Abstract
Geostatistics provides innovative tools for the radiological characterization of nuclear facilities under decommissioning as well as contaminated sites for remediation. The relevance of this approach relies on the presence of a spatial continuity for radiological contamination. In this case, geostatistics provides reliable activity estimates, uncertainty quantification and risk analysis, which are essential decision-making tools for decommissioning and dismantling projects of nuclear installations.
The experimental variogram γ(h) is calculated by averaging, within classes of distance h, the variability contribution of each couple of data; this contribution is usually quantified by the half-squared difference of the measured values. In this paper, a parallel is drawn with an alternative formula of the statistical variance to be able to break down the different variance contribution (sampling duplicates, measurement replicates, spatial variability). The application on one chemical case and one radiological case demonstrates that, the relative uncertainty on the laboratory results (analyses on destructive samples) can be seen as negligible in comparison to the sampling uncertainty on the one hand and the spatial variability on the other hand.
In addition, classification of volumes according to a threshold seems to be quite robust in comparison to the relative accuracy of the lab results. The artificial introduction of a systematic bias (± 50% for instance) leads to a corresponding limited impact on contaminated volumes (only ± 10% or ±20%).
