Comparison of alternative filtering algorithms for estimating background groundwater levels of COC for the Savannah River Site, SC
V.M. Vyas, A. Roy, M. Ouyang, P.G. Georgopoulos (EOHSI, UMDNJ - R.W. Johnson
Medical School and Rutgers University);
W. Strawderman (Department of Statistics, Rutgers University);
D. Kosson (Dept. of Civil and Environmental Engineering. Vanderbilt University)
Regulatory agencies require contaminated groundwater to be remediated to drinking water standards (DWS), in order to minimize health risk for potentially exposed individuals. However, in certain situations, background groundwater concentrations of one or more of the Constituents of Concern (COC) may be above the DWS, making it problematical to achieve DWS based clean-up standards. In such cases, it is reasonable to propose alternative remediation endpoints, based on background groundwater concentrations of the COC. Background levels are defined as concentrations arising from (a) natural geological characteristics, or (b) diffuse contaminant sources not attributable to a single waste management area or localized event.
Background groundwater concentrations of the COC at a contaminated site are generally established by monitoring the groundwater in wells sufficiently upgradient of the contaminated site. However, when dealing with the various facilities of U.S. Department of Energy's (DOE) nuclear complex, the distinction between impacted and unimpacted (background) levels of COC has to consider several complicated features of the nature of contamination: there are no recognized areas or monitoring locations on the site that have been designated as completely unimpacted; there are large spatial gaps in the monitoring of the overall groundwater quality; there are several spatially separated sources; the resulting plumes have intermingled; there are different sources for different COC; and there is contamination from activities prior to the establishment of SRS. The problem of identifying background levels falls in between defined structure and predominant randomness. There is structure in the form of plumes and spatial/temporal trends, but also randomness (or variability) due to the mixing and degradation of plumes; intermittent and fluctuating sources; and interaction of pollutants with soils (such as leaching of metals due to acid pollution). Varying sampling, analysis and reporting procedures, and a large proportion of non-detects for several critical COC, add to the complexity of analysis. Distributions for COC are neither perfectly unimodal nor clearly bimodal. The problem has to be resolved by developing methodologies that combine and adapt existing statistical and geostatistical tools, because no single conventional approach is able to achieve the necessary distinctions.
Alternative statistical methodologies for defining background groundwater levels were formulated, applied, and evaluated in response to the challenge of determining background levels. The methodologies consisted of several screening steps based on geostatistical estimation, trend analysis, regression, cluster analysis, and regression-tree methods. The final outcomes are lists of background wells and probability distributions of background levels for each COC. The sensitivity of the methodologies to alternative sets of filtering schemes was investigated. Alternative sets of mathematical filters were used in conjunction with the geostatistical methods employed to identify potentially contaminated areas, as well as with the regression-based and cluster analysis methods used to further screen out potentially contaminated wells. The results from different methodologies and filtering schemes are compared and evaluated in this presentation.
ACKNOWLEDGEMENTS: Our research has been supported by a grant to the Institute for Responsible Management, Consortium for Risk Evaluation with Stakeholder Participation, from the US Department of Energy, Instrument DE-FG26-00NT 40938. The viewpoints expressed in this report are solely the responsibility of the authors and do not necessarily reflect the views of the US Department of Energy or its contractors.