The Habitat and Dietary Preferences of the Extinct Tethythere Desmostylus, Inferred from Stable Isotope Analysis

Paleobiology, 29(4), 2003, pp. 506–519

A paleoecological paradox: the habitat and dietary preferences of the extinct tethythere Desmostylus, inferred from stable isotope analysis

Mark T. Clementz, Kathryn A. Hoppe, and Paul L. Koch

Abstract.—The Desmostylia, an extinct order of mammals related to sirenians and proboscideans, are known from the late Oligocene to late Miocene of the North Paciﬁc. Though often categorized as marine mammals on the basis of fossil occurrences in nearshore deposits, reconstructions of desmostylian habitat and dietary preferences have been somewhat speculative because morphological and sedimentological information is limited. We analyzed the carbon, oxygen, and strontium isotope compositions of enamel from Desmostylus and co-occurring terrestrial and marine taxa from middle Miocene sites in California to address the debate surrounding desmostylian ecology. The ␦13C value of tooth enamel can be used as a proxy for diet. Desmostylus had much higher ␦13C values than coeval terrestrial or marine mammals, suggesting a unique diet that most likely consisted of aquatic vegetation. Modern aquatic mammals tend to exhibit lower variability in ␦18O values than terrestrial mammals. Both fossil marine mammals and Desmostylus exhibited low ␦18O variability, suggesting that Desmostylus spent a large amount of time in water. Finally, the Sr isotope composition of marine organisms reﬂects that of the ocean and is relatively invariant when compared with values for animals from land. Sr isotope values for Desmostylus were similar to those for terrestrial, rather than marine, mammals, suggesting Desmostylus was spending time in estuarine or freshwater environments. Together, isotopic data suggest that Desmostylus was an aquatic herbivore that spent a considerable portion of its life foraging in estuarine and freshwater ecosystems.

Mark T. Clementz. Department of Earth Sciences, University of California, Santa Cruz, California 95064. E-mail: [email protected] Kathryn A. Hoppe. Department of Environmental Science, Policy & Management, University of California, Berkeley, California 94720. E-mail: [email protected] Paul L. Koch. Department of Earth Sciences, University of California, Santa Cruz, Califorina 95064. E-mail: [email protected]

Accepted: 5 March 2003

Introduction vores in coastal ecosystems has no modern an- Mammalian herbivores are a minor com- alog, so the dynamics of ancient coastal eco- ponent of the total fauna in modern marine systems may have been very different from and coastal ecosystems, limited to just four si- those today. renian species. Nonetheless, these herbivores The feeding ecology and habitat preferences are thought to play critical roles in structuring of desmostylians are not well understood, but the species richness and productivity of these before we can begin to explore what part, if ecosystems (Bowen 1997; Peterken and Con- any, desmostylians played in past coastal eco- acher 1997). These effects may have been even systems, we need such basic autecological greater in the past, when the diversity of data. Here, we reconstruct the ecology of one large-bodied, mammalian herbivores was genus, Desmostylus, through stable isotope higher (Domning and Furusawa 1992; Aran- analysis of tooth enamel. Using carbon iso- da-Manteca et al. 1994; Domning 2001). Other topes, we assessed feeding preferences, ex- groups of mammals may have exploited the ploring levels of dependence on terrestrial abundant vegetation in shallow coastal waters versus aquatic food sources. To assess the ad- that were widespread from the Eocene aptation of Desmostylus to aquatic habitats, we through the Miocene, such as the Desmostylia, used a method that relies on contrasts in ox- an extinct group of hippo-sized mammals re- ygen isotope variability between terrestrial lated to sirenians. The coexistence of several and aquatic species. Finally, we explored dif- species of large-bodied mammalian herbi- ferences in habitation of marine, estuarine,

FIGURE 1. Proposed cladogram for the mirorder Tethytheria modified after Inuzuka et al. 1995. and terrestrial systems by identifying differ- eria (Domning et al. 1986; Novacek and Wyss ences in strontium isotope composition 1987) (Fig. 1). The Desmostylia is composed of among taxa. two families, the Paleoparadoxidae, which possessed low-crowned or bunodont molars, Background and the Desmostylidae, which had high- The Desmostylia. Desmostylians are large, crowned or hypsodont molars (Inuzuka et al. hippo-sized mammals known to have inhab- 1995; Inuzuka 2000). Both families coexisted ited the Pacific coast of North America and along the northern Pacific coast of Asia and Asia from the late Oligocene (ϳ28Ma)tolate North America until the late middle Miocene. middle Miocene (ϳ10 Ma) (Barnes et al. 1985; Desmostylian ecology is currently a mys- Inuzuka et al. 1995; Inuzuka 2000). Several tery. Though most desmostylian fossils have characters of the skull suggest that desmos- been found in nearshore marine deposits, tylians share a common ancestry with sireni- there are occasional reports of desmostylians ans and proboscideans, which are classified associated with terrestrial mammals at sites with desmostylians in the mirorder Tethyth- where the depositional environment is unclear. Desmostylian morphology has also generated conflicting ecological interpretations. Early reconstructions portrayed desmostylians with a pinniped-like posture (Fig. 2A), suggesting they were fully aquatic swimmers (Repenning 1965), but recent studies compar- ing desmostylians with extant aquatic and terrestrial mammals have challenged this idea. Cole and Domning (1998) argued that desmostylian postcranial elements were most

