This project aims at developing a model to describe the statistically averaged flow behaviour of infiltration of contaminants into the soil. The contaminants considered here are dense nonaquous phase liquids (e.g. petroleum), so called DNAPLs.
Modelling of such infiltration processes is very complex, as due to the different fluid properties the flow process is influenced by capillary forces, gravity forces as well as viscous forces. The motion of a fluid phase through the soil is described by a highly nonlinear coupled system of equations. When modelling infiltration of DNAPLs into the soil, one of the major difficulties is the heterogeneous structure of the soil. The impact of heterogeneities is here more pronounced than for single-phase flow or solute transport problems. Not only the soil permeability but also the capillary entry pressure distribution of the soil has a strong influence on the flow patterns. Capillary entry pressure can act as a barrier to the infiltrating fluid or it can focus the flow. This leads to highly irrelgular flow patterns and can cause channeling of pooling of the infiltrating fluid. On the spatial scales of infiltration processes capillary forces are important and cannot be neglected.
To predict infiltration of DNAPL into soil in a deterministic manner is therefore an almost impossible task. However, based on the statistic properties of the soil parameter, the statistic average of fluid distribution can be calculated. The soil parameters are modelled as correlated random fields with given statisic properties. The mean fluid distribution as well as its variance is an important information for risk assessment. The averaged flow behaviour is modelled as an equivalent one in a homogeneous medium. The impact of heterogeneities is captured by so called effective parameters and processes. The spreading of the averaged fluid interface for instance could be described by a dispersion process with a dispersion coefficient, which quantifies the spreading.
Due to the complexity of the processes involved, averaged models cannot be derived in a closed form. Instead we have to rely on simplifications and approximations. This is in particular true for the statistic properties of the velocities. In this project we want to derive such simplified approaches in order to finally derive upscaled models for DNAPL infiltration. This is done with different methods. Laboratory experiments of infiltrations in artificially designed heterogeneous media with well defined statistic properties are carried out in order quantify the relation between the statistic soil parameter and the statistic input quantities for the averaged flow equations. The experiments are restricted to simple structures. Therefore numerical simulations will be carried out based on the experiments in order to determine the statistic input properties in a more comprehensive way. In parallel equivalent homogeneous models for the infiltration process are derived with perturbat!
ion theory and homogenization theory methods.