were relatively con- enclosures containing the captive manatees to stant over the time interval between which the attempt a calculation of an enrichment factor teeth were mineralized and shed and the diet between ␦18O of environmental water and was analyzed. Information provided by the tooth enamel. For Hugh/Buffett the mean fill animal keepers (also see Ames et al. 1996) in- water ␦18OisϪ2.1‰ (n ϭ 2) and Ϫ1.0‰ (n ϭ dicated that the diet remained constant, con- 2) for the tank water. For Snooty the mean fill sisting principally of romaine lettuce (Ͼ96%), water is Ϫ0.3‰ (n ϭ 2) and 0.0‰ (n ϭ 2) for with the remaining 4% including kale, carrots, the tank water. Despite the fact that these wa- apples, beets, broccoli, sweet potato, and com- ter samples are nonsaline, their isotopic values mercially available monkey biscuits and ele- lie close to expected values for marine water. phant chow. Our analytical results (Table 2) We interpret the slightly more positive tank FLORIDA CENOZOIC SIRENIANS 309

TABLE 3. Stable isotopic values indicating intra-tooth variation in teeth of fossil and extant Florida sirenians.

Approximate position from base of enamel, mm (proportion with respect to tooth height) ␦13C (‰) ␦18O (‰) Metaxytherium floridanum, UF JW F6912, Bone Valley-Kingsford-0, ca. 10 Ma 14.5 mm (0.86) Ϫ0.4 30.1 11.5 mm (0.64) Ϫ1.1 28.6 8.5 mm (0.47) Ϫ1.4 28.9 4.5 mm (0.25) Ϫ1.6 27.7 Mean, Standard Deviation, range Ϫ1.1, 0.5, 1.2 28.8, 1.0, 2.4 Hugh or Buffett, Trichechus manatus, captive, UF uncataloged, Recent 5.5 mm (0.61) Ϫ15.0 27.0 3.5 mm (0.39) Ϫ14.4 26.9 1.5 mm (0.17) Ϫ13.5 26.8 Mean, Standard Deviation, range Ϫ14.3, 0.8, 1.5 26.9, 0.1, 0.2 Trichechus sp., Oklawaha River I, UF uncataloged, ca. 0.025 Ma 7.5 mm (0.83) 1.3 31.0 6 mm (0.67) 1.2 30.4 4.5 mm (0.50) 1.1 30.2 3 mm (0.33) 1.5 30.3 1.5 mm (0.17) 2.6 32.2 Mean, SD, range 1.5, 0.6, 1.5 30.8, 0.8, 2.0 Trichechus sp., Oklawaha River I, UF uncataloged, ca. 0.025 Ma 8 mm (0.98) Ϫ11.0 30.5 6.5 mm (0.72) Ϫ11.7 30.0 4.5 mm (0.50) Ϫ10.2 30.4 3 mm (0.33) Ϫ9.7 30.5 1.5 mm (0.17) Ϫ8.2 30.9 Mean, Standard Deviation, range Ϫ10.1, 1.4, 3.5 30.4, 0.3, 0.9

water to represent the effect of evaporation. months to about a year in manatees and other Given the small sample size and ambiguous sirenians. results, these data are not used either to rep- To understand isotopic variation in Florida resent a good model for the ␦18O of Florida sirenians, we selected four teeth for serial natural freshwater or to calculate an enrich- sampling. Three of these represent Trichechus ment factor between the ␦18O of tooth enamel (one extant captive and two late Pleistocene relative to environmental water. specimens) and the fourth represents Miocene Isotopic Variation along the Sirenian Tooth Metaxytherium (Table 3). We selected the teeth Crown. Most specimens analyzed during from the captive individuals (Hugh/Buffett), this study consisted of one bulk sample re- realizing that these manatees are fed on an moved from an individual tooth. Recent stud- isotopically monotonous diet and live in an ies have shown that isotopic variation along enclosure with an isotopically homogeneous the crown of modern and fossil mammal teeth water source. We therefore predicted relative- can be used to explore climatic and dietary ly little variation in tooth enamel ␦13Cor␦18O. variation over the period of time during which In contrast, the late Pleistocene Trichechus the tooth mineralizes (e.g., Bryant et al. 1996; from the Oklawaha River and Miocene Metax- Cerling and Sharp 1997). The actual time of ytherium should demonstrate isotopic varia- mineralization of individual sirenian teeth is tion representing diets and habitat of wild poorly known (D. Domning personal com- populations as archived within individuals. munication 2003), but judging from eruption The stable isotopic data for the captive Tri- sequence (Domning and Hayek 1984) and chechus, i.e., Hugh/Buffett, show relatively lit- tooth development in other medium-sized tle variation in both ␦13C (range 1.5‰) and mammals (references in Bryant et al. 1996), ␦18O (range of 0.2‰). This would be expected this probably occurs over a period of many given both the monotonous diet (Table 2) and 310 BRUCE J. MACFADDEN ET AL. source of water used for the tank enclosure. sa River, was not analyzed as part of the AN- The data for the fossil specimens demonstrate OVA because its statistical ‘‘group’’ represents more variation in ␦13Cand␦18O as might be a sample size of one. The captive manatees (n predicted from diversity in diets and probable ϭ 6) also were not analyzed in this ANOVA. habitats. The dugongid Metaxytherium is gen- The ANOVA results indicate significant dif- erally believed to have had a specialized diet, ferences in the four statistical populations, 13 corroborated by the narrow range (1.2‰) of both for ␦ C(n ϭ 93, df ϭ 89, F ϭ 12.2, Fcritical ␦13C values. The corresponding ␦18Ovalues ϭ 2.71, p K 0.001) and ␦18O(n ϭ 93, df ϭ 89; show a 2.4‰ range, indicating some slight F ϭ 9.0, Fcritical ϭ 2.71, p K 0.001). The statis- variability in habitat, ranging from relatively tical differences in both ␦13Cand␦18O are marine (30.1‰) to more mixed, or freshwater therefore interpreted to have paleobiological (27.7‰). The late Pleistocene Trichechus spec- significance related to, respectively, diets and imens indicate a wide range of diets among habitat preferences. individuals (a mean of 1.5‰ indicates sea- Extant Sirenians: Wild Populations and Captive grasses feeding in one individual, whereas a Specimens of Trichechus manatus and Wild Du- mean of Ϫ10.1‰ indicates a diet of principal- gong dugon. Carbon isotopic values for Re- 18 13 ly C3 plants in the other). The mean ␦ O val- cent, wild manatees have a mean ␦ Cof ues of 30.8 and 30.4‰ suggest relatively ma- Ϫ5.7‰ and broad range between Ϫ13.9‰ rine environments and the narrow range (2.0 and 0.5‰ (Table 4), indicating a mixed diet. and 0.9‰) suggests that the manatees inhab- The corresponding ␦18O mean is 29.1‰ with ited these environments for the time that the a range between 27.3‰ and 31.2‰, indicating teeth were mineralized. Several major points a mixture of freshwater to marine habitats. are evident from the data from the serial anal- These data demonstrate that modern Triche- yses: (1) the small amount of isotopic variation chus manatus from Florida fills a broad ecolog- predicted from the captive manatee fed on a ical niche ranging from marine with a sea- monotonous diet and living in a constant grass diet, to freshwater with a C3 diet (Figs. source of water is corroborated; (2) the extinct 3, 4). Captive Trichechus marks an end-mem- dugongid and fossil manatee Trichechus dem- ber, admittedly unnatural for a freshwater, C3 onstrated more isotopic variation, but this is feeder (Fig. 3), although some specimens ap- no more than 1.5‰ for ␦13C and 2‰ for ␦18O. proach marine ␦18O values owing to the prob- This suggests that the far greater degree of to- able enrichment of water in 18O from high tal isotopic variation (22.3‰ for ␦13Cand evaporation rates in the shallow tanks. 9.2‰ for ␦18O; Table 4, Appendix) seen in the For modern wild Australian dugongs, the bulk sample data presented elsewhere in this mean ␦13C of 1.2‰ and narrow range (Ϫ0.6‰ paper cannot be accounted for within the to 1.7‰) indicate a specialized seagrass diet demonstrated limits of individual isotopic (Table 4, Fig. 4), as is also empirically known variation that would be expected during the from studies of dugong diets (Walker 1975; time interval when the tooth mineralized. Husar 1978a; Marsh et al. 1984; Neitschmann Stable Isotopes, Ancient Diets, and Habitat Pref- 1984). Likewise, the mean ␦18O of 29.8‰ and erences. To determine differences in habitat narrow range (29.6 to 30.1‰) suggest a pre- as represented by ␦13Cand␦18O data, we used dominantly marine environment. These val- ANOVA to compare four groups of tooth ues for water and diet can be assumed to rep- enamel data for Florida sirenians: (1) Eocene– resent an end-member, given that dugongs early Oligocene Protosiren (n ϭ 4); (2) Miocene generally are strictly marine and predomi- Metaxytherium, including all samples of M. nantly eat seagrasses (Fig. 3). crataegense, M. floridanum,andM. cf. floridanum The ␦13Cand␦18O values from tooth enamel (n ϭ 26); (3) extinct ?Plio-Pleistocene Triche- carbonate indicate that Recent dugongs and chus sp. (n ϭ 45); and (4) modern wild T. man- manatees demonstrate a broad range of iso- atus (n ϭ 18) (Table 4; see raw data in Appen- topic variation that correlates with known di- dix). One specimen, Corystosiren varguezi (UF ets and habitat preferences of these extant si- 18826), from the early Hemphillian Waccasas- renians. Along with the conceptual model FLORIDA CENOZOIC SIRENIANS 311