FIGURE 2. Three different skeletal reconstructions of similar to those of modern terrestrial ungu- desmostylians based on morphological data and pro- lates or extinct ground sloths, and they sug- posed lifestyle. A, Cole and Domning (1998) proposed gested a mode of locomotion and a lifestyle that Paleoparadoxia exhibited a posture and joint me- chanics similar to those of extinct ground sloths and the similar to those of the semiaquatic hippopot- extant hippopotamus. B, Repenning’s (1965) reconstruc- amus. A more radical interpretation was sug- tion of Desmostylus, with a posture similar to that of modern otariid pinnipeds, was based on interpreting gested by Inuzuka (1984). Unlike other mam- the fore- and hindlimbs as modified into flippers. C, mals, which hold their limbs directly under Inuzuka (1984) proposed a completely new style of pos- the body, Inuzaka’s reconstructed desmosty- ture for Desmostylus, termed herpetiform, which was thought to enhance stability in high-wave-energy envi- lians had a ‘‘herpetiform’’ or sprawling pos- ronments. ture similar to lizards, which would have pro- 508 MARK T. CLEMENTZ ET AL. vided a high degree of stability against lateral ences in photosynthetic physiology, sources of forces, such as those produced by waves in fixed carbon, and environmental conditions. nearshore ecosystems. It is unclear which (if In terrestrial ecosystems, differences are re- any) interpretation is correct. lated to photosynthetic pathway (O’Leary Most interpretations imply some use of 1988), resulting in relatively high ␦13Cvalues aquatic habitats by desmostylians, but the ex- (Ϫ13 Ϯ 2‰) for plants using C4 photosynthe- tent to which they foraged within these eco- sis (e.g., warm-climate grasses), low ␦13C val- systems is uncertain. Cranial and dental mor- ues (Ϫ27 Ϯ 3‰) for C3 photosynthesizers phology, as well as their phylogenetic posi- (e.g., trees, most shrubs, herbs, and cool-cli- tion, suggests that desmostylians were prin- mate grasses), and ␦13C values that can be any- cipally herbivores, though at least one study where between these extremes for plants us- has proposed that they consumed mollusks ing CAM photosynthesis (e.g., most succu- (McLeod and Barnes 1984). Their northern lents). In aquatic systems, environmental con- geographical distribution and the deposition- ditions (dissolved [CO2], mixing of the water al settings of their occurrences led Domning column, nutrient supply, etc.) have a stronger et al. (1986) to conclude that desmostylians influence on primary-producer ␦13C values, primarily ate marine vegetation (e.g., macroal- creating differences in mean ␦13C values for gae, seagrass) in cold, marginal marine envi- kelp (Ϫ17 Ϯ 4‰), seagrass (Ϫ10 Ϯ 3‰), near- ronments. However, if desmostylians were ca- shore and offshore marine phytoplankton pable of terrestrial locomotion (Cole and (Ϫ20 Ϯ 2‰, Ϫ23 Ϯ 2‰, respectively), and Domning 1998), they could have come ashore freshwater vegetation (Ϫ26 Ϯ 7‰) (Osmond to feed on terrestrial plants, filling an ecolog- et al. 1981; Boon and Bunn 1994; Hemminga ical niche similar to that of the hippopotamus. and Mateo 1996; Rau et al. 2001; Ravens et al. Thus, even the dietary preferences of this ex- 2002). tinct group of mammals are unresolved. Many studies have exploited these differ- Carbon Isotopes and Foraging Preferences. ences in primary-producer ␦13C values to trace Analysis of carbon isotopes in biogenic sub- the foraging habits of terrestrial and marine strates (e.g., bone or enamel carbonate, colla- consumers (Cerling and Harris 1999; Cle- gen) has proven useful for paleodietary recon- mentz and Koch 2001) (see Table 1). Within structions (see reviews in Schwarcz and marine habitats, consumer ␦13C values typi- Schoeninger 1991; Koch et al. 1995). Tooth cally increase toward shore, with the highest enamel carbonate is the preferred substrate ␦13C values reported for consumers within for reconstructions more than 100,000 years kelp or seagrass beds. Onshore, consumer old because it is less susceptible to diagenetic ␦13C values vary among C4 grazers (high alteration (Lee-Thorp 2000; Wang and Cerling ␦13C), C3 browsers and grazers (low ␦13C), and 1994; Koch et al. 1997). The ␦13Cvalue1 of car- freshwater foragers (very low ␦13C values). bonate in tooth enamel apatite is labeled by Overall, enamel ␦13C values provide strong ev- the ␦13C of an animal’s diet. For ungulate her- idence about the types of food mammals con- bivores, enamel ␦13C values are controlled by sume and from which food webs that food the ␦13C value of the vegetation the animal came. We used this approach to determine the consumes with a small physiological fraction- ecosystems in which Desmostylus foraged. ation of ϳϩ12–14.1‰ for wild populations Oxygen Isotopes and Aquatic Habitat Use. (Lee-Thorp et al. 1989; Cerling and Harris The ␦18O value of biogenic apatite in bones 1999). and teeth is a function of the ␦18O of body wa- The ␦13C values of primary producers at the ter, plus a temperature-dependent fraction- base of food webs vary as a result of differ- ation that is constant in homeothermic mammals (Longinelli 1984; Luz et al. 1984). Unlike carbon in enamel, which has just one source 1 ␦13 ϭ 13 12 Ϭ 13 12 Ϫ C [( C/ Csample C/ Cstandard) 1) * 1000], where (i.e., diet), body water has multiple sources of the standard is VϪPDB. ␦18O follows the same conven- tions, where the ratios are 18O/16O and the standard is oxygen, and both environmental and physio- VϪSMOW. Units are parts per thousand (‰). logical factors influence its ␦18O value. Physi- HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 509

TABLE 1. Mean ␦13Cand␦18O values for modern taxa collected from central California and Amboseli Park in Kenya. Values are reported Ϯ 1␴.

Mean Mean Taxon n Feeding zone % Aquatic ␦13C Ϯ 1␴ ␦18O Ϯ 1␴ Pinnipedia Northern elephant seal* 10 Offshore Ͼ50% Ϫ14.1 Ϯ 1.7 26.6 Ϯ 0.4 California sea lion* 6 Nearshore Ͼ50% Ϫ11.3 Ϯ 1.0 26.1 Ϯ 0.3 Harbor seal* 11 Nearshore Ͼ50% Ϫ9.2 Ϯ 1.6 26.5 Ϯ 0.3 Cetacea Pilot whale* 7 Nearshore 100% Ϫ9.7 Ϯ 1.2 28.1 Ϯ 0.2 Harbor porpoise* 11 Nearshore 100% Ϫ9.9 Ϯ 0.4 28.5 Ϯ 0.2 Bottlenose dolphin* 9 Nearshore 100% Ϫ10.1 Ϯ 0.6 27.8 Ϯ 0.2 Sirenia Dugong 12 Nearshore 100% 0.5 Ϯ 1.0 29.3 Ϯ 0.5 Manatee 17 Rivers/nearshore 100% Ϫ3.3 Ϯ 4.2 29.3 Ϯ 0.8 Carnivora Sea otter* 5 Kelp bed Ͼ50% Ϫ6.1 Ϯ 0.9 27.3 Ϯ 0.6 River otter* 10 Estuary ϳ33% Ϫ8.1 Ϯ 3.0 25.8 Ϯ 0.9 River otter* 7 Rivers/lakes ϳ33% Ϫ17.3 Ϯ 4.3 23.0 Ϯ 0.3 Coyote* 5 Terrestrial Ͻ5% Ϫ10.4Ϯ 3.4 27.4 Ϯ 3.4 Artiodactyla Black-tailed deer* 47 Terrestrial Ͻ5% Ϫ11.8 Ϯ 1.7 29.8 Ϯ 1.3 Grant’s gazelle‡ 6 Terrestrial Ͻ5% Ϫ9.2 Ϯ 1.4 33.8 Ϯ 1.7 Common wildebeest‡ 8 Terrestrial Ͻ5% 1.1 Ϯ 1.0 32.7 Ϯ 1.9 Hippopotamus‡ 3 Rivers/lakes Ͼ50% Ϫ3.0 Ϯ 0.2 26.0 Ϯ 1.4 Perissodactyla Plains zebra‡ 7 Terrestrial Ͻ5% Ϫ0.1 Ϯ 1.1 31.5 Ϯ 1.9 Black rhinoceros‡ 5 Terrestrial Ͻ5% Ϫ9.3 Ϯ 1.2 29.4 Ϯ 1.5 Proboscidea African elephant‡ 13 Terrestrial 5% Ϫ7.7 Ϯ 2.5 29.6 Ϯ 0.6

