Models simulating nutrient transport at the catchment scale are frequently used for source apportionment and thereby for finding cost-efficient management strategies for water quality improvements. One typical modelling approach at the catchment scale is the use of leaching coefficients (mass per unit flow of water) to compute the nutrient input based on land-use information. In this study two different such model approaches, the lumped Fyrismodel and the distributed HBV-N-D model, were compared based on simulations for the River Fyris catchment in central Sweden. A major difference between the models were different assumptions of specific runoff variations between different land-use classes. These differences had a considerable effect on the computed source apportionment. The higher specific runoff from agricultural areas in the HBV-N-D model compared to the Fyrismodel resulted in a larger contribution of agricultural areas to the total nitrogen export. These results demonstrate the importance of the assumptions of the spatial variation of specific runoff on source apportionment HBV-N-D model estimations.