Parameter estimation for groundwater models by indicator kriging

Earlier studies have shown that indicator kriging provides reasonable estimates of the spatial distribution of hydraulic conductivity based on both quantitative 'hard' and qualitative 'soft' data. One crucial question is the optimal choice of threshold values which define the indicator variables. Another problem is to select a representative point estimator which extracts the most suitable K-value from the local cdf. Two methods to optimize this selection will be presented: a pure statistical cross check with measured data (i) and a verification of estimates by comparison with actual now conditions (ii). The paper will illustrate these ideas by means of an actual data-set taken from pleistocene glacio-fluvial deposits in North Germany.