Over 80% of Brazil's energy demand is supplied at present by hydropower. It is therefore essential for purposes of future planning to have good estimates of reservoir inflows, and the longer the lead time of streamflow forecasts, the more distant is the planning horizon for decisions concerning the costing, production and distribution of power. Whilst it is relatively easy to forecast river basin response to precipitation that has already fallen, forecasts for periods beyond this immediate response time must be based upon estimates of future precipitation. This paper describes the results of work which combines forecasts of future rainfall given by a model of global climate with a detailed, distributed model of river basin response in the Brazilian part of the River Uruguay basin, an area of intensive hydropower production. Comparison of predicted and observed rainfall over the period of model calibration showed that it was necessary to apply a statistical correction to eliminate a spatial trend in predicted rainfall. Data from a two-year period were set aside for model validation. Results show that the combined use of models of climate and river-basin response has been successful for the River Uruguay basin, in that the root mean square error (RMSE), between observed flows and flows predicted by the combined model, is substantially less than the RMSE between observed flows and the historic mean monthly flows.