Between 1997 and 2004, the German Federal Institute of Hydrology (BfG), in cooperation with the Dutch Rijkswaterstaat Waterdienst, set up and calibrated the HBV rainfall-runoff model for the river Rhine. The model performed well for its original purpose, but less well when it was incorporated in the forecasting system FEWS in 2005. The main reason for the deteriorating performance was that the precipitation, temperature and evaporation data available for real-time applications differed from the ones used for the calibration. Another problem was that the accuracy in the low flow simulations was considered inadequate for navigation forecasts. It was thus decided that the HBV model set-up for Rhine should be updated and expanded in its functionalities primarily for use in operational forecasting. The tasks given to SMHI were: · To evaluate the evaporation calculations in HBV and recommend the best one to be used in the forecasting application. · To recalibrate the model using operationally available input data and with the aim to adequately model the whole range of flows. · To activate the HBV routine for updating model state variables before a forecast (PT updating) A new precipitation and temperature data set was provided for the calibration. This data set is consistent with the data to be used in the forecasting application, but improved as compared to the first data set used in the FEWS-DE system. To improve low flow simulations, a new model option, the contributing area approach, was used. The model was recalibrated using an automatic routine. Some minor manual parameter adjustments were made in a few sub-catchments, mainly to correct for anthropogenic influences and backwater effects on discharge measurements. The calibration was done locally for some 95 sub catchments, and verified both locally and for the total river flow. The overall model performance after recalibration with the new input data was at least as good as for the original calibration. Low flow recession and variations were reproduced to a greater degree. An evaluation with the old parameters and the new input data showed that the new data set in itself was not enough for satisfactory model performance. The recalibration was necessary. PT updating was shown to improve the forecast accuracy both for low/intermediate flows and for high flows. The effect diminishes with forecast lead time, but still remains at least up to the fifth day.