Using Herbivore Carbon Isotope Ratios to Investigate Paleoenvironments

The basis for using carbon isotope data from herbivores to investigate paleoenvironments lies in the different photosyn-thetic pathways that are utilized by most tropical trees, bushes, shrubs, and forbs (C3 plants) on the one hand, and tropical grasses (C4 plants) on the other. Biochemical and anatomical differences between C3 and C4 plants result in their having very different, non-overlapping carbon isotope ratios (Smith and Epstein, 1971; see Codron et al., 2005 for a large African dataset). Herbivores incorporate plant carbon into their tissues, and thus one can readily distinguish an animal eating vegetation from trees, bushes, shrubs or forbs (e.g., leaves, fruits) from one eating tropical grasses, based on the carbon isotope ratios of their tissues (Vogel, 1978). Furthermore, this distinction is readily preserved in dental enamel from the Pliocene (Lee-Thorp and van der Merwe, 1987; Cerling et al., 1997). Therefore, the rationale for using carbon isotope data from herbivores to investigate paleoenvi-ronments is that, at a very general level, a relationship exists between the number of C4 grass consumers at a site and the availability of palatable grasses in the local environment. For instance, if virtually every animal at a site is found to have been a C4 grass-consumer, it is a reasonable assumption that the area was dominated by grasses (we discuss potential collection biases below). Conversely, if all of the animals at a site consumed C3 vegetation, it would be fair to say that there was little if any C4 grass available for consumption.

Figure 16.1 shows the percentage of C3 and C4 consumers at 15 African game reserves, and as can be seen, this method generally distinguishes between "closed" and "open" habitats (as defined by Vrba, 1980), with the latter being isolated in the bottom right corner (few C3 specialists and many C4 specialists) (from Sponheimer and Lee-Thorp, 2003). The only "closed" area from Vrba's survey that clusters with "open" habitats is Lake Manyara, where buffaloes comprise 66% of the total bovid population. However, Lake Manyara also contains many "open" areas, so this result is at least partly a matter of definition. These data strongly suggest that when an area has fewer than 20% C3 consumers and more than 35% C4 consumers, it is likely to be an "open" environment. To a significant extent such information could be provided

Fig. 16.1 Bivariate plot showing the percentages of C3 and C4 specialists at a variety of African game reserves (see Sponheimer and Lee-Thorp, 2003). The relatively "open" environments (as defined by Vrba, 1980) are confined to the lower right hand corner of the graph.

using ecomorphology (see Reed, 1996). However, there is a decided advantage to supplementing ecomorphological with stable carbon isotope data (or other techniques that provide non-genetic information such as dental microwear analysis; see Schubert et al., 2006), as they are direct indicators of diet, while the former tells us more about the diets that challenged an animal's ancestors.

We have also applied this technique to fossil herbivores from the sites Makapansgat Limeworks (Member 3), Sterkfontein (Members 4 and 5), and Swartkrans (Members 1 and 2) (data in Lee-Thorp et al., 2007; Fig. 16.2). These are all karst sites within the Malmani dolomite formation. Conventionally, the deposits of each site have been divided into a series of Members that are believed to form a sequence from older to younger on the basis of lithostratigraphy and biostratigraphy; however, their stratigraphy is complex as a result of multiple depositional and erosional events, and as a result some of these divisions are contested (Brain, 1981, 1993; Maguire, 1985; Berger et al., 2002). The ages of the Members are based largely on biostratigraphic comparisons with well-dated sites in East Africa. For the purposes of this study we use ages of ~3 Ma for Makapansgat Member 3, ~2.2-2.5 Ma for Sterkfontein Member 4, 1.5-2 Ma for Sterkfontein Member 5, and ~1.6-1.8 Ma and 1.2-1.6 Ma for Swartkrans Member 1 and 2 respectively (based on McFadden et al., 1979; Vrba, 1982; Brain, 1993; Kuman and Clarke, 2000; Partridge, 2000; Clarke and Partridge, 2002).

