Distributed models and a good knowledge of the catchment studied are required to assess mitigation measures for nitrogen (N) pollution. A set of alternative scenarios (change of crop management practices and different strategies of landscape management, especially different sizes and distribution of set-aside areas) were simulated with a fully distributed model in a small agricultural catchment. The results show that current practices are close to complying with current regulations, which results in a limited effect of the implementation of best crop management practices. The location of set-aside zones is more important than their size in decreasing nitrate fluxes in stream water. The most efficient location is the lower parts of hillslopes, combining the dilution effect due to the decrease of N input per unit of land and the interception of nitrate transferred by sub-surface flows. The main process responsible for the interception effect is probably uptake by grassland and retention in soils since the denitrification load tends to decrease proportionally to N input and, for the scenarios considered, is lower in the interception scenarios than in the corresponding dilution zones.