Evaluation and statistical analysis of groundwater monitoring data for waste disposal facilities
Abstract
Many waste disposal facilities are subject to ground water monitoring to detect contamination. Statistical comparisons are used to determine if concentrations of chemical parameters in down-gradient wells are significantly elevated when compared to up-gradient wells. Statistical techniques for such comparisons are described in federal regulations and USEPA guidance documents. However, difficulties may occur in applying these methods to actual ground water monitoring situations. A combination of actual ground water monitoring data and artificial data sets are used to improve techniques to evaluate the presence of ground water contamination. The actual ground water monitoring data is from over 30 regulated hazardous waste disposal facilities and includes over 180,000 individual concentration measurements. Artificial data sets were created representing a variety of actual ground water monitoring situations. Characteristics of actual ground water monitoring data are evaluated for artifacts and metals contamination. Artifact contamination is usually present at concentrations below 50 parts-per-billion. Artifacts are commonly detected in multiple wells for a sampling event. The relationship between total and dissolved concentrations varied for each metal. Only for iron, was a relationship of the total metal concentration to turbidity apparent. A statistical analysis showed that for five of sixteen comparisons, there was no difference in the number of wells showing contamination for total and dissolved metals. For nine of sixteen comparisons, total metals resulted in more wells showing contamination than dissolved metals. In two cases, the number of contaminated wells was greater for dissolved metals than for total metals. In an evaluation of statistical methods where a high percentage of samples showed undetectable concentrations of the test chemical, the Kruskal-Wallis and Wilcoxon rank methods are shown to be the preferred over non-parametric and Poisson limit methods. Using artificial data sets, Shewhart-CUSUM control charts are shown to be poorly suited to ground water monitoring data. Analysis of variance methods are shown to be better suited when only the six most recent sampling events are included in the analysis. Many of these research findings are directly applicable to current practice for ground water monitoring and the evaluation of environmental contamination.
Degree
Ph.D.
Advisors
West, Purdue University.
Subject Area
Geology|Environmental science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.