We divided data for all large herbivore specimens into three broad trophic categories: C3 consumers (§13C < -9.0%), C4 consumers (§13C > -3.0%), and mixed feeders (§13C values between these two extremes). Figure 16.2 shows a general decline in the proportions of C3 consumers (browsers) through the sequence, and a concomitant rise in the proportions of C4 consumers (grazers). The two Australopithecus-bearing members, Makapansgat M3 and Sterkfontein M4, contain more than 30% C3 consumers and fewer than 40% C4 consumers. In contrast, all of the members which contain Homo (Swartkrans M1 and M2 and Sterkfontein M5) have more than 70% C4 consumers + mixed feeders. Thus, these data suggest that Australopithecus africanus inhabited more "closed," woody environments than Homo and its contemporary, Paranthropus robustus, which is in close agreement with results from non-isotopic faunal studies (e.g., Vrba, 1985; Reed, 1997).

One distinction between this and most previous faunal studies is that the carbon isotope data suggest that Makapansgat M3 was especially dominated by C3 consumers, which may suggest a densely wooded environment. Another distinction can be found in Sterkfontein Member 5. Recent excavations have shown that M5 Unit B contains sparse Oldowan tools while Unit C has yielded Acheulean technology, suggesting a more recent age for the latter (Kuman and Clarke, 2000; Clarke and Partridge, 2002).

Fig. 16.2 The relative percentages of C3, C4, and mixed feeders from a variety of South African early hominin deposits (data from Lee-Thorp et al., 2007). The percentages of C4 feeders increase sharply during the period of accumulation of Swartkrans Ml and Sterkfontein M5 (C), suggesting greater grass cover when Homo and Paranthropus were present.

Fig. 16.2 The relative percentages of C3, C4, and mixed feeders from a variety of South African early hominin deposits (data from Lee-Thorp et al., 2007). The percentages of C4 feeders increase sharply during the period of accumulation of Swartkrans Ml and Sterkfontein M5 (C), suggesting greater grass cover when Homo and Paranthropus were present.

And in fact, the carbon isotope data for herbivores in these units differ greatly (Fig. 16.2). Nearly 30% of the Unit B herbivores sampled were C3 consumers and just over 30% were C4 consumers; in contrast, fewer than 10% of the Unit C herbivores were C3 consumers and nearly 80% were C4 consumers. This suggests that Unit C accumulated when the area was dominated by grassy vegetation, and that a major environmental change occurred between the Oldowan and Acheulean deposits, most probably between 1.8 and 1.6 Ma.

We can also compare continuous, rather than categorical 513C data from each Member, which reveals a similar pattern (Fig. 16.3). In aggregate, the herbivores from Makapansgat M3 have the most negative 513C values (x = -6.8%), followed by Sterkfontein M4 (x = -5.4%), Swartkrans M1 (x = -4.4%) and Swartkrans M2 (x = -4.4%), Sterkfontein M5 Unit B (x = -3.5%), and lastly Sterkfontein M5 Unit C (x = -2.1%). Thus, once again, the §13C values of the Australopithecus-bearing members suggest more heavily-wooded environments than those bearing evidence for early Homo. Moreover, using both categorical and continuous data, Makapansgat M3 comes out as the most closed environment, with Sterkfontein M5 Unit C appearing to be the grassiest.

One potential problem with such isotope-based techniques is that, ideally, one should produce §13C values for all herbivore specimens, which is impractical as well as expensive. Thus, one either has to (1) sample a random subset of the fauna preserved in the deposit, or (2) establish mean values for all taxa, and then produce a site mean adjusted for the relative abundance of each taxon (as the site may be dominated by one or a few taxa). The latter will only be possible with well-studied faunal suites in which specimens have

