Estudo de indicadores climáticos para a previsão de longo termo de vazões na bacia do alto São Francisco

Abstract

This research had three objectives: firstly, to verify a possible influence of Southern Oscillation (SO) on the precipitation and streamflow patterns over the 51,000 km² upper São Francisco river basin, located in the tropical climatic zone of southeastern Brazil; secondly, to formulate long-range probabilistic forecasting models using anomalies of sea surface temperature in different regions of Pacific, Atlantic and Indic Oceans as predictors variables; and finally, to modify the input data on the Extend Streamflow Prediction method. The results indicate that the Southern Oscillation does not influence 6-month (Oct-Mar) precipitationvolumes. However, it seems that the Southern Oscillation has an effect on the time distribution of 6-month (Oct-Mar) rainfall. During the cold phase in OND period the cumulative rainfall depths are larger than those of other phases, while the opposite occurs during the warm phase. In JFM period the cumulative rainfall depths during the warm phaseare larger than those of the cold phase. These results may explain the tendency of low streamflows during the cold phase in the JFM period. The Southern Oscillation seems that influence 6-month (Apr-Sep) precipitation volumes, mainly in the AMJ period, when there isa tendency of higher rainfall during the warm phase. The same behavior is observed in the streamflow data. In the long-range probabilistic forecasting model for 6-month (Oct-Mar) precipitation, the anomalies of sea surface temperature of Niño1+2, Niño3.4 e ZCIT regionswere used as predictors, and in the case of the trimester steamflows there was a predominance of the anomalies of the regions of Pacific and Indic Oceans, and the Southern Oscillation Index (SOI). Considering the results of Half- Brier score, the forecasts with those models presented advantages when compared to the climatology forecast. The modification of the input data on the Extend Streamflow Prediction method was the utilization of mixed distribution 6-month (Oct-Mar) precipitation volumes, that is the sum of the distributions associated with the below normal, normal and above normal rainfall, adopting the results of long-range probabilistic forecasting model for 6-month (Oct-Mar) precipitation as the weight and applying the time distribution of precipitation associated with the South Oscillation phases. Considering the results of Half- Brier score, this methodology gave goods results when compared to the climatology forecast.