Previsão da temperatura máxima de Lavas-MG : comparação da normal climatológica com a Teoria dos Valores Extremos

• Ano de defesa: 2022
• Autor(a) principal: Ferreira, Valdeline De Paula Mequelino
• Orientador(a): Avelar, Fabrício Goecking
• Banca de defesa: Reis, Carlos José Dos, Gomes, Davi Butturi
• 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: Inferência bayesiana; Distribuição GEV; Erro médio de predição; Raiz do erro quadrático médio; Critério de informação de deviance
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA
• Link de acesso: https://repositorio.unifal-mg.edu.br/handle/123456789/2240

Agriculture accounts for approximately 3% of Brazilian GNP and it is the fourth economic activity most important in the Lavras, Minas Gerais state. It is the economic activity that most depends on climatic conditions. For this reason, the knowledge of the extreme climatic events behavior are important for the production maximization and losses minimization. In the study of climate behavior, it is used the climatological normals. Among the climatological normals there is the maximum temperature. One question that it may be done is about the use of climatological normals to predict extreme events, like the maximum temperature. In the study of extreme eventes it is used the Extreme Value Theory, where the Generalized Extreme Value Distribution (GEV) is one of the most applied. The Maximum Likelihood Estimators are usually used on the estimation of GEV distribution parameters. Alternatively, the bayesian inference has been used for this purpose. This work aims to compare the predictions to the maximum temperature of Lavras-MG obtained by climatological normals and those obtained by GEV distribution fitted by bayesian methodology with different structures of informative e non-informative priors distributions. The maximum temperature data from Machado, Minas Gerais state, were used for elicitation of the informative prioris distributions. It was concluded that the GEV distribution fitted by bayesian inference presented betters predictions to maximum temperature of Lavras and, for the most os the months, the informative priors distributions performed better the non-informative ones.