Quaternary history of atmospheric oxygen
Oxygen (O2) is perhaps the most important gas molecule for eukaryotic life on Earth, but its atmospheric concentration in Earth history is not well constrained. Unlike CO2, which has an interesting and useful acid chemistry, O2 has a large reservoir in the atmosphere with a formidable oxidizing power. This property makes O2 hard to quantify. For example, redox-sensitive proxy systems can only indicate the presence or absence of O2 with regard to a certain threshold. Attempts to directly measure past oxygen level were limited to a few geologic archives that can preserve the atmosphere, such as amber (Berner and Landis, 1988) and fluid inclusions in minerals (Blamey et al, 2016). The robustness of these archives however has been questioned (Hopfenberg et al, 1988; Yeung, 2017).
Recently, polar ice cores have been used to reconstruct Pleistocene O2 concentrations (Stolper et al, 2016), which is seemingly long overdue. After all, the first ice core drilling project in Greenland started in 1955 and Antarctic drilling began in 1968. Today, Earth historians have become familiar with the famous CO2 and CH4 records from deep ice cores such as Vostok and Dome C. Why didn't we hear about an ice-core atmospheric oxygen record for so many years?
Long story short: people tried and it was hard. In fact, the mystery of past O2 concentrations inspired some pioneering efforts to measure the elemental composition of the trapped gases in ice cores, in particular O2/N2 and Ar/N2 ratios (Sowers et al, 1989). However, these ratios were found to be 5-10‰ lower than their true atmospheric values (meaning that the δO2/N2 and δAr/N2 in the trapped gases are -5 to -10‰ against air). It was soon realized that the composition of the trapped gases inside the ice deviates from its true atmospheric value. Two processes are at play: (1) bubble close-off in firn that physically traps the air inside the ice; and (2) gas losses from the ice after it has been drilled from the ice sheets.
Figure 1. A record of atmospheric oxygen based on four deep ice cores, modified from Stolper et al (2016).
Theses processes have long prevented a clear geochemical interpretation of δO2/N2 of the trapped gases in ice cores. For example, Landais et al (2012) first observed in the trapped air of Dome C ice cores a decline of the δO2/N2 over the past 800 kyr, but the authors did not provide a conclusive explanation on the origin of this decline (i.e. a natural long-term O2 trend versus a storage artifact). Daniel Stolper, a brilliant geochemist and now an Assistant Professor at UC Berkeley, conducted a detailed time-series analysis of ice core δO2/N2 data and removed the effect of bubble close-off by its correlation with insolation and corrected for gas loss by forcing the trend line to reach 0 when δO2/N2 is regressed against time (Figure 1). After these treatments, revealed was a persistent decline of δO2/N2 in the trapped air over the past 800 kyr, which is interpretted to reflect a decreasing atmospheric O2 levels in the late Pleistocene (Stolper et al, 2016).
We found 2-million-year-old ice in Allan Hills, and began asking: can we deduce atmospheric O2 beyond 800 kyr? Turns out not as easy as it sounds. The key challenge is that million-year-old blue ice samples from Allan Hills are not stratigraphic continuous, which has two implications. First, we cannot plot the measured δO2/N2 against depth and treat it as a time-series. Second, recall that the effect of bubble close-off was removed by its correlation with insolation in Stolper et al (2016). Insolation was known because a precise age was known. When no precise chronology is available, we do not know what insolation values should be used to correct for δO2/N2 in the trapped air. That is the unfortunate case of blue ice.
Figure 2. A schematic illustration of paired δO2/N2-δAr/N2 approach (not to scale).
I came up with a solution to account for insolation-induced and gas loss fractionation in the discontinuous blue ice: Ar/N2 ratios. The basic premise is that δAr/N2 covaries with δO2/N2 in the trapped air, since both properties are fractionated during bubble close-off and gas losses, albeit with different magnitudes. This covariation has been previously documented (Bender et al, 1995), but not applied further. Importantly, as both Ar and N2 are chemically very inert, the Ar/N2 ratios should remain stable on multi-million-year timescales. Therefore, in a cross-plot where δO2/N2 is plotted against δAr/N2 (Figure 2), the offset between data of different ages should solely reflect the difference in atmospheric O2. The long-term δO2/N2 decline in Dome C ice core is successfully reproduced by this approach.
When applied to the Allan Hills blue ice gas records, this method yields a late-Pleistocene O2 decline that is statistically indistinguishable from the trend observed in deep ice cores (at 95% confidence interval), which gives us additional confidence in its validity. Intriguingly, between 1.5 Ma and 800 ka, there is no statistical difference in reconstructed atmospheric O2 (see Figure 3 below). Thus it appears that the initiation of the decline in the late-Pleistocene atmospheric O2 approximately coincided with the Mid-Pleistocene Transition (MPT). Mechanistically, we speculate that larger and more expansive glaciers after the MPT led to enhanced chemical weathering and greater oxidation of sedimentary organic carbon, which in turn increased the sink flux of O2. For a more detailed discussion, check out our latest paper (Yan et al, 2021) here: https://www.science.org/doi/10.1126/sciadv.abj9341
Figure 3. Paired δO2/N2-δAr/N2 as measured in Allan Hills ice core samples, binned according to their age (dashed lines represent the 95% confidence interval of the regression slopes).
References
Bender, M., Sowers, T. and Lipenkov, V., 1995. On the concentrations of O2, N2, and Ar in trapped gases from ice cores. Journal of Geophysical Research: Atmospheres, 100(D9), pp.18651-18660.
Berner, R.A. and Landis, G.P., 1988. Gas bubbles in fossil amber as possible indicators of the major gas composition of ancient air. Science, 239(4846), pp.1406-1409.
Blamey, N.J., Brand, U., Parnell, J., Spear, N., Lécuyer, C., Benison, K., Meng, F. and Ni, P., 2016. Paradigm shift in determining Neoproterozoic atmospheric oxygen. Geology, 44(8), pp.651-654.
Hopfenberg, H.B., Witchey, L.C., Poinar, G.O., Beck, C.W., Chave, K.E., Smith, S.V., Horibe, Y. and Craig, H., 1988. Is the air in amber ancient? Science, 241(4866), pp.717-721.
Landais, A., Dreyfus, G., Capron, E., Pol, K., Loutre, M.F., Raynaud, D., Lipenkov, V.Y., Arnaud, L., Masson-Delmotte, V., Paillard, D. and Jouzel, J., 2012. Towards orbital dating of the EPICA Dome C ice core using δO2/N2. Climate of the Past, 8(1), pp.191-203.
Sowers, T., Bender, M. and Raynaud, D., 1989. Elemental and isotopic composition of occluded O2 and N2 in polar ice. Journal of Geophysical Research: Atmospheres, 94(D4), pp.5137-5150.
Stolper, D.A., Bender, M.L., Dreyfus, G.B., Yan, Y. and Higgins, J.A., 2016. A Pleistocene ice core record of atmospheric O2 concentrations. Science, 353(6306), pp.1427-1430.
Yan, Y., Brook, E.J., Kurbatov, A.V., Severinghaus, J.P. and Higgins, J.A., 2021. Ice core evidence for atmospheric oxygen decline since the mid-Pleistocene transition. Science Advances, 7(51), eajb9934.
Yeung, L.Y., 2017. Low oxygen and argon in the Neoproterozoic atmosphere at 815 Ma. Earth and Planetary Science Letters, 480, pp.66-74.