STATISTICAL ANALYSES OF POLYCHLORINATED BIPHENYL (PCB) CONCENTRATIONS IN FISH SAMPLES COLLECTED NEAR THE LEMON LANE LANDFILL SITE MONROE COUNTY, INDIANA

Date Prepared : November 4, 2003
Prepared by : Tetra Tech EM Inc.
Tetra Tech Site Manager : Jeffrey Lifka

1.0 INTRODUCTION

Tetra Tech EM Inc. (Tetra Tech) has prepared this statistical analysis report for the U.S. Environmental Protection Agency (EPA) in partial fulfillment of the statment of work for Response Action Contract (RAC) No. 68-W6-0037 for Region 6, Work Assignment No. 941-RSBD-0529. The primary objective of the report is to evaluate the possible impact of the Lemon Lane Landfill site in Monroe County Indiana on polychlorinated biphenyl (PCB) concentrations in fish samples collected from streams near the landfill. The fish samples evaluated in this report were collected and analyzed by Viacom, Inc. (Viacom), the potentially responsible party for the site, during four sampling events that occurred between November 1996 and November 2002.

This statistical analysis report consists of six sections, including this introduction. Section 2.0 contains site background information for Lemon Lane Landfill and the four fish sampling events. Section 3.0 provides a more detailed discussion of the objectives of the statistical analyses. Section 4.0 discusses the data available for the analysis and the statistical methods used to analyze the data. Section 5.0 presents and discusses the results of the statistical analyses, and Section 6.0 summarizes the conclusions and presents recommendations for future fish sampling events. References used to prepare this report are listed after Section 6.0. Appendix A presents complete PCB concentration results for all fish samples collected during the four sampling events, and Appendix B presents the results for all statistical analyses.

2.0 LEMON LANE LANDFILL SITE BACKGROUND

This section briefly summarizes the site's history and describes the four fish sampling events that generated the data analyzed in this report. Much of the site background information presented in this section was developed fiom Tetra Tech's revised current status report for the Lemon Lane Landfill site (Tetra Tech 2001).

2.1 SITE HISTORY

Lemon Lane Landfill is located on a 10-acre parcel on the northwest side of Bloomington in Monroe County, Indiana. The landfill was opened as a refuse dump by a private owner in 1933. After 1950, Lemon Lane Landfill was operated by the City of Bloomington as a municipal waste landfill. The Westinghouse Electric Corporation (Westinghouse, later known as CBS Corporation [CBS] and now known as Viacom) discarded wastes, including electrical capacitors filled with PCB-containing oils, at the site between 1957 and l964. The landfill had no liner or cover and dumping was not controlled. Scavenging and burning were common practices at the site.

Lemon Lane Landfill was placed on the National Priorities list in October 1989. In 1985, EPA, the State of Indiana, Monroe County tlic City of Bloomington and Westinghouse (now Viacom) signed a consent decree. Under the terms of the consent decree, Viacom is to remediate six sites in the Bloomington area containing PCBs. Lemon Lane Landfill is one of the six sites covered by the consent dccree. Between May and September 1987, a series of interim remedial measures (IRM) was completed. As part of the IRMs, the site was cleared of all vegetation; all visible capacitors were removed; and the south slope grade, which had eroded, was stabilized with fill material. The entire site was covered with a temporary cap consisting of 30,000 tons of clean fill overlain by a 488,000-square-foot, 36-mil plastic membrane cover.

During the mid-1980s, site investigations were initiated to further characterize site wastes and develop a remediation plan. The remedial alternative selected by Viacom and the government parties included excavation and either off-site disposal or on-site consolidation of impacted soil and waste within and adjacent to the landfill. The landfill was then to be covered with a cap meeting Resource Conservation and Recovery Act (RCRA) Subtitle C standards.

Viacom also conducted a hydrogeologic investigation to characterize groundwater flow and determine possible impacts on local surface water. Tracer tests in the late 1980s and early 1990s had established that most of the groundwater originating at Lemon Lane Landfill drained through karst solution conduits to Illinois Central Spring (ICS), located 2,500 feet southeast of the site. As early as 1982, the spring resurgence area was known to contain PCB-contaminated sediments.

