Sensoriamento remoto orbital e não orbital no delineamento de zonas de manejo para agricultura de precisão

• Ano de defesa: 2021
• Autor(a) principal: Rosa, Helton Aparecido
• Orientador(a): Johann, Jerry Adriani
• Banca de defesa: Johann, Jerry Adriani , Andrade, Maurício Guy de , Maggi, Marcio Furlan , Souza, Carlos Henrique Wachholz de , Mercante, Erivelto
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual do Oeste do Paraná; Cascavel
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Agrícola
• Departamento: Centro de Ciências Exatas e Tecnológicas
• País: Brasil
• Palavras-chave em Português: Índices de vegetação; Fuzzy C-means; Agrupamento
• Palavras-chave em Inglês: Vegetation indices; Fuzzy C-means; Clustering
• Área do conhecimento CNPq: Sistemas Biológicos e Agroindustriais
• Link de acesso: http://tede.unioeste.br/handle/tede/5510

The general objective of the research was to outline Management Zones (MZs), which were made from average productivity data for 4 harvests (soybeans, corn, soybeans, wheat) and Vegetation Indices (VIs) data, obtained through orbital and aerial remote sensing, using the Fuzzy c-means algorithm. The monitoring was carried out throughout 4 agricultural harvests in two areas of a rural property located in the municipality of Toledo – Paraná. Sentinel-2 images (MSI sensor) were used, which were obtained through the Google Earth Engine (GEE) platform, as well as images achieved from a Remotely Piloted Aircraft System (RPA). With the productivity data for each harvest, descriptive statistical analyses were executed and, subsequently, correlation analyses among the VIs and the yields, using the Spearman (rs) correlation coefficient. The precision agriculture platform AgDataBox-MAP was used in order to outline the MZs. In it, statistical analyses, filtering of data, removal of influential points (outliers and inliers) and normalization were carried out. The created MZs were evaluated within the following metrics: smoothness index (SI), fuzzy performance index (FPI), modified partition entropy (MPE), improved cluster validation index (ICVI). The similarity of the maps was evaluated according to the Kappa statistics. For the VIs that used RPA images, the highest correlations over the harvests were found for the 2018/2019 soybean crop, with 94 DAS in area 2, and values of 0.72 (GLI and ExG), classified as a strong correlation, 0.53 for VARI and 0.47 for MPRI, as a moderate correlation. From the usage of indexes obtained through images from the Sentinel-2 Satellite, the 2018/2019 soybean crop, with 80 DAS in area 2, showed the highest correlations when it comes to productivity, therefore classified as moderate, 0.58 (NDVI and SAVI) and 0.60 (EVI2 and NDRE). For the remaining periods of time evaluated throughout the 4 harvests, the correlations were classified as: very weak or weak. The application of VIs generated by orbital and aerial remote sensing was proved to be an alternative for the creation of MZs, especially in conditions where there is no possibility of accessing data for soil attributes. In general, the VIs that also use the infrared wavelength presented better values of SI, FPI, MPE and ICVI, which implied that they were more efficient in the design of the MZs.