Comportamento espectro-temporal da soja utilizando sensores orbitais e não orbital e correlação dos índices de vegetação com a produtividade

Abstract

The cultivation of soy has an important role in the economic development of many regions of Brazil and in the world. In this sense, precision agriculture associated with remote sensing tools are excellent methods to assist in the management and improvement of productive efficiency in soybean crops. In this context, this work aimed to evaluate the spectral-temporal behavior of soybeans through vegetation indices at different phenological stages, using orbital sensors and a non-orbital sensor coupled in Unmanned Aerial Vehicles (UAV). The study was carried out in a commercial farming area,located in the municipality of Pato Branco -PR, where, during its production cycle, image collections from the Landsat 8 and Sentinel-2 satellites were performed to extract the Normalized Difference Vegetation Index (NDVI) and, five collections with RGB sensor, with a resolution of 4.0 cm/px, coupled to the UAV, to obtain the Visible Atmospherically Resistant Index (VARI). Imageprocessing and treatment were performed using the Agisoft Metashape Professional 1.5®software and ArcGIS 10.5®, in which the VARI index of the RGB sensor and the NDVI of the Satellites (Landsat 8 and Sentinel-2) were generated. A 50 x 50 m sampling grid,with 59 points, was established, in which collections of soil chemical attributes and soybean yield components were made to understand the spatial distribution of these variables. Finally, Pearson's linear correlation for soybean productivity and vegetation indexes was performed. The VARI index extracted from the RGB camera and the NDVI generated from the orbital sensors were efficient in monitoring the phenological stages of soybeans, being influenced mainly by the different concentrations of chlorophyll and biomass. Most soil chemical parameters were spatially dependent, with the exception of P which demonstrated a pure nugget effect. The area under study showed good chemical conditions, especially in the eastern portion, where the highest levels of MO, K, Ca, pH, V% and lowest Al3+content occurred. All soybean yield components (number of pods, number of grains, number of seeds, weight of a thousand grains and productivity) expressed moderate spatial dependence. The highest values of number of pods, number of grains, weight of a thousand grains and productivity occurred in the eastern part of the property, in which the best fertility conditions were observed. VARI showed a better correlation with soybean productivity, when purchased from NDVI generated from Sentinel-2 and Landsat 8. This result demonstrates that the RGB sensors associated with UAVs can be used efficiently in agricultural monitoring, providing greater productive efficiency.