From 1995 to 1998, Viacom monitored flow rate, conductivity, and PCB concentrations in ICS. Based on early findings, Viacom attempted to reduce infiltration of contaminated water into the subsurface by installing an impermeable liner over a sinkhole area in the southwest corner of the landfill. This activity was completed in December 1996.

In May 2000, Viacom began excavation, disposal, and consolidation of soil and fill materials at the site as part of a remedial action (RA) overseen by EPA, the State of Indiana, and Monroe County. Excavated material with a PCB concentration greater than 50 parts per million (ppm) was disposed of off site at a Toxic Substances Control Act (TSCA)-permitted landfill. Excavated material with a PCB concentration less than 50 ppm was consolidated underneath the landfill cap. Excavation activities were completed in October 2000, and installation of the RCRA cap was completed in December 2000. A water treatment plant located at ICS was also completed during this time period. The treatment plant began operating in May 2000, and is designed to collect and treat up to 1,000 gallons per minute of PCB-impacted water discharging to ICS.

2.2 FISH SAMPLING EVENTS

Between November 1996 and November 2002, Viacom conducted four separate fish sampling events near Lemon Lane Landfill. Viacom collected fish samples in November 1996, November 1999, November 2000, and November 2002. Fish sampling events focused on Clear Creek, which receives subterranean drainage from springs (including ICS and Quarry Springs) located southeast of the landfill area. Clear Creek flows along the west side of the City of Bloomington and then to the south. EPA and the Indiana Department of Environmental Management (IDEM) also collected a small number of fish samples in August and November 1997. However, these results cannot be used in statistical analyses because of the limited number of samples and are not discussed further in this report.

Each fish sampling event is described briefly below. The first two sampling events in November 1996 and November 1999 were conducted before Viacom started the excavation, consolidation, and disposal RA at Lemon Lane Landfill. For subsequent statistical analyses, these events are defined as "pre- remediation" sampling events. The November 2000 and November 2002 sampling events were conducted after most RA activities had been completed and are defined as "post-remediation" sampling events. Appendix A lists all fish samples collected during these sampling events, as well as the 1997 samples collected by EPA and IDEM.

November 1996. Dunng the November 1996 sampling event, Viacom collected whole fish and fillet samples from six locations in Clear Creek, ranging in distance from 1.3 to 20 miles downstream of Lemon Lane Landfill. Eleven types of fish were collected, but for reasons discussed in Section 4.0 below, data for only five types are usable for statistical analyses.

November 1999. Viacom collected whole fish and fillet samples from three sampling locations in Clear Creek during the November 1999 sampling event. Sampling locations during this event were 3, 10. and 20 miles downstream of Lemon Lane Landfill. Ten types of fish were collected, but for reasons discussed in Section 4.0 below, data for only six types are usable for statistical analyses.

Although over 300 fish samples were collected and analyzed during the four sampling events described, in Section 2.2. only about half of the sample results can be analyzed statistically. Statistical analysis of the data for any sampling location is possible only if a sufficient number of comparable data points are a\ ailable. Generally, at least three samples of a particular type of fish must be available for two or more sampling events. In addition, the statistical methods used must be appropriate for the type of data available for analysis. Section 4.1 briefly describes the types of available PCB results; Section 4.2 provides a complete inventory of fish sample results organized by sampling event, sampling location, and fish type; and Section 4.3 outlines the statistical methods Tetra Tech used to analyze the data.

4.1 PCB ANALYTICAL METHODS AND RESULTS

For all of the sampling events discussed in this report, PCB concentrations for all samples were measured by gas chromatography (GC) and reported as Aroclors. Aroclors are commercial mixtures of PCBs, characterized by the percentage of chlorine in the mixture. Variations of the GC method have been in use for over 25 years, and most historical PCB data for Lemon Lane Landfill are reported as Aroclors. In this method, the GC instrument produces a series of peaks indicating the instrument's response to the Aroclor mixture present in the sample. The analyst selects a number of peaks that will be used to determine the Aroclor concentration, compares the sample results to the peak heights and pattern for a standard Aroclor mixture used to calibrate the GC, and then uses this comparison to determine the Aroclor concentration in the sample.

