Uso da estatística multivariada para a análise de índices de vegetação para a cultura do milho (Zea mays L.)

• Autor(a) principal: Sampaio, Marco Ivan Rodrigues
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• Texto Completo: http://repositorio.ufsm.br/handle/1/19731
• topic: ACP; NDRE; Cluster; PCA; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA

Nowadays, there is a necessity for the use of tools in order to estimate yield potential during the development of corn crop. Thus, the aid of active optical sensors, or embedded in RPAS, in the generation of vegetation indices can provide significant information to the knowledge of the behavior and temporal relationship of these indices with yield parameters of agricultural crops. Normally, during the monitoring of the entire cycle of a crop, a large amount of data is generated, which is difficult to analyze and interpret because the relationships between variables are complex. The goal of this work was to use multivariate analysis techniques to better understand the relationship of corn yield under different nitrogen rates, using vegetation indices obtained from data from two distinct remote sensing platforms, one proximal and the other, the remotely piloted aircraft (RPA) at different phenological stages. The experiment was carried out by Carvalho (2019) in a crop area of the Federal University of Santa Maria with the corn cultivar. A randomized block design with five blocks and five treatments with nitrogen dose variations (N) after the culture emergence was used. Vegetation sensing was performed by two distinct platforms, sequoia sensor embedded in an RPA and Optrx sensor embedded in a bicycle, obtaining the NDVI, GNDVI, EVI2 and NDRE Vegetation Indices with the sequoia sensor embedded in the RPA, and NDVI and NDRE with the sequoia sensor embedded in the bicycle in the vegetative stages V5, V7, V9, V11 and V12. In addition to these variables, the number of plants, N levels in the plant, and yield data were obtained. In this work we used data from four blocks, containing 800 plots, but only data from 451 plots were used, due to the discarding of 349 plots, which did not vary from 14 to 18 plants at harvest and/or did not present readings for some variable, i.e., null value. The normality test was performed and the data did not present normality, the data was standardized in the software (Statistica 12). Data analysis was performed using multivariate statistics using Hierarchical Cluster Analysis (HCA), Factor Analysis (FA) and Principal Component Analysis (PCA) methods. It was concluded that two distinct groups were formed (Sequoia / ARP and Optrx), and there was a greater relationship between productivity data with IVs (NDRE, NDVI, GNDVI and EVI2) at phenological stage V9 and IV NDRE at phenological stage V12.