Long term discharge forecast for the Amazon Basin using artificial neural network

The observed annual variability in precipitation and water availability in the Amazonia, located in northeastern South America, has been shown to be influenced sea surface temperature (SST). However, the links between the large-scale SST patterns and local and regional runoff patterns are essentially complex and still not fully understood. The processes involved are believed to be highly nonlinear, spatially and temporally variable and not easily described by simple deterministic models. Artificial Neural Networks were used to develop models to forecast discharge one or two seasons in advance at 10 sites in Northeastern South America from Pacific and Atlantic sea surface temperature (SST) anomalies. Results were very encouraging. The correlation coefficient between observed and estimated discharges reached values as high as 0.96.