For the November 2002 sampling event, Viacom reported two sets of Aroclor results original results using four peaks to determine the concentration and revised results using five peaks to determine the concentration. In this report, statistical analyses are conducted using both sets of results for the November 2002 samples. The terms "original 4-peak" results and "revised 5-peak" results are used to distinguish between the two sets of results.

PCB concentrations can also be measured by high resolution GC/mass spectrometry (MS) and reported as congeners. PCBs consist of a group of 209 individual chemicals called congeners, and the GC/MS method reports results for each individual congener. Congener results can be reported in two ways. First, results of all individual congeners can be summed and reported as a total PCB congener concentration. Second, the results for 13 toxic PCB congeners (as identified by the World Health Organization) can be combined to produce a "toxicity equivalent" (TEQ) concentration.

To date, all fish samples collected from Lemon Lane Landfill have been analyzed for Aroclors using the GC method. Currently there are no accepted methods to convert individual PCB congener or TEQ results into Aroclor results and vice versa. Statistical analyses can be conducted only when PCB results are reported in the same format. Thus, if the high resolution GS/MS method is used for future fish sampling events, the resulting PCB congener and TEQ concentrations could not be statistically compared with the Aroclor results evaluated in this report.

4.2 INVENTORY OF FISH SAMPLE RESULTS

Tables 1 and 2 summarize the types of fish samples collected during the four sampling events at Lemon Lane Landfill. The tables are organized by sampling location and sampling event date. For each location and sampling event, the tables list the types of fish collected and the numbers of samples for each fish type. The final column of the tables indicates whether the data are sufficient data for statistical analysis and the types of analyses that can be conducted. Statistical analyses are discussed further in Section 4.3.

Table 1 presents the available data for whole fish samples analyzed for total PCB Aroclor concentrations, and Table 2 presents similar results for fillet samples. As the tables indicate, much of the data cannot currently be used in statistical comparisons. For example, Crayfish were collected at sampling locations 1, 3, 2.5, and 3 miles downstream during the November 1996 sampling event, but were not collected during any other sampling events. Similarly, Creek Chub fillet samples were collected at the Country Club Road location during both the November 1999 and November 2000 sampling events, but the two samples collected in November 2000 are below the minimum of three samples needed for statistical analysis.

4.3 STATISTICAL METHODS

• Changes in fish PCB concentrations from the pre-remediation period to the post- remediation period. These changes can be evaluated by comparing the combined results for the November 1996 and November 1999 pre-remediation sampling events ~ ith the combined results for the November 2000 and November 2002 post-remediation sampling events. These comparisons help answer the following question: "Are fish PCB concentrations lower after remediation than before remediation?"

• Changes in fish PCB concentrations during the post--emediation period. These changes can be evaluated by comparing the November 2000 and November 2002 results. These comparisons help answer the following question: Flare fish PCB concentrations changing over time after remediation'?"
• Trends in fish PCB concentrations over time. Trends can be evaluated when comparable data are available for three or more sampling periods at a particular sampling location. The results of these analyses can be used to answer the following questions: "Are fish PCB concentrations changing over the time period being evaluated?" and "How quickly are the concentrations changing?"

4.3.1 Two-Sample Statistical Analyses

• Two-sample analyses were conducted only when at least three data points were available for each of the two sampling events analyzed.
• Two-sample analyses were conducted using Student's two-sample l-test. The Statistix~ program produces two t-statistics for this test. The appropriate t-statistic depends on whether data from both sampling events have equal or unequal variances. This determination was made using the F-test results included in the Statistix7 t-test output. Section 4.5 of EPA QA G-9 (EPA 2000) describes the use of the F-test for this purpose.
• In one comparison, a nonparametric test (the Wilcoxon Rank Sum [WRS] test) was used to conduct the two-sample analysis. A critical assumption underlying the t-test is that the data being evaluated follow a normal (or bell-shaped) distribution. The WRS test was used as a supplemental statistical method when the data from one or both sampling events did not follow a normal distribution, as determined by the Shapiro-Wilk test for normality. The WRS test and the Shapiro-Wilk test are described in Sections 3.3 and 4.2. respectively, of EPA QA/G-9 (EPA 2000).
• For analyses, a probability value (p-value) of 0.05 was used to determine statistical significance-p-values less than or equal to 0.05 were considered statistically significant, while p-values above O.05 were not considered significant.
.

