Managed aquifer recharge (MAR) provides efficient removal for many organic compounds and sum parameters. However, observed in situ removal efficiencies tend to scatter and cannot be predicted easily. In this paper, a method is introduced which allows to identify and eliminate biased samples and to quantify simultaneously the impact of (i) redox conditions (ii) kinetics (iii) residual threshold values below which no removal occurs and (iv) field site specifics. It enables to rule out spurious correlations between these factors and therefore improves the predictive power. The method is applied to an extensive database from three MAR field sites which was compiled in the NASRI project (2002e2005, Berlin, Germany). Removal characteristics for 38 organic parameters are obtained, of which 9 are analysed independently in 2 different laboratories. Out of these parameters, mainly pharmaceutically active compounds (PhAC) but also sum parameters and industrial chemicals, four compounds are shown to be readily removable whereas six are persistent. All partly removable compounds show a redox dependency and most of them reveal either kinetic dependencies or residual threshold values, which are determined. Differing removal efficiencies at different field sites can usually be explained by characteristics (i) to (iii).