- The Paleontological Society
The Caune de l'Arago Cave (southern France) has yielded one of the best preserved and best documented sedimentary successions of the European Middle Pleistocene (Oxygen Isotopic Stages 14 to 12). Herbivorous ungulates (horse, reindeer, red deer, fallow deer, bison, musk ox, argali, and tahr) are well represented in the three major stratigraphic units CM1, CM2, and CM3. CM1 and CM3 correspond to cold and dry climate and CM2 represents temperate and humid environmental conditions. Dental microwear and mesowear analyses were performed for the ungulates from CM1–3 to test whether these methods of dental wear evaluation were suitable for detecting climate-driven changes in the dietary resources of the Arago ungulate community. We found that both dental mesowear and microwear indicate dietary traits and their relationship to climatic conditions as reflected by vegetation cover and community structure. In all units, even if some species seem to share habitats or resources, it appears that the overlap in their feeding ecology is very low. The CM1 and CM3 units, where pollen analysis indicates that the climate was cold and dry, show the lowest diversity in dietary traits. The CM2, where climate is known to be more temperate and humid, the spectrum of dietary traits is large—grazers, browsers, and mixed feeders are present.
The central hypothesis tested in this paper concerns the interrelationship between climate (through water availability and temperature) and food abrasiveness as indicated by dental wear. When several stratigraphic units from the same locality are compared, the interspecific variability in microwear and mesowear variables should reflect a climatic signal. If our hypothesis is correct, we would expect a temperate and humid climate to lead to a broader range of resources available for the herbivorous community. This should be reflected by a greater diversity of dietary traits represented in the community. In dryer and colder environments, one would therefore expect a lower range of available resources and thus a low diversity in herbivorous feeding traits.
The interrelationship between climate and dietary adaptation in ungulates revealed by using dental wear analyses is debated in many publications (e.g., Fortelius et al. 2002; Janis et al. 2002; Kaiser 2003; Kaiser and Schulz 2006; Kaiser and Rössner 2007). However, with the exception of Kaiser and Rössner (2007) dental wear analysis mostly deals with one species (or group), or a one single locality (Merceron and Ungar 2005; Merceron et al. 2006). In order to test hypotheses related to climate changes through time and dietary niche occupation of Pleistocene ungulates, relatively large samples of various mammal species are needed. If climate-related patterns are to be investigated, this approach cannot be restricted to a single species because the occurrence and abundance of a species is related not only to human hunting activity, but also to the community structure of a biome, interspecific competition, and the resource availability in a given habitat. All these factors may influence the abundance and representation of a species at a site and thus a community-based approach is preferable. Community-based approaches require community-based data. Samples from extant herbivorous mammals that represent the same locality and environment are not widely available in zoological collections. However, archaeological samples from rich localities usually fulfill this requirement as long as they are stratigraphically well provenanced, well dated, and well preserved. We selected the Caune de l'Arago sample (middle Pleistocene—Oxygen Isotope Stages [OIS] 14 to 12) for the following reasons: (1) A long stratigraphy from the middle Pleistocene to upper Pleistocene is present and well documented. (2) Sediments are well excavated and well dated. (3) The three main stratigraphic units from the Middle Complex (CM) are excellently documented. CM1 is attributed to OIS 14 (ca. 550 Ka), CM2 correlates to OIS 13 (ca. 500 Ka), and CM3 corresponds to OIS 12 (ca. 450 Ka). Units CM1 and CM3 correspond to cold and dry climate and unit CM2 represents temperate and humid environmental conditions (Fig. 1). (4) Large samples of various herbivorous ungulate species (horse, reindeer, red deer, fallow deer, bison, musk ox, argali, and tahr) are available in all three units.