* Data from Clementz and Koch 2001. All collections are from geographically constrained modern populations, chiefly in central and southern Cal- ifornia. Terrestrial plants in this region are dominantly C3. ‡ Data for populations from Amboseli Park, Kenya, discussed in Bocherens et al. 1996. Grasses in Amboseli are C4, whereas nearly all trees, shrubs, and herbs are C3. Grant’s gazelle and rhinoceros are browsers on C3 plants, wildebeest and zebra are grazers on C4 plants, and elephants are mixed feeders. ology affects body water ␦18Ovaluesbycon- cally yield ␦18O standard deviations of 0.5‰ trolling the magnitude of oxygen fluxes and or less, whereas terrestrial mammals typically isotopic fractionation that occurs as oxygen have values Ͼ1‰ (Table 1). However, there passes into and out of the body (Luz and Ko- are exceptions to this pattern. For terrestrial lodny 1985; Bryant and Froelich 1995; Kohn mammals, body size and environmental con- 1996). The ␦18O values of environmental oxy- ditions may cause ␦18O variation of popula- gen sources provide the baseline from which tions to be lower than predicted. Large mam- mammalian body water ␦18O values can de- mals (Ͼ1000 kg) obtain more of their oxygen viate via physiological effects. from drinking water (ϳ60%) than do smaller The mean ␦18O value of mammalian tooth mammals (ϳ20%) (Bryant and Froelich 1995). enamel has been proposed as a monitor of ad- If drinking-water sources are isotopically ho- aptation to freshwater or marine systems mogeneous, large-mammal ␦18Ovalues (Bocherens et al. 1996; Roe et al. 1998). How- should show less variability among individ- ever, mean ␦18O values may not always be di- uals (Amboseli elephants; Table 1). Likewise, agnostic of habitat preferences (Clementz and for animals living in humid environments Koch 2001), because physiological differences with low evaporative water loss, ␦18Ovalues among species also affect mean values. Differ- within a population may be less variable. ences in the population-level standard devia- Among aquatic species, differences in vari- tion of enamel ␦18O values, on the other hand, ability can result from ␦18O variation of oxy- can allow discrimination of aquatic/semi- gen sources, either by movement between wa- aquatic from terrestrial species (Table 1). Ter- ters of different isotopic composition (e.g., restrial mammals experience greater physio- marine vs. fresh water—sea otters, estuarine logical and environmental variability than river otters, manatees) or by ingestion of other aquatic/semiaquatic mammals, resulting in oxygen sources with distinct ␦18O values (e.g., higher variability in their body water and terrestrial, 18O-enriched plant water—Ambo- tooth enamel ␦18O values. seli hippopotamus). Though these exceptions Fully aquatic and semiaquatic species typi- do generate overlap in 1␴ values for aquatic 510 MARK T. CLEMENTZ ET AL. and terrestrial mammals, only fully aquatic, water 87Sr/86Sr ratios show long timescale strictly marine (e.g., most pinnipeds and ce- fluctuations due to shifts in these fluxes taceans), and semiaquatic mammals that for- (Armstrong 1971; Capo and DePaolo 1990). age aquatically (e.g., river otters) exhibit 1␴ The 87Sr/86Sr ratio of water in estuaries is ␦18OvaluesՅ0.5‰, providing a clear means controlled by mixing. Freshwater has a much of differentiating these habitat preferences lower [Sr] (0.006–2.94 ppm) than seawater (8.0 from fully terrestrial populations. Here, we ppm) (Capo et al. 1998). Mixing of fresh and first test this approach on fossil taxa with marine water produces a correlation of 87Sr/ known habitat preferences, and then use it to 86Sr ratio with salinity, but differences in [Sr] assess the extent of aquatic habitat use by Des- cause the marine 87Sr/86Sr signal to dominate mostylus. (Bryant et al. 1995). Strontium Isotopes and Marine versus Terres- Modern animals foraging in marine sys- trial Ecosystems. The ratio of 87Sr to 86Sr in tems have 87Sr/86Sr ratios that closely match biogenic materials has proven to be a valuable those of seawater (Schmitz et al. 1997; Venne- tool for extracting ecological information from mann et al. 2001). Modern land and freshwa- modern and fossil organisms (Koch et al. 1992, ter animals, in contrast, exhibit a greater range 1995; Kennedy et al. 1997; Hoppe et al. 1999; of ratios that depend on the geologically con- Ingram and Weber 1999). Because strontium is trolled differences at the base of the food web not fractionated measurably when it is incor- (Nelson et al. 1986; Koch et al. 1995; Ingram porated, the 87Sr/86Sr ratio of biogenic mate- and Weber 1999). Animals foraging in estu- rial is identical to that of the source of Sr and aries have not received extensive study but is passed up the food web without modifica- would be expected to show a wide range in tion (Capo et al. 1998). 87Sr/86Sr ratios among individuals in a popu- The Sr sources for an organism are the wa- lation, depending on the salinity of the water ter it drinks or inhabits and the plant or ani- they frequent. Sr isotope ratios will be our pri- mal food it ingests. On land, the 87Sr/86Sr ra- mary tool for determining whether Desmos- tios of rivers and plants are controlled by rock tylus inhabited marine, estuarine, or fully ter- and soil compositions (Miller et al. 1993). Ra- restrial ecosystems. tios of terrestrial rocks are highly variable and Diagenetic Monitoring via Control Taxa. Pri- depend on rock type and age (Capo et al. or work has shown that tooth enamel is resis- 1998). River and soil 87Sr/86Sr ratios are con- tant to diagenetic alteration of stable isotope trolled both by bedrock inputs and by atmo- ratios (Wang and Cerling 1994; Bocherens et spheric input of Sr as dust and precipitation al. 1996; Koch et al. 1997). Still, we will use (Capo et al. 1994; Kennedy et al. 1998). At control taxa to test for alteration whenever coastal sites, precipitation has a 87Sr/86Sr ratio possible. For example, we expect population- similar to that of the ocean, and if rainfall is level ␦18O variability to be higher in obviously significant, it can dominate the 87Sr/86Sr ratio terrestrial animals (e.g., horses) than in obvi- of soils (Kennedy et al. 1998); dust inputs, ously marine mammals (e.g., whales and dol- however, can affect regional soil isotope com- phins). Because diagenetic alteration would positions far from dust sources (Muhs et al. tend to homogenize isotopic signals among 1990). specimens from a single locality, retention of Today, the 87Sr/86Sr ratio of seawater is ho- consistent differences in mean and variability mogeneous globally, because the residence in control taxa provides support for the as- time of oceanic Sr is several million years, or- sumption that isotopic patterns in Desmostylus ders of magnitude longer than the timescale are preserved as well. for oceanic mixing (ϳ1000 years). The 87Sr/ 86Sr ratio of seawater is controlled by the mag- Materials and Methods nitude and isotopic ratios of Sr influxes, in- Locality and Specimen Information. Speci- cluding weathering of terrestrial rocks and mens were obtained from five sites in central oceanic basalt alteration, and by Sr removal and southern California (Fig. 3). Four sites through deposition of marine carbonates. Sea- were in a restricted, shallow marine basin that HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 511

FIGURE 3. A, Map of modern-day central California highlighting the location of sampling sites included in this project. B, Central California as it appeared at ϳ13–16 Ma, during the time that our fossil specimens were deposited (modiﬁed from Bartow 1987).

