Functional error modeling for uncertainty quantification in hydrogeology

Approximate models (proxies) can be employed to reduce the computational costs of estimating uncertainty. The price to pay is that the approximations introduced by the proxy model can lead to a biased estimation. To avoid this problem and ensure a reliable uncertainty quantification, we propose to combine functional data analysis and machine learning to build error models that allow us to obtain an accurate prediction of the exact response without solving the exact model for all realizations. We build the relationship between proxy and exact model on a learning set of geostatistical realizations for which both exact and approximate solvers are run. Functional principal components analysis (FPCA) is used to investigate the variability in the two sets of curves and reduce the dimensionality of the problem while maximizing the retained information. Once obtained, the error model can be used to predict the exact response of any realization on the basis of the sole proxy response. This methodology is purpose-oriented as the error model is constructed directly for the quantity of interest, rather than for the state of the system. Also, the dimensionality reduction performed by FPCA allows a diagnostic of the quality of the error model to assess the informativeness of the learning set and the fidelity of the proxy to the exact model. The possibility of obtaining a prediction of the exact response for any newly generated realization suggests that the methodology can be effectively used beyond the context of uncertainty quantification, in particular for Bayesian inference and optimization.

Experimental dissolution of molybdenum-sulphides at low oxygen concentrations: A first-order approximation of late Archean atmospheric conditions

The abundance of atmospheric oxygen and its evolution through Earth's history is a highly debated topic. The earliest change of the Mo concentration and isotope composition of marine sediments are interpreted to be linked to the onset of the accumulation of free O2 in Earth's atmosphere. The O2 concentration needed to dissolve significant amounts of Mo in water is not yet quantified, however. We present laboratory experiments on pulverized and surface-cleaned molybdenite (MoS2) and a hydrothermal breccia enriched in Mo-bearing sulphides using a glove box setup. Duration of an experiment was 14 days, and first signs of oxidation and subsequent dissolution of Mo compounds start to occur above an atmospheric oxygen concentration of 72 ± 20 ppmv (i.e., 2.6 to 4.6 × 10−4 present atmospheric level (PAL)). This experimentally determined value coincides with published model calculations supporting atmospheric O2 concentrations between 1 × 10−5 to 3 × 10−4 PAL prior to the Great Oxidation Event and sets an upper limit to the molecular oxygen needed to trigger Mo accumulation and Mo isotope variations recorded in sediments. In combination with the published Mo isotope composition of the rock record, this result implies an atmospheric oxygen concentration prior to 2.76 Ga of below 72 ± 20 ppmv.

Saddlepoint approximations

The saddlepoint method provides accurate approximations for the distributions of many test statistics, estimators and for important probabilities arising in various stochastic models. The saddlepoint approximation is a large deviations technique which is substantially more accurate than limiting normal or Edgeworth approximations, especially in presence of very small sample sizes or very small probabilities. The outstanding accuracy of the saddlepoint approximation can be explained by the fact that it has bounded relative error.