TABLE 4. Univariate statistical data (n, mean, standard deviation, range) for different subsets of bulk sample stable carbon and oxygen isotopic data (as discussed in the text) for the fossil and extant sirenians from Florida. See Appendix for raw data.

Group n ␦13C (‰) ␦18O (‰) Eocene–early Oligocene Protosiren 4 3.4, 1.0, 1.9 to 4.1 30.2, 1.6, 28.3 to 32.2 All Metaxytherium 26 Ϫ1.7, 4.1, Ϫ13.1 to 2.8 29.3, 1.6, 25.1 to 31.0 M. crataegense 3 Ϫ1.4, 1.5, Ϫ3.0 to Ϫ0.1 28.0, 2.2, 25.4 to 29.6 Agricola M. floridanum 8 Ϫ1.0, 1.9, Ϫ4.8 to 0.9 29.1, 2.3, 25.1 to 30.7 Gainesville M.cf.floridanum 5 0.1, 1.0, Ϫ1.0 to 1.2 30.7, 0.3, 30.4 to 31.0 Later Bone Valley M. floridanum 10 Ϫ3.4, 6.1, Ϫ13.1 to 2.8 29.0, 0.5, 28.3 to 29.9 All Florida Protosirenidae and Dugongidae 31* Ϫ0.9, 4.3, Ϫ13.1 to 5.6 29.2, 1.9, 23.0 to 32.2 ?Plio-Pleistocene Trichechus sp. 45 Ϫ7.7, 5.5, Ϫ18.2 to 1.7 27.7, 1.6, 24.7 to 32.1 Leisey 3 Ϫ6.2, 2.3, Ϫ7.7 to Ϫ3.5 27.8, 1.7, 26.5 to 29.7 Santa Fe River sites 25 Ϫ9.9, 5.3, Ϫ18.2 to 1.7 27.3, 1.6, 24.7 to 32.1 Rock Springs 5 Ϫ5.9, 7.3, Ϫ14.3 to 0.1 29.1, 1.1, 28.1 to 30.4 Oklawaha River 12 Ϫ4.4, 3.5, Ϫ11.4 to Ϫ0.3 27.7, 1.6, 25.8 to 30.4 Recent, wild Trichechus manatus 18 Ϫ5.7, 5.1, Ϫ13.9 to 0.5 29.1, 0.9, 27.3 to 31.2 All Florida fossil and wild Trichechus 63 Ϫ7.2, 5.4, Ϫ18.2 to 1.7 28.1, 1.6, 24.7 to 32.1 Captive Trichechus manatus 6 Ϫ13.8, 0.9, Ϫ14.9 to Ϫ12.3 26.5, 1.4, 25.1 to 28.5 Wild Dugong dugon 6 1.2, 0.9, Ϫ0.6 to 1.7 29.8, 0.2, 29.6 to 30.1

* Includes one specimen of late Miocene Corystosiren varguezi, UF 18826 (Appendix).

(Fig. 3), these data are used below as the ref- Eocene-Oligocene Protosiren sp. (ca. 40 to 35 erence model of extant analogs to interpret the Ma). The primitive dugongid Protosiren is diets and habitat preferences of fossil sireni- relatively rare in the middle Cenozoic lime- ans from Florida. (These data are also plotted stones of Florida. Only four individual tooth in the background of Figures 5–7 along with specimens were available for isotopic analy- those for fossil sirenians as a means of inter- ses. The three Eocene Protosiren species and preting the paleoecology of the extinct one early Oligocene cf. Protosiren species have groups.) a combined mean ␦13C value of 3.4‰ with a

FIGURE 4. Plot of ␦13C versus ␦18O for all modern sirenians sampled in this study. All measurements were made from tooth enamel carbonate. Different ecologies grade between modeled end-members of ecological variation. Modern Dugong (white diamonds) is known to be a strictly marine, seagrass feeder, and captive Trichechus (black diamonds) are strictly freshwater, C3 feeders. Wild Trichechus (gray diamonds) is known to live in either habitat, i.e., ranging between the two end-members. 312 BRUCE J. MACFADDEN ET AL.

FIGURE 5. Plot of ␦13C against ␦18O for all teeth of Protosiren sampled in this study. Data from recent sirenians are also plotted for comparison. All measurements were made from tooth enamel carbonate. narrow observed range (Table 4, Fig. 4). Given though a specialized diet and habitat are in- the isotopic enrichment factor calculated for dicated from the isotopic data, this interpre- sirenians of 14.0‰, these data indicate that tation may be an artifact of the small sample Florida Protosiren sp. fed on vegetation with a size available for analyses. Well-preserved cra- mean isotopic value of about Ϫ11‰, suggest- nia of Florida Protosiren do not exist, but if ing that their diet consisted principally of sea- they did, it is predicted that they would have grasses. These results support previous obser- a relatively well developed flexure consistent vations (Ivany et al. 1990) of a direct associa- with a seagrass diet (Domning 2001a), prob- tion with Thalassia-like seagrasses and Proto- ably similar to Protosiren fraasi from the mid- siren sp. from the Dolime Quarry, middle dle Eocene of Egypt (Fig. 1B). Eocene Avon Park Formation of northern pen- Middle to Late Miocene Metaxytherium (Bar- insular Florida (Fig. 2). stovian to ?early Hemphillian, ca. 16–9 Ma) and The mean ␦18O value of 30.2‰ (range 28.3– Corystosiren (early Hemphillian, ca. 8 Ma). 32.2‰) for these same four Protosiren speci- Samples of Metaxytherium were analyzed from mens indicates an environmental water source four different localities during this study (Ta- enriched in 18O, an observation made all the ble 4, Appendix). The mean carbon isotopic more significant in light of the relatively de- values and very small observed ranges for pleted global ocean ␦18O interpreted for Eo- three of these samples, i.e., Barstovian M. cra- cene seawater (Zachos et al. 2001). In compar- taegense (mean ␦13C ϭϪ1.4‰), early Claren- ison with the data for recent sirenians, Proto- donian M. floridanum from the Bone Valley siren plots near the values for modern du- Agricola mine (mean ␦13C ϭϪ1.0‰), and late gongs (Fig. 5), suggesting that Protosiren lived Clarendonian M. cf. floridanum from the and fed in environments similar to that of Gainesville creeks (mean ␦13C ϭ 0.1‰) indi- modern dugongs. The isotopic data indicate cate that these Metaxytherium were specialized these early sirenians were principally marine feeders with a diet predominantly consisting (also like extant dugongs) and confirm pre- of seagrasses. Most plot very close to modern vious hypotheses that early during their his- dugongs (Fig. 6). The carbon isotopic data for tory, dugongs specialized on seagrasses (e.g., Clarendonian/early Hemphillian M. floridan- Domning 1981, 2001a; Ivany et al. 1990). Al- um from the later Bone Valley sequence sug- FLORIDA CENOZOIC SIRENIANS 313

