Improved understanding of temporal and spatial Phosphorus (P) discharge variations is needed for improved modelling and prioritisation of abatement strategies that take into account local conditions . This study is aimed at developing modelling of agricultural Phosphorus losses with improved spatial and temporal resolution, and to compare the accuracy of a detailed process-based model with a rainfall-runoff coefficient-based model. The process-based SWAT model (Soil and Water Assessment Tool) was implemented for five river basins in central Sweden, and results compared with the rainfall-runoff coefficient-based model WATSHMAN (Watershed Management System) for one of these river basins. Parameter settings and attribute values were adapted to Scandinavian soil conditions, crops and management practices. Model performance regarding flow dynamics was overall satisfactory. Comparable results were achieved at several scales. The modelled P load was of high accuracy for the days when monitoring data were available for validation, generally once a month. Modelled monthly P load did not fit as well with averaged monthly monitoring load values, mainly since monthly monitoring often partly or entirely misses the peak flows. The comparison of SWAT and WATSHMAN gave slightly better results for the process-based model (SWAT). Better spatial resolution for input data such as Soil-P content and agricultural management practices will be required to reach modelling results that enable identification of measures adapted to local conditions.