Abstract

Global mining generates a large amount of mine tailings, which can produce negative effects on the environment. To counteract this, government guidelines and scientific interest have emerged to reuse tailings deposits in innovative ways, converting them from an environmental liability to an economic asset. Thus, the characterization of a tailings deposit is of great importance to analyze the content of critical raw materials, as well as for a possible revaluation of the deposit, facilitating the mining process to be carried out in a more sustainable way. To characterize their chemical composition, information from drilling campaigns can be used. However, the evaluation of resources in tailings deposits has several complexities such as the poor grade spatial continuity and its narrow geometry, reasons for which traditional geostatistics is not adapted to model this type of deposit. In contrast, transitive geostatistics can be an opportunity to tackle this taking advantage of the fully delimited domain in a tailings dam and a different structural analysis. By achieving a better characterization of tailings composition, it is possible to make better decisions for their reprocessing, thus supporting cleaner mining production.This study aims to provide a framework for the geostatistical modeling and prediction of remaining metal resources of the Haveri tailings in southwestern Finland. The modeled variables are the gold, cobalt, copper and iron grades, with cobalt being of special interest as a critical material. The grades have been measured at 165 drill holes, totaling 1201 samples composited at 1 meter.An exploratory data analysis is performed first to clean the database and to identify the statistical and spatial distributions of the data. Then a structural analysis is applied to model the grade spatial continuity. Leave-one-out cross-validation is subsequently used to validate the fitted model and to quantify the prediction errors. Finally, 3D block models of the gold, cobalt, copper and iron grade are constructed with ordinary kriging and transitive kriging and are compared.Cross-validation shows that both kriging methods yield a good precision of the predictions and perform equally well for copper and iron, but transitive kriging significantly outperforms ordinary kriging for gold and cobalt and also provides less smoothed block models than ordinary kriging. Transitive kriging thus appears as an effective alternative for assessing resources in tailings and other narrow deposits. Recommendations on the sampling design to optimize the coverage of the target area and to ease the covariogram inference are also given.