FIGURE 6. Plot of ␦13C against ␦18O for all teeth of Metaxytherium sampled in this study. Metaxytherium data are separated according to locality and/or age of the fossils. Data from recent sirenians are also plotted for comparison. All measurements were made from tooth enamel carbonate.

gest a more diverse diet for this fourth suite samples yield a mean ␦18O of 29.3‰ (Table 4). of samples. The mean ␦13C is more negative Relative to the extant Trichechus specimens an- (mean ␦13C ϭϪ3.4‰) than for the preceding alyzed here, these data indicate an isotopically samples of Metaxytherium and the observed enriched water source, suggesting a principal- range is very interesting. Seven of the speci- ly marine habitat. The majority of the data, as mens have ␦13C values close to 0‰ (Fig. 6), in- indicated by the large grouping of individual dicating a diet of seagrasses. This is consistent points between approximately 28‰ and 31‰ with previous hypotheses about the diet of (Fig. 6), support this interpretation. However, Metaxytherium based on its dentition and high three data points, i.e., 25.4‰ (UF 36446) for M. degree of rostral deflection (e.g., Domning crataegense from the Barstovian Palmetto mine, 1988, 2001a); (Fig. 1C). However, the remain- and 25.1 and 25.8‰ (UF 206653, 206655) for ing three specimens, with ␦13CvaluesofϪ10.9 M. floridanum from the early Clarendonian (JW F7502), Ϫ12.4 (JW F7407E), and Ϫ13.1‰ Agricola Fauna, are among the lowest ␦18O (JW F7405A2) (Fig. 6, Appendix) indicate a values observed during this study and are significantly different diet. Using the isotopic similar to those for the captive manatees enrichment of 14.0‰ presented above, and known to live in full freshwater (Fig. 5). taking the mean for these three specimens Given the morphology of other fossil du- (Ϫ12.1‰), then the mean value of the ingest- gongs, and known diet and habitat preferenc- ed plant foods would be Ϫ26.1‰, indicating es of extant Dugong dugon, it is perhaps not a diet consisting predominantly of C3 plants. surprising that for the most part, Metaxyther- From the isotopic results alone, and in the ab- ium from the Miocene of Florida was a spe- sence of other corroborating evidence, it is not cialized seagrass feeder and lived in marine possible to further identify what specific waters. Exceptions to this general pattern in- kinds of C3 plants these Metaxytherium indi- clude (1) three specimens indicating a diet of viduals ate when their tooth enamel mineral- C3 plants; and (2) three other specimens living ized. in more freshwater habitats, although grazing The oxygen data for all four Metaxytherium predominantly on seagrasses. 314 BRUCE J. MACFADDEN ET AL.

FIGURE 7. Plot of ␦13C versus ␦18O for all fossil teeth of Trichechus sampled in this study. Fossil Trichechus are sep- arated according to the locality from which the fossil was collected. Data from recent sirenians are also plotted for comparison. All measurements were made from tooth enamel carbonate.

Corystosiren is a late-surviving dugong at a vidual was feeding not on seagrasses (as was time when sirenian fossils are exceedingly suggested because of its thick enamel), but rare in Florida. A single specimen of C. var- rather on C3 vegetation. This same sample guezi from the early Hemphillian Waccasassa yielded a ␦18O ϭ 26.3‰, suggesting a princi- River yielded a relatively enriched ␦13Cvalue pally freshwater habitat. of 5.6‰ and a very depleted ␦18O value of Representatives of the Family Trichechidae 23.0‰ (Fig. 6 [lower right corner], Appendix). apparently dispersed into Florida either dur- These data suggest that this dugong was feed- ing the late Pliocene or early Pleistocene. The ing on seagrasses, but living in predominantly first definite, in situ occurrence of manatees, freshwater. If so, then this ecological niche is referred to here as Trichechus sp., which is not commonly found in other fossil or extant probably a species different from T. manatus dugongs. It is obvious that a single data point (Domning 1982; Hulbert 2001), comes from may not be representative of the overall aut- the 1.5 Ma Irvingtonian Leisey Shell Pit (Fig. ecology of this species. Unfortunately, more 2). The mean ␦13C value of Ϫ6.2‰ for Leisey specimens of the rare C. varguezi are not avail- Trichechus sp. indicates a mixed diet and the able for additional isotopic analyses. mean ␦18O of 27.8‰ indicates mixed fresh- ?Late Pliocene and Pleistocene Trichechus sp. water and marine habitats (Table 4). The manatees (Family Trichechidae) are be- A large sample suite (n ϭ 25) comes from lieved to have originated in South America several localities along the bed of the Santa Fe during the middle Tertiary (Domning 1982), River in northern Florida. Given the age span with the earliest fossil record of this sirenian of the associated mammalian fauna, these fos- family represented by Potamosiren magdalensis sils almost certainly represent a temporally (Domning 1997b) from the late middle Mio- mixed assemblage. The age range might be as cene (ca. 10 Ma) of Colombia. A single speci- old as late Pliocene (middle Blancan), but it men (INGEOMINAS IGM 250927; isotopic surely spans much of the Pleistocene, includ- sample BJMacF 96-70) of this early manatee ing the late Blancan and Irvingtonian NAL- from the La Venta Fauna yielded a ␦13Cvalue MAs. These samples may represent a mixing of Ϫ13.3‰ (Fig. 7), indicating that this indi- of genetically independent populations that FLORIDA CENOZOIC SIRENIANS 315 migrated into this region at different times Temporal Variation and Evolution of Diet and (Domning 1982, 2001a). Some localities along Habitat. To understand temporal variation the Santa Fe River have better temporal reso- and patterns in the stable isotopic data for the lution than others. Santa Fe River localities are samples of fossil and extant Florida sirenians divided for the purposes of plotting when a presented in this paper, we analyzed the more precise age may be estimated, but all are mean and individual ␦13Cand␦18O data with considered together in the statistical analysis. respect to time interval or zone. It should be The isotopic data have a mean ␦13C value of noted that some of the time intervals or zones Ϫ9.9‰, but the range is perhaps more inter- containing sirenian data are approximate and esting (Table 4, Fig. 7). At one end of the spec- their exact ages as plotted represent intervals trum, the minimum ␦13C value of Ϫ18.2‰ is in which ages are estimated by stage of evo- exceedingly depleted and suggests an indi- lution biochronology within a particular vidual feeding predominantly on C3 plants NALMA. As such, these age determinations that lived in closed-canopy communities (ap- should be considered approximate (to within proximately Ϫ32‰; [van der Merwe and Me- Ϯ105 to 106 years for the pre-Pleistocene data dina 1991]). At the other end of the spectrum, and 105 to 104 years for Pleistocene data). Also, 13 ␦ C values as enriched as 1.7‰ indicate in- there are two important intervals in which iso- dividuals feeding on seagrasses, a principal topic samples were rare and consequently not food source similar to that of modern Triche- available for this study (ca. 30 to 20 Ma) or it chus manatus from Florida. seems that extinct sirenians were exceedingly Fossil manatees were analyzed from two rare in the Florida fossil record (ca. 5 to 2 Ma). other late Pleistocene sites, Rock Springs and The tooth enamel carbon isotopic data for the Oklawaha River. In contrast to the tem- Eocene and Oligocene (before 30 Ma on Fig. 8) poral and possible genetic mixing that char- Protosiren yield relatively enriched ␦13Cvalues acterize the Santa Fe River specimens, these interpreted to represent a specialized seagrass other sites seem relatively homogeneous, with diet. Likewise, most of the Metaxytherium Rock Springs being about 125,000 years old specimens are relatively enriched in ␦13C, also (Wilkins 1983) and the Oklawaha being latest indicating a specialized seagrass diet, consis- Rancholabrean judging from its associated tent with the previously observed morpholog- fauna (MacFadden personal observation ical adaptation of a high degree of rostral de- 2001). The Rock Springs sample demonstrates flection (e.g., Domning 1977, 2001b). Three a broad range of ␦13C values from Ϫ14.3‰ to specimens of Metaxytherium that are relatively 0.1‰ that appears to represent a bimodal dis- negative in ␦13C indicate individuals feeding tribution of values clustering at either extreme on a principally C diet. A single specimen of (Table 4, Fig. 7). This is interpreted as some in- 3 Corystosiren with a ␦13C value of 5.6‰ is the dividuals feeding on C3 plants whereas others were feeding on seagrasses. The Oklawaha most enriched of all values analyzed in this sample likewise has a broad range of ␦13C val- study and likewise suggests a specialized sea- 13 ues, from Ϫ11.4 to Ϫ0.3‰, suggesting both grass diet. The mean ␦ C values for fossil and mixed feeding as well as some specialization extant wild Trichechus are more negative, and on seagrass. the individual data points have a broader 13 Taken together, the ␦18O data for fossil and range of ␦ C values indicating a more diverse extant wild Trichechus sp. from Florida indi- diet relative to Protosiren, Metaxytherium,or cates a considerable spread and continuous extant Dugong. variation from 24.7‰ to 32.1‰ (Table 4, Fig. The tooth enamel oxygen isotopic data for 7). These data are interpreted to indicate a Eocene and Oligocene (before 30 Ma on Fig. 9) range of habitats from full freshwater, pre- Protosiren indicate relatively enriched ␦18O, sumably peninsular rivers and springs, to full which, given the estimated values of global marine waters. This range of habitat prefer- early Cenozoic seawater (Zachos et al. 2001), ences is similar to that of extant Trichechus indicates that Protosiren was living principally manatus. in marine habitats. This also is the case for the 316 BRUCE J. MACFADDEN ET AL.

