Sistema de alarme ótimo para o modelo Tarso com aplicação na ferrugem do café

• Ano de defesa: 2011
• Autor(a) principal: Gonçalves, Luciene Resende
• Orientador(a): Sáfadi, Thelma
• Banca de defesa: Nogueira, Denismar Alves, Morais, Augusto Ramalho de, Lima, Renato Ribeiro de, Cunha, Rodrigo Luz da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: UNIVERSIDADE FEDERAL DE LAVRAS
• Programa de Pós-Graduação: DEX - Programa de Pós-graduação
• País: BRASIL
• Palavras-chave em Português: Ferrugem; Inferência bayesiana; Limiar; Sistema de alarme; Variável climática; Rust; Bayesian inference; Threshold; Alarm sistems
• Área do conhecimento CNPq: CNPQ_NÃO_INFORMADO
• Link de acesso: https://repositorio.ufla.br/handle/1/4405

Coffee crop occupies a featured position in the Brazilian economy and the country is the largest coffee producers and exporters in the world. Also, Brazil is in the second place when it considered the consumption. However, the coffee culture is very susceptible to attack diseases such as rust, existing incidence records in most of crops grown. The damage can compromise 50% of production. Thus, researches involving the limiting factor for production are intense and frequent, indicating the environmental factors as responsible for the spread epidemics, which do not occur if they are not friendly and act simultaneously. Because of these facts, the purposes of this study were to extend TARSO model with two regimes and one independent variable for two independent variables to create a function for adjusting via Bayesian inference in the R program. Besides of, applying the optimal ALARM SYSTEMS methodology for TARSO processes of p order in the development more a forecasting mechanism to coffee rust. In TARSO analysis, simulation studies were made to test the performance of the proposed extension, as well as applications the actual data to rust rates, using an independent variable and two independent variables. The analysis with the independent variable temperature average resulted in a lag d=1 and the average temperature contribution in the rust formation to temperature values below the threshold value r = 22; 30C. In the analysis using two independent variables, average temperature and leaf wetness, the lag was also d=1 and the rust rate for temperatures below the threshold r = 21; 70C was answered by the average temperature. And for values above, the rust rate was answered by leaf wetness. The alarm methodology was exemplified for a TARSO of order p=1 with an independent variable being this model adjusted via Bayesian inference. This analysis consisted to provide if any level of disease incidence crossed upper a pre-established level. Simulation studies were made and the application results showed that the system behaves efficiently at points where the catastrophe has not occurred, not having the same behavior when it occurred. And also that the system provides higher probability in higher crossing levels.