Reconstructing Past Environments: Pitfalls and Future Directions

Hominin paleoecologists often use carbon stable isotope analyses of fossil herbivore tooth enamel to reconstruct the environments our ancestors evolved within. A key assumption in these studies is that herbivore diets should reflect the vegetation structure of ancient landscapes. But is this true?

Like Comment
Read the paper

One of the most crucial aspects of understanding the evolution of our lineage is being able to reconstruct the environments within which our ancestors evolved. These environments would have exerted specific selective pressures on early hominins leading to biological and behavioral adaptations documented by the fossil and archaeological records. Hominin paleoecology has seen an explosion of geochemical techniques for reconstructing past environments. One specific method, carbon stable isotope analysis of fossil herbivore tooth enamel, has gained particular steam as it is a relatively straightforward and inexpensive approach. Carbon isotopes have been shown to faithfully reflect herbivore diets that, in turn, are used to make broad inferences about the overall vegetation structure of a fossil site. Increasingly, researchers are interested in comparing differences in the carbon values of particular taxa, such as mixed-feeders like impala, across fossil sites to understand how environments varied in space and time. This is an approach I have constantly used in my own research, but, until recently, I had thought very little about the assumptions that underlie it.

After discussions with my colleagues John Rowan from the University at Albany, Andrew Barr from the George Washington University, and Matt Sponheimer from the University of Colorado Boulder, a few things started to bother us. Key among these was our suspicion that herbivore diets might not closely track the proportion of different vegetation types in the ecosystems, which means they would be of little use for making vegetation inferences in the fossil record. Our suspicion was supported as we dug into the modern ecological literature, but there was no comprehensive analysis of how herbivore diets varied across multiple ecosystems. Our research question had emerged: do herbivore diets, as interpreted from carbon stable isotopes, actually reflect variation in vegetation structure across ecosystems? This is a fundamental assumption that many hominin paleoecologists make when generating paleoenvironmental reconstructions. We specifically focused on woody cover, as this aspect of vegetation structure has been central to hypotheses of hominin evolution for decades.

But where to start? We knew that our colleague Thure Cerling and his team had published the largest dataset of modern carbon isotope data for African herbivores in 20151, but how would we leverage this into testing our question? Furthermore, what would be an appropriate measure of woody cover? Conveniently, the Cerling et al. dataset is organized into protected wildlife parks. This allowed us to use shapefiles from the World Database on Protected Areas (WDPA) to gather spatial data for each of the samples. Our woody cover dataset came from a 2018 study by Zander Venter and colleagues2 on woody plant encroachment in Africa. Venter’s team had prepared high-resolution raster datasets of woody cover for the entire African continent and they graciously provided us access to the full datasets through Google Earth Engine. From there, we wrote a few short bits of code that allowed us to extract woody cover data for each of the parks sampled in the Cerling et al. carbon isotope dataset. 

With our dataset in hand, we analyzed the relationship between woody cover and herbivore carbon isotopes with linear models. We had hypothesized that if carbon isotope values were a reliable indicator of woody cover, we should find significant negative relationships between the two variables (i.e., more C3 consumption in woodier environments). Of eighteen herbivore lineages included in the study, eleven came back with significant relationships between woody cover and carbon isotope values, but this wasn’t quite the confirmation we were looking for. Most of these significant relationships were weak, indicating that woody cover only explained a small fraction of the variation in carbon values. Furthermore, some taxa, like mixed-feeding impala, had significant, but positive relationships between carbon isotope values and woody cover, indicating that in habitats with more woody cover impala were feeding on significantly more grasses (Figure 1). At this point, we turned to non-faunal paleoenvironmental data to dig into this issue even deeper. A large body of paleosol carbonate carbon isotope data indicate little evidence for closed forested ecosystems in the paleoanthropological record. Therefore, we re-ran our analyses by removing those ecosystems. In this truncated dataset we found even fewer significant relationships between carbon isotope data and woody cover.

Figure 1. Simplified schematic showing the expected relationship between herbivore enamel carbon isotope values and woody cover (gray dotted line) and our findings for select taxa. In the paper we predicted the direction of the relationship, but not the specific slope; slope shown in the figure is a heuristic tool. While some taxa weakly show the expected negative relationship, others show an unexpected positive relationship (meaning they consume more C4 biomass in woodier environments). Only for elephants does there seem to be a relatively strong negative relationship between ecosystem woody cover and carbon isotope values.

At this point the reader may be wondering ‘do these findings imply that enamel carbon isotope values are useless for understanding hominin paleoecology?’ To this, we respond with an emphatic no. First, carbon isotope values of tooth enamel still provide a record of past diet—this is a crucial piece of information. Second, herbivore carbon isotopes might still be useful for inferring past vegetation in a community context. In 2010, Zelalem Bedaso and colleagues3 suggested a way to combine carbon isotope values from multiple herbivore taxa into a single measure, but at the time there wasn’t a clear way to demonstrate the applicability of this approach in modern ecosystems. This approach takes into account the average amount of vegetation consumed by specific herbivores as well as their abundance in the ecosystem. Using our approach, we tested the Bedaso et al. community carbon measure on twenty-two ecosystems across Africa. While we would like to end by saying this approach is fully validated, that we cannot do. This approach revealed that the relationship between woody cover and herbivore enamel values is likely non-linear (another assumption most researchers have made), but that a non-linear model may explain > 50% of the relationship. There is a lot more work to be done! We encourage all of us to look towards the promise of new approaches that will provide accurate reconstructions of past environments and a better context for understanding our evolutionary history.

Check out the full paper here!

Poster image: An elephant feeding in Kruger National Park, South Africa. Photo Credit: John Rowan.


 1. Cerling, T. E. et al. Dietary changes of large herbivores in the Turkana Basin, Kenya from 4 to 1 Ma. Proceedings of the National Academy of Sciences USA, 112(37), 11467-11472 (2015)

2. Venter, Z. S., Cramer, M. D., Hawkins, H. J. Drivers of woody plant encroachment over Africa. Nature Communications9(1), 1-7 (2018).

3. Bedaso, Z., Wynn, J. G., Alemseged, Z., Geraads, D. Paleoenvironmental reconstruction of the Asbole fauna (Busidima Formation, Afar, Ethiopia) using stable isotopes. Geobios43(2), 165-177 (2010).

Joshua Robinson

Lecturer, Archaeology Program, Boston University