Addressing Allocation and Disparity in Methods of Life Cycle Inventory

Life cycle assessment (LCA) is an environmental management standard practiced by a wide variety of government, industrial and academic users as a tool for understanding the broader environmental consequences incurred due to provision of materials and services. Before these impacts can be classified and acted on accordingly, a catalog of resources consumed and emissions and wastes produced over a material's life cycle is made in the life cycle inventory (LCI) phase. The magnitudes of these life cycle flows are heavily influenced by decisions made through choice of arbitrary life cycle boundary and through assignment of impacts among co-products in one of the most controversial aspects of LCA, allocation.;One central contribution of this dissertation is a novel view of the allocation problem in process based LCI as the choice of solution to an underdetermined system of linear equations made under a set of side conditions. Rather than endorse any particular partitioning criterion, strategies for identification of estimable functions and guidelines for the acquisition of more data are outlined and demonstrated for a toy example and a previously published brick production case study (Marvuglia et al. 2010). When the inventory of interest is not estimable, then its final result is determined by allocation assumptions. Methods for recovering these side conditions are developed and could be used to incorporate available information in a more meaningful manner.;Fuzzy methods have become increasingly popular for error propagation in LCI. One previously published result (Heijungs and Tan 2010) asserted a monotonic property related to simultaneous substitution of lower and upper bounds of fuzzy entries in the technology matrix. In this work we state that stronger sufficient conditions are actually needed for the result to hold. Related to this argument, the a class of M-matrices have properties that may be exploited to find lower and upper bounds due to allocation via constrained optimization.;It is frequently argued that the combination of data from economic and process scales via any of three hybrid LCI overcomes weaknesses associated with an inventory prepared at one scale alone. One focus of this dissertation is the characterization of systematic discrepancies due to choice of hybrid inventory method. Conditions for the equivalence of two types of hybrid LCI are stated and heuristics for identification of extreme disparities between tiered hybrid LCI and more comprehensive methods are developed and demonstrated on a synthetic data set and a case study of coal-fired electricity production in the US in which tiered methods may understate emissions by as much as 8% compared to other hybrid inventories, a finding which has implications for the use of existing corrections for double counting.;Taking subsequent action on environmental impacts is inherently a multi-criteria decision, and trade-offs among impacts will be necessary. The tools presented in this work may provide a better understanding of the extent of trade-offs when limited data are available.