Uncertainties on the extension of a polluted zone - 59198

Data collected during the sampling of polluted sites are mainly used - through an exploratory and vario-graphic analysis, to characterize to characterize the concentration level and the spatial variability; - at fixed support, to estimate the concentrations in order to map the pollution. Kriging gives also the standard deviation of the estimation error, making it possible to delimit the zones in which the estimation is considered to lack in precision. If a proportional effect is present the map of error standard deviation has to be corrected to take into account the increase of spatial variability with the local concentration mean. A confidence interval can be derived conventionally from the kriging estimation and the associated error standard deviation. For a fixed limit threshold, the polluted site can then be divided in three areas: - the polluted zone and the not polluted zone, at a fixed statistical risk, not necessary the same for both sets; - the 'zone of uncertainty', in which the estimated concentrations are close to the threshold. Because of the estimation error, it is not possible to specify if the exact concentrations exceed or not the threshold. For an hydrocarbon soil pollution, usual and geo-statistical forecasts are compared. The effective consequences of these various forecasts on the quality of the site remediation are quantified and discussed. (authors)