inundated the interior of California during the Control taxa were also absent at the open-ma- Miocene; one site was exposed to open-marine site V3215, which was located north of the rine conditions. Site 1292 is the famous Shark other sites in the Barstovian-aged Briones For- Tooth Hill deposit within the Round Moun- mation. tain Silt Formation, which has an extensive ac- Sampling Protocol and Analyses. Stable iso- cumulation of marine mammals, shark’s teeth, tope analysis was conducted on the carbonate and rare terrestrial mammals deposited dur- within tooth enamel biological apatite. Ap- ing the Barstovian Land Mammal Age (ϳ13.5 proximately 5 to 10 mg of powder was drilled to 15.9 Ma) (Barnes 1976). We obtained several from each tooth after the surface had been Desmostylus molars and teeth from marine abraded to remove possible contamination. mammals, including the early pinniped Allo- Following the protocol in Koch et al. (1997), all desmus and small and large odontocetes powders were soaked for 24 hours in ϳ2% (toothed whales). At site 2124, Desmostylus NaOCl to oxidize organic matter, rinsed ﬁve teeth and isolated molars from the early horse times with distilled water, soaked for 24 hours Merychippus and the proboscidean Gomphoth- in 1 M calcium acetate-buffered/acetic acid to erium were obtained from the Temblor For- remove contaminating carbonate in non-lat- mation, which is estimated to be coeval with tice sites, rinsed ﬁve times with distilled wa- the Shark Tooth Hill deposit. At site V5555, ter, and then freeze-dried. Because extended Desmostylus from the Santa Margarita For- exposure time may alter enamel isotope val- mation (late Miocene, Clarendonian Age, ϳ10 ues, all samples were treated for the same Ma) were found in association with marine length of time (24 hours) to ensure the com- (e.g., Allodesmus, small odontocetes, sirenians) parability of isotope values and reduce the and terrestrial mammals (e.g., the equid Hip- risk of lab-induced variation in isotope values parion). Site V3301 is from a reef bed deposit (Koch et al. 1997). in the Temblor Formation, which has yielded Approximately 1 mg of pretreated powder a large number of Desmostylus molar frag- was analyzed using an Isocarb automated car- ments of Barstovian age, but no control taxa. bonate analysis system interfaced with a Mi- 512 MARK T. CLEMENTZ ET AL.

TABLE 2. Mean ␦13Cand␦18O, and Sr isotope values for taxa from each site. All values are reported Ϯ 1␴.

Mean Mean Mean Locality Taxon n ␦13C Ϯ 1␴ ␦18O Ϯ 1␴ 87Sr/86Sr Ϯ 1␴ 1292 Desmostylus 6 (3) Ϫ5.6 Ϯ 1.6 27.8 Ϯ 0.5 0.70850 Ϯ 0.00044 1292 Allodesmus 7 (5) Ϫ9.5 Ϯ 0.8 27.1 Ϯ 0.5 0.70861 Ϯ 0.00004 1292 Odontocete, large 4 (4) Ϫ7.4 Ϯ 1.9 28.2 Ϯ 0.3 0.70857 Ϯ 0.00015 1292 Odontocete, small 6 (2) Ϫ7.7 Ϯ 0.3 27.5 Ϯ 0.4 0.70833 2124 Desmostylus 5 (4) Ϫ7.0 Ϯ 2.5 27.2 Ϯ 0.2 0.70781 Ϯ 0.00070 2124 Gomphotherium 10 (4) Ϫ9.6 Ϯ 0.6 27.6 Ϯ 0.8 0.70763 Ϯ 0.00077 2124 Merychippus 15 (4) Ϫ8.7 Ϯ 0.7 29.0 Ϯ 1.5 0.70781 Ϯ 0.00077 V5555 Desmostylus 21 (6) Ϫ5.5 Ϯ 1.5 28.1 Ϯ 0.7 0.70850 Ϯ 0.00022 V5555 Hipparion 5 (5) Ϫ11.5 Ϯ 0.3 27.4 Ϯ 1.2 0.70789 Ϯ 0.00058 V5555 Allodesmus 3 (2) Ϫ7.1 Ϯ 1.1 28.1 Ϯ 0.8 0.70853 V5555 Odontocete, small 2 (0) Ϫ8.0 Ϯ 1.0 27.5 Ϯ 0.1 N/A V3301 Desmostylus 8 (4) Ϫ3.5 Ϯ 1.0 27.6 Ϯ 0.2 0.70808 Ϯ 0.00051 V3215 Desmostylus 4 (0) Ϫ3.0 Ϯ 0.2 26.5 Ϯ 0.3 N/A

cromass Optima gas source mass spectrome- significance of mean values by using a Stu- ter in the Deptartments of Earth and Ocean dent’s t-test for comparisons between two Sciences, University of California, Santa Cruz. populations or a parametric, one-factor anal- Samples were dissolved in 100% phosphoric ysis of variance (ANOVA) for multiple popu- acid at 90ЊC, with concurrent cryogenic trap- lations. If a significant difference was detected ping of CO2 and H2O. The CO 2 was then ad- among populations, we applied a pairwise mitted to the mass spectrometer for analysis. comparison (post-hoc Tukey test) to identify The standards used in this study were Carrera populations that were statistically distinct. To Marble and NBS 19 and values are reported use these tests, however, sample populations relative to V-PDB (for carbon) and V-SMOW must be normally distributed and equivalent (for oxygen). Precision, determined by repeat- in variance. For sample populations that didn’t ed concurrent analysis (n ϭ 19) of a modern meet these criteria, we assessed statistically elephant enamel standard, was 0.1‰ for ␦13C significant differences in median values using and 0.2‰ for ␦18O. a nonparametric Kruskal-Wallis one-factor For 87Sr/86Sr analysis, ϳ1 mg of powder was analysis of variance followed by pairwise collected from each sample, soaked in 0.5 ml comparison using the post-hoc Dunn’s meth- of 1 N acetic acid for 20 minutes, rinsed in dis- od. We note which method of comparison was tilled water, and then repeated four more used in the ‘‘Results’’ section. For compari- times. Samples were then dissolved overnight sons of variance between populations, a sim- in 2.5 N HCl, dried down, redissolved in 0.5 ple F-test was used. Spearman’s rank correla- ml of 2.5 N HCl, and injected onto a column tion method was used to assess significance of filled with a cation exchange resin to isolate Sr. correlation of different isotopic values among After washing with 20 ml of 2.5 N HCl, an ad- samples. Either Sigmastat 2.03 or Microsoft ditional 7 ml of 2.5 N HCl was passed through Excel 2000 was used for all calculations. the column and collected for analysis. Sam- ples were dried down overnight then dis- Results solved in 1 ␮l of 10% nitric acid and placed on Samples were grouped into three categories rhenium filaments. To conduct the isotope for statistical comparisons: unambiguously analyses we used the VG 354 Thermal Ioni- marine mammals, unambiguously terrestrial zation Mass Spectrometer located in the Dep- mammals, and Desmostylus. Desmostylus had tartment of Earth Sciences at the University of the highest mean ␦13Cvalue(Ϯ1␴)(Ϫ5.2 Ϯ California, Santa Cruz. Precision, determined 1.9‰), followed by marine mammals (Ϫ8.1 Ϯ by repeated measurement of the NBS 987 Sr 1.4‰), then terrestrial mammals (Ϫ9.5 Ϯ standard, was Ϯ 0.00003. 1.2‰) (Table 2, Fig. 4A). Median ␦13Cvalues Data Analysis. We assessed the statistical differed significantly among groups (Kruskal- HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 513

FIGURE 4. Carbon (a), oxygen (b), and strontium (c) isotope data collected for all taxa. Along the horizontal axis, terrestrial (squares) and marine (triangles) faunas are grouped on either side of the data for Desmostylus (circles) from all ﬁve sites. A ϭ Allodesmus;Gϭ Gomphotherium;Hϭ Hipparion;LOϭ large odontocete; M ϭ Merychippus; SO ϭ small odontocete. Enlarged, closed symbols represent the mean value for each taxon or group, and all other values are plotted as open symbols. Vertical bars represent Ϯ1␴. Estimated range in middle Miocene seawater 87Sr/ 86Sr values is based on Hodell et al. 1991. 514 MARK T. CLEMENTZ ET AL.

