In Hypericumspecies some specimens have been found to display higher levels of the toxic heavy metal Cd than the proposed guide value of 0.5mgkg. Plant and soil samples from various regions in Eastern Austria were collected to study the variability in the Cd content of the herbs and relate it to selected soil properties and soil micronutrient contents. The soils contained between 0.01 and 0.44mgkg-¹Cd in the dry soil. Altogether about 7% of the collected plant samples exceeded 0.5mgkg-¹Cd of dry shoots, the proposed guide value for the drug Herba Hyperici. Plants low in Cd were found on sites in the Alpenvorland, the Hainburger Berge, the Eastern Viennese Basin and the Northern Limestone Alps where the levels did not exceed 0.3mgkg-¹Cd of dry shoots. Higher Cd levels in the plants were present in material from the Waldviertel, the Southern Wienerwald and the Semmering region where often bioaccumulation factors for Cd above 1 could be calculated. In these regions the soil pH and carbonate content was somewhat lower than in other regions whereas the soil organic carbon varied greatly. In the Waldviertel some plants from a field were somewhat higher in Cd than nearby plants from the natural vegetation. In a regression analysis, the soil pH, soil organic carbon and soil Cd content were the main factors influencing Cd accumulation in the plant shoots.

Pedotransfer functions based on general purpose Freundlich isotherms (GPF) were derived for Cd using different combinations of input variables or development conditions. The models match the criterion of flexibility as they were derived from data of 497 soil samples, which had strongly varying soil characteristics. The prediction efficiency of the approaches was investigated by comparison of measured and simulated sorption data for 124 independent soil samples. The parameterisation of the GPF shows that models derived from data of batch sorption experiments by multiple linear regression achieve best goodness-of-fit and statistical consistency. The use of extraction data or nonlinear regression analysis results in greater deviations from the statistical optima. The best overall performance considering parameterisation and validation was observed for the models including (a) the variables 'proton activity', 'clay content', 'soil organic carbon', 'sum of dithionite-extractable Fe- and Mn-oxides' or (b) the variables 'proton activity', 'potential cation exchange capacity', 'sum of dithionite-extractable Fe- and Mn-oxides'.