Stable isotope evidence for changes in dietary niche ...sites.coloradocollege.edu/hfricke/files/2012/10/...Paleobiology,34(4), 2008, pp. 534–552 Stable isotope evidence for changes

� 2008 The Paleontological Society. All rights reserved. 0094-8373/08/3404-0007/$1.00

Paleobiology, 34(4), 2008, pp. 534–552

Stable isotope evidence for changes in dietary niche partitioningamong hadrosaurian and ceratopsian dinosaurs of theHell Creek Formation, North Dakota

Henry C. Fricke and Dean A. Pearson

Abstract.—Questions related to dinosaur behavior can be difficult to answer conclusively by usingmorphological studies alone. As a complement to these approaches, carbon and oxygen isotoperatios of tooth enamel can provide insight into habitat and dietary preferences of herbivorous di-nosaurs. This approach is based on the isotopic variability in plant material and in surface watersof the past, which is in turn reflected by carbon and oxygen isotope ratios of animals that ingestedthe organic matter or drank the water. Thus, it has the potential to identify and characterize dietaryand habitat preferences for coexisting taxa.

In this study, stable isotope ratios from coexisting hadrosaurian and ceratopsian dinosaurs of theHell Creek Formation of North Dakota are compared for four different stratigraphic levels. Isotopicoffsets between tooth enamel and tooth dentine, as well as taxonomic differences in means and inpatterns of isotopic data among taxa, indicate that primary paleoecological information is pre-served. The existence of taxonomic offsets also provides the first direct evidence for dietary nichepartitioning among these herbivorous dinosaur taxa. Of particular interest is the observation thatthe nature of this partitioning changes over time: for some localities ceratopsian dinosaurs havehigher carbon and oxygen isotope ratios than hadrosaurs, indicating a preference for plants livingin open settings near the coast, whereas for other localities isotope ratios are lower, indicating apreference for plants in the understory of forests. In most cases the isotope ratios among hadrosaursare similar and are interpreted to represent a dietary preference for plants of the forest canopy.The inferred differences in ceratopsian behavior are suggested to represent a change in vegetationcover and hence habitat availability in response to sea level change or to the position of river dis-tributaries. Given our current lack of taxonomic resolution, it is not possible to determine if dietaryand habitat preferences inferred from stable isotope data are associated with single, or multiple,species of hadrosaurian/ceratopsian dinosaurs.

Henry C. Fricke. Department of Geology, Colorado College, Colorado Springs, Colorado 80903.E-mail: [email protected]

Dean A. Pearson. Department of Paleontology, Pioneer Trails Regional Museum, Bowman, North Dakota58623

Accepted: 21 May 2008

Introduction

The Hell Creek Formation (HCF) is a fossil-rich package of terrestrial sedimentary rocksof latest Cretaceous age that is exposedthroughout Montana, North Dakota, andSouth Dakota. Detailed studies in southwest-ern North Dakota indicate that a variety ofsedimentary rock types are found throughoutthe HCF in this area, including poorly ce-mented channel and crevasse-splay sand-stones, rooted siltstones, gray to brown mud-stones, and some carbonaceous-rich shales, allof which are thought to represent meanderingand laterally accreting fluvial channel systemsand associated floodplains (Fastovsky 1987;Murphy et al. 2002). Oxygen isotope datafrom bivalves indicate that some of these larg-

er ‘trunk’ rivers drained distant high-eleva-tion areas, whereas smaller tributaries hadmore local sources of recharge (Dettman andLohmann 2000). Invertebrate assemblagesfrom the HCF reflect variable salinities thatare consistent with these rivers being some-times brackish and thus adjacent to an openmarine interior seaway (Hartman and Kirt-land 2002; Murphy et al. 2002), an interpre-tation that is supported by the occasional oc-currence of marine-tolerant animals such assharks, rays, and sawfish, as well as hesperor-nithiform birds in channel deposits (Murphyet al. 2002; Pearson et al. 2002).

This near-coastal setting offered a variety ofmicrohabitats for plants to occupy, and a di-verse megaflora dominated by angiospermshas been described for this area (Johnson

535STABLE ISOTOPE EVIDENCE FOR CHANGES

2002). Because most of these plants do nothave extant relatives, and because some de-gree of taphonomic mixing of leaves from dif-ferent habitats likely occurred (Johnson 2002),the exact nature of these Late Cretaceous plantcommunities remains somewhat obscure.Nevertheless, most of the preserved HellCreek vegetation indicates that channel andfloodplain environments were dominated bywoodland forests made up of small to medi-um-sized trees, and that small herbaceous andshrubby plants occupied point bars of largeriver systems (Johnson 2002). In addition tosuch spatial differences in plant communities,there is evidence that the floral composition offorests changed over time, most likely in re-sponse to changes in climatic conditions(Johnson 2002). Mean annual temperatures inparticular are estimated to have varied from8� to 18�C in this area during Hell Creek time(Wilf et al. 2003).

In turn, plants of the HCF supported a va-riety of large herbivorous dinosaurs, with cer-atopsians and hadrosaurids being among thebest represented in southwestern North Da-kota (Pearson et al. 2002). For example, the cer-atopsians Torosaurus latus and Triceratops hor-ridus have been found in southwestern ND(Pearson et al. 2002), and the hadrosaurids Ed-montosaurus regalis, Edmontosaurus annectens,and ?Parasaurolophus walkeri have been identi-fied in the HCF of eastern Montana (Weisham-pel et al. 2004a). With these and other herbiv-orous taxa occupying the same river channel–floodplain landscapes, important questionsarise as to how these animals interacted witheach other and their surrounding environ-ment. Did they compete for plant resources, ordid they partition resources by occupying dif-ferent microhabitats or eating different foods?Answers to these questions not only allow fora better understanding of Mesozoic ecosys-tems, but they also allow for broad compari-sons with mammal behavior during the Ce-nozoic.

Much of what is known about general di-nosaur behavior comes from the study of skel-etal morphology, biogeography, and copro-lites (Lehman 1987, 2001; Brinkman et al.1998; Dodson et al. 2004; Horner et al. 2004;Fastovsky and Smith 2004; Chin et al. 1998;

Varricchio 2001). In this paper we use a com-plementary approach of studying the dietaryand habitat preference of animals—stable iso-tope analyses of fossilized remains. This ap-proach takes advantage of the fact that carbonisotope ratios of C3 plants and oxygen isotoperatios of surface waters can vary significantlyover any given landscape, and that animals re-cord the isotopic characteristics of the part ofthe landscape they occupy when they ingestorganic material and drink from surface waterreservoirs (see below). Isotopic studies ofmodern terrestrial ecosystems (Bocherens etal. 1996), including C3-only ecosystems (Boch-erens 2003; Cerling et al. 2004), indicate thattaxonomic offsets among taxa do reflect nichepartitioning, and taxonomic offsets observedfor ancient C3 ecosystems have been inter-preted in a similar manner (Clementz et al.2003; MacFadden and Higgens 2004; Botha etal. 2005; Feranec and MacFadden 2006). There-fore it is quite likely that herbivorous dino-saurs preferring open, channel-point bar hab-itats of the HCF landscape may have differentisotopic characteristics than those occupyingthe canopy or understory of floodplain for-ests. To test the applicability of this stable iso-tope approach to studying the behavior ofHCF dinosaurs, isotopic comparisons aremade among ceratopsians and hadrosauridsfrom North Dakota, which are in turn com-pared with carbon isotope data collected fromancient plant material, as represented by bulksedimentary organic matter. These data indi-cate that these herbivores had gut physiolo-gies different from those of mammals, andprovide evidence for niche partitioningamong ceratopsians and hadrosaurids thatvaried in concert with position on the land-scape.

Background: Stable Isotope Ratios,Dinosaurs, and Niche Partitioning

As noted above, carbon isotope ratios of C3

plants and oxygen isotope ratios of surfacewaters can vary over any given landscape.Plants using the C3 (Calvin) pathway presum-ably dominated Mesozoic ecosystems (Sageand Monson 1999) and are characterized by alarge isotopic discrimination between organic

536 HENRY C. FRICKE AND DEAN A. PEARSON

material and CO2 in the atmosphere. This iso-topic discrimination can vary among taxa,with plants such as gymnosperms typicallyhaving higher carbon isotope values than oth-ers such as ferns (Tieszen 1991; Heaton 1999).Environmental conditions can cause carbonisotope ratios of specific C3 plants to vary be-cause these ratios are sensitive to the amountof CO2 in a leaf cell. In turn, concentrations ofCO2 in a leaf cell are influenced a great dealby the opening and closing of leaf stomata.Stomata are more likely to remain closedwhen environmental factors such as temper-ature, water availability, salinity, nutrientavailability and light intensity are such thatwater needs to be conserved (O’Leary 1988;O’Leary et al. 1992; Farquhar et al. 1989; Ties-zen 1991). The location of a plant under aclosed forest canopy can also affect carbonisotope ratios of plant material, as CO2 in thissetting exhibits lower carbon isotope ratiosthan the open atmosphere due to plant res-piration and decomposition on or near the for-est floor (van der Merwe and Medina 1991).