Wallis: H ϭ 53.927, p Ͻ 0.01). Pairwise com- mammals (F-test: p ϭ 0.237). Comparison of parisons revealed that Desmostylus was signif- mean 87Sr/86Sr ratios among marine mammals icantly different from both marine and terres- (t-test: t ϭ 1.358, p ϭ 0.202) and terrestrial trial mammals, but that marine and terrestrial mammals (one-factor ANOVA: F ϭ 0.011, p ϭ mammals were not different from each other 0.989) revealed no statistically significant dif- (Dunn’s method). Differences in ␦13C variance ferences. among groups were only significant between Desmostylus is the only taxon that occurs at Desmostylus and terrestrial mammals (F-test: p enough sites to warrant examination of geo- Ͻ 0.01). Within marine mammals, mean ␦13C graphic differences. Mean ␦13C values were values differed significantly between Allodes- significantly different among sites (one-factor mus and all cetaceans (t-test: t ϭ 2.413, d.f. ϭ ANOVA: F ϭ 6.09, p Ͻ 0.05) with the highest 20, p ϭ 0.026). Mean ␦13C values also differed values reported at northern sites V3301 and significantly among terrestrial mammals V3215. Mean ␦18O values were, likewise, sig- (one-factor ANOVA: F ϭ 39.591, p Ͻ 0.01); nificantly different among sites (one-factor pairwise comparison revealed statistically sig- ANOVA: F ϭ 7.73, p Ͻ 0.05). Pairwise com- nificant differences among all three groups parisons revealed that the mean for Desmos- (Tukey test: p Ͻ 0.01). tylus from site 3215 was significantly lower The mean ␦18O value for terrestrial mam- than mean ␦18O values from all other sites ex- mals (28.3 Ϯ 1.4‰) was higher than for either cept site 2124, and that mean values for sites marine mammals (27.6 Ϯ 0.6‰) or Desmos- V5555 and 2124 were significantly different tylus (27.7 Ϯ 0.7‰) (Table 2, Fig. 4B). Median (Tukey test: p Ͻ 0.05). ␦18O variance at sites ␦18O values did not differ among the three 1292 and V5555 were significantly different groups of mammals (Kruskal-Wallis: H ϭ from variance at sites 2124 and V3301. No sta- 3.823, p ϭ 0.148). Desmostylus and marine tistically significant differences were detected mammals exhibited significant differences in for mean 87Sr/86Sr values (Kruskal-Wallis: H ϭ ␦18O variance from terrestrial mammals (F- 5.67, p ϭ 0.129). test: p Ͻ 0.01), but not between each other (F- test: p ϭ 0.496). Within marine mammals, no Discussion significant difference was detected between How Committed Was Desmostylus to Aquatic means for Allodesmus and all cetaceans (t-test: Habitats? Most ecological interpretations for t ϭϪ1.029, p ϭ 0.316). Mean ␦18O values did Desmostylus favor a connection with aquatic differ significantly among terrestrial mam- habitats, but the amount of time that this mals (one-factor ANOVA: F ϭ 4.974, p ϭ mammal actually spent in the water remains 0.014); pairwise comparison revealed that contested. Possible scenarios include that Des- only Merychippus and Gomphotherium were sig- mostylus was fully aquatic, Desmostylus was nificantly different (Tukey test: p ϭ 0.031). semiaquatic and foraged onshore, or Desmos- Mean 87Sr/86Sr ratios were highest for ma- tylus was semiaquatic and foraged in the wa- rine mammals (0.70854 Ϯ 0.00014), interme- ter. Our proxy for extent of adaptation to diate for Desmostylus (0.70826 Ϯ 0.00048), and aquatic life is population-level variability in lowest for terrestrial mammals (0.70783 Ϯ ␦18O values, which we used after validation by 0.00067) (Table 2, Fig. 4C). Median 87Sr/86Sr analysis of fully aquatic and fully terrestrial values differed significantly among groups fossil taxa. (Kruskal-Wallis: H ϭ 10.462, p Ͻ 0.01), but As expected from studies of modern spe- pairwise comparison showed that only terres- cies, the standard deviation of ␦18O values for trial and marine mammals were statistically fossil fully aquatic mammal populations (of distinct (Dunn’s method: p Ͻ 0.05). Differenc- sufficient sample size, i.e., n Ն 5) was Յ0.5‰, es in 87Sr/86Sr variance were statistically sig- significantly lower than the values for fossil nificant for Desmostylus versus marine mam- fully terrestrial mammal populations (1␴ Ն mals (F-test: p Ͻ 0.01) and marine mammals 0.8‰) (Table 2). The 1␴ values for fossil fully versus terrestrial mammals (F-test: p Ͻ 0.01), aquatic and fully terrestrial mammals were but not for Desmostylus versus terrestrial comparable to those for modern mammal HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 515 populations (Bocherens et al. 1996; Clementz freshwater ecosystems) as potential habitats. and Koch 2001). We conclude that diagenetic Patterns in mean and variability in ␦18Oand alteration has not erased the in vivo differenc- 87Sr/86Sr values can be useful for discriminat- es in ␦18O values among mammals at these ing among these alternatives, even though we sites. Because Desmostylus has thicker enamel were unable to include clear isotopic end- than any of the marine and terrestrial mam- members of freshwater and estuarine fossil mals, it is even less likely to be subject to dia- aquatic mammals in our study. For ␦18O val- genetic homogenization of ␦18O values. ues, we would expect marine species to have Desmostylus populations had 1␴ values high mean values and low variability, estua- ranging from 0.2‰ to 0.7‰ (Table 2, Fig. 4B). rine environments to have lower mean values These values are lower than those for any of and higher variability, and freshwater envi- the fossil terrestrial mammals and for all but ronments to yield the lowest mean values and one of the modern terrestrial mammals (Table low variability. For 87Sr/86Sr values, we would 1). Furthermore, these 1␴ values are signifi- expect values for marine species to cluster cantly lower than values for semiaquatic, ter- near values calculated for middle Miocene restrial-foraging hippopotamus from Ambo- seawater (Fig. 4C), whereas values for fresh- seli (Table 1), suggesting that though desmos- water and estuarine species should be more tylians and hippopotamids were similar in variable. body size and basic morphology, the niches However, mean ␦18O values in modern eco- occupied by these taxa were quite distinct. systems were not always diagnostic of fresh- Low ␦18O variability, particularly 1␴ values water, let alone marine or estuarine, habitats near 0.2 or 0.3‰, provides strong support for (Table 1), and we observed a similar problem the conclusion that Desmostylus was either a for our fossil samples (Table 2, Fig. 4B), which semiaquatic mammal feeding in water, or a only yielded a 2‰ range in mean values for fully aquatic mammal. all taxa. In addition, the mean ␦18O values for However, one population (V5555) had a 1␴ marine mammals from our sample sites are value higher than that for modern fully ter- higher than expected from estimated middle restrial Amboseli elephants, which suggests Miocene seawater ␦18O values (0.5‰ below that the low ␦18O variability in Desmostylus modern seawater ␦18O (J. Zachos personal may simply be a result of large body size. This communication); Fig. 4B). However, the high hypothesis is unlikely for two reasons. First, ␦18O values we have reported for our marine Gomphotherium is larger than Desmostylus,yet mammals may reflect the regional hydrologic at the site where these two species co-occur, conditions of the basin in which these animals the difference in ␦18O variation is extreme (Ta- lived (Fig. 3). The fossils we sampled were all ble 2). An additional factor must be generating deposited within a restricted basin, which the low variability in Desmostylus. Further- may have experienced significant evaporation more no modern fully terrestrial mammal has and limited exchange with open-ocean wa- yielded ␦18O variability as low as that reported ters. If so, ␦18O values for marine waters in this for Desmostylus at the majority of sites. Thus, basin may have been enriched relative to mean we stand by the conclusion that Desmostylus seawater and could account for the high ␦18O was either semiaquatic and feeding in water, values we have reported. or fully aquatic. The slightly higher 1␴ value From the 87Sr/86Sr values of the tooth enam- at site V5555 could therefore imply use of an el, we were able to detect a clearer separation estuary with significant ␦18O variability. between terrestrial and marine mammals. What Kinds of Aquatic Habitats Did Desmos- Marine taxa had little variation in 87Sr/86Sr tylus Frequent? Though the occurrence of values (1␴Ͻ0.0002‰) when compared with Desmostylus remains in nearshore marine de- terrestrial mammals (1␴Ͼ0.0006‰) (Table 2, posits suggests an affinity for marine ecosys- Fig. 4A). As with the ␦18O data, the 87Sr/86Sr tems, the possibility of transport prior to de- values we collected for marine mammals are position means that we can’t rule out other not expected, according to the calculated aquatic environments (i.e., estuarine and range in 87Sr/86Sr values for seawater at this 516 MARK T. CLEMENTZ ET AL. time (0.70876 to 0.70877) (Hodell et al. 