Age of the oldest known Homo sapiens from eastern Africa

Efforts to date the oldest modern human fossils in eastern Africa, from Omo-Kibish1–3 and Herto4,5 in Ethiopia, have drawn on a variety of chronometric evidence, including 40Ar/39Ar ages of stratigraphically associated tuffs. The ages that are generally reported for these fossils are around 197 thousand years (kyr) for the Kibish Omo I3,6,7, and around 160–155 kyr for the Herto hominins5,8. However, the stratigraphic relationships and tephra correlations that underpin these estimates have been challenged6,8. Here we report geochemical analyses that link the Kamoya’s Hominid Site (KHS) Tuff9, which conclusively overlies the member of the Omo-Kibish Formation that contains Omo I, with a major explosive eruption of Shala volcano in the Main Ethiopian Rift. By dating the proximal deposits of this eruption, we obtain a new minimum age for the Omo fossils of 233 ± 22 kyr. Contrary to previous arguments6,8, we also show that the KHS Tuff does not correlate with another widespread tephra layer, the Waidedo Vitric Tuff, and therefore cannot anchor a minimum age for the Herto fossils. Shifting the age of the oldest known Homo sapiens fossils in eastern Africa to before around 200 thousand years ago is consistent with independent evidence for greater antiquity of the modern human lineage10.

There are only minor geochemical differences between MER silicic tephra (e.g. ref. 1 ), highlighting the importance of combining evidence from the volcanological record and the chronostratigraphy of distal sequences (the latter point is discussed in the main text) when characterising and correlating samples.
Our study includes proximal samples from Corbetti and Shala, which are the only MER systems known to have produced major eruptions between ~170 ka and ~250 ka (ref. 2 ), the timeframe relevant for the tuffs considered here. The volumes of silicic magma associated with the caldera formation of Shala (86-170 km 3 ) and Corbetti (25-63 km 3 ) are by far the largest among MER volcanoes 2 . Tephra from the major eruptions of these two volcanoes are thus the most likely to be preserved within distal sedimentary formations.
Our set of distal tephra focused on tuffs from the time interval relevant to the dating of KHS and Omo I fossils at Kibish, i.e. from ~170 ka, according to Brown et al. (2012, ref. 3 ). We were unable to locate the Nakaa'kire Tuff at its type section, nor were we able to access samples from the WAVT at Herto.
Consequently, our study focuses on geochemical characterisation of the KHS, and identification and dating its proximal counterparts. Important chronostratigraphic control is provided by the characterisation, correlation and dating of the overlying tephra ETH18-8. In addition, accessible tuffs from the Gademotta and Konso formation that appear stratigraphically consistent with the timeframe of interest were sampled and included for comparison.

