USING NATURAL CHEMICAL PARAMETERS AS TRACERS TO DEFINE KARST GROUNDWATER BASIN DRAINAGE AREAS, SPRINGFIELD, MISSOURI

In the area of Springfield, Missouri, waters from a set of 12 sites within five different sink/spring basins were sampled over a period of approximately 6 months. The local surficial aquifer (containing the waters sampled) is composed of a number of relatively flat-lying (< 3 degree dip) Mississippian-aged carbonate units with small amounts of chert or other silicates dispersed throughout the formations. Theoretically, the mixture of urban, industrial, and rural land use over the groundwater basins as well as minute differences in the local carbonate units provide enough difference in water chemistry to develop unique geochemical signatures for each drainage basin. With a dataset of over 100 samples, 20% of the data were withheld for ‘blind sample’ and the remaining 80% of the data were used to generate the ‘basin geochemical signatures.’ Classification accuracy was defined as correctly placing a ‘blind sample’ chemical analysis into the basin from which it was sampled solely by using the chemical composition of the water and statistical discriminant analysis function built using other (known) samples from multiple basins in the region. In this study classification accuracy of approximately 80% was reached using common major ion chemistry data.

Field analyses included pH, temperature, conductivity, and bicarbonate. Major ions (calcium, magnesium, sodium, chloride, sulfate and nitrate) were analyzed in the lab. The results were input into the discriminant analysis function routine of SAS 9.4. Results indicated that roughly 80% of the “blind” samples were correctly classified into their source basin by the statistical method. A stepwise analysis of the results indicated that a large majority of the variation amongst basins could be met using only five water quality variables (pH, Temperature, Conductivity, Bicarbonate, and Chloride). Running the blind sample tests using only these five parameters resulted in approximately 75% classification accuracy. Because these parameters can be readily measured and recorded using modern automated sensors or quick field methods, the geochemical signature method shows great promise in using natural conditions to estimate karst drainage basins without need to introduce artificial tracers.