Life Cycle Assessment (LCA) models are inherently uncertain due to the model structure interacting with the model inputs and modeling choices. The methods of sensitivity analysis (SA) aim at retracing the causes of the uncertainty of the results of a model. This work takes apart the methods of uncertainty propagation, SA, and LCA, and identifies the requirements of appropriate SA methods for LCA. Global input space assessment and inclusion of correlation are identified as important factors which both analytical and sampling SA methods have issues addressing. An analytical expression for covariance is formulated that combines research on the uncertainty propagation methods to address the posed requirements. Its performance is tested and shown to be promising, but further manipulation is required for practical application in SA for LCA in the future.