Modelagem do potencial mineral de bauxita a partir de dados aerogeofísicos e de sensoriamento remoto na porção sudeste da Bahia, Brasil

• Ano de defesa: 2023
• Autor(a) principal: Santos, Rodrigo Soares Vieira dos
• Orientador(a): Franca-Rocha, Washington de Jesus Sant’anna da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Feira de Santana
• Programa de Pós-Graduação: Programa de Pós-Graduação em Modelagem em Ciências da Terra e do Ambiente
• Departamento: DEPARTAMENTO DE CIÊNCIAS EXATAS
• País: Brasil
• Palavras-chave em Português: Bauxita; Aerogeofísica; Sensoriamento Remoto; Aprendizado Supervisionado; Random Forest; Google Earth Engine
• Palavras-chave em Inglês: Bauxite; Airborne geophysics; Remote Sensing; Supervised Learning; Random Forest; Google Earth Engine
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA
• Link de acesso: http://tede2.uefs.br:8080/handle/tede/1967

The present work aimed to build a bauxite mineral potential model by integrating airborne geophysics and remote sensing data in an area that corresponds to the eastern portion of the São Francisco Craton in the southeast of Bahia state, Brazil. Recent studies have demonstrated a potential for bauxite and other supergene minerals in the region, linked to the presence of well-developed lateritic profiles resulting from weathering. To build the model, maps of K, eTh, and eU concentrations were generated to identify the geochemical signature of materials associated with regolith, as well as the Total Gradient, to provide information about the magnetic susceptibility of materials. To study the geomorphological context of the area, a digital elevation model (SRTM) and a slope map were generated. Landsat 8 scenes were also used to study the spatial distribution of surface materials based on their chemical composition, along with a lithological map to provide information about the source rock of the residual materials. The Random Forest supervised learning algorithm was applied to integrate the generated evidence maps, resulting in the bauxite mineral potential model for the region. The resulting model considerably reduced the study area and obtained satisfactory performance measures, achieving an F-score of 0.81 and a kappa coefficient of 0.83. The areas suggested by the model as targets reinforce the mineral potential of areas with known occurrences, as well as suggesting new targets for prospecting. The methodology was implemented on the Google Earth Engine cloud image processing platform using a JavaScript programming language integrated development environment (IDE). In addition to suggesting prospective targets, the results also indicated that the evidence maps with the most importance within the model generation were those from the SRTM, as well as from uranium and potassium.