Student's t-test can be run as either a one-tailed test or as a two-tailed test. Tetra Tech used the two-tailed alternative for all statistical analysis conducted by t-test. The two-tailed test is used to determine whether fish PCB concentrations from two sampling events are different, without making any assumptions about whether the concentrations are increasing or decreasing over time. The two-tailed test is more conservative than the one-tailed test-a larger difference between PCB concentrations for the two sampling events is needed to achieve statistical significance. A one-tailed test is appropriate for evaluating whether fish tissue PCB concentrations have changed in only one direction-for example, whether concentrations have decreased from an earlier sampling event to a later sampling event. The one-tailed test might seem appropriate for evaluating data for Lemon Lane Landfill because fish tissue PCB concentrations are expected to decrease over time if the completed and ongoing remedial measures are effective. However, increasing PCB concentrations would be of obvious concern, and increases were observed in some of the comparisons.

Trend Analyses

Tetra Tech conducted all trend analyses following the methods described in Section 4.3 of EPA QA/G-9 (EPA 2000). These methods are described in more detail in Statistical Methods for Environmental Pollution Monitoring (Gilbert 1987). Tetra Tech adhered to the following general guidelines for all trend analyses:

• Trend analyses were conducted only when data were available for three or more sampling events, and generally when at least three data points were available for each sampling event
.
• Trend analyses were conducted using the nonparametric Mann-Kendall (M-K) test. This test can be used to evaluate whether an upward or downward trend in fish PCB concentrations exists over the sampling events evaluated. Specifically, Tetra Tech used a variation of the M-K test that evaluates multiple data points per sampling event. This variation is described in Section 4.3.4.2 of EPA QA/G-9 (EPA 2000) and in Section 16.4.3 of Gilbert (1987). Dr. John Warren of U.S. EPA's Quality Staff confirmed that this method is appropriate for evaluating fish PCB results for Lemon Lane Landfill.
• If the M-K test indicated that a statistically significant trend existed, Sen's Slope Estimator, another nonparametric test, was used to estimate the magnitude of the trend. Sen's Slope Estimator can provide an estimate of the changes in rish PCB concentrations in units of milligrams of PCBs per kilogram of fish tissue per year. Sen's Slope Estimator is further described in Section 4.3.3.1 of EPA QA/G-9 (EPA 2000) and in Section 16.5 of Gilbert (1987).
• For the M-K test, a p-value of 0.05 was used to determine statistical significance: p- values less than or equal to 0.05 were considered statistically significant, while p-values above 0.05 were not considered significant.

5.0 RESULTS AND DISCUSSION

Fish samples have been collected at eight locations in Clear Creek ranging from 1.3 to 20 miles downstream of Lemon Lane Landfill. However, based on the factors outlined in Sections 4.1 and 4.2, statistical analyses are limited to four locations that yielded a sufficient amount of comparable data. These location ~ include Allen Street (1.3 miles downstream), Country Club Road (3 miles downstream), Fluckmill Road ( 10 miles downstream), and Strain Ridge Road (20 miles downstream). Statistical analyses include comparisons of both whole fish sample and fillet sample results. This section presents the results for ( I ) two-sample statistical analyses of pre-remediation results vs. post-remediation results; (2) two-sample statistical analyses of post-remediation results vs. post-remediation results; and (3) trend analyses. Complete results for these statistical analyses are presented in Appendices B- 1, B-2, and B-3, respectively.

PRE-REMEDIATION VS. POST-REMEDIATION RESULTS

Table 3 summarizes the results of pre-remediation vs. post-remediation statistical analyses for whole fish samples. In all comparisons, whole fish PCB concentrations decreased from the pre-remediation sampling events (November 1996 and November 1999) to the post-remediation sampling events (November 2000 and November 2002). Statistically significant decreases were observed at the two sampling locations closest to Lemon Lane Landfill. At the Allen Street location (1.3 miles downstream), the average Creek Chub PCB concentration decreased from 19.2 ppm in pre-remediation samples to 8.9 ppm in post-remediation samples (using the original 4-peak November 2002 results) or 7.7 ppm (using the revised 5-peak November 2002 results). For Green Sunfish samples collected at Country Club Road (3 miles downstream), the average PCB concentration decreased from 3.56 ppm in pre-remediation samples to 1.42 ppm in post-remediation samples (using the original 4-peak November 2002 results) or 1.28 ppm (using the revised 5-peak November 2002 results).

