Inverse methods for consistent quantification of seafloor anoxia using uranium isotope data from marine sediments

Abstract

Uranium isotopes (δ238U) have quickly become one of the most widely-used redox proxies in paleoceanographic studies. The quantitative power of the δ238U proxy derives from the long marine residence time of uranium and the dominance of reduction in fractionating uranium isotopes during removal from seawater. The seawater δ238U value is therefore sensitive to the size of the anoxic sink, and by extension, the area of the seafloor overlain by anoxic waters. Leveraging the ability of carbonates to record and retain the seawater δ238U value, and the ubiquity of carbonate sediments in the geologic record, numerous studies have quantified seafloor anoxia across ocean anoxic events, mass extinctions, and global climatic changes. In most cases, forward models of marine uranium isotope mass balance have been used, illustrating potential histories of seafloor anoxia during these events.Here we show that there are multiple ways in which such forward modeling can lead to spurious inferences of anoxia, including (i) the poor sensitivity of the δ238U proxy when fractional anoxia is high, and (ii) the inherent bias in generating illustrative forward model outputs in stratigraphic sections with expected anoxic intervals. We thus explore inverse modeling approaches to constrain the most likely history of seafloor anoxia for a given δ238U dataset, and ultimately develop a framework for doing so using Bayesian inference via Markov Chain Monte Carlo simulation. We show that this approach can recover simulated trends, and further reconstruct marine anoxia for eight published δ238U datasets. We find that some previous interpretations of anoxic seafloor extent were inaccurate, either because steady state was improperly assumed, or because the illustrative forward models used were poor fits to the data. In order to overcome these issues in future work with the δ238U redox proxy, we have made this model publicly available, and also offer suggestions for the judicious use of forward models. By building on this framework, the future quantification of marine anoxia during transient environmental perturbations can be performed consistently, thereby facilitating robust comparison of anoxic extent between events.