Comparação entre os estoques de carbono total do solo medido in situ e estimado por sensoriamento remoto aéreo, orbital e próximo sob lavouras anuais

• Autor(a) principal: Faria, Ohana Cristina Oliveira
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFMT
• Texto Completo: http://ri.ufmt.br/handle/1/5347
• topic: Espectroscopia do solo; Aprendizado de máquina; Carbono orgânico do solo; Sistema remoto; VNIR_SWIR; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA; Soil spectroscopy; Machine learning; Soil organic carbon; Remot system; VNIR_SWIR

Agriculture plays an important role in global food security, although it is an activity that contributes to the increase in greenhouse gas emissions, it can also be a CO2 sink. Therefore, farming pratices and tillage need to be made more sustainable to minimize the emission of these gasses. In addition, it is important to use scalable, flexibleand reliable methods that accurately and cost-effectively measure carbon (C) stock. In this context, remote sensing stands out, a technique that is currently very widely used and of great importance for agricultural soils and environmental studies in the tropics. The objective of this study was to evaluate the C stock in three croplands (Fazendas Capuaba and Palminha in Lucas do Rio Verde MT and Fazenda Aterrado in Diamantino MT), under different soil management systems. They also aimed, to investigate the feasibility of estimating soil C levels using remote sensing techniques from proximal, aerial and orbital remote sensing techniques. For C analysis, they were collected in three different areas: I- deformed samples in the 0-30 cm layer, using dutch auger for C analysis in the elemental dry combustion analyzer (LECO®) and in the remote proximal soil sensing instrument Veris, II- undisturbed samples for soil bulk density determination were collected using koppec sampler and volumetric rings. Images were acquired using the Carcará II aerial remote sensing system, a MicaSense® brand coupled multispectral (RGB + NIR) camera, and Planet® and Sentinel 2 orbital remote sensing sensors. Of the samples designated for traditional and spectral analysis, 70% were wer for calibration and 30% were for validation. Of the three plots, two were selected for analysis in a 25x25 m field, with samples stratified into ten points (0 to 10, 10 to 20 and 20 to 30 cm). Of these ten points, five were randomly selected for sampling at depths of 30 to 50, 50 to 70, and 70 to 100 cm. The no-till system combined with crop rotation was able to increase soil carbon content, as much as native forests. Remote sensing has the potential to predict total soil carbon content. The near remote sensor was the most accurate of the sensors tested.