Inventory of field water flows for agri-food LCA: critical review and recommendations of modelling options

Purpose: In a context of flourishing eco-labelling programs and environment policy for food products, LCA applied to agricultural systems faces the challenges of being operational, accurate and exhaustive. Indeed, LCA has to include all major environmental impacts such as water deprivation. This is particularly challenging for the young water use impact assessment in LCA, with many methods recently developed. Water inventory databases (e.g. WaterStat) and agri-food LCA databases (e.g. World Food LCA Database) contain default water elementary flows for average crop and animal products. These databases should support the assessment of water use impacts, to some extent. However, these different databases are not adapted to compare specific agricultural practices with LCA. To allow all LCA practitioners fulfil their diverse agri-food LCA objectives, a proper review of available methods for field water flows inventory and recommendations are needed. Methods: We critically analysed the models on which water inventory and agri-food LCA databases rely. Then, we explored alternative modelling approaches for an inventory of field water flows in LCA aiming at discriminating cropping system practices. We finally made recommendations to help practitioners identify the most appropriate method for field water flows according to their LCA study objectives and constraints. Results and discussion: Water inventory and agri-food LCA databases provide estimates of theoretical water consumed by a crop and rely on data and methods presenting limitations, making them suitable only for background agricultural LCAs. In addition, databases do not support the application of methods assessing the impacts of both consumptive and degradative water use. For the LCA-based Eco-design of cropping systems, the inventory of water flows should be based on a model simulating evapotranspiration, deep percolation and runoff accounting for crop specificities, pedo-climatic conditions and agricultural managements. In particular, the model should account for possible water, saline and nutrient stresses; assess evaporation and transpiration separately, estimate runoff and drainage according to the systems specificities. Yield should not be estimated with a model but a primary data. Recommended and default data sources are provided for each input parameters. Conclusion: The FAO Aquacrop model represents an optimal balance between accuracy, simplicity and robustness. However, this model is not applicable for perennial crops yet.