According to the OECD, it is estimated that some 70 000 synthetic chem
icals are in daily use. This number is still growing. Some of these pr
oducts (e.g. polychlorinated biphenyls or chlorofluorocarbons) have ga
ined great public attention due to their unexpected undesirable long-t
erm impact on the ecosystem. The industrialized nations have therefore
implemented and enforced stringent laws, in order to control the use
of potentially harmful substances. It is an almost impossible task to
have the EU Base Set available for all existing substances. The EU and
other organizations like OECD have therefore put together priority li
sts of potentially harmful high-volume chemicals which are assessed by
experts. The last couple of years, a tremendous amount of work has be
en done to develop computer-based modeling systems to predict the envi
ronmental behavior and distribution of chemical substances. There are
models around (e.g. Mackay, USES, TGD EU, and others) to describe envi
ronmental distribution (Predicted Environmental Concentration (PEG)) o
f nonionic and nonpolar substances. For a first assessment, only a few
physical properties (MP, BP, SOL, and possibly pK(a)) are needed. For
ionic and/or polar substances, expert know-how is required. For envir
onmental toxicity (Predicted No Effect Concentration (PNEC)), however,
the situation is far more complex. There are different modes of actio
n for each trophic level (algae, crustacea, fish). Computer models wit
h Quantitative Structure-Activity Relationship (QSAR) offer a scientif
ically attractive tool to predict aquatic toxicity and biodegradation.
There are no models that are applicable to heterogeneous chemical cla
sses without expert know-how. Further progress can therefore be expect
ed. The result of an ERA for perchloroethylene with USES is presented,
and the problems of the adequacy of the model parameters explained. H
ere too, a further improvement is addressed. QSARs are attractive tool
s to design products with improved environmental compatibility at the
research stage even before the first synthesis. In order to achieve su
ch an ambitious aim, QSAR models that also include expert rules and ev
aluated results and their dependence on molecular structures will have
to be further developed.