Remote sensing applied to pasture management

• Ano de defesa: 2022
• Autor(a) principal: Bretas, Igor Lima
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Zootecnia
• Palavras-chave em Português: Pastagens - Manejo - Sensoriamento remoto; Aprendizado do computador; Vegetação - Índices; Satélites artificiais em sensoriamento remoto; Pastagem e Forragicultura
• Link de acesso: https://locus.ufv.br//handle/123456789/29436; https://doi.org/10.47328/ufvbbt.2022.399

For this dissertation, two chapters were prepared based on the use of satellite images and machine learning techniques for automating pasture management. In the first chapter, we hypothesized that vegetation indexes (VIs) obtained through satellites providing moderate spatial resolution (Landsat-8 and Sentinel-2), combined with meteorological data, can accurately predict the aboveground biomass (AGB) of Brachiaria (syn. Urochloa) pastures in Brazil. We used AGB field data obtained from pastures between 2015 and 2019 in four distinct regions of Brazil to evaluate: (i) the relationship between three different VIs, normalized difference vegetation index (NDVI), enhanced vegetation index 2 (EVI2), and optimized soil adjusted vegetation index (OSAVI), and meteorological data with pasture aboveground fresh biomass (AFB), aboveground dry biomass (ADB), and dry matter concentration (DMC); and (ii) performance of simple linear regression (SLR), multiple linear regression (MLR) and random forest (RF) algorithms for the prediction of pasture AGB based on VIs obtained through satellite imagery combined with meteorological data. The results highlight a strong correlation (r) between VIs and AGB, particularly NDVI (r = 0.52 to 0.84). The MLR and RF algorithms demonstrated high potential to predict AFB (R 2 = 0.76 to 0.85) and DMC (R 2 = 0.78 to 0.85). We conclude that both MLR and RF algorithms improved the biomass prediction accuracy using satellite imagery combined with meteorological data to determine AFB and DMC, and can be used for Brachiaria (syn. Urochloa) AGB prediction. In the second chapter, we aimed to develop a model for automated height classification and evaluate the accuracy of indirect estimates of forage biomass in Mombaça guinea grass (Megathyrsus maximus cv. Mombaça) pastures. This model is based on the analysis of images obtained through the Sentinel-2 satellite using machine learning techniques to support decision-making regarding grazing management and stocking rate adjustment. We used different bands from the Sentinel-2 satellite that were obtained and processed entirely in the cloud. Three forage height classes were previously defined as class 0 (<45cm), class 1 (45–80cm) and class 2 (>80cm) according to management recommendations. The random forest algorithm was used to classify forage height and predict biomass by using height and biomass field data obtained from 54 paddocks in Brazil between 2016 and 2018 as reference data. The results demonstrate precision, recall, and accuracy values of up to 83, 90, and 83%, respectively, for paddock height classification and the potential to accurately predict AFB and DMC (R2=0.69 and 0.82, respectively). We conclude that the combined use of satellite imagery and machine learning techniques makes it possible to classify height and predict the biomass of Mombaça guinea grass (Megathyrsus maximus cv. Mombaça) accurately while supporting decision-making regarding grazing management and stocking adjustment. However, more studies must be carried out to improve the models proposed in both chapters and more efforts must be made to implement the tool under field conditions. Keywords: Machine learning. Remote sensing. Satellites. Vegetation indices.