PCB homologue and congener preliminary fingerprinting and toxicity equivalents (TEQs) for two Sierra Club samples collected in the Dick’s Creek study area

May 7, 2003

Nita Nordstrom
Southwest District Office, Division of Emergency and Remedial Response, Ohio EPA

Background

I received a copy of the Sierra Club’s analysis report for two sediment samples collected in the Dick’s Creek Area in August, 2002. I do not have copies of the chain-of-custodies for these samples and am not sure of the exact date of the sampling event nor do I have a copy of their sampling plan. Although I did not observe or participate in this event, I was told that one sample was collected near the welding company located on the south bank of Dick’s Creek just west of the Monroe Ditch/Dick’s Creek confluence and east of the Yankee road bridge. Also, from what I was told, the second sample was collected in the proximity of Amanda Elementary School, downstream of the site of the first sample collection. I was also told that these samples were collected in depositional zones, as recommended by sound sediment sampling practices.

The methods employed in the actual sample collection (sampling design) are not known. The highest source of variation in sample results generally occurs in the sample collection activities. Methods used in sample collection, transport, handling, storage, and manipulation of sediments and interstitial waters can influence the physicochemical properties and results of chemical, toxicity and bioaccumulation analyses (USEPA, 2001). Sampler selection, depth of sediment penetration, sample volume, whether several samples were collected and composited or one grab sample used, and any other activities that would factor into the practical considerations for site-specific selection of sampling stations would be key in determining representativeness of the samples. Sampling designs generally fall into two major statistical categories: random or probabilistic, and targeted. However, when we have data on only one sample collected at two stations, we cannot use statistical analysis. Therefore, the confidence in the fingerprint analysis is less than if we had enough samples to analyze statistically.

According to the PCB Analysis Report, Severn Trent Laboratories (STL) received the samples August 28, 2002. Samples were extracted December 13, 2002, initial calibration was performed on December 27, 2002 and samples were analyzed January 7, 2003. Sample sizes were approximately 1 gram, no dilution factor, and concentrations of PCBs were reported in pg/g (pico grams/gram) ((parts per trillion -(ppt))(dry weight basis). These samples were analyzed using Method 1668 A, high resolution GC/MS to identify PCB homologues and congeners. I assume the samples were frozen or stored at 4 degrees C in the dark, as the holding times were extended past the 1-2 weeks generally specified in sediment sample storage laboratory protocols. However, according to current studies, high molecular weight compounds such as PCBs do not seem to vary appreciably, and longer term storage might be acceptable for those analytes (USEPA 2001). Additionally, I did not perform a QA/QC audit of the analysis report.

PCB Fingerprinting Background

PCBs are generally "fingerprinted" for two major reasons: 1) to help determine the source of the PCBs found in the environment and 2) to aid in a site specific risk assessment for this complex chemical mixture. Fingerprinting, or pattern recognition is one of the most widely used techniques for the interpretation of complex congener-specific analyses and/or large numbers of PCB determinations. Visual examination of either chromatographic patterns or numeric tables of large data sets can be daunting and/or impossible (Schwartz, Stalling 1991). Computer based models have been developed to tackle this task and they generally include pattern recognition analyses. Proper application of statistical techniques can also minimize personal or organization bias, improve the quality of data, provide data of known quality, and discourage improper assignment of Aroclor designations of weathered or metabolized PCB patterns. Jaccard coefficients can be obtained by comparing chromatograms; the coefficients are expressed as a ratio of the number of congeners in common to both chromatograms to the total number of congeners present in at least one specimen. (Erickson, 1997). However, as stated above, statistical analysis of these data is not possible. Therefore, I performed a cursory "back of the envelope" comparison. Calculations and comparisons were made as detailed below.

Sierra Club Samples PCB Fingerprinting

The "fingerprinting" analysis that I performed for these two samples is very preliminary and should be considered only as an initial analysis of what is generally a long, complicated process of fingerprinting PCBs for a complex site such as the Dick’s Creek area (e.g, the PCB fingerprinting performed on the Hudson River includes virtually thousands of data points). Therefore, these results are not definitive nor should they be interpreted as such.

The sample data (analytical report) show a listing of the sample PCB homologue groups and the concentrations found for each in each of the two samples (reference Table 1). First, I compared homologue distribution concentrations in both the Welding Company and Amanda Elementary School samples. This first step shows the shape of those list of values are very similar (we actually look at the "shape" of the list of concentrations by homologue in this case, to see if the shapes match between samples). An additional more detailed comparison would be to compare all 209 congeners individually between samples, which I did not perform. Second, for each sample, I focused on the homologue group with the highest concentration (reference Table 4 - highlighted values). From that point I calculated ratios for the other homologues detected. Third, I performed the same functions for the 15 WHO congeners shown in the PCB Congener TEQ Analysis Report) (Reference Table 4, Figures 1- 4 attached). This method shows very similar patterns of homologue and congener concentrations and ratios between the two samples (Reference Figures 1- 4). Additionally, when comparing the number of individual congeners in each homologue group that were detected in both samples, they are also very similar (reference Table 3). This shows how many individual congeners out of the total congeners in each homologue group were detected in both samples (e.g., Total mono- 3/3 = of the 3 congeners that can possibly be detected as Monochloro Biphenyls, 3 were detected in that sample).

Given the complexity of the PCB chemical, environmental "weathering," media partitioning, co-elutions, chromatogram interpretations, analytical QA/QC as well as all of the other factors involved in PCB analytical chemistry, it appears in this preliminary analysis that the PCBs in both samples are quite similar and could possibly originate from the same source(s). However, further investigation and analysis are required to definitively determine a conclusion of a high confidence "fingerprint" match.

ATTACHMENTS

TABLE 4 (Attached)

FIGURES 1- 4 (Attached)

REFERENCES