Stochastic modeling using Markov chain on the forecast standardized precipitation index

Abstract

Grain production and livestock are the basis of the state of Rio Grande do Sul/Brazil economy and favorable rainfall conditions accounts for about 20% of Brazilian production. However, the state has faced periods of drought, especially in the campaign area, decisively affecting agricultural production and hence the economy. The objective of this paper is to predict the Standardized Precipitation Index (SPI) on a monthly scale, from a series of rainfall from 1991 to 2012, for the city of Bagé/RS/Brazil, in order to extend it by over 22 years, using a Markov chain to simulate the occurrence of rainfall and the Gamma distribution for predicting the precipitated amount. The results showed that the model simulated two-state Markov sequences of dry and wet days, keeping the statistical characteristics of the series. There were underestimations of the number on events classified as extreme and severe drought, obtained by serial SPI simulated compared to observed ones.