Determination of Statistically Significant Increases over Background Concentration Levels

摘要

HDR is currently supporting its power clients to comply with the Environmental Protection Agency’s Hazardous and Solid Waste Management System;Disposal of Coal Combustion Residuals from Electric Utilities;Final Rule(CCR Rule).The groundwater sampling and analysis requirements of Part 257.93 form a critical step in the compliance process;however,the implementation is often a challenge since multiple statistical methods are possible.It is not readily apparent which method is best for testing for statistically significant increases over background levels and how the issues of non-detects,parametric versus non-parametric sample distributions and small sample sizes are tackled.HDR has developed a methodology to efficiently assess the background sample data in line with the rules found in Part 257.93 and produce sitespecific test statistics based on the background concentration levels using the upper prediction limit(UPL)for detection monitoring(Part 257.93(f)(3)).HDR has selected the UPL since this statistic has many intuitive features and can adjust for the increase in the site-wide false positive rate during detection monitoring.Attributes of HDR’s robust methodology include the statistical methods to detect for normal,lognormal and gamma distributions and methods to impute the values of non-detects with substantially less bias in the estimates than when the simple substitution method is used.The methodology draws on sound statistical principles and addresses the challenges due to small sample sizes,outliers,varying distributional forms of the samples,serial correlation,and trends to ultimately produce test statistics which can flag statistically significant increases over background to the best extent possible given the available data.