Geostatistics for Contaminated Sites and Soils: Some Pending Questions

abstract

This paper addresses three key issues when using geostatistics in soil remediation studies. Firstly, we point out the necessity of using an appropriate model when the contaminant grade distributions are highly skewed due to a large proportion of samples that do not contain the pollutants. In this case, a valuable solution may consist in combining an indicator variable for describing the presence or absence of pollutant and a Gaussian random variable for modelling the transformed contaminant grades at locations where they were present. This model is shown to reduce the uncertainty on the classification of the soils into safe or polluted. Secondly, we address the problem of change of support between the soil samples and the remediation units. A method is proposed to achieve the upscaling. But at the unit scale, average grades above some critical threshold may be explained by the occurrence, at the soil sample scale, either of a large proportion of (possibly moderately) excessive grades or by a few sample with (possibly very) high grades. In one or the other situation, the health or environmental risk is certainly different. The third issue discussed relates to the evaluation of contaminated soil volumes when those soils are affected by numerous contaminants. Multiplying the number of potential contaminants also multiplies the risk for soils to be contaminated. Better integrating the correlation between contaminants then appears essential.