Modelagem estocástica para a precipitação diária

Abstract

The stochastic daily precipitation modeling is the main objective of this dissertation. The occurrence of the process was modeled by a two-state (dry or rainy day) Markov chain and by the “wet-dry spell” approach. This second approach was considered appropriate, while the Markov chains could not describe the long droughts or the long wet spells. ln the rainy days, two assumptions were made: 1) the rainfall amounts are independents and 2) the rainfall amounts, in consecutive rainy days, are dependents and were generated by a first-order autoregressive model AR(1). The analysis of the generated and the historical data, indicated that only under the assumption of dependence, the annual series of maximum of daily precipitations, for specified durations in days, had the same probabilities distributions. This fact shows the importance of analizing extreme events and the need for some reflection on the serial correlation of the precipitation data.