Using kernel density estimation and the Dirichlet distribution for uncertainty quantification of building material emissions

Whole-building life cycle assessment (wbLCA) is typically a deterministic process that uses single-value inputs, often from environmental product declarations (EPDs), to estimate a building’s embodied carbon emissions. Several major sources of uncertainty are frequently overlooked, and existing methods of uncertainty quantification (UQ) are limited by overreliance on expert judgment and their use of unsubstantiated distributions that misrepresent irregular datasets. To address these shortcomings, this study introduces KL2, a UQ method in wbLCA that employs kernel density estimation and the Dirichlet distribution to quantify the expected range of embodied carbon for building materials. In this study, each KL2 input parameter is examined through increasingly complex hypothetical scenarios with different uncertainty characteristics. Then, as a proof-of-concept, KL2 is applied to a set of global structural steel EPDs. KL2 aims to enhance transparency in environmental impact data by identifying critical gaps that impede the development of more accurate, informative LCAs.