Partial ordering of life cycle inventory databases

Life cycle inventory (LCI) databases provide information of fundamental importance to the mainstream practice of life cycle assessment (LCA). Development and management of LCI data resources is a tremendous task. This paper seeks to develop design principles for organizing efforts around distributed database management by considering the structure of inventory networks. A number of LCI databases in wide use are placed into a partial ordering using graph theoretic techniques. First, strongly connected components, which indicate cyclic dependencies, are identified in each network. Then, those components are collapsed to single nodes, rendering the graphs acyclic. The acyclic graphs are then ordered topologically. Large databases were found to contain a single large strongly connected component, designated the background, that satisfied dependencies of other processes in the database. Processes with a lower position in the ordering than the background, designated the foreground, depended on inputs from the background in order to operate, but no background processes required inputs from any foreground processes. Processes higher in the order than the background, designated the downstream, satisfied dependencies in the background but did not themselves require the background. Databases sharing a common set of product flows could be compared on the basis of their foreground, background, and downstream segments. Making a distinction between an acyclic foreground and a strongly connected background has computational utility because the foreground and background databases can be managed independently. Describing LCA studies on the basis of their dependence on background product flows would allow them to be easily moved among distinct databases that provide the same products and would also facilitate critical review. Background databases themselves can be broken up by scope into mutually dependent systems that can be independently maintained. LCI database maintainers should consider maintaining the foreground and background components of their databases separately because of the differing implications of modeling decisions made in each of the two cases. A framework for database-independent enumeration of intermediate flows would advance distributed data management.