A snowmelt model is used for the weekly forecast of daily discharges in the Kaunas reservoir, Lithuania. The results are used to feed a risk-based decision-making model developed by the first author for dam operation during floods. Physically based calibration of a degree-day model is carried out and coupled with flow routing using Nash’s instantaneous unit hydrograph theory. Temperature forecast is used as the driving variable. Due to the relative smoothness of snowmelt over time and the considerable basin size, the model provides acceptable results. Kalman filtering is then used to merge the estimates from the snowmelt model with those from an Arima flow model, resulting in better forecasting than that using each method alone. Uncertainty analysis of the snowmelt-model results is then carried out, showing considerable influence of the main parameter degree-day and of soil moisture conditions. Therefore these must be accurately estimated for forecasting purposes during flood events.