Methods
All geochemical datasets were initially reviewed to identify any clear outliers arising from either (i) accidental incorporation of a crystal inclusion in the glass analyses, or (ii) glass shards suffering unusually high alkali mobilisation / Na-loss, indicated by Na2O values < 2 wt% and low totals <91 wt%.
Outlier removal was carried out conservatively to prevent accidental removal of shards that might represent true variability in magmatic composition. Marginal outliers were removed from plots, but have been left in Extended Data Table S1 for completeness (marked as Discarded analyses). There is evidence for greater glass alteration through alkali loss and hydrogen exchange in all of the distal tuff units, compared to the proximal ignimbrite samples ( Figure S3), in line with past studies of Quaternaryage distal peralkaline tuffs in Ethiopia 1,4 . Consequently, Na2O and K2O were considered as indicative, not diagnostic in proximal-distal comparisons. Concentrations of P2O5 were below or too close to detection limit in all samples and are therefore not reported in Extended Data Table S1.
Geochemical correlations were based on a range of major and trace element bi-plots (Figures 3 and S4), which allow the internal variability of the datasets to be observed and visual assessment of the overlap between samples be made. Principal Component Analysis was run on the major and trace element datasets (separately) to verify bi-plot interpretations and explore the data in multivariate space ( Figure   S5). To avoid miscorrelations due to differential alteration, Na2O and K2O were removed prior to normalisation for the major element PCA, and the trace element PCA was limited to immobile elements and those comfortably above their limits of detection (Y, Zr, Nb, La, Ce, Nd, Hf, Ta, Th).
The pantelleritic rhyolite nature of the glasses analysed here ( Figure 3) as well as their level of alteration leaves only a few immobile major oxide elements useful to differentiate glass compositions.
Concentrations of oxides such as CaO and TiO2 ( Figure 3, Extended Data Table S1) are typical of other silicic products of the MER systems (e.g. ref. 1 ). Large arrays of immobile trace element abundances within each sample ( Figure S4) are a common feature of peralkaline melts. For example, the peralkaline Green Tuff on the island of Pantelleria shows a wide spread of immobile trace element abundances through the c. 7-m-thick eruptive sequence, with Zr varying by 1500 ppm, Nb by 300 ppm and Th by 35 ppm (ref. 5 ). Such internal variabilities preclude straightforward correlation on element-element biplots ( Figure S4), and therefore we represent trace elements using ratio values to accentuate minor compositional differences and remove the potential effect of concentration variations in the erupted products ( Figure 3). This is justified because ratios of immobile trace elements plot as linear arrays and have been shown to be highly effective at distinguishing the different MER tephra and magmatic sources (e.g. refs. 6,7 ).

Tephra correlations
In the following, we describe the composition of each sample and discuss their correlations.

The ca. 233 ka Qi2 Shala ignimbrite and KHS tuff
Samples 17-14A1, 17-14B5 and 17-14C from the Qi2 ignimbrite display a homogeneous pantelleritic rhyolitic composition (Table S1). Samples 17-14B5 and 17-14C were analysed for their trace element abundances, whilst sample 17-14A1 did not contain suitable glass shards. Trace element abundances in 17-14C glasses reveal two populations (Table S2, Figure S4 Figure S5). 4 We note that the bimodality of 17-14C is not reflected in the KHS tuff. We stress that given the magnitude of the Qi2 eruption, geochemical heterogeneity across the different phases of the eruption is not surprising, and it is likely that tephra from distal exposures may only partially reflect the compositional spectrum of proximal deposits as observed in the Campanian Ignimbrite 9 . This could be due to (i) the erosion of the topmost part of the KHS tuff where co-ignimbrite ash from the last phase of the Qi2 eruption would have deposited 10 ; or (ii) a variation in the transport and deposition process of the co-ignimbrite ash of the Qi2 units (i.e. related to eruption intensity, plume height and prevailing wind fields).
Glasses from Kibish 18-8 and Konso TA-56 tuffs reveal similar bi-modal compositions on major element biplots (Figure 3, Table S1), with a more differentiated sub-population that overlaps the Shala Qi2/KHS clusters. Immobile trace element abundances of 18-8 and TA-56 samples overlap each other and plot on that of the Corbetti COI2E ignimbrite ( Figure S5, Table S2). Figure S5 shows that Konso TA-56 glasses display some more differentiated outliers, while a couple of Kibish 18-8 glass shards are less evolved, yet all the outliers plot on the same differentiation trend. We attribute these compositional ranges to the peralkaline nature of these samples, which show large variations in immobile element concentrations, but limited variation when ratioed (Figure 3 ; Table S7). Error bars shown are relative standard deviations derived from repeat measurements of matrix match glass secondary standard ATHO-G (n=15 ; Table S7). They are plotted in the upper right corner of each plot for clarity and rescaled to the value of the centre point. Figure S5. Ordination diagram showing plot of first two principal components of major and minor element oxide data (left, alkalis removed) and immobile trace elements (right), from the tuffs sampled in this paper. Clustering of correlated tuffs is indicative, not perfect, due to the impact of outlier glass shard analyses and geochemical variability within individual datasets, which in some cases comprise <10 hits. PCA and plots generated using the in-built analytical tools within the RESET Database 5 .