|Title||Methods for sensitivity analysis in life cycle assessment of animal production systems|
|Author(s)||Groen, E.A.; Bokkers, E.A.M.; Boer, I.J.M. de; Heijungs, R.; Boer, I.J.M. de|
|Event||WIAS Science Day 2015, Wageningen, 2015-02-05/2015-02-05|
Animal Production Systems
|Publication type||Abstract in scientific journal or proceedings|
|Abstract||Environmental impact of the agri-food industry has been of increasing concern, and in particular international awareness about the impact of animal production systems has been rising. Life cycle assessment (LCA) is a commonly applied framework that quantifies the environmental impact of a product over its entire production chain. Input parameters required to describe animal production systems, however, can be uncertain due to e.g. temporal variability or unknowns about the true value of emission factors. Analysing the effect of uncertainty in input parameters can be done by means of a sensitivity analysis (SA). Although many LCA studies have been performed over the last decade, few applied a systematic and consistent SA to address the effect of input uncertainties on the output. Two main types of SA exist. A local SA determines the effect of a (small) change in one of the input parameters at a time, whereas a global sensitivity analysis determines how much each input parameter contributes to the output uncertainty. This study aims to give insight into state-of-the-art methods for global SA in LCA and to compare several methods based on their ability to explain the output variance. This is done by applying all methods to two case studies with different sizes and input uncertainties. The first case study is a hypothetical test system, while the second case study described the production of whitefish in the northeast Atlantic.
From the limited LCA literature on global SA available, three methods were found that quantify the contribution to output variance: (standardized) regression coefficients, (Spearman) correlation coefficients and a Taylor approximation. Outside the LCA domain we found other methods, such as random balance design and the Sobol’ method, that quantify the contribution to output variance. For most methods, it is not known under which conditions they perform optimally, or if there is one method that outperforms other methods in LCA. Results showed that for case studies without any outliers (irrespective of the size of the case studies), Key issue analysis, standardized regression coefficients, and Spearman correlation coefficient were able to explain most of the output variance. For case studies with outliers (irrespective of their size) we found that the Sobol’ method was able to explain most of the output variance followed by SCC. We conclude that using sophisticated methods for global SA might only be useful for high quality data, so that the quantification of the uncertainty is as accurate as the provided input data. For LCAs in animal production systems this means that standardized regression coefficients, and Spearman correlation coefficient might be preferred because it is easier to understand and apply. This theoretical exercise shows that recommendation towards environmental improvement options in animal production systems can be made taking into account input uncertainties. For example, when comparing scenarios, the regression or correlation coefficients indicate which parameters that should be known with high confidence, before presenting results. Also, parameters with the highest regression (or correlation) coefficients indicate parameters that could contain opportunities for improvement options.