FIGURE 8. Plot of ␦13C versus time for the bulk samples of Florida sirenians presented in this paper. This includes mean ␦13C values (Table 4) for time horizons or intervals represented by large circles and individual ␦13C data points presented (Appendix). Note the change in timescale and intervals for which samples were not available for analysis. A single mean value for age and ␦13C has been calculated for all Protosiren analyses from the approximately 40 and 35 Ma levels.

Miocene (20 to 5 Ma interval) as represented tocene, as represented by Trichechus, the mean by the majority of the Metaxytherium speci- values at about 1.5, 1.25, 0.25, 0.1, and 0.025 mens analyzed. On the other hand, three spec- Ma are slightly less positive than for the Eo- imens of Metaxytherium (different from the cene through Oligocene, but the variance three outliers for the ␦13C data described around the mean is perhaps more informative above) and the one analysis for Corystosiren and interesting. The broader spread of indi- are relatively depleted in ␦18O. We interpret vidual data points is interpreted as a diversi- these four data points to represent individuals fication of habitats ranging from full marine to living in relatively freshwater environments full freshwater (spread of approximately 8‰) (although it is possible that these could rep- for Trichechus relative to either Protosiren or resent an artifact of diagenesis). For the Pleis- Metaxytherium. FLORIDA CENOZOIC SIRENIANS 317

FIGURE 9. Plot of ␦18O versus time for the bulk samples of Florida sirenians presented in this paper. This includes mean ␦18O values (Table 4) for time horizons or intervals represented by large circles and individual ␦18O data points presented (Appendix). Note the change in time scale and intervals for which samples were not available for analysis. A single mean value for age and ␦18O has been calculated for all Protosiren analyses from the approximately 40 and 35 Ma levels.

