Geostatistics for radiological evaluation: study of structuring of extreme values

Geostatistics applied to radiological evaluation of nuclear premises provides sound methods to estimate radiological activities, together with their uncertainty. Quantification and risk analysis of contaminated areas are initially performed by applying geostatistical methods relying on the multi-Gaussian assumption. However, the application of the classical bi-Gaussian model for disjunctive kriging proves sub-optimal due to the spatial structuring of high and low values. The beta model which pertains to the class of Hermitian isofactorial models is potentially better suited to radiological evaluation as it allows a continuous evolution from a mosaic to a pure diffusive model. In the test case, disjunctive kriging estimates are obtained by applying in turn the beta model and the pure diffusive model. The comparison of estimation outcomes shows rather limited differences, primarily located in and around the homogeneous contaminated areas.