#### Presentation Title

An implemented approach for estimating uncertainties for toxicological impact characterisation

#### Keywords

uncertainty, lcia, toxicity, multimedia modelling

#### Start Date

1-7-2004 12:00 AM

#### Abstract

One approach accounting for parameter and model uncertainty is implemented in the LCIA (lifecycle impact assessment) method IMPACT 2002. The uncertainty is estimated for intermediate results fromthe chemical fate, human intake fraction, and two toxicological effect modules. Overall uncertainty estimatesare then arithmetically calculated. Results are presented for impact contributions in the contexts of aquaticecosystems and human health. The approach of Hofstetter (1998) was adapted for estimating the uncertaintyrelated to chemical fate and human intake fractions. A fundamental problem when estimating uncertainties for1000’s of substances consists of the lack of uncertainty distributions for all of the input data and the need tohave a practical approach to assign distributions to each chemical. Hofstetter (1998) proposed the use of fixedfactors for clusters of substances. The choice of a factor is then dependent on the emission medium, exposureroute, and the robustness of the model relative to the chemical being considered. The factors are initiallydetermined for representative substances for each category using evaluation data, expert judgement, orapproaches such as Monte Carlo. There is then no need to repeat the Monte Carlo calculations. Multiplyingand dividing the geometric mean estimate by a factor provides an estimate of the upper and lower 95thpercentile confidence interval bounds. The human health effect factor uncertainty is similarly defined andreadily combined through addition with that of the intake fraction. Using expert judgement, three uncertaintyclasses were proposed to estimate uncertainty related to the human effects input data. These effects dataaccount for both the risk of an effect, as well as the potential consequences of population-based exposures.The uncertainty for ecotoxicological effects is currently related to the number of species tested for aquaticspecies in the water column. The more species test results available, the more robust the estimate of theecotoxicological factor is assumed to be. For estimating the ecotoxicological effect factor uncertainty, thecombined use of two distinct approaches was suggested, – the higher uncertainty estimate being adopted. Thecombination of both guaranteed more robust results compared to applying either method – both being basedon differing assumptions related to the sample versus the population distribution. The presented approachproved to be very transparent, robust but while reflecting our current level of knowledge, quick to use, and iseasily applied in practice to combine the uncertainty of the emissions inventory with those of the impactassessment phase in a life cycle assessment study.

An implemented approach for estimating uncertainties for toxicological impact characterisation

One approach accounting for parameter and model uncertainty is implemented in the LCIA (lifecycle impact assessment) method IMPACT 2002. The uncertainty is estimated for intermediate results fromthe chemical fate, human intake fraction, and two toxicological effect modules. Overall uncertainty estimatesare then arithmetically calculated. Results are presented for impact contributions in the contexts of aquaticecosystems and human health. The approach of Hofstetter (1998) was adapted for estimating the uncertaintyrelated to chemical fate and human intake fractions. A fundamental problem when estimating uncertainties for1000’s of substances consists of the lack of uncertainty distributions for all of the input data and the need tohave a practical approach to assign distributions to each chemical. Hofstetter (1998) proposed the use of fixedfactors for clusters of substances. The choice of a factor is then dependent on the emission medium, exposureroute, and the robustness of the model relative to the chemical being considered. The factors are initiallydetermined for representative substances for each category using evaluation data, expert judgement, orapproaches such as Monte Carlo. There is then no need to repeat the Monte Carlo calculations. Multiplyingand dividing the geometric mean estimate by a factor provides an estimate of the upper and lower 95thpercentile confidence interval bounds. The human health effect factor uncertainty is similarly defined andreadily combined through addition with that of the intake fraction. Using expert judgement, three uncertaintyclasses were proposed to estimate uncertainty related to the human effects input data. These effects dataaccount for both the risk of an effect, as well as the potential consequences of population-based exposures.The uncertainty for ecotoxicological effects is currently related to the number of species tested for aquaticspecies in the water column. The more species test results available, the more robust the estimate of theecotoxicological factor is assumed to be. For estimating the ecotoxicological effect factor uncertainty, thecombined use of two distinct approaches was suggested, – the higher uncertainty estimate being adopted. Thecombination of both guaranteed more robust results compared to applying either method – both being basedon differing assumptions related to the sample versus the population distribution. The presented approachproved to be very transparent, robust but while reflecting our current level of knowledge, quick to use, and iseasily applied in practice to combine the uncertainty of the emissions inventory with those of the impactassessment phase in a life cycle assessment study.