Current assessments of the environmental performance of biofuel chains are fraught by a large degree of variability and uncertainty arising from differences in life cycle inventory data for the feedstock production stage, notably crop yields and emissions of reactive nitrogen. This study sets out to improve the accuracy of these data by combining agro-ecosystem modelling and agronomic survey data, in the context of biodiesel production from winter oilseed rape in France. Crop yields and environmental emissions related to crop farming were simulated over the 2010–2030 time period in 8 administrative regions accounting for most of winter oilseed rape production, and the results were fed into a life cycle assessment (LCA) of biodiesel. The effects of changes in climate scenarios and the improvement of fertilization practices were also examined.Compared to the widely used methodology based on fixed emissions factors, ecosystem modelling led to 29–41% lower estimates for direct nitrous oxide (N2O) emissions, depending on climate scenarios, which emphasizes the importance of taking local factors into account. The life cycle GHG emissions of biodiesel from winter oilseed rape varied starkly across the studied regions, from 20.6 to 25.7 g CO2 eq MJ−1 and from 22.9 to 32.4 g CO2 eq MJ−1 depending on climate scenarios. Overall, biodiesel deployment would abate GHG emissions by 65–78% compared to fossil diesel, not accounting for land use changes effects. Improved fertilization practices mitigated the LCA impact indicators by 3–14%.