1991) (Zachos et al. 2001). With known marine and (Fig. 4C). Again, the hydrology of the basin terrestrial end-members, we can now deter- may explain these low 87Sr/86Sr values. With mine the food webs contributing carbon to restricted flow to and from the ocean, terres- Desmostylus (Fig. 4A). trial inputs of 87Sr/86Sr from rivers could have Mean ␦13C values for Desmostylus are signif- lowered the mean 87Sr/86Sr value of waters icantly higher than would be expected for a within the basin. terrestrial C3 consumer and confirm that Des- Of most significance, however, is the large mostylus must have been foraging within other difference in range of values between terres- food webs (Fig. 4A), a conclusion supported trial and marine mammals. Desmostylus 87Sr/ by our ␦18Oand87Sr/86Sr evidence. Several 86Sr variation is greater than that calculated for aquatic habitats are possible alternatives: ma- any marine taxon and similar to the degree of rine (nearshore, kelp, seagrass), estuarine variation observed for terrestrial taxa (Fig. (seagrass), and freshwater ecosystems. Iso- 4C). High variation in 87Sr/86Sr values would tope analysis of modern marine ecosystems be unlikely if Desmostylus was foraging only has found that nearshore consumers often within nearshore environments and suggests have higher ␦13C values than consumers lim- that it was spending a considerable amount of ited to terrestrial C3 resources (Hobson 1987; time in estuarine or freshwater environments. Bearhop et al. 1999). Seagrass, in particular, What Did Desmostylus Eat? From our pre- has ␦13C values that are often higher than vious analyses, we have concluded that Des- modern C4 vegetation, with a mean ␦13Cvalue mostylus was a fully aquatic or semiaquatic of ϳϪ11‰ (Hemminga and Mateo 1996). An- forager that was not restricted to marine hab- other potential resource would be marine al- itats but foraged also in freshwater and estu- gae, which also can have high ␦13C values (Ra- arine habitats. Our next step is to identify the vens et al. 2002). In particular, the kelp group food webs within which Desmostylus was feed- of marine algae (Family Laminaria), which is ing—were they nearshore marine, kelp beds, believed to have arisen along the Pacific Coast seagrass beds, or freshwater food webs? To ex- at this time (Estes and Steinberg 1988), yields plore this question, we used mean ␦13Cvalues mean ␦13CvaluesofϳϪ17‰ (Raven et al. of tooth enamel as a proxy for dietary pref- 2002). erences, basing our interpretations on values Each of these marine resources could pro- from both modern and fossil taxa (Tables 1, 2, duce the high ␦13C values observed in Desmos- Fig. 4A). tylus, either via direct consumption or indi- Fossil terrestrial taxa have ␦13C values sim- rectly via consumption of other consumers in ilar to those reported for middle Miocene these food webs. Incorporation of a multi-iso- mammals that fed on C3 vegetation elsewhere tope analysis and inclusion of ecological con- in North and South America (MacFadden et al. trol taxa limit the possible scenarios (Fig. 1994; MacFadden and Cerling 1996). Our ma- 4B,C). rine species, including the early otariid Allo- First, ␦13C values for Desmostylus are typi- desmus and two size classes of odontocetes, cally higher than values reported for near- have been interpreted as nearshore, marine shore foraging marine mammals, suggesting foragers on the basis of sedimentological and that Desmostylus was not foraging within this morphological criteria (Barnes 1972; Dupras ecosystem. This interpretation is also con- 1985). As expected, mean ␦13C values for these firmed by the high 87Sr/86Sr variation reported species were substantially more enriched in for Desmostylus, which suggests it was not 13C than values for modern offshore foragers spending much time in a strictly marine en- (Table 1), but they were also slightly more environment (Fig. 4C). The lack of a marine sig- riched than values for modern nearshore for- nal in 87Sr/86Sr values for Desmostylus also al- agers (Table 1). This offset from modern near- lows us to exclude kelp beds as a potential di- shore foragers may reflect ocean ␦13Cvalues etary resource for Desmostylus, given that kelp during the middle Miocene, which were is an exclusively marine macrophyte. ϳ1.5‰ higher than modern seawater values Seagrass, on the other hand, grows within HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 517 estuaries and can tolerate low salinity condi- desmostylians and proboscideans belong, or tions (Ͼ20 ppt). Also, seagrasses growing in did desmostylians evolve aquatic habits short- low-salinity environments often exhibit ex- ly after their divergence? This issue may be tremely low ␦13C values that are Ͼ1␴ from the addressed by applying similar techniques to mean (ϽϪ11‰). In these habitats, Desmosty- samples of early proboscideans and basal lus could also have foraged on other types of tethytheres (e.g., anthracobunids). estuarine or marsh vegetation (including Our results highlight the necessity of using some species of macroalgae, such as Ulva), multiple proxies to answer questions about which can exhibit a large range in ␦13C values. the ecology of extinct taxa. If we had limited If Desmostylus was consuming a mixture of our interpretations to only one isotopic sys- freshwater and estuarine vegetation or forag- tem, we would have generated substantially ing on other consumers within these food different conclusions about the ecology of Des- webs, then the ␦13Cand87Sr/86Sr values seem mostylus. In future work, we will incorporate more reasonable. microwear analysis into studies of other desmostylian species to identify differences in Conclusion foraging preferences among co-occurring By applying a multi-isotope approach, we taxa, and to develop a method for testing our have been able to identify significant ecologi- interpretations of the trophic level of these ex- cal differences between Desmostylus and both tinct animals. terrestrial and marine mammals in terms of aquatic affinity and dietary preferences, re- Acknowledgments ␦13 spectively. The high C values suggest that We would like to thank P. Holroyd at the Desmostylus foraged on seagrasses, probably University of California Museum of Paleon- rooting up the vegetation to consume the car- tology, L. Barnes at the Natural History Mu- bohydrate-enriched rhizomes that would have seum of Los Angeles County, and T. Deme´re´ formed dense mats within the shallow la- at the San Diego Natural History Museum for goons and estuaries along the Pacific coastline. providing access to specimens for analysis. A In addition, Desmostylus was not limited to special thank you goes to P. Holden for assist- foraging on seagrasses but likely incorporated ing in analysis of Sr samples and to D. Domn- a wide range of freshwater and estuarine ing for providing information on desmostyli- aquatic vegetation into its diet. In this way, the an phylogenetics. M.T.C. was supported by a ecology of Desmostylus was most similar to National Science Foundation (NSF) Predoctor- that of modern manatees in Florida, which al Fellowship and Achievement Rewards for seasonally forage on aquatic vegetation in College Scientists Fellowship when much of freshwater and marine ecosystems. Unlike this research was conducted. Analytical and manatees, Desmostylus was fully capable of travel costs were covered by NSF grants EAR- terrestrial locomotion and was probably sim- 9725854 and 0087742. ilar to the modern hippopotamus in terms of aquatic and terrestrial habits. Literature Cited Thus, desmostylians were a unique group Armstrong, R. L. 1971. Glacial erosion and the variable isotopic of mammals with no real modern analogs in composition of strontium in seawater. Nature 230:132–133. terms of habitat preferences, diet, or locomo- Aranda-Manteca, F. J., D. P. Domning, and L. G. Barnes. 1994. A tor capabilities. As the largest mammalian new middle Miocene sirenian of the genus Metaxytherium from Baja California and California, relationships and paleo- consumers of coastal aquatic vegetation at the biogeographic implications. In A. Berta and T. A. Deme´re´, time, their impact on coastal ecosystems must eds. Contributions in marine mammal paleontology honoring have been substantial. In addition, because Frank C. Whitmore, Jr. Proceedings of the San Diego Society of Natural History 9:191–204. they are believed to be the sister group to pro- Barnes, L. G. 1972. Miocene Desmatophocinae (Mammalia: Car- boscideans, their aquatic preferences raise in- nivora) from California. University of California Publications teresting evolutionary questions. Were the in Geological Sciences 89:1–68. ———. 1976. Outline of eastern North Pacific fossil cetacean as- aquatic habits of this group characteristic of semblages. Systematic Zoology 25:321–343. the basal members of the clade to which both Barnes, L. G., D. P. Domning, and C. E. Ray. 1985. Status of stud- 518 MARK T. CLEMENTZ ET AL.