Concluding Remarks Florida (Ivany et al. 1990) and the presence of Fossil discoveries and morphological inter- shallow marine or estuarine faunas. Stable iso- pretations over the past several decades have topic analyses of tooth enamel add another done much to advance our understanding of line of evidence that allows further testing of fossil sirenians, including those from the rich these paleoecological hypotheses. Cenozoic sequence in Florida. The diet of ex- The isotopic data presented here (Table 4) tinct sirenians previously has been interpreted for the Eocene to late Miocene Protosiren and from morphological characters such as the de- dugongs (i.e., extinct species of the Family Du- gree of rostral deflection (e.g., Domning gongidae; Table 1) from Florida differ signifi- 2001a). The habitat preferences of extinct si- cantly from those of fossil and wild Trichechus renians have been interpreted from associated for both ␦13C(T-statistic ϭ 5.7, T-critical ϭ 1.7, fossils such as Eocene seagrasses from central p K 0.001) and ␦18O(T-statistic ϭ 2.8, T-crit- 318 BRUCE J. MACFADDEN ET AL. ical ϭ 1.7, p ϭ 0.006). These isotopic data are conspicuous in their paucity of sirenians. This interpreted to represent differences in the eco- evidence suggests that other than a few rare logical adaptations of Protosiren and extinct occurrences (Corystosiren varguezi and unde- Florida dugongs relative to extinct and extant scribed dugongids, sensu Domning 2001a and manatees. As a general pattern, Protosiren and Hulbert 2001), dugongids had essentially be- the extinct Florida dugongs had a more spe- come extinct in Florida by about 5 Ma (i.e., the cialized diet of seagrass and a predominantly age of the Upper Bone Valley Formation). marine habitat. In contrast, fossil manatees Despite the rich fossil record of sirenians in and modern Trichechus manatus in Florida have Florida, it also is unfortunately inadequate been more generalized feeders, with a broader during other critical time intervals. The max- isotopic range representing diets of C3 plants imum diversity of dugongs in Florida oc- to seagrasses. In general, the isotopic data re- curred during the late Oligocene and early ported here confirmed previous hypotheses Miocene with three coexisting taxa, Metaxy- based on morphological and associated faunal therium sp., Crenatosiren olseni,andDioplother- evidence. There were, however, a few surpris- ium manigaulti, as represented at localities es; e.g., Metaxytherium crataegense fed on sea- such as White Springs in northern Florida grass despite a low degree of rostral deflec- (Domning 1989a). Although all three species tion, and three specimens of M. floridanum are considered to have fed principally on sea- principally fed on C3 plants. grasses, judging from degree of rostral deflec- The results presented here indicate that Pro- tion, tusk morphology, and estimated body tosiren and extinct dugongs from Florida were size, Domning (2001a) interprets some niche relatively conservative in their adaptations, differentiation for Metaxytherium sp., C. olseni, maintaining diet and habitat preferences sim- and D. manigaulti. With regard to diet, these ilar to those of modern Dugong dugon.Incon- specializations pertain to the size of the sea- trast, Trichechus has been considerably more grass rhizome upon which each dugong spe- generalized both in diet and habitat prefer- cies would have fed. If specimens were avail- ence since it first appeared in Florida about 2 able for these extinct dugongs, stable carbon Ma. Both dugongs and manatees have filled isotopes could confirm that they fed upon sea- the ‘‘aquatic megaherbivore niche’’ (Domning grasses but would not be able to further re- 2001a) or paleoguild (sensu van Valkenburgh solve the hypotheses about specific speciali- 1995). It therefore has been interesting to spec- zation on different-sized rhizomes. ulate whether these two sirenian groups com- After dugongids became rare in the late peted, or whether the manatees filled an Miocene, there is a roughly 3-Myr gap (ca. 5 ‘‘empty’’ niche created by the extinction of the to 2 Ma) in which sirenians are virtually ab- late Miocene dugongs. Although the fossils of sent in Florida, despite the presence of appro- sirenians in Florida are abundant, the critical priate-aged (Blancan) faunas containing time interval, i.e., late Miocene to Pliocene (ca. aquatic elements. The available fossil evidence 10 to 3 Ma) is poorly represented and there- indicates that primitive trichechids originated fore does not allow testing of the competition in the New World, with fossil occurrences of versus replacement hypotheses for these Potamosiren magdalensis from the late Miocene aquatic megaherbivores. We suspect, however, La Venta Fauna of Colombia and Rhibodon that this niche was vacated by dugongs and from the Pliocene of Argentina and North then filled several million years later by man- Carolina (Domning 1982, 1997b). By the time atees. Our reasons for this relate to the fact that Trichechus first appears in the Florida se- that Hemphillian marine/terrestrial mam- quence by the early Pleistocene, it is similar to malian faunas are very well represented in the T. manatus, although Domning (1982) consid- Florida sequence, including those of the Up- ers the fossil manatee a different, heretofore per Bone Valley (e.g., Palmetto Fauna [Morgan undescribed species. 1994]). These faunas contain both diagnostic The stable isotopic analyses were successful Hemphillian land mammals and a diverse in reconstructing the paleobiology of Florida marine mammal fauna (Morgan 1994) but are sirenians. As a modern baseline, studies of FLORIDA CENOZOIC SIRENIANS 319 captive Trichechus manatus were important in data on the food eaten by Hugh and Buffett. establishing a carbon isotopic enrichment val- D. Colbert of the Mote Marine Laboratory, Sar- ue (␧*) of 14.0‰, essentially the same calcu- asota, provided samples of naturally shed lated for medium to large terrestrial mam- teeth and food of Hugh and Buffett. S. C. malian herbivores (Cerling and Harris 1999). White and C. Audette of the South Florida Likewise, the extant sample of Australian Du- Museum, Bradenton, gave us naturally shed gong dugon formed a baseline for comparison teeth and food of Snooty. Most of the fossil with extinct Florida dugongs. Carbon isotopic specimens analyzed here are from the Uni- analyses were useful in interpreting the an- versity of Florida, Florida Museum of Natural cient diets of extinct sirenians, as described History Vertebrate Paleontology collection. In above. In addition, although there is a com- addition, we thank J. Waldrop of Lake Wales plex isotopic source, which is dependent upon and R. Madden, Duke University, for other several environmental and physiological fac- fossil sirenian specimens analyzed during this tors, the resulting ␦18O analyzed in tooth study. We have benefited greatly from the ad- enamel can be used to discriminate habitat vice, consultations, and assistance of C. A. 18 preferences. For Florida sirenians, ␦ Ovalues Beck, U.S. Geological Survey Sirenian Project, range from relatively freshwater (depleted Gainesville; T. E. Cerling, C. Cook, and B. Pas- 18 ␦ O), representing inland rivers and springs, sey, University of Utah; J. Curtis, University of 18 to full marine (enriched ␦ O), representing Florida; D. P. Domning, Howard University; P. shallow, coastal habitats. Koch, University of California-Santa Cruz; C. Although fossils provide unparalleled clues McCaffrey and L. Wilkins, Florida Museum of to ancient life, the fossil record is far from Natural History Mammal Collection; and B. complete, and new discoveries can always ad- Beatty, R. C. Hulbert, and J. O’Sullivan, Flor- vance our understanding of any group. Such ida Museum of Natural History Vertebrate Pa- is the case with fossil sirenians from Florida. leontology Collection. Isotope samples were Additional specimens of the late Oligocene– prepared in the laboratory of T. Cerling (Uni- early Miocene diversity, better knowledge of versity of Utah), D. Dilcher (University of Flor- the rare, late-surviving Mio-Pliocene du- ida), and P. Koch (University of California- gongs, and discovery of in situ Pliocene tri- Santa Cruz) and analyzed in the University of chechids would all go a long way toward ad- Florida Department of Geological Sciences vancing our understanding of the phylogenet- Stable Isotope Laboratory; SIRFER laboratory ic history and paleobiogeography of this in- in the Department of Biology, University of teresting group. In addition, the application of Utah; or Earth Sciences, University of Califor- stable isotopic analyses of tooth enamel car- nia-Santa Cruz. The manuscript was im- bonate could be used in conjunction with these future discoveries to further refine our proved thanks to the comments from the two understanding of the diets and habitat pref- reviewers. This research was partially sup- erences of the extinct Sirenia of Florida. ported by funds provided from National Sci- ence Foundation grant EAR 0087742 and Geo- Acknowledgments logical Society of America grant 1998. This is Teeth from extant Trichechus manatus, either University of Florida Contribution to Paleo- naturally shed or in existing collections, were biology number 529. analyzed under the permission of U.S. Fish Literature Cited and Wildlife permit MA033974-0 to the Flor- ida Museum of Natural History (B. J. Mac- Ambrose, S. 1998. Implications of 24 controlled diet experi- Fadden, permit director). We thank M. Farris ments for diet reconstruction with carbon isotopes of bone collagen and carbonate. Journal of Vertebrate Paleontology, for her assistance with this permit. D. Kwan of Abstracts of Papers 18:24A. James Cook University, Townsville, Australia, Ames, A. L., E. S. Van Vleet, and W. M. Sackett. 1996. The use kindly provided specimens of modern du- of stable carbon isotope analysis for determining the dietary habits of the Florida manatee, Trichechus manatus latirostris. gong. A. Ames of the University of South Flor- Marine Mammal Science 12:555–563. ida, Tampa, provided us with unpublished Andersen, S. H., and E. Nielsen. 1983. Exchange of water be- 320 BRUCE J. MACFADDEN ET AL.

