Sensitivities of flow and transport parameters in fractured porous media using automatic differentiation

Abstract

SUMMARYThe hydraulic complexity of hardrock aquifers with, for example, fast fluid flow in fractures, storage effects in the rock matrix, high contrasts in permeability, and a high spatial variability of structures requires a high standard of experimental and numerical techniques. This paper focuses on the interpretation of tracer tests. The sensitivity of tracer breakthrough curves with respect to flow and transport parameters is analysed for different model types in order to identify the relevant flow and transport processes. This leads to a better understanding of the system behaviour and facilitates the interpretation of experimental results. The technique of automatic differentiation is used to evaluate the sensitivities. The main advantage of this technique is that the sensitivities are calculated without truncation error.

First-order uncertainty analysis using Algorithmic Differentiation of morphodynamic models

“…Our approach, FOSM/AD, utilizes Algorithmic Differentiation (AD) to perform a first-order second-moment uncertainty analysis (Melching, 1992). AD was previously applied by Vogel et al (2006) for a sensitivity analysis in environmental flow models. More information about AD methods can be found (Griewank and Walther, 2008;Naumann, 2012, www.autodiff.org).…”