Likewise, oxygen isotope ratios of waters instreams, lakes, and leaves also vary signifi-cantly in response to environmental factorssuch as temperature and aridity, and to the hy-drological ‘‘history’’ of air masses that aresupplying precipitation to these surface waterreservoirs (e.g., Epstein and Meyada 1953;Dansgaard 1964; Rozanski et al. 1993; Gat1996). At present, oxygen isotope ratios ofglobal precipitation ranges from �0‰ to�30‰ (Dansgaard 1964; Rozanski et al.1993). The primary causes of isotopic vari-ability in surface waters are (1) the preferen-tial incorporation of 18O into condensate aswater is precipitated and removed from cool-ing air masses, and (2) incorporation of 16Ointo vapor as evaporation of water bodiestakes place. Lastly, precipitation from large ar-eas and over long periods of time can bemixed together during the formation of lakes,soil and ground waters, and larger rivers.Thus, depending on the mix of plants, watersand local environmental conditions, a mosaicof isotopic ‘‘domains’’ can exist over any givenlandscape (Fig. 1), including that associatedwith the Hell Creek Formation.

Animals living in any given isotopic do-

main record the isotopic characteristics of itwhen they ingest organic material and drinkfrom surface water reservoirs and then formbioapatite [Ca5(PO4, CO3)3(OH, CO3)]), whichis a major component of tooth enamel, toothdentine, bone, and body scales of some fish.Carbon found in the carbonate phase of bio-apatite is related to ingested organic material,such as plants in the case of herbivores (Kochet al. 1994; Koch 1998; Cerling and Harris1999; Passey et al. 2005), and carbon isotoperatios are influenced by how and which or-ganic compounds are utilized during diges-tive and metabolic processes (DeNiro and Ep-stein 1978; Gannes et al. 1998; Hedges 2003;Jim et al. 2004; Passey et al. 2005). Oxygen invertebrate bioapatite has sources primarily iningested water and atmospheric oxygen thatcontribute to blood/metabolic water (Longi-nelli 1984; Luz and Kolodny 1985; Bryant andFroelich 1995; Kohn 1996; Kohn and Cerling2002). The oxygen isotope ratio of atmosphericoxygen has remained relatively constant overtime and space (Kohn 1996); thus, it does notinfluence oxygen isotope variations in bioapa-tite of vertebrates living in different places ordrinking different waters. Oxygen isotope ra-tios of carbonate are, however, more temper-ature dependent than carbon isotope ratios;thus, variations in animal body temperatureare important to consider. For the purposes ofthis study, it is assumed that dinosaurs uti-lized carbon in the same basic way as extantvertebrates such as birds, reptiles, and mam-mals, and that they were homeothermic, re-gardless of the exact way in which these bodytemperatures were regulated (e.g., Spotila1980; Fricke and Rogers 2000; Amiot et al.2006). It is also assumed that for a single kind ofdinosaur, controls on the utilization of carbonand oxygen did not vary significantly overtime or space.

Because dinosaur bioapatite should providea record of the ‘‘isotopic systematics’’ of thepart of the landscape they occupy, isotopiccomparisons among populations can be usedto address questions of niche partitioning ofresources that are otherwise difficult to ad-dress using fossil morphology alone. In par-ticular, it is possible to interpret relative dif-ferences in isotope ratios among Mesozoic


FIGURE 1. Schematic illustration of how different parts of a single C3 ecosystem can be characterized by distinctisotopic domains (after Fricke 2007). These areas include: coastal freshwater swamps (A); riparian forest canopy(B); streams with precipitation from high elevation (C); open shrubland (D); mixed forest/shrubland (E); zone ofmarine and freshwater mixing (F). Using area E as a baseline, plants living in areas A, C, D, and F are likely to havehigher �13C, while those living in area B are likely to have lower �13C. Reasons for relatively higher �13C valuesinclude slow rates of diffusion of CO2 through water (A); lower concentrations of CO2 in the atmosphere at higherelevations (C); enhanced aridity and thus less moisture availability in open environments (D); osmotic stress inbrackish waters along with the possibility of evaporative tidal flat settings (F). Reasons for relatively lower �13Cvalues include recycling of CO2 in the understory of dense, closed-canopy forests, and the abundant availability offreshwater. Using area E as a baseline, water from A, D, and F are likely to have higher �18O, whereas that from Band C is likely to have lower �18O. Reasons for relatively higher �18O values include evaporation of standing water,particularly in open vegetation or arid settings (A); enhanced evaporation of leaf water in sunnier, windier opensettings (D); mixing of high �18O ocean water with freshwater sourced in precipitation (F). Reasons for relativelylower �18O values include reduced evaporation of leaf water in humid, shady, and still understory settings (B) andcollection of precipitation from higher elevations having lower �18O values (C).

data sets, such as those among taxa from thesame locality as reflecting primary (i.e., an-cient) isotopic differences in ingested foodand water, and thus microhabitat preference.Unfortunately, isotopic overlap among taxa ismore difficult to interpret; it is possible thatthese populations each consumed plants andwater from different sources where isotope ra-tios happened to be similar to one other, orthat they consumed the same plants and wa-ter.

Background: Diagenesis

In order to use stable isotope data to inves-tigate dinosaur ecology, it is important to de-termine if diagenesis, defined here as thechemical alteration of bioapatite after the

death of an animal, obscured original behav-ioral information. Unfortunately, no methoddescribed to date can provide unambiguousevidence whether isotopic alteration has orhas not occurred in fossil bioapatite (Kohnand Cerling 2002). As a result, our goal hereis not to demonstrate that isotopic alteration isabsent. Rather, our goal is to demonstrate thatdiagenesis has not entirely obscured originalpaleoenvironmental, paleoecological, and/orpaleobiological information reflected in stableisotope ratios of biogenic apatite. In this case,we feel that by comparing isotopic data (1) be-tween enamel and dentine from the same fos-sil element, and (2) among vertebrate taxa,and (3) among different microsite localities, itis possible to make a strong case that original


isotopic information is still preserved to somedegree in enamel carbonate.

To understand the rationale for comparingdifferent isotopic data sets, a basic knowledgeof diagenetic processes, and how these relateto physical and chemical variations withinbioapatite, is necessary. In general, diageneticalteration of bioapatite occurs by two end-member processes: (1) isotopic exchange be-tween biogenic apatite and surrounding fluidscontaining H2O, HCO3

�, CO2, CH4, and (2)dissolution and/or addition of secondary ap-atite and carbonate (e.g., Zazzo et al. 2004).The former requires that C–O bonds of anioniccomplexes within apatite be broken and thenreformed so that isotope exchange may occur.In addition, for the isotope ratio of carbonatein recrystallized apatite to differ significantlyfrom initial ratios, the temperature of this pro-cess must be significantly different from thatof formation, or isotopic exchange must occurin the presence of C and O from an externalsource that has an isotope ratio much differentfrom that found in the primary carbonatecomplex. In the case of secondary mineral pre-cipitation, biogenic apatite may retain its orig-inal isotope ratios, but this primary signal canbe overwhelmed. Secondary carbonate min-erals may precipitate from ground waters thatare isotopically much different from body wa-ter, and at temperatures that are much differ-ent from those in the body of an animal. As aresult, isotope ratios of secondary mineralsmay not be the same as those of unaltered bio-genic apatite, and the degree of isotopic alter-ation observed will depend on the percentageof secondary mineral present.

What follows from this overview is thatskeletal remains with high porosities may besubjected to greater fluxes of exogenous fluids,whereas those with smaller apatite crystalshave much more surface area available to un-dergo isotopic exchange and more volumeavailable for precipitation of secondary phos-phates and carbonates. Of the common skele-tal materials, bone has very small apatite crys-tals tens of nanometers in length, and a frame-work of organic collagen that makes up �30%of unaltered bone (Hillson 1986). Tooth den-tine and the dentine-like material underlyinggarfish scales are characterized by similar

crystal sizes, but less collagen. In contrast,enamel and the homologous ganoine of somefish scales are made up of larger apatite crys-tals hundreds of nanometers in length, andonly contain �3% original organic material(Hillson 1986; Zylberberg et al. 1997). Organiccollagen is likely to be altered or removed ear-ly after burial, thus providing a pathway forfluids. As a result, bone is most likely to besusceptible to diagenetic processes (Nelson etal. 1986; Kolodny et al. 1996; Kohn and Cerling2002; Trueman and Tuross 2002; Trueman etal. 2004), tooth dentine is less so, and toothenamel and scale ganoine are least likely to beaffected.

For these reasons, enamel is generally con-sidered the best material to analyze in orderto obtain primary isotopic information. Nev-ertheless, anecdotal concerns have been raisedthat the ‘‘old’’ and ‘‘thin’’ enamel of dinosaurteeth makes it more prone to interaction withfluids and thus more susceptible to alterationthan the ‘‘thick’’ enamel of Cenozoic mam-mals. At present, such concerns are unfound-ed, as no stable isotope data have been pub-lished that demonstrate the existence of tem-poral or thickness ‘‘thresholds’’ at which al-teration of enamel becomes problematic. Infact, indirect evidence suggests the opposite.In the case of time, geochemical evidence fromtrace elements suggests that when diagenesisoccurs, it does so thousands to hundreds ofthousands of years after burial (Trueman andTuross 2002; Trueman et al. 2004). Thereforeall pre-Quaternary fossil material is equallysusceptible to diagenetic processes, not justmaterial of Mesozoic age. In the case of thick-ness, no authors publishing isotopic data fromCenozoic teeth have reported that they dis-card surface enamel, and thus ‘‘altered’’ sur-face enamel should have negatively affectedall such studies. Given the success of stableisotope methods in addressing paleobiologi-cal questions, however, such widespread con-tamination due to altered surface enamel doesnot appear to be a problem.