tween the harbor porpoise, Phocoena phocoena, and the envi- leontology honoring Frank C. Whitmore, Jr. Proceedings of ronment. Experientia 39:52–53. the San Diego Society of Natural History 29:177–189. Billups, K., and D. P. Schrag. 2002. Paleotemperatures and ice ———. 1997a. Sirenia. Pp. 383–391 in R. F. Kay et al., eds. Ver- volume of the past 27 Myr revisited with paired Mg/Ca and tebrate paleontology in the Neotropics: the Miocene Fauna of 18O/16O measurements on benthic foraminifera. Paleocean- La Venta, Colombia. Smithsonian Institution Press, Washing- ography 17:1–11. ton, D.C. Bocherens, H., P. L. Koch, A. Mariotti, D. Geraads, and J.-J. Jae- ———. 1997b. Fossil Sirenia of the West Atlantic and Caribbean ger. 1996. Isotope biogeochemistry (13C, 18O) of mammalian Region. VI. Crenatosiren olseni (Reinhart, 1976). Journal of Ver- enamel from African Pleistocene hominid sites. Palaios 11: tebrate Paleontology 17:397–412. 306–318. ———. 1999. Fossils explained 24: Sirenians (seacows). Geology Boutton, T. W. 1991. Stable carbon isotope ratios of natural ma- Today 15:75–79. terials. II. Atmospheric, terrestrial, marine, and freshwater ———. 2001a. Evolution of the Sirenia and Desmostylia. Pp. environments. Pp. 173–195 in D. C. Coleman and B. Fry, eds. 151–168 in J.-M. Mazin and V. de Buffre´nil, eds. Secondary Carbon isotope techniques. Academic Press, San Diego. adaptation of tetrapods to life in the water: proceedings of the Bryant, J. D. 1991. New early Barstovian (middle Miocene) ver- international meeting, Poitiers, 1996. Verlag Dr. Friedrich tebrates from the upper Torreya Formation, eastern Florida Pfeil, Munich. panhandle. Journal of Vertebrate Paleontology 11:472–489. ———. 2001b. Sirenians, seagrasses, and Cenozoic ecological Bryant, J. D., and P. N. Froelich. 1995. A model of oxygen isotope change in the Caribbean. Palaeogeography, Palaeoclimatolo- fractionation in body water of large mammals. Geochimica et gy, Palaeoecology 166:27–50. Cosmochimica Acta 59:4523–4537. ———. 2001c. The earliest known fully quadrupedal sirenian. Bryant, J. D., P. N. Froelich, W. J. Showers, and B. J. Genna. 1996. Nature 413:625–627. A tale of two quarries: biological and taphonomic signatures Domning, D. P., and L. C. Hayek. 1984. Horizontal tooth re- in the oxygen isotopic composition of tooth enamel phos- placement in the Amazonian manatee (Trichechus inunguis). phate from modern and Miocene equids. Palaios 11:397–408. Mammalia 48:105–127. Cerling, T. E., and J. M. Harris. 1999. Carbon isotope fraction- Domning, D. P., G. S. Morgan, and C. E. Ray. 1982. North Amer- ation between diet and bioapatite in ungulate mammals and ican Eocene sea cows (Mammalia: Sirenia). Smithsonian Con- implications for ecological and paleoecological studies. Oec- tributions to Paleobiology 52:1–69. ologia 120:347–363. Ehleringer, J. R., R. F. Sage, L. B. Flanagan, and R. W. Pearcy.

Cerling, T. E., and Z. Sharp. 1997. Stable carbon and oxygen iso- 1991. Climate change and the evolution of C4 photosynthesis. tope analysis of fossil tooth enamel using laser ablation. Pa- Trends in Ecology and Evolution 6:95–99. laeogeography, Palaeoclimatology, Palaeoecology 126:173– Farquhar, G. D., J. R. Ehleringer, and K. T. Hubrick. 1989. Carbon 186. isotope discrimination and photosynthesis. Annual Review of Clementz, M., and P. Koch. 2001. Differentiating aquatic mam- Plant Physiology and Plant Molecular Biology 40:503–537. mal habitat and foraging ecology with stable isotopes in tooth Gremillion, P. T. 1994. Separation of streamflow components in enamel. Oecologia 129:461–472. the Econlockhatchee River System using environmental stable Coplen, T. B. 1994. Reporting of stable hydrogen, carbon, and isotope tracers. Ph.D. dissertation. University of Central Flor- oxygen isotopic abundances. Pure and Applied Chemistry 66: ida, Orlando. 273–276. Hemminga, M. A., and M. A. Mateo. 1996. Stable carbon iso- Coplen, T. B., and C. Kendall. 2000. Stable hydrogen and oxygen topes in seagrasses: variability in ratios and use in stable iso- isotope ratios for selected sites of the U.S. Geological Survey’s topes. Marine Ecology Progress Series 140:285–298. NAS-QAN and benchmark surface-water networks. Open-file Hobson K. A., J. F. Piatt, and J. Pitocchelli. 1994. Using stable Report 00–160. U.S. Geological Survey, Reston, Va. isotopes to determine seabird trophic relationships. Journal of Dienes, P. 1980. The isotopic composition of reduced organic Animal Ecology 63:786–798. carbon. Pp. 329–406 in P. Fritz and J. C. Fontes, eds. Handbook Hoefs, J. 1997. Stable isotope geochemistry. Springer, Berlin. of environmental isotope geochemistry. 1. The terrestrial en- Hui, C. A. 1981. Seawater consumption and water flux in the vironment. Elsevier, Amsterdam. common dolphin Delphinus delphis. Physiological Zoology 54: Domning, D. P. 1977. An ecological model for late Tertiary si- 430–440. renian evolution in the North Pacific Ocean. Systematic Zo- Hulbert, R. C. 1988. Cormohipparion and Hipparion (Mammalia, ology 25:352–362. Perissodactyla, Equidae) from the late Neogene of Florida. ———. 1981. Sea cows and sea grasses. Paleobiology 7:417–420. Bulletin of the Florida State Museum (Biological Sciences) 33: ———. 1982. Evolution of manatees: a speculative history. Jour- 229–338. nal of Paleontology 56:599–619. ———, ed. 2001. The fossil vertebrates of Florida. University ———. 1988. Fossil Sirenia of the West Atlantic and Caribbean Press of Florida, Gainesville. Region. I. Metaxytherium floridanum. Journal of Vertebrate Pa- Husar, S. L. 1978a. Dugong dugon. Mammalian Species 88:1–7. leontology 8:395–426. ———. 1978b. Trichechus manatus. Mammalian Species 93:1–5. ———. 1989a. Fossil sirenians from the Suwannee River, Florida Ivany, L. C., R. W. Portell, and D. S. Jones. 1990. Animal-plant and Georgia. Pp. 54–60 in G. S. Morgan, ed. Miocene pale- relationships and paleobiogeography of an Eocene seagrass ontology and stratigraphy of the Suwannee River basin of community from Florida. Palaios 5:244–258. north Florida and south Georgia. Guidebook 30, Southeastern Keeley, J. E., and D. R. Sandquist. 1992. Carbon: freshwater Geological Society, Tallahassee. plants. Plant, Cell and Environment 15:1021–1035. ———. 1989b. Fossil Sirenia of the West Atlantic and Caribbean Koch, P. L. 1998. Isotopic reconstruction of past continental en- Region. II. Dioplotherium manigaulti Cope, 1883. Journal of Ver- vironments. Annual Review of Earth and Planetary Sciences tebrate Paleontology 9:415–428. 26:573–613. ———. 1990. Fossil Sirenia of the West Atlantic and Caribbean Koch, P. L., J. C. Zachos, and P. D. Gingerich. 1992. Correlation Region. IV. Corystosiren varguezi, gen. et sp. nov. Journal of between isotope records in marine and continental carbon Vertebrate Paleontology 10:361–371. reservoirs near the Palaeocene/Eocene boundary. Nature 358: ———. 1994. A phylogenetic analysis of the Sirenia. In A. Berta 319–322. andT.A.Deme´re´, eds. Contributions in marine mammal pa- Koch, P. L., N. Tuross, and M. L. Fogel. 1997. The effects of sam- FLORIDA CENOZOIC SIRENIANS 321

