Set-aside comprising rotational (Rot_fal) or continuous green fallow (Cont_fal) and recultivated in autumn or spring was assessed using models at different spatial scales (field, catchment). Nitrogen (N) dynamics and nitrate leaching in sandy arable soils were simulated using digitized maps of soil properties, land use distribution and crop management for a 6-yr period. The rotational and Cont_fal scenarios took 17 to 22% of the arable land out of production, relative to a nonfallow (Non_fal) scenario, with the removals randomly distributed over the catchment. The scenarios provided information on the relative importance of set-aside for hydrology, nitrogen loss and conservation. The sample simulations for the Non_fal crop rotation at the field scale agreed with measurements for comparable sites. The Cont_fal decreased drainage more than the rot_fal (32 vs. 14%) relative to the nonfallow (Non_fal) scenario. At the field scale, Cont_fal met the EC-standard for nitrate in drinking water giving a concentration of 48 mg l - 1 , but under Rot_fal, nitrate concentration was very similar to Non_nfal conditions ( 130 mg l - 1 ) because of simultaneous drainage reduction. At the catchment scale, the mean of nitrate concentration for the Cont_fal sites was only half of that for the Rot_fal and spring tillage reduced nitrate concentrations by another 50% relative to autumn tillage. For the total catchment, however, the contribution of Rot_fal and Cont_fal to the total nitrate load were similar because more area was involved in the Rot_fal treatment. N accumulation and mineralization after Cont_fal exceeded the N demand of subsequent crops. Con_fal may therefore lead to groundwater pollution because mineralizable N was positively related to nitrate leaching at the catchment scale. The sensitivity of set-aside parameters for the simulation results was analyzed.