Fig. 16.3 Mean S13C values for East (and Central) African and South African herbivore assemblages. The East African deposits have generally more positive S13C values than South African sites from similar time periods. This may reflect true environmental differences, but could also be at least partly a reflection of the sampling strategies utilized. C5 is Kossom Bougoudi, Chad (~5.3 Ma), C4 is Kolle, Chad (~5-4 Ma), C3 is Koro Toro, Chad (3.5-3.0 Ma), G4 is Gona, Ethiopia (~4.3 Ma), LT is Laetoli, Tanzania (~3.5 Ma), M3 is Makapansgat M3, South Africa (~3Ma), ST4 is Sterkfontein Member 4, South Africa (~2.5Ma), SK1 is Swartkrans Member 1, South Africa (~1.8 Ma), SK2 is Swartkrans Member 2, South Africa (~1.6 Ma), and ST5 is Sterkfontein Member 5, South Africa (~1.8 Ma) (Data from Lee-Thorp, 1989; Sponheimer, 1999; Zazzo et al., 2000; Luyt, 2001; Levin et al., 2004; Kingston and Harrison, 2007).

Fig. 16.3 Mean S13C values for East (and Central) African and South African herbivore assemblages. The East African deposits have generally more positive S13C values than South African sites from similar time periods. This may reflect true environmental differences, but could also be at least partly a reflection of the sampling strategies utilized. C5 is Kossom Bougoudi, Chad (~5.3 Ma), C4 is Kolle, Chad (~5-4 Ma), C3 is Koro Toro, Chad (3.5-3.0 Ma), G4 is Gona, Ethiopia (~4.3 Ma), LT is Laetoli, Tanzania (~3.5 Ma), M3 is Makapansgat M3, South Africa (~3Ma), ST4 is Sterkfontein Member 4, South Africa (~2.5Ma), SK1 is Swartkrans Member 1, South Africa (~1.8 Ma), SK2 is Swartkrans Member 2, South Africa (~1.6 Ma), and ST5 is Sterkfontein Member 5, South Africa (~1.8 Ma) (Data from Lee-Thorp, 1989; Sponheimer, 1999; Zazzo et al., 2000; Luyt, 2001; Levin et al., 2004; Kingston and Harrison, 2007).

been precisely classified to genus or species and for which relative abundance data are available. In contrast, selecting a random sample of herbivores should be quite easy, although it is rarely, if ever, done in practice. Thus, it is difficult to compare our data with published datasets from other sites in Africa (Fig. 16.3), since none of the studies (some of ours included) are explicit with regard to sampling strategy. For instance, the mean §13C value for herbivores from Chad suggests that at ~5 Ma (Kossom Bougoudi; X = -3.8%), the area was already as open as the areas inhabited by Homo in South Africa about 1.8 Ma, and that by ~3 Ma (Koro Toro; X = +0.3%) the area was virtually pure grassland (Fig. 16.3; data from Zazzo et al., 2000). This might indicate a real paleoenvironmental difference between the two regions, with Chad having generally more open environments, but it might also partly reflect differences in sampling strategy. Indeed, additional sampling and analysis by Jacques (2007) intimates that C4 grasses may have been slightly less extensive than the original study suggested.

It is important to note that we are not suggesting that any of the isotopic studies of herbivore enamel in South, Central, or East Africa were flawed, but only that they were not necessarily concerned with determining the eating habitats of the herbivore faunas on the whole. If herbivore carbon isotope data are to be used for paleoenvironmental purposes in the future, however, the results will be more useful when obtained from collections for which the alpha taxonomy and relative abundance studies are well advanced, or from randomly selected specimens.

Another problem, especially at the South African hominin sites, is the unknown length of time in which a fossil assemblage accumulated. Therefore, the assemblages may or may not fairly represent the fauna found in the vicinity of the cave at any given time. Moreover, the relative percentages of taxa found in each deposit may not accurately reflect the living community at any given moment due to collection biases (Brain, 1981; Behrensmeyer, 1991; Lyman, 1994). These difficulties plague all faunal and palynological analyses, however, so are of no special concern for stable carbon isotope studies. And importantly, a recent taphonomic study of the Sterkfontein Valley sites discussed here revealed that while there are taphonomic biases among these assemblages (e.g., more craniodental remains relative to postcranial material in calcified compared to decalcified/uncalcified sediments), there was no evidence that they influenced the taxonomic composition of the faunas (De Ruiter et al., 2008).

Was this article helpful?

0 0

Post a comment