Our feeling is that the geologic history ofsediments hosting fossilized remains is farmore important when considering diagenesis.If there is evidence that sediments have un-dergone deep burial or have experienced a


TABLE 1. Fossil microsite localities sampled as part ofthis study, as described in Pearson et al. 2002. Meter lev-els are below the K/T boundary in the areas, and ap-proximate ages were calculated using the sedimentationrates and ages described in Hicks et al. 2002. Althoughfossil leaves are not found associated with any of theselocalities, they can be placed in a megafloral zone (John-son 2002) based on their stratigraphic position.

Site number Meter level Age (Ma) Floral zone

V87006 28.7 65.855 HC IIV86002 28.8 65.86 HC IIV92067 38.9 65.977 HC IbV92025 73 66.387 HC IaV89004 81.7 66.491 HC Ia

great deal of fluid flow (e.g., veins, secondarymineralization), then there is a greater prob-ability that fossils in these sediments mayhave undergone isotopic modification, regard-less of the exact age or type of fossils present.Because the movement of fluids through sed-iments can vary to a large extent even oversmall areas, researchers must consider dia-genesis on a case-by-case basis using methodssuch as those described above and those de-scribed by Kohn and Cerling (2002), ratherthan relying on a priori assumptions and ar-bitrarily defined thickness and age thresholds.

Geologic Setting and Materials

Fossil remains were collected from withinthe Hell Creek Formation in the general areaof Marmarth, North Dakota (Sheehan et al.1991, 2000; White et al. 1998; Pearson et al.2001, 2002). The formation is approximately100 m thick in this area, representing the last�1.4 Myr of the Maastrichtian Epoch of theCretaceous Period (Hicks et al. 2002). Verte-brate fossil samples were obtained primarilyby surface collection of microsite bonebeds,which typically consist of small resistant ele-ments such as dinosaur and crocodilian teeth,fish scales, and crocodilian and turtle scutes,and which occur as lags associated with chan-nel and crevasse-splay deposits in the part ofthe HCF investigated here (Pearson et al.2002). These types of hardpart accumulations,with their great abundance of dissociatedskeletal material, are ideal for the comparativeapproach we followed, because from a taph-onomic perspective (see Badgley 1986) eachspecimen analyzed can be assumed to repre-sent a distinct individual unless associationamong elements can be demonstrated (e.g.,matching breaks on two separate specimens).Furthermore, the most taxonomically com-plete sampling for representation of a paleo-community is found within channel and near-channel assemblages because they incisethrough and most efficiently collect the flood-plain deposits (Sheehan et al. 1991; Pearson etal. 2002). Particular fossils collected includeshed and fragmentary teeth of hadrosaurianand ceratopsian dinosaurs, although the exacttaxonomic identification of shed teeth is dif-ficult to determine. For the study area in gen-

eral, hadrosaurian remains are indeterminatebeyond the family level, whereas the ceratop-sians Torosaurus and Triceratops are the mostcommon identifiable taxa. In addition to teeth,we collected ganoid scales of fish (family Lep-isosteidae, referred to as ‘‘gar’’ in this paper)as well as multiple samples of bulk sediment.Plant fossils are generally preserved as im-pressions, and therefore we did not sampleleaf cuticle. No authigenic carbonates (e.g.,spar, micrite) were found in any of the othersediments or fossils studied, and thus couldnot be collected.

Sample collection was focused on five sep-arate microsites that are located from 28 to 81meters below the K/T boundary and within�50 km of each other (Pearson et al. 2002) (Ta-ble 1). Given the limited exposure of outcropboth laterally and stratigraphically that ischaracteristic of this part of North Dakota, itis not always possible to resolve subtle differ-ences in depositional environment (e.g., chan-nel of a large versus a small river; mud of alake, floodplain, or swamp) on the basis ofsedimentological and paleontological obser-vations from each site. There are no floral lo-calities directly associated with these five mi-crovertebrate localities, which most likely re-flects different taphonomic processes for ver-tebrate hardparts than for more delicate plantremains. It is possible to place the five locali-ties in association with known floral zonesthat are defined for the HCF in general by thepresence and absence of certain plant mor-photypes (Johnson 2002) (Table 1), but eventhese zones are associated with anywherefrom 10 to 50 morphotypes. Therefore neither


the sediments nor megaflora associated withour vertebrate localities can provide detailedinformation regarding dinosaur dietary orhabitat preferences. Although these limita-tions deprive us of an independent ‘‘tests’’ ofthe behavioral interpretations we make belowusing stable isotope data, they highlight ex-actly why such geochemical data are of value.

Methods and Results

Using a Dremel drill with diamond-tippedbits, we took samples for analysis from den-tine and enamel from dinosaur teeth, and den-tine and ganoine from gar scales (the latter be-ing homologous to tooth enamel [Zylberberget al. 1997]). Enamel from ceratopsian teeth isgenerally thicker than that from hadrosaurteeth and gar ganoine, which ranges from�0.5 to �1.5 mm. Enamel and ganoine thick-ness, however, varies from element to element.Carbon and oxygen isotope ratios of ingestedplant matter and water likely varied season-ally by a few per mil (Fricke and O’Neil 1996;Fricke et al. 1998; Sharp and Cerling 1998;Kohn et al. 1998; Straight et al. 2004), and bulksamples of tooth enamel collected from frag-mentary material may record only some ofthis variability. To help overcome this poten-tial problem, we collected and analyzed mul-tiple teeth from each locality. Such samplingof multiple bulk samples allows all the sea-sonal variability experienced by a single pop-ulation to be ‘‘captured’’ (Clementz and Koch2001).

Carbon and oxygen isotope ratios of milli-gram-sized enamel and dentine samples weresoaked for 24 hours in 0.1 N acetate-buffer so-lution, rinsed four times in distilled water, anddried (Koch et al. 1997). Stable isotope ratiosare reported as �13C and �18O values, where �� [Rsample/(Rstandard � 1)] ·1000‰, and thestandard is VPDB for carbon and VSMOW foroxygen. �18O and �13C of tooth enamel car-bonate were measured with an automated car-bonate preparation device (KIEL-III) coupledto a Finnegan MAT 252 isotope ratio massspectrometer at the University of Arizona andat the University of Iowa. Powdered sampleswere reacted with dehydrated phosphoricacid under vacuum at 70�C (UA) or 75�C (UI)the presence of silver foil. The isotope ratio

measurement is calibrated on the basis of re-peated measurements of NBS-19, NBS-18 andin-house powdered carbonate standards. An-alytical precision is �0.1‰ for both �18O and�13C (1s). The carbonate –CO2 fractionation forthe acid extraction is assumed to be identicalto calcite.

Both isolated organic fragments and bulksediment samples were soaked in 0.1M HClfor three hours, rinsed in distilled water fourtimes, and dried. �13C was measured on a con-tinuous-flow gas-ratio mass spectrometer(Finnegan Delta PlusXL). Samples were com-busted with an elemental analyzer (Costech),which was coupled to the mass spectrometerat the University of Arizona. Standardizationis based on NBS-22 and USGS-24. Precision isbetter than �0.06 for �13C (1), based on re-peated internal standards.

Carbon and oxygen isotope data for dino-saur tooth enamel and dentine, and gar scaleganoine and dentine, are given in Appendix 1(see online supplementary material athttp://dx.doi.org/10.1666/08020.s1). Stableisotope data for sedimentary organic materialare presented in Appendix 2. Statistical com-parisons of variance between sample popula-tions were made using a simple F-test, and theappropriate student t-test was then used tocompare mean values for sample populations.We used Microsoft Excel for these statisticalanalyses, as well as for determining correla-tion coefficients. Carbon and oxygen isotopiccomparisons were made (1) between enamel/ganoine and dentine from the same type ofanimal from each horizon; (2) between enameland ganoine from coexisting ceratopsians,hadrosaurs, and gar at each locality; and (3)between enamel and ganoine from ceratop-sians, hadrosaurs, and gar grouped by in-ferred environmental preference (see below).Results are summarized in Appendix 3.

Discussion

Comparisons of isotopic data obtained fromenamel, ganoine, dentine, and organic mate-rial can be used to address questions regard-ing diagenesis, dinosaur physiology and di-nosaur behavior. Each of these topics is dis-cussed in more detail below.

Diagenesis. In an effort to demonstrate that


diagenesis has not entirely obscured originalbehavioral information that is reflected in sta-ble isotope ratios of bioapatite, comparisons ofisotopic data are made (1) between enameland dentine from the same fossil element, (2)among vertebrate taxa, and (3) among differ-ent microsite localities.

