Fate modelling of chemical compounds with incomplete data sets

Impact assessment of chemical compounds in Life Cycle Impact Assessment (LCIA) and Environmental Risk Assessment (ERA) requires a vast amount of data on the properties of the chemical compounds being assessed. These data are used in multi-media fate and exposure models, to calculate risk levels and other indicators. ERA typically addresses one specific chemical, but in an LCIA, the number of chemicals encountered may be quite high, up to hundreds or thousands. This study explores the development of meta-models, which are supposed to reflect the “true”multi-media fate and exposure model in an approximate way. The idea is that not all data needed in a multi-media fate and exposure model are completely independent and equally important, but that there are physical-chemical and biological relationships between sets of chemical properties. A statistical model is constructed to underpin this assumption, and to provide simplified proxies for the more complicated “real”model relationships. In the presented study two approaches for the reduction of the data demand associated with characterization of chemical emissions in USEtoxTM are tested: The first approach yields a simplified set of mode of entry specific meta-models with a data demand of app. 63 % (5/8) of the USEtoxTM characterization model. The second yields a simplified set of mode of entry specific meta-models with a data demand of 75 % (6/8) of the original model. The results of the study indicate that it is possible to simplify characterization models and lower the data demand of these models applying the presented approach. The results further indicate that the second approach relying on 75 % of the original data set provides the meta-model sets which best mimics the original model. An overall trend observed from the 75 % data demand meta-model sets, is that except for fate factors covering the route from soil emission (natural and agricultural soil) to aquatic compartments, good correlation between the predicted fate factors (derived from the meta-models) and the observed fate factors (modeled in USEtoxTM) are obtained. The regression coefficients obtained for the predicted fate factors plotted against the observed fate factors, excluding the fate factors covering the route from emission to soil to aquatic compartments, were all in the range 0.6818≤ R2 ≤0.9470