Multiple batch extraction test to estimate contaminant release parameters using a Bayesian approach

Industrial activities produce vast amounts of weakly contaminated materials which are commonly reused as filling materials on natural ground. There is a strong demand to define guidelines for the application of these materials, to estimate the leaching potential of contaminants from the materials, and to assess the potential hazard for groundwater pollution. We present a multiple batch experiment, where measurements of liquid-phase concentrations at varying liquid/solid ratios are used to estimate the total mass of contaminant that can be extracted from a contaminated material with a mild extractant like water. Furthermore, the experiment yields estimates of the isotherm describing the partitioning of the contaminant between the solid and liquid phases, and a concentration that might be expected under soil hydraulic conditions representative for the field situation. Model parameters are estimated from liquid-phase concentrations within a Bayesian framework by applying the Shuffled Complex Evolution Metropolis Algorithm (SCEM-UA), an efficient Markov Chain Monte Carlo sampler. A sensitivity analysis and inversions of synthetically generated data corrupted with noise show the general suitability of the proposed method. An uncertainty analysis for model parameters and model predictions shows the expected accuracy of the estimates. An application to concentration measurements obtained from a multiple batch extraction test illustrates the applicability of the approach for a real situation.