Detecção de doenças e estimativa da produtividade em plantações de palma de óleo na Amazônia brasileira.

Abstract

The potential of remote sensing in agricultural management in the Brazilian Amazon has been little explored despite being highly promising with the evolution of sensors and the arrival of drone platforms. Especially in the detection and spatial modeling of diseases, which in recent decades has had a strong impact on the productivity of regional cultures. Anticipating impacts and negative trends in oil palm production, supported by remote sensing and applicable methodologies, are among the objectives of this study at the doctoral level. In recent years, several factors such as pests, diseases and severe droughts have interfered in the productivity of oil palm in the region, generating the need to adopt new techniques for detecting and monitoring these problems. In the first article of this work, successful spectral analysis tests were carried out to detect diseases in the oil palm at Fazenda Companhia Palmares da Amazônia (CPA), belonging to the Agropalma S.A group, in the city of Acará-PA. The experiments were carried out in plots L30, L36, H25 and H27, with individual areas of 36 hectares of mature conventional type plantations with about 16 years of age, arranged in an equilateral triangle with 9 x 9 meter spacing. Five image bands of the Sentinel-2A system, MSI sensor, were used, referring to the visible and infrared spectra acquired on July 25, 2017. The reflectance images, spectral curves and vegetation indices were generated for the areas of the palm stands of Oil. The results allowed the identification of expressive minimum and maximum reflectance patterns of the studied plots, correlating them with the occurrences of diseases registered in the area. The results showed that the EVI index showed an excellent correlation with real occurrences of diseases. However, the NDVI and SAVI indices also showed good adjustments with the occurrence of diseases in the year 2017. The areas corresponding to plots L36 and H27 showed higher occurrences of diseases, based on the analysis by vegetation indices. Thus, it could be concluded that the enhancements in the reflectance images, NDVI, SAVI and EVI, are efficient in detecting diseases in the plots. The results allowed the identification of diagnostic anomalies of stresses in the plots, either due to a disease or other factor, which allows decision making in a timely manner, avoiding large-scale eradication in the extensive areas of commercial oil palm plantations in the region. The second article of this work referred to the calculation of palm oil productivity by means of orbital images, evaluating its correlation with actual productivity and with the infestation by Fatal Yellowing (FA). The productivity estimation by remote sensing can be used as an important support for forecasting harvests, supporting the traditional methods adopted when harvesting in the field, which in turn can be imprecise, time consuming and with high cost of execution. In this context, spectral enhancement through the calculation of vegetation indices has great potential to estimate the productivity of palm oil in the Amazon. For the calculation of productivity, the bands of the Rapideye sensor image for the year 2014 were used, covering an experimental area with eight conventional plots of Fazenda CRAI (Companhia Real Agroindustrial SA) in the municipality of Thailand-PA, state of Pará. spectra derived from the vegetation indices: NDVI, EVI, ARVI and RENDVI. For the productivity estimate, the statistical method of linear regression was used for the data for the year 2014, based on the correlation with the productivity in the field. Preliminarily, the EVI presented the best correlation with the productivity observed in the field, showing the potential to estimate productivity. The validation of the method was performed using the mean Quadratic-RMSE error, which considered the estimates obtained satisfactory, confirming the potential effectiveness of orbital remote sensing for research related to productive or crop prediction.