Quantifying the trade-off between parameter and model structure uncertainty in life cycle impact assessment.

To enhance the use of quantitative uncertainty assessments in life cycle impact assessment practice, we suggest to quantify the trade-off between parameter uncertainty, i.e. any uncertainty associated with data and methods used to quantify the model parameters, and model structure uncertainty, i.e. the uncertainty about the relations and mechanisms being studied. In this paper we show the trade-off between the two types of uncertainty in a case of maize production with a focus on freshwater ecotoxicity due to pesticide application in The Netherlands. Parameter uncertainty in pesticide emissions, chemical-specific data, effect and damage data, and fractions of metabolite formation of degradation products was statistically quantified via probabilistic simulation, i.e. Monte Carlo simulation. Model structure uncertainties regarding the concentration-response model to be included, the selection of the damage model, and the inclusion of pesticide transformation products were assessed via discrete choice analysis. We conclude that to arrive at a minimum level of overall uncertainty the linear concentration-response model is preferable, while the transformation products may be excluded. Selecting the damage model has a relatively low influence on the overall uncertainty. Our study shows that quantifying the trade-off between different types of uncertainty can help to identify optimal model complexity from an uncertainty point of view.