Predição digital do carbono orgânico do solo no bioma Caatinga auxiliado por sensoriamento remoto

• Autor(a) principal: Lourenço, Valéria Ramos
• Data de Publicação: 2017
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/31131
• topic: Landsat-8 TM; Carbono orgânico do solo; Árvores de regressão

The study and determination of carbon stocks in the soil as well as their spatial and temporal distribution is essential for the understanding and quantification of the soil storage potential and, from this, develop management plans for this Carbon to reduce environmental impacts. Methods involving remote sensing such as DMS (Digital Soil Mapping) enable rapid and low-cost effective analysis of large areas from small in loco sampling. In face of that, the objective of this study was to evaluate the viability of using orbital sensor images (Landsat-8 TM) associated to environmental variables through regression tree models for the mapping of soil organic carbon (SOC) distribution in layer (0-20 cm) in a watershed in the Caatinga biome. The experiment was carried out in the Bengué Representative Basin (BRB), located in the southwest of the state of Ceará, which has an area of 1000 km². As a validation area, it was selected the Experimental Basin of Aiuaba (BEA) inserted in BRB and with an area of 12 Km². The soil collection processes were carried out in March 2017. There were collected 48 samples in the BRB and 12 in the BEA, the samples were analyzed in the soil laboratory of EMBRAPA Tropical Agroindustry for the determination of SOC and clay contents. As predictive environmental variables, soil texture, classes of use and occupation, precipitation, topographic aspects, bands 1 to 6 of Landsat-8 TM Satellite and NDVI were selected. Three regression tree models were constructed using software R, the first tree consisting of all variables, the second only by climatic and topographic data, and the last only by data obtained through geoprocessing. The models after their construction were submitted to indices to evaluate the efficiency of their predictions. From the construction of the prediction models based on regression trees, the determination of the Total Organic Carbon content in the (0-0.20 m) layer can be carried out for a characteristic area of Caatinga vegetation. The association of environmental variables with Landsat-8 satellite images proved to be very promising as a tool for mapping and predicting TOC levels. The main variables that had influence in the prediction models were the reflectance data of the multispectral images, besides the data of precipitation. The prediction model constructed only with multispectral image reflectance data can be considered as promising as the complete model in predicting TOC contents for the soils of a catchment area of the Caatinga, showing itself as a promising tool for TOC management in this biome.