Flood forecasting in the upper Uruguay River basin

Floods are common in the upper Uruguay River, and they may occur almost at any moment, because there are not defined rainfall seasonal patterns. Moreover, there is not an official model for flood forecasting in this basin. This study developed a methodology for 1-month flood forecasting in the upper region of the Uruguay River basin (a parts per thousand yen3000 m(3) s(-1)), from the headwaters to the Ita reservoir. The monthly version of the SMAP (Soil Moisture Accounting Procedure) model was presented and used to describe the transformation of rainfall into runoff, and the CFSv2 (Climate Forecast System version 2) model was used to provide rainfall forecasts. Twenty-five 1-month-lead rainfall forecasts were used to calculate 25 flow predictions for every month. Ensembles with different number of members were compared among them and with the official model currently used for 1-month flow forecasting in the upper Uruguay River. The best accuracy was achieved with the average of the first seven members, which showed a mean relative error of 10.8 % during the floods, while the official model presented 64.0 %, predicting remarkably lower flows. Furthermore, during the period assessed, the correlation between the natural flow and the first-seven-member ensemble was > 0.77, while with the official model was 0.34. Thus, coupling SMAP and CFS is a valid approach that can be useful to anticipate mitigating actions to decrease the effects of severe floods in the upper Uruguay River and, probably, in other Brazilian basins.