As noted above, the rationale behind a com-parison of isotopic data from enamel and den-tine is that isotope ratios of dentine shouldrepresent a more diagenetically altered ma-terial. Thus isotopic differences between den-tine and enamel can be used to address thepreservation of the latter. In general, dentine ischaracterized by higher �18O and �13C, partic-ularly carbon isotope ratios, relative to enamel(Fig. 2); lines drawn between average valuescan illustrate these relations more clearly (sol-id lines; Fig. 2). Assuming unaltered enameland dentine of teeth had the same isotope ra-tios, the trend in dentine toward higher iso-tope ratios is interpreted to represent a vari-able exposure of dentine to diagenetic fluids,more isotopic exchange, and/or secondarymineral formation in dentine compared toenamel, and thus isotope ratios closer to thediagenetic end-member (which in this areamust be characterized by moderately high�18O and very high �13C). It is important tonote that although dentine is more susceptibleto diagenetic alteration, there is no reason toexpect that all diagenetically altered sampleswill obtain the exact same isotope ratios, orthat they will necessarily have isotope ratiosthat fall outside the range for enamel. It is pos-sible that variable porosity and permeabilityof surrounding sediment variably inhibitedthe flow and amount of diagenetic fluids thatinteracted with different tooth fragments.Furthermore, the precipitation of secondarycarbonate can itself gradually reduce porosityand permeability of skeletal materials (True-man and Tuross 2002; Trueman et al. 2004),thus making it unlikely that all dentine willhave the exact same diagenetic history andthus the exact same isotope ratio. Overall,these general isotopic differences betweenenamel and dentine indicate that originalranges in �18O and �13C of enamel were notmodified to the same degree as for dentine,

and thus diagenetic alteration of enamel isminimal.

Comparisons of enamel-ganoine data onlyamong HCF taxa provide more convincing ev-idence for limited effect of diagenesis onenamel isotope ratios. In particular (1) signif-icant offsets in average �18O and/or �13C areobserved between at least two taxa from everymicrosite, (2) different correlation coefficientsfor �18O and �13C are observed for coexistingtaxa, and (3) intra-populational variance in�18O and �13C is different for coexisting taxa(Fig. 3). Similar kinds of taxonomic differenc-es in average �18O and/or �13C, and in isotopicvariance, have been observed in studies ofmodern and Cenozoic mammalian toothenamel (Bocherens et al. 1996; Bocherens 2003;Clementz et al. 2003, 2006; Cerling et al. 2004;MacFadden and Higgins 2004; Feranec andMacFadden 2006), and they have been inter-preted to reflect physiological and/or ecolog-ical differences among taxa. They are not un-expected in Mesozoic ecosystems because al-though enamel of hadrosaurian and ceratop-sian teeth and ganoine of gar scales aremineralogically and structurally similar, theanimals are characterized by different physi-ologies and/or ecological behaviors.

The simple observation that offsets in aver-age �18O and/or �13C and different correlationcoefficients are preserved among taxa isstrong evidence that any isotopic exchange be-tween groundwaters and enamel-ganoine orsecondary precipitation of carbonate in enam-el-ganoine was not extensive enough to ob-scure original paleoecological and paleobio-logical information. If diagenetic alterationwas a significant occurrence, then the resultshould be uniform isotope ratios, similar cor-relation coefficients, and similar variances forall remains in a single microfossil bonebed, re-gardless of taxonomic affinity. It should be not-ed that it is not necessary to have an a prioriunderstanding of the exact physiological/eco-logical meaning of these taxonomic offsets inorder for them to be useful in addressing theissue of diagenetic alteration (although specif-ic physiological and ecological interpretationsof these offsets are provided below). It is sim-ply enough to know that isotopic offsets anddifferences in correlation coefficients among


FIGURE 2. Comparisons of isotope data from enamel-ganoine (closed symbols) and dentine (open symbols) forhadrosaurians (A), ceratopsians, (B) and gar (C). Regression lines are dashed and average isotope ratios are shownwith an ‘‘.’’ In all cases, dentine is characterized by higher carbon-to-oxygen isotope ratios. These results areconsistent with isotope ratios of dentine having been affected to a larger degree by diagenetic processes, and in-dicate that carbonate from enamel is more likely to have preserved primary isotopic information. Solid lines con-necting average isotope ratios for enamel-ganoine and dentine illustrate the general isotopic trends characteristicof diagenetic alteration.


FIGURE 3. Isotopic differences among enamel and ganoine of coexisting ceratopsians, hadrosaurs, and gar fromfour microsites. Average isotope ratios are represented by a star and the corresponding initial for each animal (C,H, G), and isotopic variability is represented by �2 uncertainty bars. Correlation coefficients (CC) are also pro-vided for each taxon. Limited ceratopsian data from site V87006 (shaded area) are included along with data fromthe nearby site V86002. Level below the K/T is given in meters for each site. Extensive overprinting of primaryisotope signals by diagenetic carbonate should result in uniform isotope ratios. Thus, the within-site differences inaverages and isotopic variability, and in correlation coefficients, are evidence that complete isotopic resetting didnot occur. Note that in the case of site V89004, a single hadrosaurian enamel sample had anomalously high isotoperatios compared to other hadrosaurians, but falls within the range for ceratopsians. This sample may be a ceratop-sian tooth that was incorrectly identified, and it is not included in the average and uncertainty calculations forhadrosaurian data.

animals can be preserved only if isotopic al-teration does not obscure original isotopic in-formation. It should also be noted that isotopicoverlap that is observed among taxa at eachsite does not necessarily imply diagenetic al-teration; instead these particular animalscould have consumed plants and water fromdifferent sources where isotope ratios hap-pened to be similar to one other, or they couldhave fractionated carbon and oxygen in dif-ferent ways that happened to result in similar�18O and �13C of bioapatite.

The last critical observation to make regard-ing diagenesis is that offsets in average �18Oand/or �13C, correlation coefficients, and iso-topic variances among taxa are not the samefor each microsite; rather two basic offset pat-terns can be observed. In one, composed ofsamples from V86002 (29 m) and V92067 (39m), gar ganoine has the highest mean �13C ofall taxa and ceratopsian enamel has the lowestmean �13C and �18O values and these data arecharacterized by a low correlation coefficient(Fig. 4A). In contrast, the general pattern for


FIGURE 4. Two distinct groupings of stable isotope data for ceratopsians, hadrosaurs, and gar from the five sampledmicrosites. Average isotope ratios for each grouping are represented by a star and the corresponding initial for eachanimal (C, H, G). Level below the K/T is given in meters for each site. In one group (A), ceratopsians have higher�18O and �13C than associated gar, as well as a high correlation coefficient (0.91). In the other (B), ceratopsians havelower �13C than associated gar, and a low correlation coefficient (0.27). It is difficult, if not impossible, to providea plausible diagenetic scenario to explain differential fluid flow that could preserve isotopic differences among taxa,yet cause a change in direction and nature of these differences for sites from the same formation located so closeto each other.

sites V87006 (29 m), V92025 (73 m), andV89004 (81 m) consists of gar ganoine with thelowest mean �13C of all taxa and ceratopsianenamel with the highest mean �13C and �18Ovalues (Fig. 4B). Furthermore, there is a strongcovariation between �18O and �13C for theseceratopsians. All of the sampled micrositesare relatively close together, separated strati-graphically by less than �50 m, and separatedlaterally by less than �50 km. In particular,sites V86002 and V87006 are within a meter ofeach other from the same outcrop area (Pear-son et al. 2002). It is difficult, if not impossible,to provide a plausible diagenetic scenario toexplain differential fluid flow that could pre-serve isotopic differences among taxa, yetcause a change in direction and nature of thesedifferences for sites from the same formationlocated so close to each other. Instead, it ismore likely that original paleoecological in-formation is preserved, and that these twopatterns accurately represent the environ-ments or behaviors of animals, which were infact different (see below).

To summarize this discussion of diagenesis,we admit that any one of the isotopic com-parisons described above does not provide

conclusive evidence that original paleoenvi-ronmental, paleoecological, and/or paleobio-logical information is retained in enamel-gan-oine isotope data. However, when isotopic off-sets from all materials, from all taxa, and fromall microsites are considered together as awhole rather than as separate ‘‘tests,’’ themost parsimonious interpretation of thesedata is that primary isotopic signals have beenpreserved.

High Carbon Isotope Ratios. It has been not-ed in other studies of Late Cretaceous herbiv-orous dinosaurs, in particular hadrosaurs,that �13C values are unusually ‘high’ whenconsidered in the context of data from mam-mals living in C3 ecosystems (Stanton-Thomasand Carlson 2003; Fricke et al. 2008), and thiscould be due in part to diagenetic processes(Stanton-Thomas and Carlson 2003). In an ef-fort to determine whether diagenesis is a pos-sible reason for these higher values in spite ofarguments such as those made above, Fricke etal. (2008) compared �13C of hadrosaur toothenamel from two Campanian-aged sites with�13C of associated sedimentary organic matterand paleosol carbonate nodules with the goalof estimating the carbon isotope offset be-


FIGURE 5. �13C of herbivore tooth enamel and �13C of bulk sedimentary organic matter collected from sedimentsassociated with microfossil bonebeds of the Hell Creek Formations. The range in �13C of HCF organic matter issimilar to that of modern C3 plants, although absolute values are several per mil higher than modern values in partbecause �13C of atmospheric CO2 was higher during the Late Cretaceous. The average offset between enamel andpresumed diet is �18‰. This same offset exists regardless of exact age, location, depositional environment, orherbivore species (see also Fricke et al. 2008), and suggests that herbivorous dinosaurs utilized organic carbon ina manner different than modern mammals.

tween diet and tooth enamel (��13Cdiet-enamel).They concluded that ��13Cdiet-enamel was �18‰,which is in fact larger than the �12 to 15‰observed for modern mammals (Koch 1998;Cerling and Harris 1999; Kohn and Cerling2002; Hoppe et al. 2004; Passey et al. 2005), aconclusion supported by isotopic data fromHCF sediments studied here (Fig. 5). Alto-gether, herbivorous dinosaur material and as-sociated organic matter from 8 other micro-fossil bonebeds ranging in age from late Cam-panian to late Maastrichtian and in space fromwestern Montana to western North Dakotahas been analyzed, and the average��13Cdiet-enamel is always �18‰ (Fricke et al.2008: Fig. 5). Such remarkable consistency in-dicates that the observed ��13Cdiet-enamel for di-nosaurs is a primary signal, as the possibilityof ‘convergent diagenesis’ leading to similaroffsets despite differences in time and geo-graphic location is highly unlikely. Thereforecarbon isotope systematics of modern mam-mals do not appear to provide a valid contextfor interpreting data from dinosaurs, and‘high’ �13C values of tooth enamel alone

should not be considered as evidence for in-tense diagenetic alteration of primary signa-tures.

