Modelagem de séries temporais aplicada à previsão da precipitação nas microregiões do estado da Paraiba

Abstract

The general objective of this work was to verify the rainfall data collected daily from January 1996 to December 2016 in Paraíba municipalities, in order to analyze the ability of different time series models to predict monthly precipitation for the Paraíba micro-regions. The work's methodology included the use of the following models: Auto-regressive Models (AR), Moving Average Models (MA), Auto-Regressive and Moving Average Models (ARMA), Integrated Auto-Regressive Moving Average Models (ARIMA) and Seasonal Integrated Mixed Models (SARIMA) in 11 different test-based structures to evaluate the ability of these models to predict monthly precipitation, using R software, version 4.0.5. The Box & Jenkins methodology used proved to be adequate for modeling climatological data. It is confirmed that the best method for adjusting the series and forecasting was described by the SARIMA model. Therefore, based on the interpretation of the results, it is possible to plan various economic activities for the micro-regions of the state of Paraíba, especially in the agricultural sector.