ies on fossil marine mammals. Marine Mammal Science 1:15– timate marine and terrestrial protein content in gull diets. Ca- 53. nadian Journal of Zoology 65:1210–1213. Bartow, J. A. 1987. Cenozoic evolution of the San Joaquin Valley, Hodell, D. A., P. A. Mueller, and J. R. Garrido. 1991. Variations California. U.S. Geological Survey Report OF 87-0581. in the strontium isotopic composition of seawater during the Bearhop, S., D. R. Thompson, S. Waldron, I. C. Russell, G. Al- Neogene. Geology 19:24–27. exander, and R. W. Furness. 1999. Stable isotopes indicate the Hoppe, K. A., P. L. Koch, R. W. Carlson, and S. D. Webb. 1999. extent of freshwater feeding by cormorants Phalacrocorax Tracking mammoths and mastodons: reconstruction of mi- carbo shot at inland fisheries in England. Journal of Applied gratory behavior using strontium isotope ratios. Geology 27: Ecology 36:75–84. 439–442. Bocherens, H., P. L. Koch, A. Mariotti, D. Geraads, and J. Jaeger. Ingram, B. L., and P. K. Weber, 1999. Salmon origin in Califor- 1996. Isotopic biogeochemistry (13C, 18O) of mammalian nia’s Sacramento-San Joaquin river system as determined by enamel from African Pleistocene hominid sites. Palaios 11: otolith strontium isotopic composition. Geology 27:851–854. 306–318. Inuzuka, N. 1984. Skeletal restoration of the desmostylians: her- Boon, P. L., and S. E. Bunn. 1994. Variations in the stable isotope petiform mammals. Memoirs of the Faculty of Science, Kyoto composition of aquatic plants and their implications for food University, Series of Biology IX:157–253. web analysis. Aquatic Botany 48:99–108. Inuzuka, N. 2000. Primitive Late Oligocene desmostylians from Bowen, W. D. 1997. Role of marine mammals in aquatic ecosys- Japan and phylogeny of the Desmostylia. Bulletin of Ashoro tems. Marine Ecology Progress Series 158:267–274. Museum of Paleontology: 91–123. Bryant, J. D., and P. N. Froelich. 1995. A model of oxygen isotope Inuzuka, N., D. P. Domning, and C. E. Ray. 1995. Summary of fractionation in body water of large mammals. Geochimica et taxa and morphological adaptations of the Desmostylia. Is- Cosmochimica Acta 59:4523–4537. land Arc 3:522–537. Bryant, J. D., D. S. Jones, and P. A. Mueller. 1995. Influence of Kennedy, B. P., C. L. Folt, J. D. Blum, and C. P. Chamberlain. freshwater flux on 87Sr/86Sr chronostratigraphy in marginal 1997. Natural isotope markers in salmon. Nature 387:766–767. marine environments and dating vertebrate and invertebrate Kennedy, M. J., O. A. Chadwick, P. M. Vitousek, L. A. Derry, and faunas. Journal of Paleontology 69:1–6. D. M. Hendricks. 1998. Changing sources of base cations dur- Capo, R. C., and D. J. DePaolo. 1990. Seawater strontium isotopic ing ecosystem development, Hawaiian Islands. Geology 28: variations from 2.5 million years ago to the present. Science 1015–1018. 249:51–55. Koch, P. L., A. N. Halliday, L. M. Walter, R. F. Stearly, T. J. Hus- Capo, R. C., O. A. Chadwick, and D. M. Hendricks. 1994. Con- ton, and G. R. Smith. 1992. Sr isotopic composition of hy- straining atmospheric inputs and in situ weathering in soils droxyapatite from recent and fossil salmon: the record of life- developed along a climate gradient using Sr isotopes. In M. time migration and diagenesis. Earth and Planetary Science A. Lanphere, G. B. Dalrymple, and B. D. Turin, eds. Abstracts Letters 108:277–287. of the Eighth International Conference on Geochronology, Koch, P. L., J. Heisinger, C. Moss, R. W. Carlson, M. L. Fogel, and Cosmochronology, and Isotope Geology. U.S. Geological Sur- A. K. Behrensmeyer. 1995. Isotopic tracking of change in diet vey Circular 1107:1–47. Denver, Colo. and habitat use in African Elephants. Science 267:1340–1343. Capo, R. C., B. W. Stewart, and O. A. Chadwick. 1998. Strontium Koch, P. L., N. Tuross, and M. L. Fogel. 1997. The effects of sam- isotopes as tracers of ecosystem processes: theory and meth- ple treatment and diagenesis on the isotopic integrity of car- ods. Geoderma 82:197–225. bonate in biogenic hydroxylapatite. Journal of Archaeological Cerling, T. E., and J. M. Harris. 1999. Carbon isotope fraction- Science 24:417–429. ation between diet and bioapatite in ungulate mammals and Kohn, M. J. 1996. Predicting animal ␦18O: accounting for diet implications for ecological and paleoecological studies. Oec- and physiological adaptation. Geochimica et Cosmochimica ologia 120:347–363. Acta 60:4811–4829. Clementz, M. T., and P. L. Koch. 2001. Differentiating aquatic Lee-Thorp, J. A. 2000. Preservation of biogenic carbon isotopic mammal habitat and foraging ecology with stable isotopes in signals in Plio- Pleistocene bone and tooth mineral. In S. H. tooth enamel. Oecologia 129:461–472. Ambrose and K. A. Katzenberg, eds. Biogeochemical ap- Cole, M., and D. P. Domning. 1998. Locomotor and positional proaches to paleodietary analysis. Advances in archaeological adaptations in the desmostylian genera Desmostylus and Pa- Science 5:89–116. Plenum, New York. leoparadoxia. Journal of Vertebrate Paleontology 18(Suppl. to Lee-Thorp, J. A., J. C. Sealy, and N. J. van der Merwe. 1989. Stable No. 3):35A. carbon isotope ratio differences between bone collagen and Domning, D. P. 2001. Sirenians, seagrasses, and Cenozoic eco- bone apatite, and their relationship to diet. Journal of Ar- logical change in the Caribbean. Palaeogeography, Palaeocli- chaeological Science 16:585–599. matology, Palaeoecology 166:27–50. Longinelli, A. 1984. Oxygen isotopes in mammal bone phos- Domning, D. P., C. E. Ray, and M. C. McKenna. 1986. Two new phate; a new tool for paleohydrological and paleoclimatolog- Oligocene desmostylians and a discussion of Tethytherian ical research? Geochimica et Cosmochimica Acta 48:385–390. systematics. Smithsonian Contributions to Paleontology59:1– Luz, B., and Y. Kolodny. 1985. Oxygen isotope variations in 56. phosphate of biogenic apatites. IV. Mammal teeth and bones. Domning, D. P., and H. Furusawa. 1992. Summary of taxa and Earth and Planetary Science Letters 75:29–36. distribution of Sirenia in the North Pacific Ocean. Island Arc Luz, B., Y. Kolodny, and M. Horowitz. 1984. Fractionation of ox- 3:506–512. ygen isotopes between mammalian bone-phosphate and en- Dupras, D. L. 1985. Life beneath the Temblor Sea: Sharktooth vironmental drinking water. Geochimica et Cosmochimica Hill, Kern County, California. California Geology 38:147–154. Acta 48:1689–1693. Estes, J. A., and P. D. Steinberg. 1988. Predation, herbivory, and MacFadden, B. J., and T. E. Cerling. 1996. Mammalian herbivore kelp evolution. Paleobiology 14:19–36. communities, ancient feeding ecology, and carbon isotopes: a Hemminga, M. A., and M. A. Mateo. 1996. Stable carbon iso- 10 million-year sequence from the Neogene of Florida. Journal topes in seagrasses: variability in ratios and use in ecological of Vertebrate Paleontology 16:103–115. studies. Marine Ecology Progress Series 140:285–298. MacFadden, B. J., Y. Wang, T. E. Cerling, and F. Anaya. 1994. Hobson, K. A. 1987. Use of stable-carbon isotope analysis to es- South American fossil mammals and carbon isotopes: a 25 HABITAT AND DIETARY PREFERENCES OF THE EXTINCT TETHYTHERE DESMOSTYLUS 519

