Data strategy for environmental assessment of agricultural regions via LCA: case study of a French catchment

Various approaches have been carried out to extrapolate environmental assessments of farms to the regional level, some of them oversimplified and thus leading to high uncertainty. Key challenges include selection of a representative sample, construction of a farm/land use typology, the extrapolation strategy and dealing with data limitations. This work proposes a method for addressing these issues by means of statistically supported approaches. We applied a novel approach combining a sampling strategy, estimation of farm-level environmental impacts via life cycle assessment (LCA), a farm typology based on principal component analysis, a statistical method for extending the farm sample given data constraints and finally linear extrapolation based on regional production and land use, taking into account the regional import-export balance. The approach was applied to a French case study, the Lieue de GrSve catchment in the dairy-intensive Brittany region. A decision flowchart was developed to generalise the approach for similar applications dealing with farm and LCA data constraints. Additionally, innovative farm practices were modelled and their impacts propagated to the regional level. The typology developed identified "dairy", "beef", "dairy + beef" and "swine" farms as the dominant farm types in the region. While swine farms had the highest mean impacts per hectare, dairy and dairy + beef farms had impacts two to five times as high as those of beef and swine farms, when extrapolated to the entire catchment. Multiple linear regressions based on an extended farm and LCA dataset were used to predict environmental impacts of dairy farms lacking LCA results, thus increasing their sample size before extrapolation. The inclusion of farm and LCA data from a neighbouring region did not contribute to the accuracy of predicted impacts, as determined by comparing them to those of the farm closest to the dairy cluster's centre, but rather produced significantly larger coefficients of variation. Results of tests of including two extra-regional farm and LCA datasets helped determine decision rules for the decision flowchart. Modelling of innovative agricultural practices yielded regional impacts consistent with previous estimates. This approach provides a generalisable approach for farm typologies, data handling and regional extrapolation of farm-level LCAs, applicable to estimate environmental impacts of any agricultural area if requirements of a representative farm sample are met. We demonstrate the utility of the method for estimating effects of innovative agricultural practices on a region's impacts by modelling practices on virtual farms and extrapolating their results.