A climatic model for predicting seasonal rainfall

This study presents a methodology based on a model proposed by Silva for predicting the seasonal rainfall in the State of Pernambuco, Brazil. Seventy-two rain-gauge stations were employed, all of them distributed in three mesoregions in Pernambuco State. A rainy season (RS) with different subdivisions was established for each mesoregion. The Zi proportions – the ratio between the cumulative rainfall of the first RS period (Xi) and the rain that fell during the whole RS - were made to fit the Beta probabilistic model used for calculating the first and fourth quintis and the probability of rainfall above the climatologic average for the RS second period (Yi). The performance of the prognostic model for the period 1996-2000 was evaluated. In the period 1996 to 2000, with rainfall above Yi average, the observed error was less than 20 %, except for the RS8 in 2000. In 1998 - a very dry year - the observed error was over 20 %; this for the prognostic of the accumulated minimal rainfall. The methodology adopted proved very accurate for forecasting droughts in Pernambuco State.