Espacialização da probabilidade de ocorrência de veranicos extremos em regiões de Minas Gerais: uma abordagem Bayesiana

• Ano de defesa: 2020
• Autor(a) principal: Reis, Claudiana Goulart Dos
• Orientador(a): Beijo, Luiz Alberto
• Banca de defesa: Liska, Gilberto Rodrigues, Nogueira, Denismar Alves
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Alfenas
• Programa de Pós-Graduação: Programa de Pós-Graduação em Estatística Aplicada e Biometria
• Departamento: Instituto de Ciências Exatas
• País: Brasil
• Palavras-chave em Português: Distribuição (Teoria da probabilidade); Precipitação (Meteorologia)
• Área do conhecimento CNPq: PROBABILIDADE E ESTATISTICA::PROBABILIDADE
• Link de acesso: https://repositorio.unifal-mg.edu.br/handle/123456789/1755

The occurrence of extreme dry spells, that is, long periods of consecutive days without rain, tends to cause damage and losses to society in general. The climatic variability of a region has an important influence on the various socioeconomic activities. Therefore, the prediction of the occurrence of extreme dry spells has great importance for human activities, mainly for the agricultural sector, subsidizing agricultural planning to optimize the efficient use of water resources in cultivable areas, as well as for the proper management of plantations and pastures. Thus, the present work has as main objective to spatialize the probability of occurrence of maximum summer for the rainy and dry seasons in different regions of Minas Gerais using Bayesian inference, initially comparing different prior distribution structures for the adjustment of the Generalized Distribution of Extreme Values (GEV) to the annual maximum dry spells data sets. The historical series were constructed from the accumulated daily precipitation data of pluviometric stations in the study region, being obtained from the records of the Meteorological Database for Teaching and Research (BDMEP) of the National Institute of Meteorology (INMET). To obtain the marginals of the posterior distributions, the Markov chain Monte Carlo method was applied using software R and freeware OpenBugs. Subsequently, the most efficient informative prior structure was applied to calculate the probability of occurrence of extreme dry spells for different return periods in each season in the region to build maps from the spatialization of this variable. It was observed that the insertion of prior information from Belo Horizonte or Lavras made it possible to obtain better results in the analysis of maximum dry spells in most of the analyzed cities. The most significant values of probability of occurrence of maximum summer in the dry season were observed predominantly in the portion close to the limit between the regions Central Mineira and Metropolitan de Belo Horizonte. The highest probability values for the rainy season were concentrated in the regions Campo das Vertentes, Metropolitana de Belo Horizonte and Sul/Sudoeste de Minas. The areas with the greatest vulnerability for agricultural production due to the occurrence of dry spells over 60 and 70 days without a rainfall greater than 1 and 5 mm during the dry season, respectively, were estimated for the region close to Bom Despacho, Curvelo and Sete Lagoas.