Modelagem em séries temporais : aplicação em dados de precipitação na região do sertão de Pernambuco - Brasil

Abstract

One of the major concerns of climate scientists and meteorologists in recent years is climate change and its impacts on mankind. This work presents an analysis of time series of monthly precipitation data, using models of the class ARIMA (Integrated Autoregressive of Moving Averages). The objective was to model and predict the precipitation of three municipalities (Arcoverde, Ouricuri and Petrolina), located in the hinterland region of Pernambuco State, Brazil. SARIMA models were tested, that is, ARIMA models that consider the seasonal feature of the series. The selected models were those that obtained the lowest values the criterion of selection of models Medium Square Error (EQM). The selected models were ARMA (3.3) municipality of Arcaverde, SARIMA (3.2) (1.1) municipality of Ouricuri and SARMA (2,3) (1,1) municipality of Petrolina, which obtained a good adjusted the series of precipitation data studied. The selected and duly verified models were used to perform precipitation forecasts of the next months, the results obtained are in agreement with the fact that the series are all stationary, so that their mean and variance are constant over time.