Abstract
Global and local sensitivity analyses are essential for identifying key parameters in life cycle assessment models. However, due to limited information on parameter uncertainty, they are often overlooked. This paper's objective is to address this gap by proposing a methodological framework for defining input sensitivity, for midpoint and end point indicators, and a quantitative approach for determining input uncertainties. Applied to a case study on xylitol production as a phase change material, the methodology uses Monte Carlo for uncertainty propagation and Python's SALib to calculate Sobol indices. Results show a 2% relative error in midpoint indicators, aligning with pedigree matrix methods. While accuracy depends on choosing the appropriate distribution function, both global and local sensitivity analyses showed consistent outcomes. This structured, user-friendly approach offers decision-makers a simplified yet effective way to prioritize inputs, either by verifying multiple indicators individually or focusing on damage-oriented indicators. Future studies could refine database coefficients and explore their influence on overall uncertainty, as well as the nonlinearity of the model if the parameters are correlated, offering opportunities to enhance accuracy.