Modelling spatially explicit impacts from phosphorus emissions in agriculture

Excess phosphorus from fertilizer application and mobilised soil phosphorus from erosion are partially lost to the aquatic environment where they might cause eutrophication. Phosphorus emissions vary spatially and it is the goal of this study to broaden the scope of the existing inventory to the global scale and to increase the spatial resolution by accounting for relevant environmental processes. Phosphorus emissions were estimated globally at a resolution of 5 arc-minutes for 169 crops. Two models were coupled for that purpose. First, the Universal Soil Loss Equation (USLE) model was used to determine soil erosion which is the dominant process inducing phosphorus emissions. Second, the Swiss Agricultural Life Cycle Analysis (SALCA) model was applied to estimate the phosphorus emissions from four different processes with erosion being one of them. The emissions as inventory were compared to the ecoinvent database and subsequently translated into environmental impacts on biodiversity via characterisation factors. Additionally, sensitivity and contribution to variance analyses were carried out. Our results suggest that the data in the ecoinvent database, which is widely used for life cycle assessments, underestimate phosphorus emissions by up to an order of magnitude. Furthermore, the contribution to variance analysis highlighted the importance of regionalising both, inventory results and characterisation factors. Since the ecoinvent database provides a poor representation of global conditions, we highly recommend using regionalised estimates of phosphorus emissions provided in this study.