Nitrous oxide (N2O) is a potent fertilizer-derived greenhouse gas which plays a predominant role in the life-cycle assessment of biofuels. The use of generic emissions factors to estimate its emissions in such context has been widely criticized because of its lack of accuracy, since N2O emissions are highly dependent on local soil and climate conditions, as well as crop management. This paper proposes an alternative method using an agroecosystem model (CERES-EGC) at interconnected geographical scales (from plot to regional). This method was applied to a case study comparing 1st and 2nd generation bioethanol made from sugar beet and Miscanthus feedstock, respectively, in the Picardy region (France), using a spatially-explicit approach and simple land-use options. This new method made it possible to capture the variability in direct N2O emissions in relation to pedoclimatic conditions. Biomass supply options, minimizing N2O emissions or minimizing the land area used by maximizing high yielding units, were used to select the most suitable arable fields on which to grow the feedstock. There were few differences between the two options in terms of area of land cultivated for both crops. However for sugar beet, the effect of minimizing total direct N2O emissions was more noticeable. Regional yield and N2O emission maps resulted in supply curves useful to optimize the placement of biorefinery facilities and feedstock production fields. This method should be further completed with the estimation of changes in soil carbon stocks to take into account land-use change effects and with landscape-scale simulation to consider indirect downstream emissions.