Other decreases in whole fish PCB concentrations from the pre-remediation period to the post- remediation period were not statistically significant. These analyses included Creek Chub samples collected at Country Club Road, as well as whole fish samples collected at locations farther downstream of Lemon Lane Landfill-Green Sunfish and White Sucker at Fluckmill Road (10 miles downstream) and Longear Sunfish at Strain Ridge Road (20 miles downstream).

The results in Table 3 also show that substituting the revised 5-peak November 2002 PCB results for the original 4-peak results does not affect the outcome of the statistical analyses. On average, the revised 5-peak results are slightly lower than the original 4-peak results, as indicated by the slight decreases in the average post-remediation PCB concentrations listed in Table 3. The original 4-peak results and the revised 5-peak results produce identical statistical conclusions.

Table 4 summarizes the results of pre-remediation vs. post-remediation statistical analyses for fillet samples. The differences between pre-remediation and post-remediation fillet PCB concentrations were not statistically significant in any of these analyses. At Country Club Road (3 miles downstream), White Sucker fillet concentrations increased slightly. Minor decreases in PCB concentrations were observed for Rock Bass fillet samples collected at Fluckmill Road (10 miles downstream) and for Shorthead Redhorse fillet samples collected at Strain Ridge Road (20 miles downstream).

The results of these pre-remediation and post-remediation comparisons suggest that observable impacts of PCB discharges from Lemon Lane Landfill are limited to sampling locations closest to the landfill. Statistically significant differences in whole fish PCB concentrations were seen only at the ICS/Quarry Springs and Country Club Road sampling locations, 1.3 and 3 miles downstream. The results for Creek Chub samples collected at these locations further suggest that the impact of continuing PCB discharges from Lemon Lane Landfill decreases rapidly as downstream distance increases. For pre-remediation samples, average Creek Chub PCB concentrations were about 12 times higher at the Allen Street location (1.3 miles downstream) than at the Country Club Road location (3 miles downstream). For post- remediation samples, average Creek Chub PCB concentrations were about 8 times higher at the Allen Street location than at the Country Club Road location. Data for additional fish types and downstream sampling locations would be needed to confirm this observation.

5.2 POST-REMEDIATION POST-REMEDIATION RESULTS

Table 5 summarizes the results of post-remediation vs. post-remediation stanstical analyses. Post-remediation samples were collected in November 2000, just after completion of the excavation, consolidation, and disposal RA, and in November 2002. These analyses cover whole fish samples collected at three sampling locations-Creek Chub samples collected at Country Club Road (3 miles downstream) and at Fluckmill Road (10 miles downstream), and Longear Sunfish samples collected at Strain Ridge Road (20 miles downstream). Fillet samples were collected in November 2000 but not in November 2002.

Average PCB concentrations for Creek Chub samples collected at Country Club Road decreased from 1.68 ppm in the November 2000 samples to 0.79 ppm using the original 4-peak November 2002 results or 0.78 ppm using the revised 5-peak November 2002 results. Although this decrease appears large (a factor of two), it was not statistically significant because the sample results during each event were highly v ariable. Concentrations for Creek Chub samples collected in November 2000 ranged from 0.8 to 2.4 ppm (a factor of three), while concentrations for the November 2002 samples ranged from 0.35 to 1.20 ppm (a factor slightly greater than three).

At the remaining two locations, average whole fish PCB concentrations showed a slight increase between the November 2000 and November 2002 sampling events. These increases were not statistically significant.

5.3 TREND ANALYSES

Table 6 summarizes trend analysis results. Three trend analyses were conducted for whole fish samples. These included Creek Chub samples collected at Country Club Road (3 miles downstream), White Sucker samples collected at Fluckmill Road (10 miles downstream), and Longear Sunfish samples collected at Strain Ridge Road (20 miles downstream). A trend analysis was also conducted for Rock Bass fillet samples collected at Fluckmill Road. Trend analyses were conducted using the original 4-peak PCB results for November 2002 because there were only minor differences between these results and the revised 5-peak results.

