The Rainy Seasons of the 1990s in Northeast Brazil: Real-Time Forecasts and Verification

Based on circulation diagnostics in the tropical Atlantic sector and the equatorial Pacific, empirical methods have been developed to forecast anomalies in the March–June rainy season of northeast Brazil from observations to the end of the preceding January. Techniques include stepwise multiple regression, neural networking, and linear discriminant analysis. The methods were developed from the dependent training period 1921–57, and their performance was validated on the independent record 1958–89, prior to real-time application. Real-time forecasts have been regularly issued during the 1990s. These forecasts were in close agreement with the observed rainfall, except for the extreme El Niño year of 1998. A possible cause of this failure is seen in the lack of comparably extreme Pacific warm events within the training period used for the development of the empirical methods.