Identification of Background Levels from Soil Constituent Data with a High Proportion of Non-Detects, at the USDOE Savannah River Site in South Carolina
Vikram M. Vyas (EOHSI, UMDNJ - R.W. Johnson Medical School and Rutgers
University)
Brian Hough (Savannah River Ecological Laboratory)
Saichiu Tong (EOHSI, UMDNJ - R.W. Johnson Medical School and Rutgers University)
Christopher Romanek (Savannah River Ecological Laboratory)
Panos G. Georgopoulos (EOHSI, UMDNJ - R.W. Johnson Medical School and Rutgers
University)
David Kosson (Vanderbilt University)
Identification of Background levels of Constituents of Potential Concern (COPC) is an essential step in USEPA mandated Remedial Investigations and Feasibility Studies. In cases where risk based state and federally designated MCLs fall below background levels, the latter constitute a more feasible remediation endpoint for site cleanup activities.
For many COPC, soil and groundwater concentrations are often close to or below method detection limits (MDL). For data sets with few values below MDL, the non-detects may be adjusted to a fixed number such one half of the MDL, without introducing distortions in the probability distribution of the measured data. However, in cases where non-detects constitute a significant portion of data, using substitutions with fixed values can significantly skew the actual probability distribution. Background levels are linked to measurements through parametric or non-parametric statistics. In parametric statistics, a high proportion of non-detects skew the mean; in non-parametric tests, if the proportion of non-detects is more than 50%, the median of the distribution can not be identified. Thus, a high proportion of non-detects complicates the identification of background levels.
This study evaluated two methods for estimating non-detect values for soil concentrations of some COPC at the USDOE's Savannah River Site facility. The methods were evaluated for robustness to high proportion of detects and non-normal distributions, and an optimum method was identified. Soil background levels of COPC were estimated from data adjusted for non-detects, and compared to the risk based screening levels.
ACKNOWLEDGEMENTS: This work is funded by grant DE-FG26-00NT 40938 to the Consortium for Risk Evaluation with Stakeholder Participation, from the USDOE. The views expressed in this work are solely the responsibility of the authors and do not necessarily reflect the views of the USDOE or its contractors.