million-year sequence from the Bolivian Andes. Palaeogeog- non-diffusive inorganic carbon uptake by phytoplankton in raphy Palaeoclimatology Palaeoecology 107:257–268. an upwelling environment. Deep-Sea Research I 48:79–94. McLeod, S. A., and L. G. Barnes. 1984. Fossil desmostylians. Pp. Ravens, J. A., A. M. Johnston, J. E. Kubler, R. Korb, S. G. McInroy, 39–44 in B. Butler, J. Grant, and C. J. Stadum, eds. The natural L. L. Handley, C. M. Scrimgeour, D. I. Walker, J. Beardall, M. science of Orange County. Memoirs of the Natural History Vanderklift, S. Fredriksen, and K. H. Dunton. 2002. Mecha- Foundation of Orange County. Natural History Foundation of nistic interpretation of carbon isotope discrimination by ma- Orange County, Newport Beach, Calif. rine macroalgae and seagrasses. Functional Plant Biology 29: Miller, E. K., J. D. Blum, and A. J. Friedland. 1993. Determination 355–378. of soil exchangeable-cation loss rates using Sr isotopes. Na- Repenning, C. A. 1965. Stanford fossil; studied by U.S.G.S. Min- ture 362:438–441. eral Information Service 18:124–125. Muhs, D. R., C. A. Bush, R. R. Tracy, and C. C. Stewart. 1990. Roe, L. J., J. G. M. Thewissen, J. Quade, J. R. O’Neil, S. Bajpai, A. Geochemical evidence of Saharan dust parent material for Sahmi, and S. T. Hussain. 1998. Isotopic approaches to un- soils developed on Quaternary limestones of Caribbean and derstanding the terrestrial-to-marine transition of the earliest western Atlantic islands. Quaternary Research 33:157–177. cetaceans. Pp. 399–422 in J. G. M. Thewissen, ed. The emer- Nelson, B. K., M. J. DeNiro, M. J. Schoeninger, D. J. DePaolo, and gence of whales. Plenum, New York. P. E. Hare. 1986. Effects of diagenesis on strontium, carbon, Schmitz, B., S. L. Ingram, D. T. Dockery III, and G. Aberg. 1997. 87 86 nitrogen, and oxygen concentration and isotopic composition Testing Sr/ Sr as a paleosalinity indicator on mixed marine, brackish-water and terrestrial vertebrate skeletal apatite in of bone. Geochimica et Cosmochimica Acta 50:1941–1949. late Paleocene–early Eocene near-coastal sediments, Missis- Novacek, M. J., and A. R. Wyss. 1987. Selected features of the sippi. Chemical Geology 140:275–287. desmostylian skeleton and their phylogenetic implications. Schwarcz, H. P., and M. J. Schoeninger. 1991. Stable isotope anal- American Museum of Novitates 2870:1–8. yses in human nutritional ecology. Yearbook of Physical An- O’Leary, M. H. 1988. Carbon isotopes in photosynthesis. Bio- thropology 34:283–322. science 38:328–336. Vennemann, T. W., E. Hegner, G. Cliff, and G. W. Benz. 2001. Iso- Osmond, C. B., N. Valaane, S. M. Haslam, P. Uotila, and Z. Rok- topic composition of recent shark teeth as a proxy for envi- ␦13 sandic. 1981. Comparisons of C values in leaves of aquatic ronmental conditions. Geochimica et Cosmochimica Acta 65: macrophytes from different habitats in Britain and Finland: 1583–1599. some implications for photosynthetic processes in aquatic Wang, Y., and T. E. Cerling. 1994. A model of fossil tooth and plants. Oecologia 50:117–124. bone diagenesis: implications for paleodiet reconstruction Peterken, C. J., and C. A. Conacher. 1997. Seed germination and from stable isotopes. Palaeogeography Palaeoclimatology Pa- recolonisation of Zostera capricorni after grazing by dugongs. laeoecology 107:281–289. Aquatic Botany 59:333–340. Zachos, J., M. Pagani, L. Sloan, E. Thomas, and K. Billups. 2001. Rau, G. H., F. P. Chavez, and G. E. Friederich. 2001. Plankton Trends, rhythms, and aberrations in global climate 65 Ma to 13C/12C variations in Monterey Bay, California: evidence of present. Science 292:686–693.