Speaker: Dr. Grégoire Mariéthoz
Abstract:
Many hydrological processes taking place at the Earth’s surface are directly observable, either through in-situ measurements or by using remote sensing techniques. In contrast, understanding properties of the subsurface poses formidable observational challenges. Firstly, measurements of the Earth’s interior are either difficult and expensive to acquire (e.g., boreholes) or indirect (e.g., measurement of subsurface flow and tracer concentrations). Secondly, the Earth’s interior is 3-dimensional, resulting in very empty data spaces. For example, even intensive drilling campaigns and geophysical surveys of an aquifer often sample only a small fraction of the total rock volume.
This talk will present stochastic aquifer modeling approaches that allow quantifying uncertainty in such data-poor problems. For many practical questions, the uncertainty in subsurface hydraulic properties further propagates into uncertainty in water resources management problems—for example, the transport of contaminant to a drinking water well, the intake of a groundwater-based desalination plant, or the behavior of a tracer in a subglacial drainage system. Stochastic aquifer models allow using a statistical description of the unobserved system to formulate ensemble predictions, resulting in a distribution of possible outcomes. The use of stochastic models will be illustrated in different environments, focusing on applications that integrate indirect data through inverse approaches.