Sensoriamento remoto e inteligência artificial no monitoramento de capim-tanzânia
Nenhuma Miniatura disponível
Data
2025-09-24
Autores
Bonadiman, Paula Alberti
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal do Espírito Santo
Resumo
Remote sensing (RS), a tool of digital agriculture, allows obtaining data from a specific object without direct contact and in a non-destructive way, generating information that assists in management. Thus, the objective was to evaluate RS as a tool to monitor the development, nitrogen (N) use efficiency, and productivity prediction of Megathyrsus maximus (syn. Panicum maximum) cv. Tanzania. To this end, the study was conducted in the experimental area of UFES, Alegre, ES, in two areas: one managed with fertilization (50 kg.ha-1 of N) in each crop cycle, while the other was without fertilization. Eight surveys were conducted at georeferenced points, obtaining agronomic and spectral variables. The spectral behavior of the plants was analyzed using aerial and proximal remote sensors. The Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Red Edge Index (NDRE), spectral bands, and the nitrogen sufficiency index (NSI) were analyzed in association with productivity and climatic conditions, and were also used in multiple linear regression and Random Forest models to predict forage productivity. The results showed that the area managed with nitrogen fertilizer presented higher average productivity, apparent leaf chlorophyll indices, and vegetation índices compared to the area without nutritional management, and the strongest correlations were obtained between productivity and vegetation indices. The highest productivity values were observed in locations where reflectance was higher in the near-infrared and lower in the visible bands. The ISN demonstrated potential as a fertilizer recommendation tool, with thresholds <0.95 for ISN-NDVI and ISN-NDVI Proximal, and <0.90 for ISN NDRE, especially in the months of September to March, a period in which agronomic efficiency and the partial productivity factor were also higher. In predicting productivity through Multiple Linear Regression and Random Forest, it was possible to generate models from input data obtained only through aerial remote sensing, highlighting NDRE, NDVI, and NIR as the most important variables in the prediction, in addition to predictive maps that represented the spatial pattern of actual productivity. Thus, this work contributed to the consolidation of remote sensing as a tool for management and decision-making, since, combined with data analysis and machine learning, it made it possible to monitor, identify spectral response patterns, and predict Tanzania grass productivity in a non-destructive and precise way, contributing to efficient and sustainable grazing systems
Descrição
Palavras-chave
Agricultura digital , Forragem , Sensor , Produtividade , Nitrogênio , Digital agriculture , Forage , Sensor , Productivity , Nitrogen