Premium
Incremental Material Flow Analysis with Bayesian Inference
Author(s) -
Lupton Richard C.,
Allwood Julian M.
Publication year - 2018
Publication title -
journal of industrial ecology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.377
H-Index - 102
eISSN - 1530-9290
pISSN - 1088-1980
DOI - 10.1111/jiec.12698
Subject(s) - material flow analysis , computer science , markov chain monte carlo , reuse , bayesian probability , industrial ecology , uncertainty analysis , bayesian inference , material flow , production (economics) , inference , scenario analysis , markov chain , data mining , operations research , econometrics , machine learning , sustainability , artificial intelligence , mathematics , statistics , engineering , economics , simulation , ecology , macroeconomics , biology , waste management
Summary Material flow analysis (MFA) is widely used to study the life cycles of materials from production, through use, to reuse, recycling, or disposal, in order to identify environmental impacts and opportunities to address them. However, development of this type of analysis is often constrained by limited data, which may be uncertain, contradictory, missing, or over‐aggregated. This article proposes a Bayesian approach, in which uncertain knowledge about material flows is described by probability distributions. If little data is initially available, the model predictions will be rather vague. As new data is acquired, it is systematically incorporated to reduce the level of uncertainty. After reviewing previous approaches to uncertainty in MFA, the Bayesian approach is introduced, and a general recipe for its application to material flow analysis is developed. This is applied to map the global production of steel using Markov Chain Monte Carlo simulations. As well as aiding the analyst, who can get started in the face of incomplete data, this incremental approach to MFA also supports efforts to improve communication of results by transparently accounting for uncertainty throughout.