Forecasting River Uruguay flow using rainfall forecasts from a regional weather-prediction model

The use of quantitative rainfall forecasts as input to a rainfall-runoff model, thereby extending the lead-time of flow forecasts, is relatively new. This paper presents results from a study in which real-time river flow forecasts were calculated for the River Uruguay basin lying within southern Brazil, using a method based on observed rainfall, quantitative forecasts of rainfall given by a regional numerical weather-prediction model, and rainfall-runoff simulation by a distributed hydrological model. The performance of discharge forecasts was evaluated over a continuous 167-day period and from one selected flood event, using rainfall forecasts at three spatial resolutions. The performance of these forecasts was also compared with that of forecasts obtained (a) by assuming that no further rain would fall, and (b) by assuming that rainfall forecasts were equal to the rainfall actually recorded, this representing a surrogate for 'perfect' rainfall forecasts. The results show that for the basin considered, there is plenty of scope for improving usefulness of rainfall forecasts.