Quantitative uncertainty and sensitivity analysis of the OMNIITOX Base Model algorithm

Abstract: This thesis analyzes the quantitative uncertainty and sensitivity of an intermediate parameter in the OMNIITOX Base Model algorithm which can be used for calculation of characterisation factors to carry out Life cycle assessments (LCA) and Environmental risk assessments(ERA). The quantitative uncertainty of LCA and ERA has been widely recognized, but thereexists no estimation of the quantitative uncertainty of the OMNIITOX BM. The purpose ofthis thesis is to give an example of how large the quantitative uncertainty of an intermediate parameter of OMNIITOX BM can be and which input parameters that contributes themost to this uncertainty. This example could inspire to go further and make a quantitativeuncertainty and sensitivity analysis for the complete algorithm, involving all parameters. Themethodology that is used in this thesis can be modified to address the issue in more generalterms.First, the intermediate parameter Rate coefficient from air to fresh water in Europe was chosenfor further analysis. The uncertainty analysis was limited to the nature specific parameters.For further investigation of the statistical properties of these parameters, the reference chemical Toluene was chosen for the chemical specific input parameters, provided as single pointvalues. The distribution functions for the nature specific input parameters were estimatedfrom information given in literature. A simulation program was implemented in Matlab whichruns the algorithm for the chosen intermediate parameter with random numbers from the estimated input parameter distribution functions. The output data was later analyzed with theMatlab-tool dfittool to measure the uncertainty and fit a distribution function to the chosenintermediate parameter. The simulation program also measured the sensitivity of the inputparameters with normalized correlation coefficients. The distribution function for the inter-mediate parameter was later applied to similar intermediate parameters in another simulationprogram which runs the complete algorithm of OMNIITOX BM. With this program, the uncertainty of the characterisation factors was estimated.The quantitative uncertainty of the Rate coefficient air to fresh water Europe is remarkablyhigh, with a 95%-confidence interval of [-0.0130, 0.0270], mean = 0.00679 and std = 0.01060if a normal distribution cut-off fit is chosen for the output data. The input parameters thatcontributes the most to this uncertainty is Mixing height of air and Particle dry depositionvelocity in air. When randomly generating numbers from the distribution function of Ratecoefficient air to fresh water Europe in the calculation of characterisation factors, they wereinsignificantly affected. If an equal distribution was assumed for all rate coefficients as theanalyzed intermediate parameter, the uncertainty of the characterisation factors became extremely high.