We use micro- and mesowear analyses to infer the dietary behavior in the herbivores represented in the three stratigraphic units, CM1, CM2, and CM3. Taken alone, the dietary signal would allow inferences about the community structure of herbivores and the dietary resource partitioning within the community. Food availability and habitat structure are closely linked parameters in terrestrial ecosystems. Because dental durability in ungulates is linked to the abrasiveness of foods comminuted, dental wear analysis evaluates a major habitat interface, which reflects the overall abrasiveness of foods eaten. By heavily increasing abrasiveness, phytoliths present in many angiosperms, both monocotyledons and dicotyledons, constitute an important system of mechanical defense (MacNaughton et al. 1985; Piperno et al. 2002). In addition to phytoliths, airborne grit and grit deriving from attached soil potentially make up the second major source of abrasives eaten by herbivorous mammals (Kay and Covert 1983). Additionally some very general habitat-driven mechanisms can increase abrasiveness of food available to herbivores.
The Caune de l'Arago
The Caune de l'Arago is a large karstic cave naturally developed in the Urgo-Aptian limestone of the Corbières Massif in the Eastern Pyrenees (France). The cave has been excavated each year since 1964 by teams directed by Professor Henry de Lumley. The deposits are almost 15 m thick and cover a period from 690,000 to 100,000 years ago (Fig. 1). All units are rich in faunal materials, which are housed at the Centre Européen de Recherches Préhistoriques of Tautavel. Most assemblages are associated with abundant stone tools, and human remains (Homo heidelbergensis) are found in some levels. Numerous remains of large mammals have been collected, such as reindeer (Rangifer tarandus (Linnaeus 1758)), red deer (Cervus elaphus Linnaeus 1758), fallow deer (Dama clactoniana Falconer 1868), argali (Ovis ammon (Linnaeus 1758) antiqua Pommerol 1880), tahr (Hemitragus bonali Harlé and Stehlin 1913), bison (Bison priscus (Bojanus 1827)), musk-ox (Praeovibos priscus Staudinger 1908), horse (Equus ferus mosbachensis von Reichenau 1903), and rhinoceros (Stephanorhinus hemitoechus (Falconer 1868)). In addition, the following carnivores are present in some units: Deningeri bear (Ursus deningeri von Reichenau 1904), brown bear (Ursus arctos Linneaus 1758), wolf (Canis mosbachensis Soergel 1925), dhole (Cuon priscus Thenius 1954), fox (Vulpes vulpes Linnaeus 1758), leopard (Panthera pardus Linnaeus 1758), cave lion (Panthera leo spelaea Goldfuss 1810), and lynx (Lynx spelaeus Boule 1906).
Within sublevels of the sequence, assemblages are distinct in their taphonomic history and comprise time frames ranging from long-duration habitats (level CM3/G), temporary seasonal accumulations (levels CM3/F and CM2/J), and accumulations related to short-term hunting stopovers (levels CM1/K and L).
Mesowear analysis was performed to characterize dietary traits of ungulates following the protocol developed by Fortelius and Solounias (2000). Mesowear and its application to extinct species is discussed further by Kaiser et al. (2000), Kaiser and Solounias (2003), Kaiser and Fortelius (2003), Franz-Odendaal and Kaiser (2003), and Kaiser and Rössner (2007). Mesowear describes the prolonged type of cumulative dental wear that represents a comparatively long period of an individual animal's lifetime. The method reduces ungulate tooth wear to two variables: occlusal relief and cusp shape. Occlusal relief is classified as high or low, depending on how high the cusps rise above the valley between them. The second mesowear variable, cusp shape, includes three attributes: sharp, round, and blunt according to the degree of facet development. A mesowear score (unpublished method developed by Matthew C. Mihlbachler), based on the upper second molar (M2), was constructed for each fossil sample (i.e., each species in a given unit). Individual molar cusp shape and relief scores were converted to a single mesowear score as follows: a combination of sharp cusps and high relief was assigned a score of 0; a combination of rounded cusps and high relief was assigned a score of 1; a combination of rounded cusps and low relief was assigned a score of 2; a combination of blunt cusps and low relief was assigned a score of 3. Mesowear scores for each individual within a sample were then averaged to obtain an average mesowear score for that sample. See an example of application in Rivals et al. (2007a) and Semprebon and Rivals (2007). Because mesowear is sensitive to the age of individuals (Rivals et al. 2007b), only adult individuals were sampled.
