Article · International Journal of Mining Reclamation and Environment · November 19, 2025
Hybrid machine learning and factor-based simulation for geostatistical modelling of residual resources in a tailings storage facility
Aidyn Tileugabylov, Nasser Madani, Mohammad Maleki, Annika Parviainen
Abstract
Mining tailings storage facilities (TSFs) contain significant residual mineral resources due to past processing limitations. As reprocessing becomes economically viable, accurately modelling spatial variability is crucial. Traditional cosimulation struggles with computational inefficiency and limitations of the Linear Model of Coregionalization. This study integrates machine learning-based imputation with Minimum/Maximum Autocorrelation Factor and turning bands simulation (MAF-TBSIM) to address uneven sampling. Using an Au-Cu TSF case study, Gaussian Process Regression yielded superior imputation performance. Compared to traditional methods, MAF-TBSIM offers improved spatial representation and efficiency, supporting favourable resource estimates and highlighting the reprocessing potential of TSFs.
More by Nasser Madani
- Conditional Cosimulation of Multivariate Gaussian Random Fields with Unknown but Related MeansMathematical Geosciences · 2026
- Development of Application-Focused Fuzzy Membership Functions for Geomechanical Parameters of CO2 Storage SystemsEnergies · 2026
- Geochemically constrained hierarchical co-simulation for modeling Cu–Au associations in mine tailingsApplied Geochemistry · 2026
- Improved assessment of in-situ and recoverable resources in brine deposits of the Lithium Triangle: A geostatistical perspectiveOre Geology Reviews · 2026