Mapping the network of industrial processes from a Life Cycle Inventory database
Résumé
In spite of rising climatic and environmental concern, modeling of Input-Output theory,
Industrial Ecology, Life Cycle Assessment, and Material Flow Analysis are conspicuously
under-represented in complex networks research. To fill that gap we compare various network
representations of interactions occurring in the aforementioned domains, following [1], and
extending the scope (e.g. to chemical reactions networks, and the Von-Neuman’s growth
problem).
First the different methods to construct such networks are summarized, starting from available
data. Then, a classical empirical study of network’s properties is conducted and shows a high
heterogeneity. More precisely, networks stemming from Life Cycle Inventory databases are
compared to networks built using mappings with actual industrial process operated by
economic actors, such as hybrid Life Cycle Analysis, based on Input-Output Analysis at
international, national, and even regional scales.
The benefits of various pre-processing steps such as monopartite projection are assessed and
higher-order formalism (e.g, hypergraphs) used for example in omics research are discussed.
Existing applications, for example community finding [2] are exposed, and limitations of
conducting a network-oriented analysis are debated (such as data availability, uncertainty in
databases, or lack of spatial information).
Special emphasis is drawn on criticisms raised over the interest of the network formalism
with respect to scale-dependence. The way this topic is dealt with by linear algebraic tools [3]
that are standard in those fields is analyzed in that light. The interest of mitigating this issue
using recent multi-scale representations [4] and aggregation techniques in complex networks
research is assessed.
Origine | Fichiers produits par l'(les) auteur(s) |
---|