Uma aplicação de conjuntos difusos na otimização do prognóstico de consenso sazonal de chuva no Nordeste do Brasil

Abstract

This study presents the application of fuzzy sets theory as a tool for sazonal rain forecast of Semi-Arid Northeast Region of Brazil for the period of 1985-1996. Thermodynamic parameters in Tropical Atlântic and Pacific Oceans were used as input variables. The results have shown that qualitatively, in an annual scale, this technique was able to forecast at least one of the categories of the output variable (total February/May rain in the of Semi-Arid Northeast Region of Brazil). Quantitatively, the smallest errors have been observed for the years classified according to the output variable in the categories of Normal (N), Rainy (C), and Very Rainy (MC), with correlation coefficients ranging fron 0.8 to 0.85, depending on the defuzzification method used. This technique allows for the unification of all the climatic information used in the rain foreceast of Semi-Arid Northeast Region of Brazil and leads to a forecast grouped in more than one category, informing the one most likely to occur as a function of its membership value.