Modelagem agrometeorológica para a previsão de produtividade de cafeeiros na região sul do Estado de Minas Gerais

• Ano de defesa: 2015
• Autor(a) principal: Victorino, Euler Cipriani
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-Graduação em Recursos Hídricos; UFLA; brasil; Departamento de Engenharia
• Palavras-chave em Português: Agrometeorologia; Meteorology, Agricultural; Café; Coffee; Déficit hídrico; Water deficit; Seleção backward; Backward selection; Recursos hídricos
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/10499

The knowledge of effective crop forecasting techniques is of great importance for the coffee market, enabling better planning and making this activity more sustainable. Agrometeorological crop forecasting models can be developed based on the relations of climate changes, especially soil water availability, with coffee phenological phases, given that these relations directly impact productivity and the final quality of coffee. Therefore, this study aimed at developing a predictive model for coffee yield based on water availability, for the municipalities of Lavras and Varginha, in southern Minas Gerais, Brazil. The models were generated from the multiple linear regression of productivity loss (Ye/Yp) as a function of the previous year productivity (Ya/Yp) and water deficit in the different phenological phases, represented by relative evapotranspiration (ETR/ETP)i. The (ETR/ETP)i variables were calculated as quarterly and bimonthly averages, generating 12 different phenological sequences (7 bimonthly and 5 quarterly). During the parameterization, we obtained the water deficit response coefficients (Ky) and the previous year production coefficient (Ky0) for each sequence. Non-significant coefficients were then excluded by means of backward selection methodology, until the models presented only significant coefficients. During this process, in general, the models were highly sensitive to the rainy season (from November to April), and variables related to important periods, such as flowering, were not significant. At the end of parameterization, we concluded that the models have good potential for coffee crop forecasting. Yields of previous years should be considered. The phenological sequence with best performance was Sep./Oct, Nov./Dec., Jan./Feb., Sep. /Apr.