Technical Comments

Response to Comment on the Paleoenvironment of Ardipithecus ramidus

Science  28 May 2010:
Vol. 328, Issue 5982, pp. 1105
DOI: 10.1126/science.1185466


Cerling et al. contest our interpretation of the woodland habitat preference of Ardipithecus ramidus. However, their reconstruction of a predominantly open grassy environment with riparian woodlands is inconsistent with the totality of the fossil, geological, and geochemical evidence. In the Middle Awash, Ar. ramidus fossils are confined to the western portion of the sampled Pliocene landscape where the species is associated with woodland to grassy woodland habitat indicators.

The comment by Cerling et al. (1) challenges our interpretation of the woodland habitat preference of Ardipithecus ramidus. We are pleased that the authors agree that “open savanna grassland was not the environmental context of Ardipithecus.” However, their statement that our interpretation “relies heavily on the carbon isotopic composition of paleosol organic matter and carbonate” is inaccurate and overlooks much of the integrated, high-resolution evidence underlying our inferences about Ardipithecus habitat preference (2). Our papers presented the totality of the relevant physical and biological evidence, including structural geology, geochronology, sedimentology, soil isotope geochemistry, paleoflora, taphonomy, terrestrial invertebrates, micromammals, and birds. We also detailed the relative abundances of larger mammals, as well as their ecomorphology, diet, and enamel isotopic composition, including those of Ardipithecus itself (35).

Cerling et al. (1) conflate regional “environmental context” with the specific habitat inferred for Ardipithecus. They posit a “savanna” habitat with woody grasslands and restricted areas of riparian forest and rely on speculations about carnivore behavior to explain the abundance of large arboreal and terrestrial woodland mammal species consistently found with Ardipithecus. They also rely on a modern reference set that induces a bias toward more open habitats to reinterpret the paleosol isotopic evidence.

The large size and diverse content of the new physical, chemical, and biological evidence directly associated with Ardipithecus distinguish these Ethiopian occurrences from others in Pliocene Africa. The taphonomic integrity and tight temporal resolution of this evidence uniquely allowed our targeted assessment of actual hominid habitat preference at Aramis. We made it clear that the regional and local environmental mosaics included grasslands as potential habitats available to Ardipithecus but concluded that “Its ecological habitat appears to have been woodland-focused” (6) rather than grassland-based.

In 1994, Kingston et al. (7) described an analogous Kenyan rift valley Neogene setting as “...a heterogeneous environment with a mix of C3 and C4 plants [that] has persisted for the last 15.5 million years.” They concluded that “...interpretations of the origin of hominids in East Africa during the Late Miocene should be considered within the context of a heterogeneous mosaic of environments.” It would be ecologically naive to assume that Ardipithecus equally exploited all of the habitats available to it within the Afar rift. The challenge lies in linking hominid with preferred habitat.

Almost all hominid-bearing occurrences in Pliocene Africa yield faunal assemblages derived from temporally and spatially shifting mosaics of open and closed vegetation settings (5, 8). As a consequence, it has been nearly impossible to accurately place the relatively rare hominids from most of these sites in any one of the habitats available within the tectonically active and ecologically dynamic rift valley settings they necessarily sample. We presented clear evidence that even the local Aramis landscape of the regional Afar environment at 4.4 million years ago (Ma) supported such an ecological mosaic: “...the early Pliocene Afar included a range of environments, and the local environment at Aramis and its vicinity ranged from forests to wooded grasslands...” (5). We demonstrated environmental heterogeneity even within the ~9-km transect available for the Lower Aramis Member. We have no means to assess how extensive or dominant the relatively open environments were in the 4.4-Ma Afar rift and hence did not attempt such assessment. Rather, we focused on determining which of these available habitats Ardipithecus preferred, concluding from the biology (anatomy, dental wear, and isotopic compositions), faunal and flora associations, and sedimentology that this organism preferred the woodland habitat abundantly documented at Aramis by its associated birds, large and small mammals, gastropods, and fossil wood (35).

Cerling et al. (1) focus on our soil isotopic data to conclude that Aramis localities record an open wooded grassland, which they characterize as a “tree or bush savanna, with 25% or less woody cover” that likely included “habitats ranging from riparian forest to grassland.” They agree with our floral habitat terminology but then fail to adhere to this United Nations Scientific and Cultural Organization (UNESCO) system, using the terms “tree or bush savanna.” The UNESCO (9) classification notes that savanna is too broadly and variably defined for formal characterization of African floral habitats. In figures 1 and 2 in (1), the term “savanna” is used for habitats ranging from >60% (e.g., “Brachystegia woodland savanna” in Fig. 1B) to less than 10% woody cover. The appropriate terms in the UNESCO classification are simply “Brachystegia woodland” and “grassland”, respectively (9).

Fig. 1

Paleosol carbonate δ13C and δ18O values along a west-east transect of fossil localities in the lower Aramis Member of the Sagantole Formation in the Middle Awash Valley. Regression formulae and one-tailed P values are indicated. The trend toward increasing values for both isotopes to the east show that less wooded and drier habitats predominated on the eastern half of this transect, findings consistent with the floral and faunal content of the Ardipithecus-free sediments on this pole of the natural transect.

