Reply to Cerling et al.

Abstract

As we previously noted (White et al. 2010), Cerling and colleagues’ (2010, 2011) model for interpreting the relationship between paleosol carbon isotope values and percentage of woody plant cover differs substantially from that used in all previous studies, including those by the coauthors of Cerling et al. 2010 (Cerling 1984; Fox and Koch 2003; Kingston 2007; Levin et al. 2008; Wynn 2004). Their new model has two serious deficiencies. First, they use data from sites with sandy soils that are known to support lower grass biomass than finegrained fertile soils (Bell 1982; Sankaran et al. 2005). This results in lower δC values. Second, they include deep forest sites that have extremely negative δC values due to the closed canopy effect, which shifts the C3 end-member value to lower values than applicable to drier settings relevant here. Thus, their model underestimates percentage of woody cover when applied to Ardipithecus ramidus–bearing sites characterized by fine-grained distal floodplain soils formed on fertile volcanic parent materials. In their current critique (Cerling et al. 2014, in this issue; Dominguez-Rodrigo 2014, comment by Tim D. White), they use a variety of configurations of eastern African soil isotopic data, but they still have not considered the relationship between soil properties and tree-grass ratios. In Cerling et al.’s (2011) modern African data, soil δC values appear on average 2‰ lower on sites that probably represent sandy soils, such as those formed on weathered Basement Complex rocks, coastal sands, Kalahari sands, or high-energy floodplains. Cerling et al. (2014, in this issue, fig. 1) show a “cumulative probability distribution” of woody cover from Aramis that differs from their previous reconstruction (Cerling et al. 2010). Their original presentation had Aramis woody cover “average” 17%, with a 2–standard deviation (SD) range of 0%–35%, which means virtually all localities would have been grassland. In the revised estimate based on eastern African control data, Aramis woody cover “averages” 25%, with a 2-SD range of 5%–45%. This is a modest but meaningful difference, because the latter suggests a greater mosaic of habitats. However, as discussed above, it is probable that their model still underestimates woody cover in the Aramis member sites. We have consistently concluded that the range of habitats