Upscaling flow and transport processes in the subsurface is mostly based on statistic approaches. The soil parameters are modelled as random fields with a given correlation structure. Based on these statistic properties, mean and variance of measurable quantities (fluid content, solute concentration, pressure heads or flow velocities) are derived. In general this has to be carried out in a perturbation approximation, where only the first and second moments of the random parameters are considered. The underlying assumption is here, that the parameters are mutli-Gaussian fields, which are uniquely quantified by the first two moments. Multi-Gaussian fields are smooth and continuously changing.
Spatial parameter distributions of real soil hardly resemble multi-Gaussian fields. Instead, soil is composed of different separated zones of material, which are well or badly connected with each other. Connectivity properties of certain zones of the soil can have a very strong influence of the averaged flow. Information about the geometric properties of the soil structures on the large scale might be more important for upscaling of flow in the unsaturated zone than stochastic properties.
In this project upscaled models for the flow and transport in the unsaturated zone are developed using different methods (homogenization and percolation theory approaches) and tested for materials with different geometric properties. The methods are tested using numerical simulations and compared to experiments in the laboratory in collaboration with the soil physics group of the Institute of Terrestrical Ecology of the ETH Zurich.
10/2004 - 09/2007
Bundesministerium für Bildung und Forschung (BMBF)