Reasons for a larger ��13Cdiet-enamel for her-bivorous dinosaurs of this study compared tomodern mammals are likely due to taxonomicdifferences in (1) the biogeochemical process-es that take place as carbon from plants is in-corporated into bioapatite, particularly meth-ane production during digestion, and/or (2)which organic compounds in a plant (i.e., pro-teins, carbohydrates, lipids) are actually uti-lized by the animal when forming bioapatite(Gannes et al. 1998; Hedges 2003; Jim et al.2004; Passey et al. 2005). These factors arethought to explain the observations that��13Cdiet-enamel is actually variable for largemammalian herbivores themselves, rangingfrom �12‰ to 15‰ (Koch 1998; Cerling andHarris 1999; Kohn and Cerling 2002; Hoppe etal. 2004; Passey et al. 2005), and that��13Cdiet-eggshell for birds is even greater(�16‰; Johnson et al. 1998). Thus it is possi-ble that hadrosaurs and ceratopsians utilizedorganic compounds in ways different than


herbivorous mammals and birds, or produceda larger percentage of methane in their stom-achs during longer rumination periods, andthese factors led to a larger ��13Cdiet-enamel. Suchimplied differences in dinosaur gut biologyimplied by a larger offset could in turn explainthe apparent ability of hadrosaurs and cera-topsians to get adequate nutrition from low-quality, fibrous plant material thought to beavailable to them (e.g., Dodson et al. 2004; Fas-tovsky and Smith 2004; Horner et al. 2004). Al-though not the main focus of this study, thepaleobiological implications of these resultsare intriguing and are the focus of ongoing re-search.

Partitioning of Hell Creek Habitats among Di-nosaurs. Isotopic offsets between ceratopsianand hadrosaurid tooth enamel (and betweendinosaurs and gar) occur at all stratigraphiclevels, although the nature of these offsets, aswell as patterns in ceratopsian data, differsfrom site to site (Figs. 3, 4). Because oxygenand carbon in bioapatite have a source in in-gested water and food, respectively, these iso-topic patterns do not represent random pat-terns influenced by depositional environ-ments or taphonomic processes; rather, at themost basic level these offsets must reflect ei-ther physiological or ecological differencesamong coexisting animals.

Considering dinosaurs only, different aver-age body temperatures could affect isotopicfractionation of oxygen during bioapatite for-mation, whereas differences in how organiccarbon is utilized could affect carbon isotoperatios of enamel carbonate. It is unlikely, how-ever, that such physiological factors can ac-count for the observed taxonomic offsets.First, hadrosaurs have relatively constant �18Oand �13C values throughout the section, indi-cating that any physiological changes for thistaxon were minimal. In contrast, average �18Oof ceratopsians ranges from ��8‰ to��12‰, and this in turn would require bodytemperatures of ceratopsian populations tovary up to �16�C. It has been suggested thatlarge dinosaurs such as ceratopsians werelikely effective homeotherms because of theirlarge body mass (e.g., Spotila 1980) or becauseof endothermic metabolic activity (Fricke andRogers 2000; Amiot et al. 2006); thus, such

large physiological differences among cera-topsian species are implausible. In the case ofcarbon, average �13C of ceratopsians rangefrom ��9‰ to ��6‰, and these changes in�13C of ceratopsians would imply a dramaticchange in organic carbon utilization amongrelated species that is much larger than differ-ences in ��13Cdiet-enamel observed among taxaeven at the order level in modern mammals(Cerling and Harris 1999; Passey at al. 2005).For these reasons, the observed isotopic off-sets among taxa must be primarily conse-quences of behavioral differences among had-rosaurs and ceratopsians.

Because differences in �13C and �18O oftooth enamel carbonate are assumed to reflect�13C and �18O of ingested plants and surfacewaters, respectively, the most parsimoniousinterpretation of dinosaur data is that hadro-saurs and ceratopsians were utilizing differentmicrohabitats that made up the Hell Creekecosystem. What is especially interesting isthat the isotopic systematics, and thus habitatpreferences, of ceratopsians appears to varyover time. As illustrated in Figure 4, ceratop-sians from sites V89004 (81 m), V92025 (73 m),and V87006 (28 m) have higher �13C and �18Ovalues with a high correlation coefficient rel-ative to hadrosaurs, whereas ceratopsiansfrom sites V92067 (39 m) and V86002 (29 m)have lower �13C and �18O than associated had-rosaurs. The nature of these habitat preferenc-es, and how they may have changed over time,can be inferred from relative differences inisotope ratios and their patterns for each taxacombined with general sedimentological andpaleobotanical observations from the HCF (asnoted above it is not possible to provide in-dependent sedimentological and paleobotan-ical evidence to support these interpretationson a microsite-by-microsite basis).

In the case of sites V89004, V92025, andV87006, higher �18O and �13C and a high cor-relation coefficient (0.91) for ceratopsians arethought to reflect their preference for drinkingand feeding in open habitats—channel bars oflarge rivers, swamps, tidal flats—that mayhave had some connection to the marine en-vironment (Fig. 6, lower box). Sedimentolog-ical data indicate that such settings were char-acteristic of the HCF in general (Fastovsky


FIGURE 6. Schematic illustration of the Hell Creek landscape, the environments occupied by dinosaurs and fish thataccount for isotopic patterns illustrated in Figure 4, and how the location of these environments could shift overtime. For both times 1 and 2, areas closer to the shoreline have a mix of vegetation types in the river channels,riparian forests, freshwater swamps, and tidal marshes. Because of their location relative to the coast, these areascan have a hydrological connection with the marine environment as well as with local precipitation. Farther inland,meandering trunk rivers are associated primarily with forested areas only. Rivers have a hydrological connectionwith upland drainages as well as local precipitation. The inferred relations between these two general kinds ofenvironments and herbivore behavior are highlighted by the upper and lower boxes, which are linked to groupedisotopic data from Figure 4. Stratigraphic-temporal changes in environments, and thus isotopic patterns, can beexplained by changes in sea level, and hence the position of the shoreline, and/or by changes in the position ofrivers and associated distributary networks over time (such as those represented here by the landscape reconstruc-tions for times 1 and 2).

1987; Hartman and Kirtland 2002; Murphy etal. 2002; Pearson et al. 2002), and paleobotan-ical data indicate that sediment bars of largeopen river systems of the HCF were charac-terized by a shrubby and herbaceous flora dif-ferent from that of floodplain forests (Johnson2002). Lacking a forest canopy, plants in thesesettings, as well as those living in other coastalsettings such as tidal marshes or swamps,could have been more susceptible to evapo-rative loss of water from ponds, soils, orleaves, which would shift �18O of these watersto higher values (Sternberg 1989; Gat 1996).Ocean water typically has much higher �18Othan precipitation even in coastal areas (Roz-anski et al. 1993; Gat 1996), and so a mixing ofprecipitation and ocean water in coastal areascould also result in brackish surface waterswith higher �18O values. Evidence that surfacewaters with a wide range in �18O did in factexist for ceratopsians to drink from is provid-

ed by the range in �18O of gar for these sites(Fig. 4A), which is similar to that observed forceratopsians. In the case of carbon, marine in-fluences and open coastal settings can also re-sult in higher �13C of C3 plants. For example,coastal mangroves affected by salt and nutri-ent stress today have �13C values up to 5‰higher than those that are not (McKee et al.2002; Wooller et al. 2003), an effect also ob-served for Cretaceous plant fossils of the sameorder as those found in the HCF (Nguyen Tuet al. 1999; Aucour et al. 2008). Furthermore,leaves of C3 plants exposed to direct sunlight,such as would be found in more open settings,typically have �13C values that are 2–3‰ high-er than those that are not (Lockheart et al.1998). Lastly, aquatic vegetation growing inocean waters typically has higher �13C valuesthan land and aquatic vegetation from terres-trial settings (Boutton 1991; summary in Cle-mentz and Koch 2001). Because both �18O of


surface waters and �13C of plants increase inresponse to more open and marine-influencedconditions, a positive correlation betweenthem, such as that observed for ceratopsians(Fig. 4A), is expected.

