Application of Geostatistics to 3D Modelling of Contaminated Sites: A Case Study

Abstract

This paper presents, through an actual case study, a general approach for the application of geostatistics to the diagnosis of polluted sites. Diagnoses are preliminary work prior to deciding about remediation plans. The main objectives are to delineate the soil pollution for some concentration thresholds, to derive contaminated soil volumes, and to assess the uncertainty attached to them. Volumes can then be converted into treatment costs which provide support for decision making regarding remediation or complementary investigations.

For such a diagnosis task, geostatistics provide many tools that allow better integrating different types of information and give direct access to all kind of uncertainty measurements. In this study, the geostatistical approach relies on the following three-step procedure: 1) model the geometry of the site which here consists of infill materials delimited by a topographic surface and underlying alluvium; 2) fill in, through conditional simulation, the whole site domain with the different soil types which have been identified, each soil type being associated to a specific contamination degree (discrimination of concentration populations); 3) generate, from the simulated soil type images, probability maps that provide the local probability of exceeding given concentration thresholds.

In this case study, the available information consist of contaminant concentrations measured on soil samples, soil descriptions recorded on several drill-holes, and resistivity measurements along several profiles over the site. Resistivity measurements are not directly related to soil contamination. It is proven, however, that they are correlated to some specific soil types and allow detecting their presence or absence.