Uncertainty Information in LCI-Databases and Its Propagation Through an LCA Model

This article deals with uncertainties in particular with the uncertainty of inventory data analysis. Uncertainties cannot be avoided in LCA studies. Therefore, they should be analysed and interpreted. One problem is that in many LCA studies uncertainties are not noted. In ecoinvent, which is an example for an inventory database, the lognormal distribution is choosen as the dataset’s standard distribution type. One reason is that many quantities found in nature can only take positive values. In most cases, however, the normal distribution can as well be taken as the default distribution type. In this article the uncertainty information of the ecoinvent datasets is explained and uncertainty analysing methods like the Monte Carlo simulation, der pedigree matrix or the sensitivity analysis are described. Furthermore, the convolution is mentioned as a method for analyzing the sucsessive uncertainty propagation through an LCA model. For using the convolution, the LCA data should be independent, continuous and normal distributed. In addition to that the LCA model should be a linear and not a complex system. Since the convolution is a new approach further research will be required.