In contrast, lower �18O and �13C values ofhadrosaurs from V89004, V92025, and V87006(Fig. 4A) indicate that these animals preferreda microhabitat different from that of coexist-ing ceratopsians, in particular river and flood-plain forests. Again, sedimentological data in-dicate that such depositional environmentswere common on HCF landscape (Fastovsky1987; Murphy et al. 2002), and paleobotanicaldata indicate associated that forests weremade up of small to medium-sized trees(Johnson 2002). Lower �18O and �13C valuessuggest that these forests were not influencedby marine waters and were characterized bymore dense vegetative cover (‘‘fingers’’ of in-terfluve vegetation in Figure 6). Nevertheless,the weak covariation in �18O and �13C impliesthat C3 plants in this habitat experienced arange of environmental conditions, perhapsfrom more humid, shady understory (lower�18O and �13C) to a more exposed canopy topwhere water loss via transpiration was morelikely (higher �18O and �13C). Overall, thesetaxonomic differences in isotope ratios amongceratopsians and hadrosaurs from V89004,V92025, and V87006 are indicative of a spatial,or horizontal, partitioning of food resourcesfrom adjacent areas characterized by differenttypes of marsh, swamp, and forested vegeta-tion (Fig. 6). This interpretation is consistentwith the biogeographic distribution of cera-topsian remains, which suggest that these an-imals preferentially occupied coastal marshesand swamps (Lehman 1987, 2001; Brinkmanet al. 1998)

Focusing now on sites V92067 and V86002,evidence for dietary niche partitioning is stillquite evident, but the nature of this partition-ing is different than for the other sites. In thesecases, ceratopsians generally have lower �18Oand �13C values than coexisting hadrosaurs(Fig. 4B), whose average isotope ratios remainrelatively unchanged with stratigraphic posi-tion (Fig. 4). Because hadrosaurs show thisisotopic consistency, we infer that they occu-pied a similar river-floodplain forest micro-

habitat. Lower �18O and �13C values for cera-topsians suggest that they also occupied sucha closed-forest setting; however, they relied onwater and plant resources from the understo-ry of this forest rather than from the canopy(Fig. 6 upper box). In the case of modernclosed-canopy forests located in tropical re-gions, �13C of C3 plants varies by up to 10‰depending on the plant’s position in the can-opy, with lower plants having lower �13C dueto recycling of CO2 or to more humid, shadedconditions (van der Merwe and Medina 1991;von Fisher and Tieszen 1995; Martinelli et al.1998; Cerling et al. 2004). This pattern in �13Cof C3 plants is mirrored by herbivores, withthose feeding preferentially in the understoryhaving lower �13C of tooth enamel (Cerling etal. 2004). Furthermore, herbivores occupyingunderstory microhabitats are also character-ized by lower �18O compared to those utilizingcanopy plants (Cerling et al. 2004), presum-ably because evaporation of water is reducedby higher humidity and more shade. Overall,taxonomic differences in isotope ratios amongceratopsians and hadrosaurs are again indic-ative of a partitioning of food resources, onlyin this case it is a vertical partitioning withinthe same area (Fig. 6), rather than a horizontalpartitioning between adjacent areas. This in-terpretation is consistent with previous mor-phological studies of these dinosaurs, whichsuggest that in forested settings the size andorientation of ceratopsian skeletons wouldhave led to a preference for understory (�1–2m) vegetation, whereas hadrosaurs wouldhave browsed at low to intermediate heights(2–4 m) (Dodson et al. 2004; Horner et al. 2004;Fastovsky and Smith 2004).

Relation of Niche Partitioning, Available Vege-tation, and Landscape Change. The change inisotopic offsets and thus ceratopsian diet overtime described above begs the question of thecause of this change in behavior. The most ob-vious answer is that the behavior of ceratop-sians simply reflects the type of vegetationavailable to them in any one part of the HellCreek landscape. Reasons for a change in thetype of vegetation present in any one place in-clude sea level rise or fall, and/or a change inthe position of river channels and networks ofdeltaic distributaries. Even a small change in


sea level would result in a broad migration ofdepositional environments either landward orseaward over this broad shallow landscape,and this in turn determines the location ofopen tidal marshes, wide river channels, andconnection to marine water (times 1 and 2;Fig. 6). Alternatively, a change in the positionof river channels and associated landformssuch as deltas and distributaries could pro-duce a similar change in vegetation types inthe absence of a change in sea level (times 1and 2; Fig. 6).

Sedimentological and paleontological evi-dence for such subtle, and apparently short-lived, variations in depositional environmentis difficult to obtain in our field area; never-theless, isotopic data from gar scales are con-sistent with their existence. Just as for terres-trial vertebrates, isotopic ratios of gar scalesreflect those of ingested water and food, anda large difference in �18O, and particularly�13C, for gar between sites V89004-V92025-V87006 and sites V92067-V86002 (Fig. 4) in-dicates that these animals lived in differentsettings. In particular, higher �18O and lower�13C values from V89004-V92025-V87006 aresuggestive of local precipitation, perhapsmodified by evaporation and ocean mixing, asa source of water, and local organic matter asa source of carbon in the rivers and swampsthat were likely occupied by these fish (Fig. 6lower box). In contrast, lower �18O and higher�13C for gar from V92067-V86002 indicate thatwater there had a source in precipitation fromhigher elevation areas and a contribution ofcarbon from the dissolution of marine carbon-ates, both of which are consistent with thesefish living in meandering trunk streams witha direct hydrologic connection to upland areas(Fig. 6, upper box). It is interesting to note thatthe �18O of river water inferred to be in equi-librium with gar bioapatite at 10�C (averagemean annual temperature [MAT] for timestudied time interval [Wilf et al. 2003]) rangesfrom ��16‰ to �7‰ for all sites, and this issimilar to the range in �18O of river waters in-ferred from �18O of bivalves shells collectedfrom HCF localities in Montana (Dettman andLohmann 2000).

Limits of Taxonomic Resolution. Because ofour inability to identify shed teeth beyond the

family level, isotopic data from this study canbe used only to delineate broad dietary dif-ferences and habitat preferences among had-rosaurian and ceratopsian dinosaurs. There-fore, it is possible that individual dinosaurspecies had dietary or habitat preferences thatwe are not able to distinguish with these data.For example, the hypothesized change in cer-atopsian dietary preference with environmentdoes not require that a single ceratopsian tax-on underwent a change in behavior; rather itcould be that different genera or species maybe present during each specific time period,each with its own unique behavior patterns. Infact, two ceratopsian genera, Torosaurus andTriceratops are found in HCF in this area, andon the basis of biogeographic distributionsLehman (1987) suggested that Triceratops pre-ferred tidal marsh/swamp settings more thanTorosaurus, and would thus be the more com-mon taxon in a setting such as those inferredfor sites V89004-V92025-V87006. In the case ofhadrosaurs, it is not possible to determine if asingle species with a generalist behavior is as-sociated with the entire range in isotope ra-tios, or whether multiple species with uniquedietary preferences and smaller isotopic rang-es are instead grouped together. Future stableisotope research involving identifiable speci-mens is required to test these possibilities.

Summary

This study provides the first direct geo-chemical evidence for dietary niche partition-ing among coexisting herbivorous dinosaurs,specifically ceratopsians and hadrosaurs. Fur-thermore, changes in isotopic offsets amongthese taxa suggest that the nature of this nichepartitioning varies in response to changes invegetation structure, with ceratopsians prefer-ring open marsh settings when they are pres-ent along with forests, but preferring under-story vegetation in forest-only environments.The interpretations of stable isotope data pre-sented here are consistent with ceratopsianand hadrosaur behaviors inferred from mor-phological characteristics of these dinosaurs,and demonstrate the potential of stable iso-tope studies to (1) complement other kinds ofpaleontological research, and (2) provide nov-el information regarding late Cretaceous


plant-animal interactions, resource partition-ing among herbivores, and vertebrate com-munity structure as a whole. The ability of sta-ble isotope data to reveal behavioral differ-ences among herbivorous dinosaurs, even inthe face of low taxonomic resolution, bodeswell for the use of this approach in areas withfossils that can be identified to the genus orspecies level.

Acknowledgments

Thanks to K. Johnson for introducing HCFto this topic and field area, and for comment-ing on this research. A. Bercovici made theHell Creek come alive in the last figure, andwe are grateful for his effort. Constructive re-views of an earlier version of this manuscriptwere provided and D. Fastovsky, D. Fox, andan anonymous reviewer. Financial supportwas provided by National Science Foundationgrant EAR-0319041, and American ChemicalSociety–Petroleum Research Fund grant38141-GB8.

Literature CitedAmiot, R., C. Lecuyer, E. Buffetaut, G. Escarguel, F. Fluteau, and

F. Martineau. 2006. Oxygen isotopes from biogenic apatitessuggest widespread endothermy in Cretaceous dinosaurs.Earth and Planetary Science Letters 246:41–54.

Aucour, A.-M., B. Gomez, S. M. F. Sheppard, and F. Thevenard.2008. �13C and stomatal number variability in the Cretaceousconifer Frenelopsis. Palaeogeography, Palaeoclimatology, Pa-leoecology 257:462–473.

Badgley, C. 1986. Counting individuals in mammalian fossil as-semblages from fluvial environments. Palaios 1:328–338.

Bocherens, H. 2003. Isotopic biogeochemistry and the paleo-ecology of the mammoth steppe fauna. Deinsea 9:57–76.

Bocherens, H., P. L. Koch, A. Mariotti, D. Geraads, and J. J. Jae-ger. 1996. Isotopic biogeochemistry (13C, 18O) and mamma-lian enamel from African Pleistocene hominid sites. Palaios11:306–318.

Botha, J., J. Lee-Thorp, and A. Chinsamy. 2005. The palaeoecol-ogy of the non-mammalian cynodonts Diademodon and Cy-nognathus from the Karoo Basin of South Africa, using stablelight isotope analysis. Palaeogeography, Palaeoclimatology,Paleoecology 223:303–316.

