Comments on Water Quality Issues at Lemon Lane, Neal's Landfill, and Bennett's Quarry

EPA Response 8-21-00

COPA
March 20, 2000

Mr. Thomas Alcamo
Bloomington Project Manager
Region V
USEPA
77 West Jackson Blvd.
Chicago, IL

Dear Mr. Alcamo,

I am summarizing some of the issues that the TAG consultants have raised on the data reports for the Bloomington Superfund sites. Although most of the observations and comments focus on Lemon Lane, these problems were first noticed as similar issues and questions for stream and spring samples at Neal’s Landfill, and the field reports for the site remediation at Bennett’s Quarry. We are concerned whether these issues are ongoing and pervasive in all of the work. The data sets which we have specifically examined are a) the Admin Record Update 7/29/99 for Water Sampling at Neal’s 2/24 – 5/4, 1999, b) the Summary of Lemon Lane Landfill Analytical Results for Split Samples Collected by EarthTech, February, 1997, and c) the 1/19/99 Conduit Investigation Progress Report for July to October, 1998. Where a problem is described, we also try to give an example from the data sets. We also provide a specific question for which we need a response that addresses the problem and help provide a resolution.

1. Reporting results from the wrong analytical method. The method cited in the Quality Assurance Project Plan (QAPP) for the analysis of PCBs is SW-846 Method 8082, which is correct. However, the laboratory reports for all of the LL data cite Method 8081 as having been used. Method 8081 is only for Organochlorine Pesticides such as chlordane, and not PCBs. Although the two methods are similar, I cannot imagine that the laboratory could make such an error. I suspect that this is a typographical error, but it is so obvious a question that a data validator should spot and verify. It is more an indication of poor attention to detail, and should have been checked by CBS to ensure that the correct method was used, and have revised reports prepared.

Could you confirm that analyses done by all laboratories doing work for the Bloomington Superfund sites were appropriate for polychlorinated biphenyls and explain why Heritage Laboratories was using an incorrect method?

2. Consistently using comparable methods. It is unclear whether all analytical programs are using the same methods. For example, at Lemon Lane CBS, EarthTech, and EPA had samples analyzed. The EPA laboratory would have used a contract laboratory program participant. They typically use different methods than used under RCRA. EarthTech submitted CERCLA analyte lists in its Summary Report (item b). CBS used Heritage and they use RCRA methods. EPA needs to confirm that the methods and quality control used are consistent and comparable. I raise this question because the CBS and EarthTech split samples yield widely varying results and we need to be sure that it was not caused by the analytical methods.

Could you provide evidence that the analytical methods used for the samples at the Bloomington Superfund sites were all consistent with the QAPP, that is used SW-846 Method 8080 and, if not, provide an analysis of the effect of any differences to ensure comparability of the data?

3. Lack of quality control data The QAPP calls for a number of quality control checks to be done along with the samples to demonstrate that the results are accurate within the project’s data quality requirements. The important quality control (QC) reports are the conti9nuing calibration verification (CCV) reports, the results of internal standard checks, method blank run, laboratory control sample checks, and matrix spike and duplicate runs. None of these are presented with the data (items a and c). It is impossible to say whether the data are usable.

Could you provide the quality control data for the analytical data for the site characterization at Neal’s Landfill, Bennett’s Quarry, and Lemon Lane Landfill, and direct all that all future reports contain this information?

4. Inadequate surrogate recovery information. The one piece of QC data that is supplied is the surrogate recovery for each sample (items a and c). However, the control limits are not given. Control limits are the range of acceptable recoveries that indicate that the extraction and analysis were done correctly. Some explanations. Surrogates are compounds that are similar in chemistry to the target compounds and will be affected the same by all processing steps. The surrogates are added to each sample at the beginning of processing to track the effectiveness of the extraction, concentration, and analyzing the target compounds. If the result shows that 100% of what was "spiked" into the sample at the beginning was recovered at the end, then it is very likely that 100% of the target compounds in the sample were also extracted and measured. Anything less, indicates either inefficiencies in extraction or losses during processing. What affects the surrogates also affects the targets, and low surrogate recovery indicates undermeasurement of the target compounds.

