Towards a sampling protocol for the resource assessment of critical raw materials in tailings storage facilities

Resource estimates are crucial to assess the economic potential of tailings storage facilities (TSF) for re-mining and the extraction of critical raw materials. However, a lack of consensus exists on best practices in sampling for this purpose. This study aims to address this gap by assessing different sampling schemes for the resource classification of TSFs. To do so, one layer of a TSF was sampled with regular and nested grids of varying sizes and additional random holes. Systematic spatial trends in the data were fitted with low-order polynomial functions of the coordinates and the grayscale values of a historical aerial photograph of the same layer of the TSF. Variography was performed on the trend residuals, and exponential or Gaussian variogram models were fitted. Universal Sequential Gaussian Simulation was then used to produce 1000 realisations of the potential ground truth. The optimum sampling strategy was investigated by re-sampling these realisations with varying sample densities and configurations and using geostatistical modelling of the re-sampled data to assess uncertainties on the estimated metal tonnages. Robust estimates of metal tonnages can be achieved at relatively low sampling densities, particularly with regular grids. When historical image information is used, spatial variability is better reproduced, and a lower number of samples is required to reach a certain confidence level. Furthermore, an approach to approximate expected errors on grades/tonnages estimated with a given sampling scheme was developed to assess whether further sampling is required. Overall, the findings of this study have clear and transferrable implications for the best-practice sampling and modelling of TSFs and their critical raw materials resource.