Microwear analysis was performed following the procedure developed by Solounias and Semprebon (2002). Microwear features observed on the occlusal enamel surfaces are the result of single occlusal events, and thus reflect diets within a much shorter time frame (Walker et al. 1978), perhaps days or even hours before the individuals death. Semprebon et al. (2004) have validated the robustness of the low magnification light microscopy microwear method when applied to different measurement sites. High-resolution epoxy casts were examined at 35× magnification using a Leica MZ16 stereomicroscope. Examination of microwear was done on the second enamel band of the paracone of the upper second molar or alternatively the protoconid of the lower second molar. Relative pit sizes and scratch textures were assessed on the basis of differential light refraction properties as outlined by Solounias and Semprebon (2002) and Semprebon et al. (2004). Counts of pits and scratches were made on a standard 0.4 × 0.4 mm square area on each tooth.
Typical grazers have the highest numbers of scratches and the lowest numbers of pits; typical leaf browsers have lower numbers of scratches and more disparate numbers of pits. Differentiation between trophic groups was made using a bivariate plot of scratches versus pits, and also by calculating the percentage of scratch counts in a taxon that fall into a low scratch range, i.e., the percentage of counts that fall between 0 and 17 scratches (Solounias and Semprebon 2002; Semprebon and Rivals 2007). There is no overlap between the low scratch ranges of the extant browsers (72.73–100% of counts fall between 0 and 17) and grazers (0–22.2% of counts fall between 0 and 17) studied. No overlap is seen in the ranges of browsers and mixed feeders and very little overlap is seen between grazers and mixed feeders (20.93–70% of scratches fall between 0 and 17).
Mesowear and microwear summary results are presented in Tables 1 and 2. Mesowear scores of fossil species from Arago are compared with those of extant ungulates in Figure 2. Microwear variables are given in bivariate plots for each stratigraphic complex independently (Fig. 3A–C). Dietary assignments for each species are summarized in Table 3. A total of eight species were sampled, 604 specimens were molded. From these, 54 dental specimens had to be excluded from analysis because of taphonomic alterations such as flaking of the ectoloph and severe weathering, leaving 550 teeth used for mesowear analysis. An even more reduced set of 309 specimens was suitable for microwear evaluation.
Equus ferus mosbachensis
In all three units (CM1, CM2, and CM3), E. ferus displays mesowear and microwear signals similar to those of extant grazers. Mesowear scores are very similar in the three units (ranging from 1.93 to 2.00) and are the highest scores recorded in our samples (Fig. 2). Such mesowear scores reveal highly abrasive diets. Compared to extant relatives (E. quagga and E. grevyi), the samples from the Caune de l'Arago have slightly lower scores.
The microwear signal reveals a similar trend (Fig. 3A,B,C). The number of scratches is relatively high (equal or superior) relative to extant grazers. The numbers of pits always fall in the range of extant grazers. The percentage of scratch counts potentially falling in the range of 0–17 is null in all three samples (i.e., more than 17 scratches per count on all specimens). In all units, both mesowear and microwear classifies E. ferus as a grazer. In the CM1 our fossil sample has more pits and scratches than extant African equids. Samples from the CM2 and CM3 both only have only slightly more scratches than their extant relatives.
R. tarandus is present in the two units corresponding to cold and dry climate (CM1 and CM3). In the CM1, it has a lower mesowear score than in the CM3 (Fig. 2). We observe an increase in abrasion control in the mesowear equilibrium from mixed feeding to grazing without significant climatic change because both the CM1 and CM3 correspond to cold phases of the Pleistocene. Both samples, however, and particularly in the CM3, have higher mesowear scores than the extant R. tarandus.
Microwear reveals an opposite shift in dietary traits from grazing to a more mixed feeding behavior in the CM3, where both fossil and extant species plot near to each other (Fig. 3A,C). In the CM1, R. tarandus plots near Ovis canadensis (Oc), an open-habitat mixed feeder. Thus, in both units we observe a discrepancy between mesowear and microwear signatures in R. tarandus.
