Life Cycle Inventory Data: Foreground, Background, Datasets and Uncertainty
Foreground versus background
Foreground data are the processes you control and model specifically - your recipe, your plant's energy use, your transport legs. Background data come from databases covering generic upstream chains: ecoinvent, Sphera MLC, Agribalyse, USLCI, ELCD, and EXIOBASE for input-output screening.
Dataset selection and quality
Match a dataset on four dimensions: technology, geography, time, and system model. ecoinvent offers cut-off, APOS (allocation at the point of substitution) and consequential models; results differ, so name the model and its version. Pedigree matrices score reliability, completeness, temporal, geographical and technological correlation; these scores feed uncertainty estimates, typically as a lognormal distribution described by a geometric standard deviation.
A common convention: flows below 1% of mass and energy and 1% of environmental relevance may be omitted, with cumulative exclusions kept under about 5% and every exclusion recorded.
Worked example
You need low-voltage electricity for a plant in Spain for a 2023 production year. A 2015 EU-28 average dataset fails both the temporal and the geographical test. Use a Spanish low-voltage market dataset for the closest year, note the voltage transformation and whether the mix is market-based or location-based, then rerun with the EU average to see whether the study conclusion moves.
Traps
Using global datasets for country-specific processes; ignoring the dataset reference year; treating aggregated datasets as transparent; not recording dataset name, version and UUID; mixing market-based and location-based electricity inside one model; double counting co-products already allocated upstream.