Mapping hidden risks in mine tailings: A multi-variable geostatistical framework for environmental management.

The growing volume of mine tailings presents both environmental risks and opportunities for secondary resource recovery. Accurate spatial estimation of both valuable and potentially harmful elements within these deposits is essential not only for evaluating reprocessing potential but also for assessing environmental liabilities and guiding remediation strategies. Univariate geostatistical approaches, such as linear kriging, are limited in their ability to capture non-linear relationships and to quantify local uncertainty-particularly the probability or extent of threshold exceedance-across spatially variable and heterotopically sampled datasets, as commonly encountered in legacy tailings deposits. They also struggle to represent inter-variable correlations critical for joint risk and resource assessments. This study introduces a novel extension of multi-Gaussian kriging (MGK) to a multivariate framework-termed multi-Gaussian cokriging (MGCOK)-which enables the joint estimation of local recoverable functions, including mean concentrations, estimated metal/element quantity, and the volume of material exceeding specified environmental or regulatory threshold levels. The methodology is applied to the Haveri tailings deposit in Finland, focusing on sulfur (S), iron (Fe), and cobalt (Co), elements that are relevant to environmental monitoring. MGCOK leverages cross-variable correlations and a linear model of coregionalization to improve predictions of local recoverable functions by incorporating information from more densely sampled ones. Compared to standard MGK; MGCOK yields reduced estimation variance, and improve delineation of zones with elevated concentrations that may warrant environmental attention. Cross-validation confirms its superior performance, particularly for under-sampled and highly variable elements. This work highlights the utility of multivariate non-linear geostatistical modeling as a tool for environmental risk assessment in complex mine tailings environments, as well as, even though not studied in here, it has potential for resource evaluation.