ple treatment and diagenesis on the isotopic integrity of car- Ortner, P. B., T. N. Lee, P. J. Milne, R. G. Zidka, M. E. Clarke, G. bon in biogenic hydroxylapatite. Journal of Archaeological P. Podesta, P. K. Swart, P. A. Testa, L. P. Atkinson, and W. R. Science 24:417–429. Johnson. 1995. Mississippi River flood waters reached the Kohn, M. J. 1996. Predicting animal ␦18O: accounting for diet Gulf Stream. Geophysical Research Letters 100:13,595– and physiological adaptation. Geochimica et Cosmochimica 13,601. Acta 60:4811–4829. Reinhart, R. H. 1976. Fossil sirenians and desmostylids from Langeland, K. A. 1990. Hydrilla: a continuing problem in Florida Florida and elsewhere. Bulletin of the Florida State Museum waters. Circular No. 884. Cooperative Extension Service/In- (Biological Sciences) 20:187–300. stitute of Food and Agricultural Sciences, University of Flor- Roe, L. J., J. G. M. Thewissen, J. Quade, J. R. O’Neil, S. Bajpai, A. ida, Gainesville. Sahni, and S. T. Hussain. 1998. Isotopic approaches to under- Lear, C. H., H. Elderfield, and P. A. Wilson. 2000. Cenozoic standing the terrestrial-to-marine transition of the earliest deep-sea temperatures and global ice volumes from Mg/Ca whales. Pp. 399–422 in J. G. M. Thewissen, ed. The emergence in benthic foraminiferal calcite. Science 287:269–272. of whales. Plenum, New York. MacFadden, B. J., and T. E. Cerling. 1996. Mammalian herbivore Savage, R. J. G., D. P. Domning, and J. G. M. Thewissen. 1994. communities, ancient feeding ecology, and carbon isotopes: a Fossil Sirenia of the west Atlantic and Caribbean region. V. 10 million-year sequence from the Neogene of Florida. Journal The most primitive known sirenian, Prorastomus sirenoides of Vertebrate Paleontology 16:103–115. Owen, 1855. Journal of Vertebrate Paleontology 14:427–449. MacFadden, B. J., T. E. Cerling, and J. Prado. 1996. Cenozoic ter- Schoeninger, M. J., and M. J. DeNiro. 1982. Carbon isotope ratios restrial ecosystem evolution in Argentina: evidence from car- of apatite from fossil bone cannot be used to reconstruct diets bon isotopes of fossil mammal teeth. Palaios 11:319–327. of animals. Nature 297:577–578. MacFadden, B. J., T. E. Cerling, J. M. Harris, and J. Prado. 1999. Simenstad, C. A., D. O. Duggins, and P. D. Quay. 1993. High turnover of inorganic carbon in kelp habitats as a cause of ␦13C Ancient latitudinal gradients of C3/C4 grasses interpreted from stable isotopes of New World Pleistocene horse (Equus) variability in marine food webs. Marine Biology 116:147–160. teeth. Global Ecology and Biogeography 8:137–149. Simpson, G. G. 1932. Fossil Sirenia of Florida and the evolution Marsh, H. 1986. The status of the dugong in Torres Strait. Pp. of the Sirenia. Bulletin of the American Museum of Natural 53–76 in A. K. Haines, G. C. Williams, and D. Coates, eds. Tor- History 59:419–503. res Straits fisheries seminar, Port Moresby. Australian Gov- Swart, P. K., L. Sternberg, R. Steinen, and S. A. Harrison. 1989. Controls on the oxygen and hydrogen isotopic composition ernment Publishing Service, Canberra. of waters from Florida Bay. Chemical Geology 79:113–123. Marsh, H., and W. K. Saalfeld. 1991. The status of the dugong Swart, P. K., R. E. Dodge, and H. J. Hudson. 1996. A 240-year in Torres Strait. In D. Lawrence and T. Cansfield-Smith, eds. stable oxygen and carbon isotope record in a coral from South Sustainable development of traditional inhabitants of the Tor- Florida: implications for the prediction of precipitation in res Strait region. GBRMPA Workshop Series 10:187–194. southern Florida. Palaios 11:362–375. Marsh, H., G. E. Heinsohn, and L. M. Marsh. 1984. Breeding cy- Tedford, R. H., M. F. Skinner, R. W. Fields, J. M. Rensberger, D. cle, life history and population dynamics of the dugong, Du- P. Whistler, T. Galusha, B. E. Taylor, J. R. Macdonald, and S. gong dugon (Sirenia: Dugongidae). Australian Journal of Zo- D. Webb. 1987. Faunal succession and biochronology of the ology 32:767–788. Arikareean through Hemphillian interval (late Oligocene McCrea, J. M. 1950. On the isotopic chemistry of carbonates and through earliest Pliocene epochs) in North America. Pp. 153– a paleotemperature scale. Journal of Chemical Physics 18:849– 210 in M. O. Woodburne, ed. Cenozoic mammals of North 857. America: geochronology and biostratigraphy. University of McKenna, M. C., and S. K. Bell. 1997. Classification of mammals California Press, Berkeley. above the species level. Columbia University Press, New York. van der Merwe, N. J., and E. Medina. 1991. The canopy effect, Meyers, J., P. K. Swart, and J. Meyers. 1993. Geochemical evi- carbon isotope ratios and foodwebs in Amazonia. Journal of dence for groundwater behavior in an unconfined aquifer, Archaeological Science 18:249–259. South Florida. Journal of Hydrology 148:249–272. van Valkenburgh, B. 1995. Tracking ecology over geological Morgan, G. S. 1994. Miocene and Pliocene marine mammal fau- time: evolution within guilds of vertebrates. Trends in Ecol- nas from the Bone Valley Formation of central Florida. In A. ogy and Evolution 10:71–75. Berta and T. A. Deme´re´, eds. Contributions in marine mam- Walker, E. P. 1975. Mammals of the world, 3d ed. Johns Hopkins mal paleontology honoring Frank C. Whitmore, Jr. Proceed- University Press, Baltimore. ings of the San Diego Society of Natural History 29:239–268. Wilkins, K. T. 1983. Pleistocene mammals from the Rock Springs Morgan, G. S., and R. C. Hulbert. 1995. Overview of the geology Local Fauna, central Florida. Brimleyana 9:69–82. and vertebrate biochronology of the Leisey Shell Pit Local Yoshida, N., and N. Miyazaki. 1991. Oxygen isotope correlation Fauna, Hillsborough County, Florida. Bulletin of the Florida of cetacean bone phosphate with environmental water. Jour- Museum of Natural History 37:1–92. nal of Geophysical Research 96(C1):815–820. Nietschmann, B. 1984. Hunting and ecology of dugongs and Zachos, J., M. Pagani, L. Sloan, E. Thomas, and K. Billups. 2001. green turtles, Torres Strait, Australia. National Geographic Trends, rhythms, and aberrations in global climate 65 Ma to Society Research Report 17:625–651. present. Science 292:686–693. 