There is a growing interest in the characterization of mining residues, both for environmental assessments and critical raw materials recovery. The lack of sufficient in situ samples hampers an effective geostatistical modelling of material concentrations variability. This paper proposes a method to characterize the aluminum spatial variability in a mine residue from remote sensing data and imprecise information from daily dumping procedures. The method is proposed for the mapping of aluminum within a Greek bauxite residue, using Sentinel-2 imagery. The spatial correlation between metal concentrations and remote sensing indicators (e.g., spectral band ratios) is the premise for mapping aluminum varieties. The proposed method is based on Conditional Gaussian Co-Simulation, where Sentinel-2 images can be used as auxiliary variables. Simulation results are compared with the Co-kriging estimation method. To perform the Co-kriging estimation, the same conditions as simulation are used (same inputs, models, and neighborhoods). Simulation results quantified the metals variability in mining residues, presenting the metal concentration of piled materials in two time periods. For results validation and selecting the best map, fourteen validation samples were used. For the best representative maps of aluminum concentration, a correlation coefficient of about 0.7 between the validation data and obtained aluminum concentration map was obtained.
Read full abstract