PREDICTING THE MAJOR-ION COMPOSITION OF GROUNDWATER IN THREE DIMENSIONS ACROSS THE CONTERMINOUS UNITED STATES

Measurements of salinity and total dissolved solids are functions of the major-ion composition of water. The major ion composition of water also affects the mobility, aqueous speciation, and toxicity of metals and other solutes, and determines the potential for galvanic corrosion and encrustation. As such, the major ion composition of water can determine its suitability for specific uses and water treatments, and its potential to adversely affect aquatic organisms. Where comprehensive water-chemistry data are available, aqueous geochemical models can explain the occurrence and evolution of the ionic composition and can provide accurate estimates of salinity. These models cannot, however, be used to estimate water-quality conditions where water-chemistry data are incomplete or not available. The major ion composition of water can be categorized into water types based on the predominant cations and anions. Here we demonstrate the use of ensemble-tree machine learning to predict five water type categories (Ca+Mg – HCO₃, Na+K – HCO₃, Ca+Mg – SO₄, Na+K – SO₄, and Cl). A sixth category was defined as “mixed” indicating no predominant anions (>50%). Water type predictions were made in three dimensions across the conterminous United States. The ensemble-tree model makes predictions through pattern recognition and supervised statistical learning using selected environmental factors known to influence the ionic composition of groundwater. This approach allows for estimates of major-ion composition in unmonitored locations. Understanding the major ion composition of groundwater, its influence on calculating salinity, and the relevant factors controlling its spatial distribution can assist water managers in identifying alternative supplies or appropriate treatment strategies for specific end uses.