Analysis of agricultural production with life cycle based methodologies is data demanding. To build comprehensive life cycle inventories, secondary datasets are commonly used when primary data are not available. However, different inventory data and modelling approaches are used to populate secondary datasets, leading to different results. The present study analyses the features of twelve secondary datasets to support datasets selection and proper interpretation of results. We assess twelve datasets for arable crop production in France, as modelled in three databases often used in the LCA field (Agri-footprint, ecoinvent and AGRIBALYSE). First, we compared system boundaries and general assumptions. Second, we focused on foreground systems comparing, inventory data, data sources and modelling approaches. Third, we performed a contribution analysis of impact assessment results to identify modelling choices that contribute most to differences in the results. Nine relevant elements were identified and assessed: definition of system boundaries and modelling of agricultural practices, characteristics of inventory data, agricultural operations, fertiliser application and fate, plant protection products application and fate, heavy metals inputs to the agricultural system and fate, irrigation assumptions, land use and transformation. The datasets differ greatly with respect to these elements. Hence, recommendations are drawn from the datasets comparison, supporting the selection of the datasets coherently with the goal and scope of a study and interpretation of results.
Keywords: Agriculture; Arable crops; Databases; Life Cycle Assessment; Life cycle inventory.