Cerling et al. (1) envision strips of riparian (river or lake margin) forest or thicket at Aramis within a “bush savanna.” However, expected sedimentological, taxonomic, and taphonomic evidence for such settings is absent at the hominid-bearing fossil localities at Aramis (3, 4). The widespread massive carbonate horizons with preserved fossil wood, seeds, and terrestrial invertebrate traces in the Lower Aramis Member (3) likely formed from shallow groundwater carbonates and springs. These wetter soils most likely supported patches of more densely wooded vegetation in an otherwise grassy woodland habitat. These dense patches do not constitute riparian habitats, and there is no local evidence for a major channel or of the aquatic fauna expected in the kind of setting postulated by Cerling et al.

We concluded that fossil soil (paleosol) carbon isotope data from Aramis “reflect woodland to grassy woodland savanna floral habitats with 30% to 70% C4 plants” (3). Cerling et al. effectively illustrate this [figure 2 in (1)]. However, their interpretation of the isotopic data in terms of the floral habitats leads to a systematic overestimation of proportion of grassland. This bias first stems from their inclusion of modern comparative habitats ranging to deep multistory Amazonian and Australian rainforests. These extremely low δ13C forest values are irrelevant for interpreting the isotopic signature of Aramis. This is because the C3 plant biomass δ13C values in such humid multistory dense forests are substantially more negative than the average for C3 plants and, in particular, for drier African forests (10, 11). Indeed, Cerling et al. have not themselves previously used such low carbon isotope forest end-members to reconstruct the floral environments of African fossil sites (7, 8, 1215).

Cerling et al. (1) appropriately note that the presence of carbonates in Aramis paleosols indicate a semiarid environment with less than 1000 mm of rainfall. Their figure 1B and table 1 correctly show that drier African forests have δ13C values of –25.5 ± 1.4 per mil (‰). Using the curvilinear regression line in figure 1A in (1) as a guide, such forests would be classified as supporting merely 50 to 60% woody cover. The biased selection of ranges of isotope ratios for woodland habitats is further illustrated by the low average δ13C value (–24.9‰) presented in table 1 and figure 1B for the Brachystegia woodland of Mongu Zambia (1). Mongu has been reported with 40 to 65% tree canopy cover [under canopy mean soil organic matter (SOM) δ13C = –25.4 ± 2.6‰], but a substantial portion of this landscape has an intercanopy mean δ13C of –18.2 ± 0.9‰ (16). This intercanopy mean SOM δ13C value for the Mongu grassy woodland is similar to that of –17.6 ± 1.6‰ (n = 57) for the Lower Aramis Member [tables S1 and S2 in (3)] and less negative than the Aramis mean adjusted for the fossil fuel combustion effect to –19.1‰ for comparisons with modern values (15). We therefore conclude that the convex second-order polynomial regression line in figure 1A in (1) is substantially biased away from the δ13C value ranges of Aramis. We take their regression with caution, because the relationship of percentage of woody cover with soil δ13C values is known to vary substantially depending on a multitude of factors, including soil texture, fire intervals, grazing, and agricultural effects (16, 17). Furthermore, because present-day tree cover must be influenced to varying degrees by human activity, the direct applicability of modern measures of canopy cover and soil carbon (and phytoliths) to past ecosystems and paleosols must be more carefully considered.

Cerling et al. (1) contest our carbon isotopic evidence for an east-west gradient from more wooded biotopes in which Ar. ramidus is consistently found, to more open biotopes, where the species and its associated ecological signals are absent. They state that “Consideration of each site separately shows that there is no isotopic gradient from west to east” [figure 2 in (1)]. This is incorrect. Using data from tables S1 and S2 in (3), Fig. 1 shows that there is a statistically significant increase in carbonate δ13C along this transect that indicates an increase in C4 grass biomass accompanied by a weaker trend of increasing δ18O, reflecting higher temperatures and/or evaporation.

We calculated a provisional “evaporation deficit” of 1500 mm between the most evaporation-sensitive folivores (Giraffa and Sivatherium) and evaporation-insensitive water-dependent grazer (Hippopotamus), which can be considered a maximum of such estimates (18). Cerling et al. suggest that this indicates a riparian setting nearly as arid as the Lake Turkana basin of northern Kenya and southern Ethiopia—one of today’s hottest and driest regions of eastern Africa (1). However, if more species were used to calculate this index, the apparent evaporation deficit at Aramis would be substantially lower. Furthermore, the application of the aridity index in South African Plio-Pleistocene hominid sites has produced results that contradict associated environmental evidence (19). Until these discrepancies can be satisfactorily explained, the promise of the enamel oxygen isotope aridity index for paleoclimatic reconstruction should be considered with caution.