From the CM2 to CM3, i.e., from temperate to cold climatic phases, C. elaphus shows an increase in mesowear score (from 0.52 to 0.87). C. elaphus has a mesowear score similar to that of C. elaphus canadensis (Cd) in the CM2, and similar to C. unicolor in the CM3 (Fig. 2). The shift observed in C. elaphus diet is to be interpreted as a shift in diet toward more abrasive food items in the CM3.
The microwear data classify both samples as mixed feeders. Both samples plot close to the extant Rangifer tarandus (Fig. 3B,C). However the difference in numbers of scratches shows a tendency toward more grazing in the CM3, as concordantly supported by the mesowear signal.
In the CM2, D. clactoniana displays the lowest mesowear score (MWS) recorded in this study (0.42). This classifies the species among extant browsers (Fig. 2). The microwear signal confirms the mesowear observation. The high percentage of scratches in the low range (0–17) suggests that D. clactoniana was a browser (Fig. 3B). Unfortunately, no data for extant conspecifics were available for comparison.
In the CM3, the mesowear score of B. priscus plots among the extant grazers, and indicates a highly abrasion-dominated feeding trait (MWS = 1.61). However, compared with the extant Bison bison (bb), the mesowear score is lower, which indicates a lower level of abrasion (Fig. 2).
Microwear also reveals a grazing trait, and certainly the same preference for grasslands habitats as indicated by the mesowear signal. B. priscus from the CM3 has fewer scratches than its extant relative, B. bison, suggesting fewer abrasives in its diet, as demonstrated by mesowear; however, it also has more pits than B. bison (Fig. 3C).
In the CM3, P. priscus has a mesowear score of 1.36 reflecting a moderately abrasive diet (Fig. 2). P. priscus can be either referred a mixed feeder or a grazer. The mesowear score recorded in our sample is higher than for the extant relative, Ovibos moschatus.
Looking at microwear, P. priscus falls in the grazing morphospace (Fig. 3C). Its microwear signal, however, differs from that of O. moschatus in having a lower proportion of pits.
Ovis ammon antiqua
O. ammon has mesowear scores in the range for extant mixed feeders but close to the transition with extant grazers (Fig. 2). Mesowear scores are similar in the three units and reflect intermediate levels of abrasion. Moreover scores are similar to their extant relative O. ammon poli from Mongolia.
The microwear signal shows more variation than mesowear between the three units (Fig. 3A,B,C). In the CM1 and CM3, the percentages of counts with a low scratch range (0 and 5.6 respectively) classify O. ammon among the grazers. In the CM2, the scratch distribution indicates lower abrasive levels and classifies O. ammon among the extant mixed feeders.
In the CM3, H. bonali has a mesowear score slightly higher but in the same range as recorded for O. ammon (Fig. 2). H. bonali can be assigned a mixed feeder or a grazer. However, the microwear signal is more typical of an extant grazer (Fig. 3C). As we observed for Praeovibos priscus, the mesowear score for the fossil species exceeds what is recorded for its extant relative, H. jemlahicus.
Both mesowear and microwear analysis of feeding traits result in very similar dietary assignments for the fossil species investigated from the Caune de l'Arago (Table 3). The widest range of dietary adaptations is observed in the CM2 unit where all three major dietary traits described by Hofmann (1972) are represented. A reduced diversity in feeding traits is observed in the CM1 and the CM3 units where herbivores representing only two categories were present, grazers and mixed feeders. CM1 and CM3 units correspond to dry and cold climate conditions (OIS 14 and 12, respectively), whereas CM2 reflects rather temperate and humid climate. Thus our central hypothesis, that cold and dry climate conditions should support a comparatively less diverse range of feeding traits, is supported by our data.
In the CM1 unit, we observe a reduced variety of dietary traits which corresponds to only one category given for ungulates by Hofmann and Stewart (1972). The three species (Equus ferus mosbachensis, Ovis ammon antiqua, and Rangifer tarandus) were classified as grazers (or graze-dominated mixed feeders). According to sediment and pollen analyses (Table 3), CM1 unit corresponds to dry and cold climates and large open environments (Renault-Miskovsky 1995; de Lumley et al. 1984). We find that these results further correspond to the presence of large areas of grasslands in the surroundings of the cave. Such open environments would well correspond to the feeding traits of the three species considered here.
