Author(s)

A. Dan & P. Jamet

Abstract

This article presents an original method for the estimation of pollutant source-term chemical composition and of the parameters governing their groundwater transport in a given brownfield site. This method is based on field data exhaustion (primarily the pollutant mass fractions measured in sampling points, in groundwater). We develop site-specific metamodels (like kriging and neural networks)—regression models carried out on sampled data from simulation model outputs—to extract the functional relation between the chemical compositions computed with a transport code at monitoring locations and the source composition. Applying the determined metamodels to new chemical compositions observed in groundwater samples, we directly estimate the source composition of the sampled pollution. Thus, it becomes possible to estimate the pollutant source terms (source chemical composition and past industrial activities or past uses of the site) that have led to the observed contamination. We illustrate this method on chemical compositions from sites polluted by chlorinated solvent. The aim of this research is to determine, by cost efficient means, the nature and the dynamics of the pollutants involved in a brownfield site under investigation. Keywords: pollutant source identification, source-term chemical composition estimation, transport parameters estimation, metamodel, chlorinated solvents, kriging, neural networks.

Keywords

pollutant source identification, source-term chemical composition estimation, transport parameters estimation, metamodel, chlorinated solvents, kriging, neural networks.