Mapping cadmium concentration and the risk of exceeding a local sanitation threshold using indicator geostatistics

In the north-east of Belgium an extensive area has been contaminated with cadmium as a The performances of two geostatistical algorithms (ordinary kriging with a global trend model and indicator kriging) for estimating the topsoil Cd content at 276 test locations are compared. Better re-estimation scores are obtained for the ''E-type'' estimate, that is the mean of the conditional cumulative distribution function provided by indicator kriging. Besides an estimate of Cd content, indicator kriging allows one to assess the probability to exceed the sanitation threshold for vegetable gardens (most risky land use). That threshold varies across the region depending on local clay and organic matter contents. The comparison of estimated probabilities with actual Cd concentrations at test locations indicates a good assessment of the risk by indicator kriging