The frequently observed discrepancy between estimations of N 2 O emissions at regional or global scale based either on field data or inventories (bottom-up) or on direct atmospheric observations (top-down) suggests that riparian areas and river surfaces play a significant role as hot spots of emission. We developed a modeling procedure to assess N 2 O emissions occurring during the transfer of water masses from the subroot water pool of the watershed to the outlet of the river drainage network, including their passage through riparian wetlands. The model was applied to three river basins of increasing size located in the sedimentary geological area of the Paris basin (France) and validated by its capability to predict river N 2 O concentrations and fluxes across the river-atmosphere interface. At the scale of the Seine watershed, indirect emissions, i.e. emissions linked to agricultural practices but occurring elsewhere than directly at the field plot, are estimated to represent approximately 20% of the direct emissions from the watershed soils, in good agreement with previous estimates based on empirical accounting approaches. Denitrification in riparian zones is responsible for the largest share of these indirect emissions. The model results are very sensitive to the value of the ratio of N 2 O versus (N 2 ? N 2 O), in the final products of denitrification in rivers and wetlands. By calibration on river N 2 O concentrations, a value of 0.015 ± 0.05 is proposed for this ratio, in agreement with recent studies. This represents the main uncertainty factor of the model. In basins with conditions prone to increasing the value of this ratio, higher proportions of indirect N 2 O emissions might possibly be observed.