Boutton, T. W. 1991. Stable carbon isotope ratios of natural ma-terials. II. Atmospheric, terrestrial, marine, and freshwaterenvironments. Pp. 173–185 in D. C. Coleman and B. Fry, eds.Carbon isotope techniques. Academic Press, New York.

Brinkman, D. B., M. J. Ryan, and D. A. Eberth. 1998. The paleo-geographic and stratigraphic distribution of ceratopsids (Or-nithischia) in the upper Judith River Group of western Can-ada. Palaios 13:160–169.

Bryant, J. D., and P. N. Froelich. 1995. A model of oxygen isotopefractionation in body water of large mammals. Geochimica etCosmochimica Acta 59:4523–4537.

Cerling, T. E., and J. M. Harris. 1999. Carbon isotope fraction-ation between diet and bioapatite in ungulate mammals and

implications for ecological and paleoecological studies. Oec-ologia 120:347–363.

Cerling, T. E., J. A. Hart, and T. B. Hart. 2004. Stable isotope ecol-ogy in the Ituri Forest. Oecologia 138:5–12.

Chin, K., G. M. Tokaryk, G. M. Erickson, and L. C. Calk. 1998.A king-sized theropod coprolite. Nature 393:680–682.

Clementz, M. T., and P. L. Koch. 2001. Differentiating aquaticmammal habitat and foraging ecology with stable isotopes intooth enamel. Oecologia 129:461–472.

Clementz, M. T., K. A. Hoppe, and P. L. Koch. 2003. A paleo-ecological paradox: the habitat and dietary preferences of theextinct tethythere Desmostylus, inferred from stable isotopeanalysis. Paleobiology 29:506–519.

Clementz, M. T., A. Goswami, P. D. Gingerich, and P. L. Koch.2006. Isotopic records from early whales and sea cows: con-trasting patterns of ecological transition. Journal of VertebratePaleontology 26:355–370.

Dansgaard, W. 1964. Stable isotopes in precipitation. Tellus 16:436–468.

DeNiro, M. J., and S. Epstein. 1978. Influence of diet on the dis-tribution of carbon isotopes in animals. Geochimica et Cos-mochimica Acta 42:495–506.

Dettman, D. L., and K. Lohmann. 2000. Oxygen isotope evi-dence for high-altitude snow in the Laramide Rocky Moun-tains of North America during the Late Cretaceous and Pa-leogene. Geology 28:243–246.

Dodson, P., C. A. Forster, and S. D. Sampson. 2004. Ceratopsi-dae. Pp. 494–516 in Weishampel et al. 2004b.

Epstein, S., and T. Mayeda. 1953. Variations in the 18-O contentof waters from natural sources. Geochimica et CosmochimicaActa 4:213–224.

Farquhar, G. D., J. R. Ehleringer, and K. T. Hubick. 1989. Carbonisotope discrimination and photosynthesis. Annual Review ofPlant Physiology and Plant Molecular Biology 40:503–37.

Fastovsky, D. E. 1987. Paleoenvironments of vertebrate-bearingstrata during the Cretaceous-Paleogene transition, easternMontana and western North Dakota. Palaios 2:282–295.

Fastovsky, D. E., and J. B. Smith. 2004. Dinosaur Paleoecology.Pp. 614–626 in Weishampel et al. 2004b.

Feranec, R. S., and B. J. MacFadden. 2006. Isotopic discrimina-tion of resource partitioning among ungulates in C3-domi-nated communities from the Miocene of Florida and Califor-nia. Paleobiology 32:191–205.

Fricke, H. C. 2007. Stable isotope geochemistry of bonebed fos-sils: reconstructing paleoenvironments, paleoecology, andpaleobiology. Pp. 437–490 in R. R. Rogers, D. A. Eberth, andA. R. Fiorillo, eds. Bonebeds: genesis, analysis, and paleobi-ological significance. University of Chicago Press, Chicago.

Fricke, H. C., and J. R. O’Neil. 1996. Inter- and intra-tooth var-iations in the oxygen isotope composition of mammaliantooth enamel: some implications for paleoclimatological andpaleobiological research. Palaeogeography, Palaeoclimatolo-gy, Paleoecology 126:91–99.

Fricke, H. C., and R. R. Rogers. 2000. Multiple taxon-multiplelocality approach to providing oxygen isotope evidence forwarm-blooded theropod dinosaurs. Geology 28:799–802.

Fricke, H. C., W. C. Clyde, J. R. O’Neil, and P. D. Gingerich. 1998.Intra-tooth variation in �18O of mammalian tooth enamel as arecord of seasonal changes in continental climate variables.Geochimica et Cosmochimica Acta 62:1839–1851.

Fricke, H. C., R. R. Rogers, R. Backlund, C. N. Dwyer, S. Echt.2008. Preservation of primary stable isotope signals in dino-saur remains, and environmental gradients of the Late Cre-taceous of Montana and Alberta. Palaeogeography, Palaeocli-matology, Paleoecology (in press).

Gannes, L. Z., C. M. d. Rio, and P. Koch. 1998. Natural abun-dance variations in stable isotopes and their potential uses in


animal physiological ecology. Comprehensive BiochemicalPhysiology A 119:725–737.

Gat, J. R. 1996. Oxygen and hydrogen isotopes in the hydrologiccycle. Annual Review of Earth Planetary Sciences 24:225–262.

Hartman, J. H., and J. I. Kirtland. 2002. Brackish and marinemolluscs of the Hell Creek Formation of North Dakota: evi-dence for a persisting Cretaceous seaway. Pp. 271–296 inHartman et al. 2002.

Hartman, J. H., K. R. Johnson, and D. J. Nichols, eds. 2002. TheHell Creek Formation and the Cretaceous-Tertiary boundaryin the northern Great Plains: an integrated continental recordof the end of the Cretaceous. Geological Society of AmericaSpecial Paper 361.

Heaton, T. H. E. 1999. Spatial, species, and temporal variationsin the 13C/12C ratios of C3 plants: implications for paleodietstudies. Journal of Archaeological Sciences 26:637–649.

Hedges, R. E. M. 2003. On bone collagen—apatite-carbonate iso-topic relationships. International Journal of Osteoarchaeology13:66–79.

Hicks, J. F., K. R. Johnson, J. D. Obradovich, L. Tauxe, and D.Clark. 2002. Magnetostratigraphy and geochronology of theHell Creek and basal Fort Union Formations of southwesternNorth Dakota and a recalibration of the age of the Cretaceous-Tertiary boundary. Pp. 35–56 in Hartman et al. 2002.

Hillson, S. 1986. Teeth. Cambridge University Press, Cam-bridge.

Hoppe, K. A., R. G. Amundson, M. Vavra, M. P. McClaran, andD. L. Anderson. 2004. Isotopic analysis of tooth enamel car-bonate from modern North American feral horses: implica-tions for paleoenvironmental reconstructions. Palaeogeogra-phy, Palaeoclimatology, Paleoecology 203:299–311.

Horner, J. R., D. B. Weishampel, and C. A. Forster. 2004. Had-rosauridae. Pp. 438–463 in Weishampel et al. 2004b.

Jim, S., S. Ambrose, and R. Evershed. 2004. Stable carbon iso-topic evidence for differences in the dietary origin of bonecholesterol, collagen and apatite: implications for their use inpalaeodietary reconstruction. Geochimica et CosmochimicaActa 68:61–72.

Johnson, K. R. 2002. Megaflora of the Hell Creek and lower FortUnion Formation in North Dakota: vegetational response toclimate change, the Cretaceous-Tertiary boundary event, andrapid marine transgression. Pp. 329–392 in Hartman et al.2002.

Johnson, B. J., M. Fogel, and G. H. Miller. 1998. Stable isotopesin modern ostrich eggshell: a calibration for paleoenviron-mental applications in semi-arid regions of southern Africa.Geochimica et Cosmochimica Acta, 62:2451–2461.

Koch, P. L. 1998. Isotopic reconstruction of past continental en-vironments. Annual Review of Earth and Planetary Sciences26:573–613.

Koch, P. L., M. Fogel, and N. Tuross. 1994. Tracing the diet offossil animals using stable isotopes. Pp. 63–94 in K. Lajtha andR. Michener, eds. Stable isotopes in ecology and environmen-tal science. Blackwell Science, Oxford.

Koch, P. L., N. Tuross, and M. L. Fogel. 1997. The effects of sam-ple treatment and diagnosis on the isotopic integrity of car-bonate in biogenic hydroxyapatite. Journal of ArchaeologicalSciences 24:417–429.

Kohn, M. J. 1996. Predicting animal �18O: accounting for dietand physiological adaptation. Geochimica et CosmochimicaActa 60:4811–4829.

Kohn, M. J., and T. E. Cerling. 2002. Stable isotope compositionsof biological apatite. Reviews in Mineralogy and Geochem-istry 48:455–488.

Kohn, M. J., M. J. Schoeninger, and J. W. Valley. 1998. Variabilityin herbivore tooth oxygen isotope compositions: reflections ofseasonality or developmental physiology? Chemical Geology152:92–112.

Kolodny, Y., B. Luz, M. Sander, and W. A. Clemens. 1996. Di-nosaur bones: fossils of pseudomorphs? The pitfalls of phys-iology reconstruction from apatitic fossils. Palaeogeography,Palaeoclimatology, Palaeoecology 126:161–171.

