Previsão Hidroclimática de Vazão de Curto Prazo na Bacia do Rio São Francisco

Abstract

The flow forecast based in precipitation forecast permits larger antecedence than other methods, but more sophisticated models and investments in measurements and transmission of data are needed. Only with the improvement of the atmospheric and hydrologic deterministic models, in the last few years, consistent opportunities were generated to amplify the forecasts lead times, by the integrations of these models. The main objective of this research is to improve the integration methodology between atmospheric and hydrologic models (hydroclimatic forecast), to short-term flow forecasts (up to 14 days) and long-term flow forecast (up to 6 months). To this, the MGB-IPH distributed hydrologic model was adjusted to the São Francisco River basin (639000 km2). Precipitation forecasts generated by atmospheric general circulation model (AGCM) and ETA regional model, from Centro de Previsão de Tempo e Estudos Climáticos – CPTEC, were used as input data for the hydrologic model. Forecasts were made for the entire basin, but the results were analyzed in details to Três Marias and Sobradinho dams. Precipitation forecasts of ETA model up to 10 days were used for the short-term flow forecasts. Weekly flow forecasts were produced up to 2 weeks for Três Marias and 4 weeks for Sobradinho, after 10 days of forecast the precipitation was considered equal to zero. The results show that forecasts are very good to both dams and the model is able to estimate events that stochastic models have great difficult to represent. The errors of short-term forecasts were modeled by stochastic and transfer functions models to improve the forecasted flow, with very satisfactory results to both dams. For the long-term flow forecasts, 6 months lead time climatic forecasts of AGCM and ETA models, both from CPTEC, were used. The flow forecasts were made for 6 months lead too and compared in monthly intervals, or larger, with the PREVIVAZM stochastic model from the Operador Nacional do Sistema Elétrico (ONS). The results to Três Marias show that the hydroclimatic model, with AGCM rainfall, does not have advantages in relation to PREVIVAZM. To Sobradinho the hydroclimatic model, with AGCM rainfall, presents little advantage, but it is significant only for 2 months lead time. Using precipitation generated by ETA model, the results were worse than PREVIVAZM in both dams and for all the lead times.