International Journal of Mining and Geo-Engineering

International Journal of Mining and Geo-Engineering

Integrated geochemical modeling of gold in legacy tailings at La Cienega, Peru: comparing geostatistics, machine learning and copulas for recovery screening

Document Type : Research Paper

Authors
1 Mining Engineering School, Faculty of Engineering, Universidad Privada del Norte, Trujillo, Peru.
2 Department of Mining Engineering, Faculty of Engineering, National University of Trujillo, Trujillo, Peru.
3 Laboratory for Mine Design and Planning-DIPLAMIN, National University of Trujillo, Trujillo, Peru.
4 Research Group in Advanced Modeling in Mining (GAMM), National University of Trujillo, Trujillo, Peru.
10.22059/ijmge.2026.409498.595355
Abstract
Legacy gold tailings are both an environmental challenge and a potential source of valuable metals. We studied Tailings Deposit I at La Cienega (Peru) using an integrated comparative workflow that evaluates geochemistry, geostatistics, machine learning, and copula-informed models to screen areas with higher recovery potential. We analyzed 109 samples from six drillholes and one trench. Gold (Au) and related elements (Ag, Cu, Fe) were evaluated using ordinary kriging as an exploratory spatial baseline, four common regressors Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Regression (SVR), and Artificial Neural Networks (ANN), and Gaussian/Student’s-t copula-informed normal-score models. The average Au grade is 1.42 ppm, with strong correlations with Ag (ρ = 0.73), Fe (ρ = 0.64), and Cu (ρ = 0.56). Under leave-one-out validation, RF and XGB performed best (R = 0.72–0.71; RMSE = 0.25 ppm). ANN tended to overestimate the highest grades, while the Student’s-t dependence model gave conservative estimates at the upper end. The compared model outputs outline coherent high-grade domains. Under mean conditions, the contained gold is about 1.19 thousand ounces (average = 1.18 ppm), with close agreement among the strongest models. This data-driven, multi-model approach helps rank target zones for follow-up sampling and metallurgical tests in complex tailings.
Keywords
Subjects


Articles in Press, Accepted Manuscript
Available Online from 22 November 2026