Re-engineering LCA for simplicity and flexibility
Weidema B P (2011)
Publication info
Presentation for the SETAC Europe Annual Meeting, Milano, 2011.05.15-19.
Introduction: Database design strategy
A design strategy for LCI databases was proposed by Weidema (2003), containing three elements: 1) Database completeness 2) Unlinked and unallocated unit processes 3) Markets as separate unit processes.
Database completeness is required to ensure that different kind of models can be applied without being hampered by lacking data availability. Data completeness of course also ensures that comparisons are not biased by differences in completeness of the compared systems.
Having datasets available as unlinked, unallocated unit processes is a requirement for allowing different models to apply different algorithms for linking the datasets. The difference in linking algorithms is what distinguishes attributional and consequential models. Linking algorithms differ in the extent to which they take into account constraints (thus excluding specific unit processes from the supply chain), and in the way they deal with by-products (by substitution or allocation; the latter with many different allocation keys). This also provides a clear distinction between the verifiable, unlinked, unallocated process models that can be checked directly against their real life counterparts, and the linked, mono-product systems that necessarily require the introduction of assumptions, either on how markets react to changes in demand and supply (the economic assumptions of consequential models) or on what constitutes a fair attribution of the exchanges of unit processes to their products (the normative assumptions of attributional models).
Modelling markets as separate unit processes provides a simple way of combining the same (non-market) unit processes in many different ways depending on the system model applied, without changing the flows in each of the processes supplying and being supplied by the market. Furthermore, this makes it possible to document different market conditions using the same data format as for all other unit processes.
While Weidema (2003) reported two examples of databases that had applied this design strategy, the full potential of the strategy was not realised at the time, partly because the market activities were not systematically implemented as separate unit processes, partly because of technical limitations in the LCI data formats and LCA softwares available (e.g. inability to assign more than one property to a flow, inability to handle negative product flows, lack of machine-interpretable technological and geographical identifiers to delimit markets and locations of unit processes).
With the implementation of the ecoSpold 2 data format in the context of the ecoinvent database version 3, these identified limitations have been removed.