The objective of the work reported here was to reduce the uncertainty on the greenhouse gas balances of biofuels using agro-ecosystem modeling at a high resolution over the Ile-de-France region in Northern France. The emissions simulated during the feedstock production stage were input to a life-cycle assessment of candidate biofuel pathways: bioethanol from wheat, sugar-beet and miscanthus, and biodiesel from oilseed rape. Compared to the widely-used methodology based on fixed emission factors, ecosystem modeling lead to 55-70% lower estimates for N2O emissions, emphasizing the importance of regional factors. The lifecycle GHG emissions of first-generation biofuels were 50-70% lower than fossil-based equivalents, and 85% lower for cellulosic ethanol. When including indirect land-use change effects, GHG savings became marginal for biodiesel and wheat ethanol, but were positive due to direct effects for cellulosic ethanol.