A nowcasting system for generation of short-range precipitation forecasts has been developed at the Swedish Meteorological and Hydrological Institute (SMHI). The methodology consists of utilising a time-series of radar reflectivity composites for deriving an advection field, which will give a better representation of the motion of the precipitation pattern compared to a model wind field. The advection field is derived applying a 4-dimensional variational data assimilation technique. The resulting field is then used for a semi-Lagrangian advection of the latest available reflectivity field forward in time. During the forecast, the advected field is gradually replaced by a numerical weather prediction forecast in order to include the onset of convection and advection into the radar coverage area. In an idealised example with simulated observations the functionality of the method is demonstrated. For a case study of a full scale example the resulting precipitation forecast shows large improvements compared to the operational numerical weather prediction model used at SMHI, especially for forecasts up to three hours, where the largest influence from the radar advection occurs. In an objective validation of the structure, amplitude and location of modelled precipitation, where the forecasts are compared to radar observations, these findings are confirmed. The same validation of model runs over a longer time period also clearly indicates that the amplitude, structure and location of the precipitation patterns are significantly improved as compared to a short-range forecast from the operational forecast model used at SMHI. (C) 2010 Elsevier B.V. All rights reserved.