Mapping of critical raw materials in bauxite mining residues using geostatistics and remote sensing

In remote sensing analysis, bands information and indices are used to map different regionalized variables, for different applications of earth and environmental sciences. Moreover, the classifications methods can be used to differentiate the areas, and then with kriging tools to estimate the target variable values and variances. Often, these analyses are enriched by the validation of the obtained estimation maps using values from in-situ samples. On the other hand, to get effective and reliable maps, there is the need of high amount of data. In this research, remote sensing studies (statistical studies, spectrum view and unsupervised classifications) applied to Copernicus Sentinel-2 images have been combined with advanced geostatistical approaches (Gaussian simulation using Turning Bands -TBs- algorithm) to map the distribution of one critical raw material (Vanadium element-V2O5). The approach has been applied to a Bauxite tailings case study in Greece. Simulation results have been obtained for the Vanadium grade variability maps in the Bauxite tailings for 1000 realizations using infield samples as direct and Sentinel-2 images as an auxiliary variable. To test the simulation results, the reproduced experimental variograms of the realizations are compared with the selected variogram model of the Vanadium concentration and they have shown a coherent convergence. Hence, despite the lack of band-ratio existence for Vanadium identification in remote sensing analysis and, on the other hand, the limited number of initial sampling of data for geostatistical analysis, the integration of both approaches has generated appropriate maps of Vanadium grade distribution, within the Bauxite tailings case study.