E. ferus would then have exploited a highly abrasive food resource in this open and dry landscape compared to the other species from CM1. However, the comparison with modern-day relatives (E. quagga and E. grevyi) reveals that the fossil species had a less abrasive diet. The higher proportion of pits observed for the CM1, however, suggests that compared to extant relative's habitat, more airborne grit infested the grass dominated diet of the fossil horse from the Caune de l'Arago. It seems clear that E. ferus inhabited large open spaces with grassland in the plain next to the cave. The presence of such habitat settings, a steppe with grass and forbs, is also indicated by the presence of 78% of non-arboreal pollen (Renault-Miskovsky 1995).
For O. ammon and R. tarandus the food resources reflect exactly the same level of abrasion as indicated by the mesowear signal. This suggests that the same source for forage was exploited by the two species, which were probably bound to open habitats on either the plains or the plateau, or both. Competition between the two species is not likely to have occurred because O. ammon and R. tarandus were demonstrated to be adapted to live in different habitats. Their behavior and morphology suggest that O. ammon inhabited the plateaus above the cave (Rivals 2004; Fedosenko and Blank 2005), whereas R. tarandus more likely occupied the plains (Klein 1992). Niche segregation could thus explain the use of similar food resources in the local environment around the cave. It has been shown that both populations were seasonally hunted (Rivals et al. 2004) by the cave occupants. The discrepancy between mesowear and microwear signals observed for R. tarandus in CM1 and CM3 would most likely be related to this seasonality in hunting behavior. Differences observed are certainly related to the fact that microwear records the diet of the last days of the animal's life, because mesowear represents a temporally more averaged signature. It would further indicate that both mesowear and microwear add complementary dietary information because they cover times frames of different duration, and thus help resolve seasonality patterns.
The competition between the various species known in the CM1, however, does not seem very high. This relatively low level of competition might be related to the fact that each ungulate seems to have occupied its own niche and feeding strategy.
In the CM2 unit, the three major dietary categories described by Hofmann and Stewart (1972) are present. One species is a typical grazer (Equus ferus mosbachensis), two are mixed feeders (Ovis ammon antiqua andCervus elaphus), and one is a browser (Dama clactoniana). The high variety of dietary traits represented corresponds here to a temperate and humid climate where a forest (thermophilous taxa) was well developed, but open grasslands were also present in the vicinity of the cave (Renault-Miskovsky 1995; de Lumley et al. 1984). The four species seem to have exploited different dietary resources in a habitat, supporting a comparatively high diversity of food resource.
C. elaphus and D. clactoniana were both frequent prey animals of Pleistocene hunters at the cave and thus most likely were also abundant species in the Arago paleohabitat vicinity as indicated by the high number of individuals represented by the CM2 sample, in particular in level “J” (Moigne et al. 2006). However, dental wear signatures give evidence that, even if both species were probably foraging in forested habitats, C. elaphus and D. clactoniana did not exploit the same resource. This is probably related to interspecific competition between the two species. E. ferus seems to have focused on food items representing the same level of abrasiveness as the species also did in CM1 times. Pollen analysis revealed that grassland was still present in CM2 even if forested areas were abundant in the area (Renault-Miskovsky 1995).
