Uncertainty Estimation in Matrix-based Life Cycle Assessment Models

<p>Life Cycle Assessment (LCA) has been applied to help decision-makers understand quantitative environmental effects and impacts through the life stages of a product or process. Matrix-based LCA models are widely incorporated to LCA software tools to simplify the assessment and provide straightforward results. However, these tools do not sufficiently provide the uncertainties that arise from the inventory data as well as from the matrix-based models. To address this problem, in this thesis I use a range method to explore three types of uncertainties (parameter, scenario, and model uncertainties) present in matrix-based LCA models. These three types of uncertainties are assessed separately for two different types of LCA models: the Input-Output- based LCA model, and the process-based LCA models analyzed with matrix methods. IO-based LCA models are studied with the Environmental Input-Output Life Cycle Assessment (EIO-LCA) model, and the US LCI database incorporated to matrix methods is used as an example of process-based LCA models. I selected two demonstrate the results with two environmental effects (greenhouse gas emissions and energy consumptions) and five environmental impacts (global warming, ozone depletion, acidification, eutrophication and ecotoxicity). First, I analyzed the parameter uncertainty in the EIO-LCA model. Publicly available data sources and assumptions are used to estimate the parameter uncertainties of the direct energy consumption in the US industrial sectors. The direct and indirect energy consumption ranges are estimated through the EIO-LCA model. The results show that the parameter uncertainties are generally within -40% to 40% from the default values, with several outliers. Second, I examined the scenario uncertainties in total carbon dioxide emissions by using alternative inputs in the US LCI database. I found that the US LCI database fails to take full advantage of matrix-based methods; when incorporated to matrix-based LCA models, less than 10% of the processes contribute to the indirect environmental effects. The results of scenario uncertainty estimation in the US LCI database show that on average, the total carbon dioxide emissions across all processes are between -30% to -30%. Finally, I addressed the model uncertainty by using different Life Cycle Inventory Assessment (LCIA) methods incorporated in the matrix-based models. The results show that when the US LCI inventories are applied, the uncertainties due to choosing different impact methods are within 5%. This is possibly caused by the incompleteness of the inventories: more than 50% of the characterized substances are excluded in the inventory, resulting in the neglect of some impact values. The results from this study emphasize the importance of estimating uncertainties in matrix-based LCA models. The variability in the LCA results is caused by all three types of uncertainties, as well as the incomplete inventories embedded in the matrix-based LCA models. Future LCA database and software should focus on including uncertainty estimation in the features and improving the inventory data to take full advantage of the matrix-based LCA models.