Previsibilidade da precipitação na região semiárida no Nordeste do Brasil, durante a estação chuvosa, em função do comportamento diário das chuvas na pré-estação.

Abstract

One of the most relevant features of the rainfall interannual variability over the Northeast Brazil (Nordeste) semi-arid region is its irregular spatial and temporal distribution, particularly in extremely dry or wet years. Pre-season rainfall totals over Nordeste have been revealed a powerful predictor, specially when it includes a set of additional meteorological features, such as the meridional component of the surface wind over the tropical Atlantic, and the sea surface temperature anomalies in the Equatorial Pacific. This work is aimed at researching the skill of pre-season rainfall as a predictor to Nordeste rainy season (FMAM). Among other results, this work suggests that the correlations between pre season total rainfall and rainfall during northern Nordeste rainy period (FMAM) are not stationary; they were lower during the first half of the time series used, and higher for the second half of it. Nordeste farmers have observed regularly the weather since generations. They formulated empirical predictions about the quality of the rainy season (i.e., whether dry, normal, or wet) based, among other meteorological events, on the thunder periodicity during the pre-season months (September to January). The most important motivation for this work was to verify statistically these empirical considerationsSixty-two (62) rainfall stations distributed over the states of Ceara, Rio Grande do Norte, Paraiba and Pemambuco for the period 1926 to 1985 were considered. Statistical correlation, linear and multiple discriminant analysis and composite analysis were used in this approach. Daily rainfall indexes relative to several pre-season intervals were correlated. Statistical correlation between pre-season August-January precipitation and February- May (FMAM) precipitation presented the most relevant results. In addition nearly 54% of the season rainfall for the 59 years analyzed were well-predicted by the multiple discriminant analysis model. Otherwise 26% of the season rainfall predictions failed for the analyzed period, 19% of them in years presenting anomalies of the oceanic and atmospheric fields.