Modelagem da produtividade de plantios de café utilizando parâmetros, bioclimáticos, geomorfométricos e dados espectrais

• Autor(a) principal: Rocha, Thainá Guimarães
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/55101
• topic: Modelo de regressão linear; Campo das Vertentes; Linear regression model; Recursos Florestais e Engenharia Florestal

In Brazil, the largest coffee producer in the world, it is estimated that the area destined for coffee planting is 2.16 million hectares. The productivity of these areas is capable of supplying not only national demand, but also international markets. Half of the Brazilian production is located in Minas Gerais, an activity that contributes to the technological, cultural and economic development of the state. Finding new ways to determine and characterize factors that influence and help to estimate the productivity of coffee plantations is a necessity that increasingly allows for the growth of the culture. Knowing that all vegetation is influenced by the environment in which it is inserted, and in view of the amount of easily accessible data available today, this work aims to: (1) determine the correlation of bioclimatic, spectral and terrain variables with the productivity; (2) from the regression method, obtain a productivity model that can represent the productivity of the region; (3) evaluate and understand the variables presented in the model. Bioclimatic data from meteorological stations in the region were used to obtain the climatic variables. Spectral data derived from the Sentinel – 2 satellite, with 10 meters of spatial resolution and geomorphometric variables derived from the Digital Elevation Model SRTM, with a spatial resolution of 30m, were also used. For a pre-selection of variables, Pearson's correlation was used. With the pre-selected variables, a global linear regression model was fitted. Based on the Bayesian information criterion, the 5 best linear models were selected, the final model was the one that presented the best values for the coefficient of determination and the adjusted coefficient of determination (R2 = 0.71 and R2adjus = 0.54). The variables that composed the final model were: NDWIs; NDVIs; B3u; B2u; Profile Curvature; and Valley Depth. The results found in our study, despite being satisfactory, demonstrate the need to continue studies to better understand the relationship between coffee and variables that may influence the productivity of plantations.