Abstract
The preparation of the models forecasting the concentrations of toxic chemicals in water and sediment by the use of their chemical and watershed properties was examined to apply the models to the risk-assessment of them. The database on 28 chemicals and 64 stations was constructed and applied for calibrating and validating the statistical models i.e. multi-regression analysis and quantification theory type I. In the both models and irrespective of analysis type, there observed within one order of magnitude difference in the case of water and 1.5 orders of magnitude difference in the case of sediment between the forecasted and the observed values for almost all the data. The models on an each chemical basis or on an each station basis, the dominant input variables, and the number of samples for constructing the model having sufficient accuracy were also investigated and discussed.