In spite of the attempts by the laboratory to minimize interferences with extraction (noted below as an additional questionable practice) surrogate recoveries are frequently quite low (see for example item c, samples LL1447 and LL1455). If they exceed the tolerance limits (high or low) in the method, the sample needs to be re-extracted and re-analyzed.

Surrogate recoveries exceeding 100% are also questionable. It is impossible to actually get more than was put in unless it was initially present, or a serious analytical error has occurred. When most of the data suggests recoveries are 50-60%, a recovery of 149% (item c, sample LL1453) suggests that the surrogate was spiked twice into the sample. Errors like this can erroneously qualify a result as acceptable. None of the ancillary detail is provided to allow determination of acceptable recoveries for the surrogate.

Although much of what is addressed in this issue appears to be covered by the request for the quality control reports, in order to assess the quality of the data, the acceptance limits are also needed. Could you direct that the Quality Assurance Project Plan that contains the acceptance limits for quality control results be immediately included in the information record at the Monroe County Public Library?

5. Discarding a critical portion of the sample. Evident in the reports for water samples in items a and c is the recurrence of a very questionable practice of only using a portion of the supplied sample (numerous examples in items a and c – any sample less than 1,000 ml.). Because the final volume used at the instrument is fixed, using more or less of the original sample in the first extraction step will change what ends up in the final volume. The method for PCBs calls for a 1,000-ml sample at the extraction step. From the field sheets, this is what has been supplied from the field. But, the laboratory has been using only a portion of the sample. This causes two problems. First, if the water sample is free of sediment, a smaller than required volume causes higher detection limits than required by the QAPP. This can miss low, but significant occurrences of PCBs. The laboratory is not correcting its detection limits to reflect the lower sample volume. Second, for samples with sediment, the laboratory appears to be letting the sediment settle out, and pouring off a "cleaner" portion and discarding the sediment. Considering the relationship of PCB transport sorbed on the sediment, any practice that reduces or eliminates the most critical portion of the sample results in a serious biasing of the result to lower values.

Could you explain the discrepancies among the volume of sample required for analysis in the Method, the volume available for analysis as indicated by field records, and the volume actually used in the analysis suggesting a discarding of a critical component, and the relatively low levels of PCBs detected?

6. Unnecessary dilutions. The laboratory also incorrectly uses a practice that is acceptable to bring concentrations within the range of calibration, but not to avoid the impact of sediment in the sample. Instead of discarding only15-20% of the sample, they frequently discard up to 95% for processing (item c samples LL1450 and LL1451), supposedly to reduce matrix interferences. This elevates detection limits from 0.1 ug/l to 2.0 ug/l. At this level many samples are run without the ability to detect significant concentrations of contaminants.

They ascribe this practice to a need to minimize interference. This practice is actually justifiable only when target concentrations are so great that they exceed the quantitation limit for the calibration standards. However, the laboratory cannot tell this will occur unless an initial analysis is done. This initial result should be presented to justify the subsequent dilution of the extract. This allows an evaluation of the surrogate recovery in the extraction step to determine if the processing was correctly done and verifies that concentrations exceeded calibration limits. The dilution is done on the extract, not on the original sample. Yet, data are presented on diluted samples with apparently good surrogate recoveries and low PCB levels indicating that a smaller sample was taken initially.

From the "dilution " problem noted here, and this discarding of "dirty water" in #5, it appears that the laboratory does not follow the proper execution of the extraction method. The method requires a total transfer of the supplied sample, plus a solvent rinse of the sample container that is added to the extraction vessel. Any debris carried into the extractant is removed in cleanup steps. To work with only "clean" water expedites sample processing, but falsifies the result. The only acceptable reason for using less than 1,000 ml is that insufficient sample was available (and then, detection levels have to be raised in the report). The field notes do not indicate any problem getting sufficient water. Sample LL1445 was taken from a 55-gallon drum and only 900n ml was used. LL1442 had three 55-gallon drums as a source, yet about 500 ml was used.

Another problem with using a smaller volume is that it distorts surrogate recovery data. Spiking the surrogate into a volume lower than required, artificially elevate recovery results. This falsely states the recovery and implies a higher quality of performance than was actually achieved.

