Modelo probabilístico de previsão climática: aplicações à pluviometria do Estado de Pernambuco.

Abstract

In this research the performance of a probabilistic model, proposed by Silva (1988) and applied to monthly rainfall of Pernambuco state, was analyzed. For this proposal were defined four rainy stations (EC), subdivided in two periods - initial (X) and final (Y), for three geographical mesorregioes of Pernambuco: two for the Interior, one for the Rural and one for the Zona da Mata/Litoral. The methodology proposed by the model consists of adjusting proportions Z i , ratio o f accumulated rainfall of the first period (Xi) and the total rainfall of the EC ( X i + Y i ) , to the distribution Beta. Based on the first and fourth quintis o f Zi the maximum and minimum values for the period Y of the EC were determined. The Kolmogorov-Smirnov test was used to evaluate the adjustment of the model for each set of values related to each EC, respectively. The validation o f the model was verified with data collected from 1994 to 2002 for each EC. In more than 80 % of the stations the model performed with success in the prognostics of the minimum precipitation (Ymin) and maximum (Ymax), when analyzed regionally for each EC. In extremely dry or rainy years there is a tendency for fails associated with the Ymin and Ymax forecasted values. Maps with the expected minimum precipitation at the level o f 80% were elaborated, since they provide strong contribution to the planning o f several activities, especially agricultural and hydrological. In obtaining the probabilities of rainfall above the historical average of Y was verified that when X values are very superior to the climatological value, there were great differences between observation and predicted values. It means that there is a need for establishing a limit for X, before proceeding to the calculation o f such probability.