LCA cartography using network analysis
Main Presenter: Artur Donaldson
In this talk you will learn about applying network analysis to life cycle assessment databases, and create a map of the processes that are most integral, which influence results most strongly, and note where data can be further developed.
This talk is of interest for LCA data developers and practitioners looking to improve the accuracy and reliability of LCA results, find weak spots, or visualize the structure of databases (Artur Donaldson, 2020). Techniques from network science may also help data creators to prioritize which data to develop.
Network science or network analysis is the analysis of a system of interconnected components. The connections can be almost anything: lines between stations on a metro system, friendships between users on social media, or trade routes exchanging materials between economic actors. Network science is used regularly in correspondingly diverse fields: from neurology to logistics, epidemiology to ensuring phone network robustness.
Network analysis has been applied previously to co-authorship networks within the LCA community (de Souza and Barbastefano, 2011), which explored the structure of the community and graphed sub-communities. A disaggregated life cycle database is naturally viewed as a network where each process is a vertex, connected by product flows. Indeed, some LCA software allow you to see the network structure for a given LCA dataset at calculation.
The significance of particular datasets within a LCA model or database can be quantified by analyzing the network structure, using measures such as the betweenness centrality, or weighted distances between nodes on the graph. For instance, a vertex with a high centrality has many upstream processes that rely on it, or takes part in circular cycles of product flows.
Moreover, including attributes of vertices, such as the data quality score, or references to literature can give further insights into data gaps and enriching linkages to other fields of research beyond LCA. For instance, one can highlight nodes which have high uncertainty, but which are relied upon by a significant number of other datasets. Network analysis can give a bird-eye view of complex systems, just as cartography gives a bird’s eye view of a city. This talk is therefore a form of cartography for LCA.