Multivariate statistical analysis of hydrogeochemical data to identify and delineate primary influences on groundwater chemistry

This study uses multivariate statistical analyses to identify and delineate mine-influenced groundwater from other types of groundwater near a former copper mine in Yerington, NV, USA. This statistical approach was used because some analytes found in the mine-influenced groundwater are also present in naturally-mineralized and agriculturally-influenced groundwater in the vicinity of the mine. Uni- or bi-variate graphing and mapping of only a few analytes does not provide a unique delineation of the extent of MIW in this groundwater system.Unsupervised and supervised Exploratory Data Analysis methodologies were used in this statistical evaluation of 426 groundwater samples. The methodologies were Hierarchical Cluster Analysis (R-mode), Principal Component Analysis, K-Means Cluster Analysis (Q mode) and Linear Discriminant Analysis.All four statistical methods were able to identify separate groupings of analytes that are consistent with the Goldschmidt Classification framework for lithophile, siderophile and chalcophile elements, a framework that is relevant to solute mobility and rock-water interactions among the three water types. Maps of the statistically based groupings of analytes are consistent with known areas of mining, agriculture, and areas where natural mineralization remains a dominant influence on groundwater quality.The results of the statistical methods are used to delineate the extent of mine-influenced groundwater. The statistical methods described in this paper have broad applicability for characterizing the extent of various natural and anthropogenic geochemical influences on the chemical composition of groundwater as well as many other types of environmental investigations involving large datasets. These methods may be particularly important when allocation or apportionment of impact is necessary.