PATTERNS IN GROUNDWATER GEOCHEMISTRY: UTILIZING COMBINED MULTIVARIATE STATISTICS TO ASSESS GROUNDWATER SYSTEMS

Efforts to quantify groundwater properties and behavior often require the use of complex methodologies to uncover the temporal and spatial patterns between different components of the hydrogeologic system. The use of a variety of multivariate statistical techniques has become a standard procedure in studies on groundwater hydrochemistry. Often only one multivariate method is applied to a dataset, however, which may limit how the results can be interpreted. We analyzed hydrochemical data compiled from 48 groundwater sites (29 springs and 19 wells) from around Kentucky using three multivariate techniques—principal components, non-metric multidimensional scaling, and cluster analysis—to assess the ability of these methodologies to work alone and in combination to improve our understanding of groundwater system characteristics.

Each method provides a different insight into the data. Non-metric multidimensional scaling provides a visualization of similarity, cluster analysis provides indications of a clear relationship between sites, and principal components provides the strongest evidence for which hydrogeologic factors drive similarities and differences. Cluster analysis was able to group groundwater samples based on similarities of all geochemical constituents. Through hierarchical clustering, the Dunn’s index values of best fits for the data indicate that the spring and well samples represent six distinct groundwater systems. Principal component results showed that these six groundwater systems are primarily distinguished by variability in conductivity, sodium, and potassium, and secondarily by calcium and magnesium concentrations.