EPA interim final report on WTC background study, April 2003

NYC 9/11 Public Portal Document

CERCLA Cleanup Program (OSWER 9285.6-07P, April 2002), and Calculating the Upper Confidence Limits for Exposure Point Concentrations at Hazardous Waste Sites (OSWER 9285.6-10, December 2002).

Evaluating environmental data that are collected to establish background concentrations of chemicals can be done using a variety of endpoints, such as reporting ranges, arithmetic or geometric means, or point-estimates based on a statistical computation. The method that is used to evaluate and interpret the data is dependent on the application of the background concentration. For this study, a point-estimate that represented an upper-bound of the population mean was the desired endpoint. This resulted in using a statistical approach, specifically calculating a upper confidence limit (UCL) based on the mean, for estimating an upper-bound point estimate for each analyte in its associated media.

Several different statistical methods can be used to estimate an UCL of a data set, depending on the data distribution. Two key steps were completed to estimate the UCL for these data sets:

1) Determined the distribution of the data (i.e., normal, lognormal, or neither^) and.

2) Computed the UCL using the appropriate confidence interval'* (e.g., 95%, 9'7.5%, 99%) and statistical method for the data distribution.

These two steps were performed with the ProUCL statistical software developed for EPA (Lockheed Martin 2001). Based on EPA guidance (Risk Assessment Guidance for Superfimd - Part A, 1989) and standard EPA Region 2 procedures, the samples that were reported as being below the detection limit were assigned a value equal to one-half the laboratory-reported detection limit for the statistical calculations’.

’If the distribution of the data is neither normal nor lognormal, it is recommended to use a non-parametric (i.e., distribution-free) statistic to estimate the UCL

*rhe confidence interval used to estimate the UCL was determined by the skewness of the data. For mild to moderately skewed data sets (e.g., o in the interval of 0.5 to 1) a 95% confidence interval was used, for moderate to highly skewed data sets (e.g., o in the interval of 1 to 2) a 97.5% confidence interval was used, and for highly to extremely highly skewed data sets (e.g., o in the interval of 2 to 3) a 99% confidence interval was used.

’A UCL was calculated for any data set that had variability. In situations where all of the samples were below the detection limit, but the detection limit varied, the UCL that was estimated was identified as a theoretical value because the mean that was calculated, fi-om which the UCL was derived, was based only upon the detection limit and it does not include actual detected concentrations. Therefore, the UCL associated with these data sets are theoretical in nature and are reported as such. WORLD TRADE CENTER BACKGROUND STUDY 20

NYC-WTC_000145895