Sources of uncertainty and their importance in life cycle assessment and material flow analysis

: (Alternative Format Thesis)

Abstract

The challenges of the circular economy require us to make assessments at an increasing level of detail and disaggregation. But at higher levels of detail, data and knowledge limitations become more significant, and analysis is only practical if uncertainty is considered. Properly reflecting and understanding the impact of uncertainty is important. It can be used to focus data-collection efforts on the most influential parameters to achieve the greatest reduction in uncertainty, thereby efficiently improving the robustness of the results. The parameters identified as most influential in reducing uncertainty can then inform actionable recommendations for policy, for example by guiding decisions on which data should be monitored and collected and by supporting the introduction of appropriate measures. In this frame, the goal of this study is to develop methods in material flow analysis (MFA) and life cycle assessment (LCA) to incorporate and understand a wider range of uncertainty sources by combining recent advances in uncertainty and sensitivity analysis methods in MFA and LCA and by building new models.

There are sources of uncertainty that are currently not systematically analysed for uncertainty in MFA and LCA and whose impacts are not well understood on results. For example, in MFA, conversion factors used to harmonise data to the unit of the study are not modelled as uncertain. The key limitation in existing MFA methodologies is the lack of a flexible modelling approach that can propagate uncertainty in a wide range of model inputs and simultaneously handle overdetermined systems. In LCA, several recent developments allow uncertainty and global sensitivity analysis for model inputs which present opportunities to be used in combination. Ecoinvent provides characterised uncertainty for certain sources of uncertainty. But there are sources of uncertainty not characterised by the Ecoinvent database, which never show up in sensitivity analysis results and whose impacts on model outcomes remain unknown.

Three case studies demonstrate how recent advances in uncertainty and sensitivity analysis in LCA and MFA can be further developed to understand the importance of sources of uncertainty detailed in the previous paragraph. The first use case is about mapping wood flows in the UK with a high level of disaggregation on finished products to support material efficiency strategies. A deterministic statistical reconciliation MFA method is extended to take data on conversion factors, propagate uncertainty by the Monte Carlo method, and analyse the uncertainty by global sensitivity analysis (GSA). This new ``ProbPACTOT'' model is used to understand the impact of uncertainties in 39 conversion factors, 56 process recipes, and 117 data observations. The second use case is about making comparative statements in LCA to support early design developments under high scenario uncertainty: comparing the climate change impact of additive versus conventional manufactured treatments for knee osteoarthritis and understanding important model inputs that shift conclusions. This case study provides a comprehensive use case of combining existing tools, showing how complementary approaches, such as Monte Carlo contribution analysis and GSA, can be used to include and interpret uncertainty in LCA. Finally, the third use case is about understanding the impact on life cycle impact assessment (LCIA) results of modelling uncertainty in market datasets of the Ecoinvent database. A new GSA method is developed for LCA for screening uncertain market mixes, not covered by existing methods, and is demonstrated on product systems of the Ecoinvent database, screening 434 market mixes in one of the examples shown.

The three case studies are aimed at database developers and modellers. The case studies demonstrate how combining methods in uncertainty and sensitivity analysis can be used to support detailed assessments that can be used to bring nuances in comparative analysis, to inform policy recommendations, enhance data transparency, and target data collection efforts.

Date of Award	18 Feb 2026
Original language	English
Awarding Institution	University of Bath
Supervisor	Rick Lupton (Supervisor), Stephen Allen (Supervisor) & Marcelle McManus (Supervisor)