USE OF PROXIMAL SENSORS FOR THE ANALYSIS OF POTENTIALLY TOXIC ELEMENTS AND SOIL PROPERTIES IN IRON AND GOLD MINING AREAS IN THE EASTERN
AMAZON
pXRF; Nix Pro™; soil quality; machine learning; Carajás mineral province; Serra Pelada.
This thesis demonstrates the effective application of proximal sensors for analyzing soils in areas impacted by mining activities. The first chapter highlights the use of portable X-ray fluorescence (pXRF) technology to quickly and cost-effectively assess the total and extractable contents of potentially toxic elements (PTEs) in regions affected by artisanal gold mining in the eastern Amazon, Brazil. The
study compared pXRF with standard acid digestion methods, showing that pXRF is a viable, eco-friendly alternative for quantifying PTEs and predicting their extractable forms using robust machine learning
algorithms. The best model fits for total PTE contents were achieved for elements like Cu, Fe, Mn, and Pb, and the predictions for extractable Cu and Zn showed strong performance. The second chapter focuses on the rehabilitation of iron mining areas in the Carajás Mineral Province, where proximal sensors such as pXRF and Nix Pro™ were employed to characterize mining pit substrates. The research involved sampling and analyzing substrates to obtain elemental content and color parameters, and applying machine learning algorithms to
predict soil fertility and texture attributes. The results indicated that pXRF and Nix Pro™ can effectively characterize iron mining substrates, with the fusion of data from both sensors providing slightly improved predictive models for several soil attributes. The findings underscore the potential of these proximal sensors to facilitate efficient and sustainable soil analysis and rehabilitation practices in mining-affected regions.