All of the data used in the trend analyses was also evaluated using a two-sample pre-remediation vs. post- remediation statistical analysis (see Section 5.1 and Tables 3 and 4). In these two-sample comparisons, none of the observed decreases in PCB concentrations was statistically significant. Three trend analyses confinned this findingÄthat is, a decreasing trend in PCB concentrations was observed, but the trend was not statistically significant.

In contrast, the trend analysis for Creek Chub samples collected at Country Club Road showed a small but statistically significant decreasing trend. Sen's Slope Estimator indicates a 0.16 ppm-per-year decrease for these results. The difference in the trend analysis and two-sample test results can be partially explained by the test methods used. A t-test was used for the two-sample analysis, and this method is sensitive to the variability in the sample results. If samples within either the pre-remediation group or the post-remediation group are highly variable, differences betwec ~ the two groups will be harder to detect. The M-K test used for trend analysis considers only whether one sample result is higher or lower than another and is much less sensitive to such variability.

6.0 CONCLUSIONS AND RECOMMENDATIONS

Viacom collected fish samples at several locations downstream of Lemon Lane Landfill during four sampling events between November 1996 and November 2002. As discussed in Section 4.O, over 300 samples have been collected and analyzed for PCB concentrations, and approximately half of these samples have been included in statistical analyses. The statistical analyses presented in Section 5.0 address six different fish types (Creek Chub, Green Sunfish, Longear Sunfish, Rock Bass, Shorthead Redhorse, and White Sucker) collected at four sampling locations in Clear Creek. These locations include Allen Street ( 1.3 miles downstream of Lemon Lane Landfill), Country Club Road (3 miles downstream), Fluckmill Road (10 miles downstream), and Strain Ridge Road (20 miles downstream).

Statistical analyses (both two-sample comparisons and trend nalyses) were used to compare the results of fish samples collected both before and after the RA was implemented during 2000 at Lemon Lane Landfill. The primary purpose of these comparisons was to evaluate whether remedial measures were effective in reducing fish tissue PCB concentrations. The results of the statistical analyses indicated that some reductions have been observed. Statistically significant decreases between pre-remediation samples (November 1996 and November 1999) and post-remediation samples (November 2000 and November 2002) were observed only at the two sampling locations closest to the landfill:

• Average PCB concentrations in Creek Chub (whole fish) samples collected at Allen Street decreased from 19.2 ppm to 8.9 ppm (original 4-peak November 2002 results).
• Average PCB concentrations in Green Sunfish (whole fish) samples collected at C ountry Club Road decreased from 3.56 ppm to 1.42 ppm (original 4-peak November 2002 results).
• A trend analysis for Creel: Chub (whole fish) samples collected at Country Club Road showed a small but significant decrease of 0.16 ppm per year.

No statistically significant changes in PCB concentrations were observed for other pre-remediation vs. post-remediation comparisons involving whole fish or fillet samples or for any of the post-remediation vs. post remediation comparisons that were conducted.

Table 7 presents recommendations concerning the fish types and sampling locations that should be considered for future fish sampling events at Lemon Lane Landfill. These recommendations focus on two areas. First, the recommendations identify data that should be collected to extend the statistical analyses summarized in this report. For example, Rock Bass fillet samples were collected at Fluckmill Road during the November 1996, November 1999, and November 2000 sampling events in numbers sufficient to support statistical analyses. Comparable samples should be collected during future sampling events. Second, the recommendations identify new data that could be combined with existing results to support additional statistical analyses. For example, Crayfish samples were collected at the Allen Street, Country Club Road, and Fluckmill Road locations only during the November 1996 sampling event. Collecting Crayfish samples from these locations during future events would allow statistical comparison of pre- remediation and post-remediation PCB concentrations.

Finally. as noted in Section 4.1, all fish samples collected from Lemon Lane Landfill have been analyzed for PCB Aroclor concentrations using the GC method. Samples collected during future sampling events should be analyzed by the same method if they are to be used to extend the statistical analyses discussed in this report. Analysis of future samples for PCB congener and TEQ concentrations by high resolution GC/MS should be considered only as a supplement to, and not a substitute for, PCB Aroclor analysis.