Lehman, T. M. 1987. Late Maastrichtian paleoenvironments anddinosaur biogeography in the western interior of NorthAmerica. Palaeogeography, Palaeoclimatology, Paleoecology60:189–217.

———. 2001. Late Cretaceous dinosaur provinciality. Pp. 310–328 in D. Tanke and K. Carpenter, eds. Mesozoic vertebratelife. Indiana University Press, Bloomington.

Lockheart, M. J., I. Poole, P. F. Van Bergen, and R. P. Evershed.1998. Leaf carbon isotope compositions and stomatal char-acters: important considerations for palaeoclimate recon-structions. Organic Geochemistry 29:1003–1008.

Longinelli, A. 1984. Oxygen isotopes in mammal bone phos-phate: a new tool for paleohydrological and paleoclimatolog-ical research? Geochimica et Cosmochimica Acta 48:385–390.

Luz, B., and Y. Kolodny. 1985. Oxygen isotope variations inphosphates of biogenic apatites. IV. Mammal teeth and bones.Earth and Planetary Science Letters 75:29–36.

MacFadden, B. J., and P. Higgins. 2004. Ancient ecology of 15-million-year-old browsing mammals within C3 plant com-munities from Panama. Oecologia 140:169–182.

Martinelli, L. A., S. Almeida, I. F. Brown, M. Z. Moreira, R. L.Victoria, L. S. L. Sternberg, C. A. C. Ferreira, and W. W. Thom-as. 1998. Stable carbon isotope ratio of tree leaves, boles, andfine litter in a tropical forest in Rondonia, Brazil. Oecologia114:170–179.

McKee, K. L., I. C. Feller, M. Popp, and W. Wanek. 2002. Man-grove isotopic (d15N and d13C) fractionation across a nitro-gen vs. phosphorous limitation gradient. Ecology 83:1065–1075.

Murphy, E. C., J. W. Hoganson, and K. R. Johnson. 2002. Litho-stratigraphy of the Hell Creek Formation in North Dakota. Pp.9–34 in Hartman et al. 2002.

Nelson, B. K., M. J. DeNiro, and M. J. Schoeninger. 1986. Effectsof diagenesis on strontium, carbon, nitrogen, and oxygenconcentration and isotopic composition of bone. Geochimicaet Cosmochimica Acta 50:1941–1949.

Nguyen Tu, T. T., H. Bocherens, A. Mariotti, F. Baudin, D. Pons,J. Broutin, S. Derenne, and C. Largeau. 1999. Ecological dis-tribution of Cenomanian terrestrial plants based on 13C/12Cratios. Palaeogeography, Palaeoclimatology, Paleoecology145:79–93.

O’Leary, M. H. 1988. Carbon isotopes in photosynthesis. Bio-science 38:328–336.

O’Leary, M. H., S. Mahavan, and P. Paneth. 1992. Physical andchemical basis of carbon isotope fractionation in plants. Plant,Cell and Environment 15:1099–1104.

Passey, B. H., T. F. Robinson, L. K. Ayliffe, T. E. Cerling, M. Spon-heimer, M. D. Dearing, B. L. Roeder, and J. R. Ehleringer. 2005.Carbon isotope fractionation between diet, breath CO2, andbioapatite in different mammals. Journal of ArchaeologicalSciences 32:1459–1470.

Pearson, D. A., T. Schaefer, K. R. Johnson, and D. J. Nichols. 2001.Palynologically calibrated vertebrate record from North Da-kota consistent with abrupt dinosaur extinction at the K-Tboundary. Geology 29:39–42.

Pearson, D. A., T. Schaefer, K. R. Johnson, D. J. Nichols, J. P.Hunter. 2002. Vertebrate biostratigraphy of the Hell CreekFormation in southwestern North Dakota and northwesternSouth Dakota. Pp. 145–168 in Hartman et al. 2002

Rozanski, K., L. Araguas-Araguas, and R. Gonfiantini. 1993.Isotopic patterns in modern global precipitation. Pp. 1–36 inP. K. Swart, K. C. Lohmann, J. McKenzie, and S. Savings, eds.Climate change in the continental isotopic records. Geophys-


ical Monograph 78. American Geophysical Union, Washing-ton, D.C.

Sage, R. F., and R. K. Monson. 1999. C4 plant biology. AcademicPress, San Diego.

Sharp, Z. D., and T. E. Cerling. 1998. Fossil isotope records ofseasonal climate and ecology: straight from the horse’smouth. Geology 26:219–222.

Sheehan, P. M., D. E. Fastovsky, R. G. Hoffman, C. B. Berghaus,and D. L. Gabriel. 1991. Sudden extinction of the dinosaurs:Latest Cretaceous, upper Great Plains, USA. Science 254:835–839.

Sheehan, P. M., D. E. Fastovsky, C. Barreto, and R. G. Hoffman.2000. Dinosaur abundance was not declining in a ‘‘3 m gap’’at the top of the Hell Creek Formation, Montana and NorthDakota. Geology 28:523–526.

Spotila, J. R. 1980. Constraints of body size and environment onthe temperature regulation of dinosaurs. Pp. 233–252 in R. D.K. Thomas and E. C. Olson, eds. A cold look at warm-bloodeddinosaurs (American Association for the Advancement of Sci-ence Selected Symposium 28). Westview, Boulder, Colo.

Stanton-Thomas, K., and S. J. Carlson. 2003. Microscale d18Oand d13C isotopic analysis of an ontogenetic series of the had-rosaurid dinosaur Edmontosaurus: implications for physiologyand ecology. Palaeogeography, Palaeoclimatology, Paleoecol-ogy 206:257–287.

Sternberg, L. S. L. 1989. Oxygen and hydrogen isotope ratios inplant cellulose: mechanisms and applications. Pp. 124–141 inP. W. Rundel, J. R. Ehleringer, and K. A. Nagy, eds. Stable iso-topes in ecological research. Springer, Heidelberg.

Straight, W. H., R. E. Barrick, and D. A. Eberth. 2004. Reflectionsof surface water, seasonality and climate in stable oxygen iso-topes from tyrannosaurid tooth enamel. Palaeogeography,Palaeoclimatology, Paleoecology 206:239–256.

Tieszen, L. L. 1991. Natural variations in carbon isotope valuesof plants: implications for archaeology, ecology, and paleo-ecology. Journal of Archaeological Sciences 20:227–248.

Trueman, C. N., and N. Tuross. 2002. Trace elements in recentand fossil bone apatite. Reviews in Mineralogy and Geochem-istry 48:489–521.

Trueman, C. N., A. K. Behrensmeyer, N. Tuross, and S. Weiner.

2004. Mineralogical and compositional changes in bones ex-posed on soil surfaces in Amboseli National Park, Kenya: dia-genetic mechanisms and the role of sediment pore fluids.Journal of Archaeological Science 31:721–739.

van der Merwe, N. J., and E. Medina. 1991. The canopy effect,carbon isotope ratios and foodwebs in Amazonia. Journal ofArchaeological Science 18:249–259.

Varricchio, D. J. 2001. Gut contents for a Cretaceous tyranno-saur: implications for theropod digestive tracts. Journal of Pa-leontology 75:401–406.

von Fischer, J. C., and L. T. Tieszen. 1995. Carbon isotope char-acterization of vegetation and soil organic matter in subtrop-ical forests in Luquillo, Puerto Rico. Biotropica 27:138–148.

Weishampel, D. B., P. M. Barrett, R. A. Coria, J. LeLoeuff, X.Xing, Z. Xijin, A. Sahni, E. M. P. Gomani, and C. R. Noto.2004a. Dinosaur distribution. Pp. 517–606 in Weishampel etal. 2004b.

Weishampel, D. B., P. Dodson, and H. Osmolska, eds. 2004b. TheDinosauria. University of California Press, Berkeley.

White, P. D., D. E. Fastovsky, and P. M. Sheehan. 1998. Taphon-omy and suggested structure of the dinosaurian assemblageof the Hell Creek Formation (Maastrichtian), eastern Montanaand western North Dakota. Palaios 13:41–51.

Wilf, P., K. R. Johnson and B. T. Huber. 2003. Correlated terres-trial and marine evidence for global climate changes beforemass extinction at the Cretaceous-Paleogene boundary. Pro-ceedings of the National Academy of Sciences USA 100:599–604.

Wooller, M., B. Smallwood, U. Scharler, M. Jacobson, and M. Fo-gel. 2003. A taphonomic study of d13C and d15N values inRhizophora mangle leaves for a multi-proxy approach to man-grove palaeoecology. Organic Geochemistry 34:1259–1275.

Zazzo, A., C. Lecuyer, and A. Mariotti. 2004. Experimentally-controlled carbon and oxygen isotope exchange betweenbioapatites and water under inorganic and microbially-me-diated conditions. Geochimica et Cosmochimica Acta 68:1–12.

Zylberberg, L., J. Y. Sire, and A. Nanci. 1997. Immunodetectionof amelogenin-like proteins in the ganoine of experimentallyregenerating scales of Calamoichthys calabaricus, a primitiveActinopterygian fish. Anatomical Record 249:86–95.

Documents

Stable isotope evidence for changes in dietary niche ...sites.coloradocollege.edu/hfricke/files/2012/10/...Paleobiology,34(4), 2008, pp. 534–552 Stable isotope evidence for changes