Kingston (8) notes that “Paleoecological data most relevant and significant to interpretations of hominid evolution are those collected directly in association with hominid fossils—from the same local stratigraphic horizon” and that “Associated vertebrate fauna provide some of the most direct evidence of early hominid paleoecology.” Ironically, Cerling et al. (1) seem to overlook the hominid-associated abundance data for larger mammals and birds. We remain unconvinced that the proportions of kudus and monkeys can be explained by unidentified “specialized ambush predators.” The Cerling et al. argument that the Aramis faunal assemblage does not provide evidence for extensive dense closed forest because of the absence of duikers is particularly unconvincing given the available taphonomic and relative abundance data (4).

The abundance and relative proportions of Aramis bovid and monkey taxa contradict Cerling et al.’s rendering of Ardipithecus habitat. Cerling et al. propose that the dominance of kudu (Tragelaphus) in the Aramis herbivore assemblage is consistent with riparian thickets. However, the extreme rarity of reduncine and alcelaphine bovids in the Aramis fauna argues strongly against riparian settings grading into grasslands. More important, the Aramis bovid dental and postcranial ecomorphological evidence is dominated by clear signals of closed habitats (5). Cerling et al. (1) also seem to overlook the high abundance of leaf-eating colobine and cercopithecine monkeys at Aramis. Both their postcranial ecomorphology and enamel isotopes indicate that the dominant Aramis monkeys were arboreal and in part forest-dwelling (see below). Consequently, Cerling et al.’s conclusion that Ardipithecus occupied a habitat with less than 25% woody cover is incompatible with its close association with the numerically dominant Aramis kudus and monkeys, especially in conjunction with the hominid’s own craniodental and postcranial anatomy.

Cerling et al. (1) contend that our interpretation of tooth enamel isotopic evidence of closed woodland to forest habitats proximal to Aramis is “misleading.” We concluded that “small patches of closed canopy forests were present, although woodlands to wooded grasslands probably dominated” (5). Evidence for forest patches is found in δ13C values for some Aramis Member Kuseracolobus individuals low enough (–14.2 and –14.8‰) to have formed their enamel in a closed canopy forest similar to the Kibale Forest in western Uganda [see supporting online material for (5)]. These low values suggest that forest patches large enough to encompass the home range of a colobine monkey troop (0.5 to 1.6 ha) (20) were locally sustained in the Lower Aramis Member. The carbon isotope composition of Ar. ramidus tooth enamel also seems to have been overlooked by Cerling et al. in evaluating its habitat preferences. It has a lower mean δ13C value than all previously analyzed hominids from South and East Africa (5, 21). Compared with all later hominid species in open habitats, Ar. ramidus consumed substantially less C4/CAM-based food, a fact consistent with our inference of a woodland habitat for this taxon at Aramis.

Because of the unique integrity and resolution of its depositional context and taphonomic content, the Lower Aramis Member has finally afforded multiple, independent lines of tightly associated evidence to meet the challenge of resolving the habitat of Ardipithecus within the environmental mosaics typical of rift valleys. We documented taxa ecologically tied to more open grasslands (alcelaphine, hippotragine, antilopine bovids, xeric-adapted rodents, and birds) and to aquatic habitats (fish, waterfowl, and hippos) at Aramis, but all are relatively very rare (35). At the Aramis hominid-bearing localities, the most abundant taxa are morphologically adapted and/or isotopically tied to the woodland habitat, as is Ardipithecus itself.

The vision of apes trekking bipedally between increasingly isolated forest patches has maintained its allure across decades of research. The repeated discovery of obligatorily bipedal, megadont Australopithecus in later open habitats erroneously reinforced this notion of bipedality’s beginnings. It was perhaps inevitable that proxy records reflecting global shifts in carbon isotope values would be postulated as the missing piece of the puzzle of hominid origins (22).

By focusing too coarsely on the regional environment, Cerling et al. (1) seem to overlook evidence that differentially and consistently links Ardipithecus to a woodland habitat and thereby distinguishes it ecologically from Australopithecus. We contend that compared with Ar. ramidus, Australopithecus was more ecologically flexible, probably ranged more frequently and further into the open environments that Cerling et al. term “tree or bush savanna,” and evolved remarkably distinct and highly derived dietary and locomotor adaptations to this end.

After assessing the totality of the pertinent environmental and ecological evidence, we concluded that Ar. ramidus preferred the more wooded habitats among the available spectrum in the regional geography: “...the integration of available physical and biological evidence establishes Ar. ramidus as a denizen of the closed habitats along this continuum” (5). The Aramis evidence is not easily accommodated by an environmentally deterministic view that involves globally shrinking forests spawning savanna-striding hominids. We contend that this narrative is now undermined by the totality of data from 4.4-Ma Aramis. These rich, diverse data are spatially and chronologically intimately associated with Ardipithecus, thereby providing an unparalleled view of the early hominid niche within the larger geographic setting.

References and Notes

  1. We employ “habitat” in the ecological sense, wherein the species or population is the subject: the place where an organism or a biological population normally occurs and is most likely to be found.
  2. The responding authors note that the comment by Cerling et al. (1) was directed at multiple papers, each with a different authorship, published in the 2 October 2009 issue of Science. Due to space and time allowances, this smaller set of authors has taken responsibility for the response, while at the same time acknowledging the wider authorship of our original papers under comment here.
View Abstract


Navigate This Article