Could you explain why the laboratory is diluting samples, which leads to elevated detection limits in cases where PCB concentrations are quite low and the motivation appears to be to limit sediment incorporation in the extracted volume?

7. Inadequate information to assess data problems. An indication of the severity of this problem of discarding a portion of the sample might be available if accurate measures of suspended solids were supplied for each sample, or if field sample data sheets showing the sample appearance were included with the laboratory sample reports, but they are not. Drilling activity field notes (which have many discrepancies, not the least of which are missing days and samples (item c LL1460-ll1484).

Could you direct that all reports on samples collected at Neal’s Landfill, Bennett’s Quarry, and Lemon Lane Landfill include sampling records showing purging data, field measurements, turbidity, specific conductivity, etc. for each sample be provided to the information repository?

8. Improper quantitation technique. The laboratory, under the guise of a legitimate confirmation-column check, is using two analyses and choosing the lower result to report (item a). This biases data to the low side and is an error. If calibration was correct for each method, the results of two analyses should be the same. Therefore the only legitimate use of two analytical results is to report both, and possibly claim that the average of the two is the "better" number. If the laboratory is actually calibrating two methods, i.e., two columns and two separate detectors, this is an extraordinary effort that requires a level of time and effort that can only be justified by and compensated with higher per sample charges. If CBS is compensating them for this type of analysis, I suspect that there is a collusion to bias the results to lower levels. This is also apparent in the discard of sediment, but may not come at a higher per sample cost because it actually reduces the cost of analyses by reducing the processing steps for the samples. The laboratory’s productivity goes up and that is its own reward. As with much of the environmental analytical process, it pays to do it poorly when poorly produces "better" results for the responsible party.

Could you verify that the quantitation technique used for these samples meets the requirement of the QAPP and is consistent with good analytical technique and EPA policy for these analytical methods?

9. Concerns about the usability of the 1989 Westinghouse High Flow Tracer Results, which appear as a supporting document to the proposed plan for Lemon Lane. The issue involves a recurring contention based on that study that Lemon Lane is drained by three major springs and under storm conditions, a few others. To our knowledge, this report does not, and cannot be used to support that contention. No quantification of the dye detections was possible with the methods used, nor was the data presented to support that bromide recoveries could support it either.

Could you or CBS provide an explanation of the data in that study which is being used to conclude that 98.9% of the site drainage is occurring through Illinois Central, Quarry, and ICG-1 springs?

10. Considering the questions on data quality for water and sediment samples raised above, data from past investigations of PCB releases may need review to confirm its accuracy and any conclusions resulting from the data. This includes the statements in the proposed plan that contends PCB releases have only occurred in Illinois Central, Quarry, and Slaughterhouse Springs.

Could EPA please provide the field notes, the data and quality control reports and any analysis of the usability of this data for public review and confirmation of its usability?

11. Because the EPA has decided that complete removal is not necessary to meet the objectives of the Superfund cleanup of the Bloomington sites, it is critically important to the community to understand and be convinced that the treatment plant will meet the goal of clean water from the sites. The assumptions provided for review are appreciated. However, the system design goes beyond these to the actual calculations and sizing of facility equipment. Unless we can review the actual design details, it is difficult to assure the community that the system is what is needed to do the job.

Could EPA please provide the community with the actual design plans for the treatment facility that show the development of equipment, capacities, monitoring, and system controls, which will meet the needs at Illinois Central Spring and the others which now, or may show evidence of contamination?

12. Finally, because the information repository at the Monroe County Public Library does not meet the standards for an Administrative Record under Superfund, a number of difficulties have arisen in accessing information necessary to meet the need of the community to understand the clean-up process at the Bloomington sites. In addition to the difficulty with locating specific documents due to the cataloging system, data is included in a very difficult form (if present, at all).

Could the EPA require all existing and future reports and data sets, including maps and drawings in CAD form be made available in electronic format?

We are preparing to meet with EPA and its consultants to discuss the issues addressing water treatment and monitoring. It is necessary that we have these questions of data quality, system design, and electronic data addressed in order to make progress in that area. Your immediate attention to these questions is appreciated.

Sincerely yours,

 

Michael R. Baker

COPA