Within the EU 5th framework OMNIITOX project a simple base model (SBM) is currently being developed to calculate characterisation factors for life cycle impact assessment of toxic releases. This SBM will be derived from a multi-media base model (BM) using statistical techniques combined with mechanistic insights in fate and effect relationships. The SBM must be able to cover a much higher number of substances than the BM and hence it must be based on fewer and more available substance properties than the BM. Preparatory work has been performed to gain insight into the structure of the BM e.g. in terms of how the input parameters enter the regression equation. In the absence of a final OMNIITOX BM a model of similar complexity USES-LCA, has been used as surrogate BM. We have applied partial least square of latent structure regression (PLSR) and combined insights from this with knowledge on data availability limitations to select key parameters that explain much of the variance and at the same time are relatively easily available. Further, PLSR was used to derive linear SBM models. In further investigations multiple linear regression (MLR) will be used to derive predictive equations for SBM characterisation factors. The result of this will be tested on common sense and environmental knowledge and a mechanistically understandable SBM will be developed by rounding off the coefficients of the regression equations. Preliminary results including PLSR derived linear SBM’s of this work is presented.
|Title of host publication||SETAC Europe 14th Annual Meeting Prague, Czech Republic, 18-22 April 2004: Abstracts|
|Place of Publication||Brussels|
|Publication status||Published - 2004|
|Event||SETAC 2004 - Prag, Czech Republic|
Duration: 14 Nov 2004 → 18 Nov 2004
|Period||14/11/2004 → 18/11/2004|