Search for: All records

Abstract. Deposition of sulfuric acid in ice cores is important both for understanding past volcanic activity and for synchronizing ice core timescales. Sulfuric acid has a low eutectic point, so it can potentially exist in liquid at grain boundaries and veins, accelerating chemical diffusion. A high effective diffusivity would allow post-depositional diffusion to obscure the climate history and the peak matching among older portions of ice cores. Here, we use records of sulfate from the European Project of Ice Coring in Antarctica (EPICA) Dome C (EDC) ice core to estimate the effective diffusivity of sulfuric acid in ice. We focus on EDC because multiple glacial–interglacial cycles are preserved, allowing analysis for long timescales and deposition in similar climates. We calculate the mean concentration gradient and the width of prominent volcanic events, and analyze the evolution of each with depth and age. We find the effective diffusivities for interglacial and glacial maximums to be 5±3×10-9 m2 a−1, an order of magnitude lower than a previous estimate derived from the Holocene portion of EDC (Barnes et al., 2003). The effective diffusivity may be even smaller if the bias from artificial smoothing from the sampling is accounted for. Effective diffusivity is not obviously affected by the ice temperature until about −10 ∘C, 3000 m depth, which is also where anomalous sulfate peaks begin to be observed (Traversi et al., 2009). Low effective diffusivity suggests that sulfuric acid is not readily diffusing in liquid-like veins in the upper portions of the Antarctic Ice Sheet and that records may be preserved in deep, old ice if the ice temperature remains well below the pressure melting point. 

The fairness of machine learning-based decisions has become an increasingly important focus in the design of supervised machine learning methods. Most fairness approaches optimize a specified trade-off between performance measure(s) (e.g., accuracy, log loss, or AUC) and fairness measure(s) (e.g., demographic parity, equalized odds). This begs the question: are the right performance-fairness trade-offs being specified? We instead recast fair machine learning as an imitation learning task by introducing superhuman fairness, which seeks to simultaneously outperform human decisions on multiple predictive performance and fairness measures. We demonstrate the benefits of this approach given suboptimal decisions.