In the CM3 unit, the seven species analyzed appear to be grazers or graze-dominated mixed feeders throughout. The diversity of dietary traits of the CM3 ungulates seems to be fairly reduced compared to CM2. Pollen analysis revealed that as in CM1, plant species indicative of open and dry landscapes predominated in CM3 times. A steppe environment with grass and forbs is attested in the pollen record by the presence of 78% of non arboreal pollen, indicating a dry and cold climate (Renault-Miskovsky 1995). The absence of browsers in the ungulate community is thus in accordance with these climate conditions and subsequent resource availability. Except for E. ferus, dietary assignments in CM3 are not definite and vary between mixed feeding and grazing. The discrepancy observed here between mesowear and microwear is explained by the fact that the microwear method is more sensitive to seasonal changes than is the mesowear signature (Solounias and Semprebon 2002; Rivals and Semprebon 2006). The short-term and more immediate wear captured by the microwear technique reveals the presence of these seasonal fluctuations, whereasthe long-term and more cumulative wear captured by mesowear reveals an intermediate pattern. The discrepancy in the signals recorded by mesowear and microwear testifies to mixed feeding habits, one method revealing more grazing habits because it captures a small time frame (the last meals), and the other tending toward mixed feeding (because it captures a larger time frame). Such discrepancies in results are more useful than limiting because they help to gain an insight into both the seasonal shifts in diet and the long-term trends (Rivals and Semprebon 2006). It is known that all species in all levels were seasonally hunted (Rivals et al. 2004).
During the deposition of the CM3, the resources available to ungulates were certainly reduced compared to CM2, which represents temperate and more humid conditions. The pollen record (Renault-Miskovsky 1995) indicates that structural diversity in the vegetation was reduced in CM3 and ungulates were probably constrained to closer feeding niches in the local environment. Because open habitats were widely represented in the valley and the plateau around the cave, we suggest that resource partitioning among these four species was achieved by differential habitat preferences rather than by different foraging strategies. This notion is supported by adaptations of O. ammon and H. bonali, which indicate that O. ammon was certainly bound to the plateau above the cave (Rivals 2004; Fedosenko and Blank 2005; Caughley 1969), whereas H. bonali was restricted to the plain below (Straus 1981; Gebert and Verheyden-Tixier 2001; Feldhammer et al. 1988). For B. priscus the higher number of pits, compared to its extant relative, is related to an open and arid environment where airborne dust and grit accumulate on the leaves. This is another convincing piece of evidence also supported by the pollen data (Table 3).
Our study is the first to examine the dietary adaptations of an ungulate assemblage and their variations through time using samples from a single locality. At the Caune de l'Arago, dental wear analyses show that diversity of dietary traits was high during the temperate and humid phase (CM2) and decreased during the cold and dry phases (CM1 and CM3). Our data also demonstrate the importance of studying communities, instead of isolated species, in dental wear studies. This community-based approach yielded unique evidence on resources and niche partitioning of ungulates as well as on the diversity of food resources available in a given environment and time.
Our data indicate that diets for P. priscus, R. tarandus, and H. bonali in the middle Pleistocene were more abrasive than those of their extant close relatives. The reverse situation applies for the two larger species, E. ferus and B. priscus, which had less abrasive diets in the past. No dietary shift was found for O. ammon. Such shifts in dietary niches were already found in the North American bison (Bison bison) (Rivals et al. 2007a). As observed at Arago, the bison's diet was more abrasive in the Pleistocene, once more demonstrating the species' flexibility and the climate sensitivity at the same time. This study adds more evidences to the notion that modern ungulates may occupy very different dietary niches than their fossils conspecifics. A shift in dietary niche occupation during a period of just 500,000 years as indicated by the majority of ungulates in the Arago assemblage also underscores the need for caution when eco-profiling fossil habitats using taxonomic uniformitarianism.
We acknowledge H. de Lumley for access to the Caune de l'Arago collections, and A.-M. Moigne and N. Boulbes for their help at the Centre Européen de Recherches Préhistoriques in Tautavel. We are grateful to S. Holman for help with the manuscript. We acknowledge N. Solounias, G. M. Semprebon, and two other anonymous reviewers for their helpful comments on an earlier draft of this paper. This research was supported by a postdoctoral fellowship of the Alexander von Humboldt Foundation to F.R. while at the Biocentrum Grindel and Zoological Museum– University of Hamburg (2005–2006). T.M.K. acknowledges the Deutsche Forschungsgemeinschaft (DFG) for funding related work (grant KA 1525/4-1, 4-2, 6-1), which greatly benefited this manuscript.
- Accepted 20 September 2007.