322 BRUCE J. MACFADDEN ET AL. rsity of O (‰) 30.0 30.3 28.3 29.6 29.1 25.4 30.3 30.5 30.5 30.7 25.1 29.6 25.8 30.5 29.9 28.7 29.3 29.4 28.5 28.7 28.6 28.8 28.3 29.3 30.4 30.7 30.5 31.0 31.0 18 in particular ␦ C (‰) 4.1 4.1 1.9 0.1 1.1 3.0 0.6 0.2 0.7 0.9 4.8 2.1 1.4 0.9 0.4 0.1 1.3 2.8 0.1 0.6 0.7 0.7 1.0 0.2 1.2 1.2 13 10.9 12.4 13.1 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ ␦ Ϫ Ϫ Ϫ ion, now part of UF; UF, Unive O data calibrated to V-PBD were corrected to (presented V-SMOW 18 ␦ ozoic Florida sirenians analyzed during the present study. Ages, in Waccasassa River Dolime Quarry Waccasassa River La Camelia La Camelia Bone Palmetto Valley, Bone Agricola Valley, Fauna Bone Agricola Valley, Fauna Bone Agricola Valley, Fauna Bone Agricola Valley, Fauna Bone Valley-4 corners-0 Bone Valley-4 corners-0 Bone Valley-IMC-Hopewell Bone Agricola Valley, Fauna Bone Tiger Valley, Bay-1 Bone Valley-Kingsford-0 Bone Valley-Kingsford-0 Bone Valley-Kingsford-0 Bone Valley-Kingsford-0 Bone Valley-Kingsford-0 Bone Valley-Kingsford-0 Bone Tiger Valley, Bay-0 Bone Phosphoria-0 Valley, Bone Phosphoria-0 Valley, Gainesville creeks Gainesville creeks Gainesville creeks Gainesville creeks Gainesville creeks mples of Cen wing Hoefs (1997), Appendix 1 LM? rm3 M2 or M3 molar molar molar partial tooth RM3 frag cheek tooth molar frag molar frag RM3 RM3 molar frag molar molar molar molar molar molar molar molar molar molar partial tooth partial molar molar partial tooth molar ion. Follo 3(.91). ϩ egense egense egense ) floridanum floridanum floridanum floridanum floridanum cf. cf. cf. cf. cf. V-PDB sp. tooth frag Terra Mar 3.3 32.3 O 18 sp. sp. sp. ␦ Protosiren 1.03091 ( Protosiren Protosiren Protosiren Metaxytherium crata Metaxytherium crata Metaxytherium crata Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium floridanum Metaxytherium Metaxytherium Metaxytherium Metaxytherium Metaxytherium ϭ V-SMOW O 18 ␦ ocality, age, and isotopic data for the tooth enamel carbonate bulk sa Museum# Taxon Tooth Locality UF 24806 UF 115722 UF 25708 UF uncatalogedUF 182432 UF 182433 UF 36446 cf. FGS V-3512 FGS V-10056 UF 28829 UF 28833 UF 206653 UF 206654 UF 206655 UF uncataloged UF 28816 JW F7502 JW F7502 JW F6912 JW F6914 JW F7502A JW F6914 JW F7407E JW F7405A1 JW F7405A2 UF 97338 UF 115953 UF 8023 UF 58098 UF uncataloged Taxonomic, l million years (Ma), areinterval. approximate Abbreviations: mid-points JW, within private given collectionFlorida intervals. of Vertebrate Paleontology These Collection; John UF-M; mid-point UF Waldrop, Lake ages Mammal Collect are Wales, Florida; not FGS-V, to Florida be Geological taken Survey as Collect precise ages for each specimen with below) as follows: Middle Eocene, Clairbornian, ca. 40 Ma Early Oligocene, ca. 35 Ma Middle Miocene, Barstovian, ca. 16 Ma Middle Miocene, early Clarendonian, ca. 12 Ma Late Miocene, Clarendonian/early Hemphillian, ca. 10 Ma Late Miocene, late Clarendonian, ca. 9 Ma FLORIDA CENOZOIC SIRENIANS 323 O (‰) 29.7 27.2 26.5 32.1 26.6 26.3 27.4 26.9 27.2 28.5 29.3 25.1 24.7 27.6 29.0 27.3 28.1 28.7 27.5 28.1 27.6 28.3 26.7 26.1 26.2 26.1 25.9 25.3 28.7 30.2 28.1 28.1 30.4 18 ␦ 7.3 3.5 7.7 3.6 6.9 1.7 7.5 5.1 1.6 4.7 7.2 5.8 5.0 8.6 0.1 0.1 2.0 C (‰) 10.9 18.2 12.2 17.3 11.1 13.5 13.3 14.3 11.4 10.2 13.2 18.1 15.3 14.5 14.3 13.5 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ 13 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ ␦ Leisey Shell Pit Leisey Shell Pit Leisey Shell Pit Santa Fe River 1B Santa Fe River 1B Santa Fe River 1B Santa Fe River 1B Santa Fe River 1B Santa Fe River 1B Santa Fe River 1B Santa Fe River 1A Santa Fe River 1A Santa Fe River 1A Santa Fe River 1 Santa Fe River 1 Santa Fe River 2 Santa Fe River 2 Santa Fe River 2 Santa Fe River 2 Santa Fe River 2 Santa Fe River 2 Santa Fe River 2 Santa Fe River 4A Santa Fe River Santa Fe River Santa Fe River Santa Fe River Santa Fe River Rock Springs LF Rock Springs LF Rock Springs LF Rock Springs LF Rock Springs LF incisorlm? lm? lm? cheek tooth Waccasassa Rivercheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth 5.6cheek tooth cheek tooth cheek tooth cheek tooth 23.0 cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth abrean, mixed, ca. 1.25 Ma sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. Crystosiren varguezi Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus abrean, ca. 0.25 Ma abrean ca. (‘‘Sangamonian’’), 0.1 Ma Continued. Museum# Taxon Tooth Locality UF 10710 UF 10710 UF 182418 UF 182419 UF 182420 UF 18826 UF 156878 UF 87227 UF 93273 UF 182425 UF 182426 UF 182427 UF 182428 UF 182429 UF 182430 UF 182431 UF uncataloged UF uncataloged UF uncataloged UF 182421 UF 182422 UF 182423 UF 182424 UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF/FGS V-4404 UF/FGS V-4409 UF/FGS V-4451a UF/FGS V-4451b UF 48982 Appendix 1. Late Miocene, early Hemphillian, ca. 8 Ma Early Pleistocene, Irvingtonian, 1.5 Ma Late Pliocene–Pleistocene, late Blancan–Ranchol Late Pleistocene, middle Ranchol Late Pleistocene, Ranchol 324 BRUCE J. MACFADDEN ET AL. O (‰) 27.4 27.5 30.2 28.8 30.4 27.8 28.1 26.0 28.0 25.8 29.5 26.9 29.4 31.2 30.4 29.4 29.7 27.3 28.9 29.1 29.7 29.4 28.4 28.5 29.5 29.4 29.4 28.7 28.0 27.9 28.5 25.3 25.1 25.5 27.6 27.0 29.6 29.8 29.9 30.1 29.6 18 ␦ 5.5 8.4 4.4 3.5 0.7 3.3 0.3 0.6 0.7 6.7 6.9 3.6 0.9 0.1 1.2 0.7 3.7 0.0 5.9 0.5 9.9 4.2 2.3 1.3 0.6 1.7 1.6 1.5 C (‰) 11.4 11.0 11.0 13.9 11.9 11.0 12.6 12.3 14.9 13.8 14.1 13.2 14.5 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ 13 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ ␦ Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Oklawaha River 1 Brevard County St. John’s River Brevard County Duval County Duval County Volusia County Brevard County Lee County Dade County Dade County Dade County Broward County Dade County Collier County Lee County Brevard County Lake County Lake County Captive Captive Captive Captive Captive Captive Australia Australia Australia Australia Australia cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth cheek tooth U cheek tooth l cheek tooth l cheek tooth l cheek tooth cheek tooth U cheek tooth tusk tusk tusk tusk tusk tusk Australia 1.7 29.8 sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. sp. Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Trichechus manatus Dugong dugon Dugong dugon Dugong dugon Dugong dugon Dugong dugon Dugong dugon abrean, ca. 0.025 Ma Continued. Museum# Taxon Tooth Locality UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF uncataloged UF M 15110 UF M 15113 UF M 15114 UF M 15115 UF M 15120 UF M 15122 UF M 15141 UF M 15160 UF M 15172 UF M 15173 UF M 15176 UF M 15195 UF M 15196 UF M 15134 UF M 19135 UF M 20602 UF M 20773 UF M 23993 Hugh or Buffett Hugh or Buffett Hugh or Buffett Hugh or Buffett Hugh or Buffett Snooty MD-142 MD-19 MD-146 MD-16 MD-22 MD-116 UF 182412 UF 182413 UF 182416 UF 182417 UF uncataloged Appendix 1. Late Pleistocene, late Ranchol Recent (osteological and naturally shed from capive individuals), 0 Ma