Premium
An evaluation of the Confidence Removal Goal approach for making remediation decisions at Superfund sites
Author(s) -
Schultz Brad,
Singh Ashok K.,
Singh Anita
Publication year - 2002
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.563
Subject(s) - environmental remediation , superfund , population , sample (material) , statistics , confidence interval , environmental science , contamination , computer science , mathematics , chemistry , ecology , hazardous waste , biology , demography , chromatography , sociology
The Confidence Removal Goal approach was developed by the Gradient Corporation for the cleanup decisions at the METCOA Superfund Site. The CRG approach is a statistical approach to meet site cleanup goals based on the arithmetic mean of the contaminant concentration distribution. The CRG approach involves computing a threshold value c * (called the CRG value) from observed sample values in cases where the data distribution is log‐normal and the UCL of the population mean calculated for the sample exceeds the Cleanup Goal (CUG). The cleanup would then consist of removing all soils from the site with observed concentrations higher than the CRG value c * . At the METCOA site, the cadmium (Cd) and nickel (Ni) concentrations passed the test of log‐normality, and the UCL of the means exceeded their respective CUG values. The CRG values were then computed for observed Cd and Ni concentrations. In both cases, the CRG values turned out to be much higher than the respective observed maximum concentration. Based on these CRG values, it would appear that no cleanup would be needed at the METCOA site. A Monte Carlo simulation experiment was conducted by two of the authors of this article to investigate the performance of the CRG approach. The simulation experiments showed that the CRG approach had a rather high percentage of false negatives, i.e. it recommended no cleanup a large number of times when the simulation was performed for contaminated site conditions. The present article investigates the performance of the CRG approach for normal, uniform and exponential distributions, in addition to the log‐normal distribution. The CRG approach gives reasonable results for all models considered in this article, except for the log‐normal model, suggesting the inadequacy of the log‐normal distribution for modeling skewed environmental data. Copyright © 2002 John Wiley & Sons, Ltd.