Accurate and timely hydrological forecasts are key to ensure water safety. These forecasts highly depend on the meteorological input. Current numerical weather prediction models do not have sufficiently high accuracy, spatial and temporal resolutions for adequate use on short lead times (up to 6 hours). Radar rainfall nowcasting, the statistical extrapolation of rainfall observations to the future, can be a way forward for these short lead times. This thesis has analysed the potential and possible improvements of radar rainfall nowcasting, and its application in operational (flash) flood forecasting systems, for different algorithms, regions and rainfall types on different scales. In addition, a blending approach and an alternative source for nowcasting have been introduced to advance the possibilities of nowcasting for (flash) flood forecasting.