Sustainable water management in mineral processing by using multivariate variography to improve sampling procedures

Funding Information: The authors would like to thank Boliden Kevitsa mine for industrial support and access to the data. Prof. Kari Heiskanen is acknowledged for his advices, comments, and suggestions. The authors thank Prof. K.H. Esbensen for the very valuable comments provided as part of the pre-examination of TMK Le's PhD manuscript. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Publisher Copyright: © 2021 The Authors Due to the rising pressure from governmental regulations, the water recycling rate has increased significantly in mining operations over the last decades, resulting in a high variation of process water quality, which could potentially impact the plant performance. The current effort to assess water quality in mining is shifting from managing water to fulfill environmental regulations (focus on the effluents) to controlling water quality to maintain the operating performance (focus on the water within the process). However, minimal effort has been made to design a dedicated sampling procedure for process water. This study investigates the use of multivariate variography and principal component analysis (PCA) for improving the process water sampling procedure at the Kevitsa Mine, Finland. The aim is to design a sampling procedure for evaluating water quality using two different types of datasets and illustrating the impact of the dataset structure on the sampling design. The results showed that the common spot sampling procedure generated a very high sampling error and was not the best practice for process water. The weekly sampling frequency used at the mine site, suitable for fulfilling environmental regulations was too low to capture the process water variation. Therefore, it is not recommended to use environmental water datasets for operating control purposes. The multivariate variographic analysis revealed the hidden cyclic variation through its ability to summarize the time variations and the correlation between multiple variables that were not visible through the classical univariate variogram approach. However, the number of increments recommended by the global multivariogram became impractically high. Hence, an alternative approach combining PCA to the mutivariogram was used to filter noise from the data and keep the relevant information. This study highlights the benefits of using multivariate variography to improve water sampling procedures in the mining industry and to reduce both operational and environmental risks associated with water quality variability. Thus, this method has the potential to be used in worldwide mining operations as a standard procedure for sampling water to provide reliable results.