Previsão estatística da pluviometria da estação chuvosa na costa este do Nordeste do Brasil.

Abstract

For the east coast of the northeastern Brazil region it were pre-established four rainny seasons between January and August. Monthly rainfall values of 34 locations of the region were used for determining the contribution of the first and second periods to each pre-established rainny season. The Beta probability model was used in the adjustment of the ratios between the first period rainfall and the total of the rain occurred in the rainny season, with the degree of adjustment been verified through the Kolmogorov-Smirnov statistical test. The methodology proposed by Sih/a (1985, 1988) was used for forecasting the rainfall second period of the pre-established rainny seasons, based upon the first period rainfall and at statistic parameters of the locations rainfall time series. The Beta probability distribution model showed to be statistically efficient for adjusting the North and central of the region rainfall time series and least efficient in the southern region at a 0.20 significance level. The methodology proposed by Sifva (1985, 1988) showed to be efficient for predicting the maximum and minimum rainfall values of the pre-established rainny season second period for the North and Central region and least efficient in the South, according to projections made for the last 10 years of each studied series which wasn't used in the determination of the statistical parameters. The analysis of the crack number in the prognostics that surmount the established for Sifva model (1985, 1988), particularly in the South of the region, showed evidences to be associated to the occurrences of El Nino and or with the establishment of the rainny season.